Space Flight Dynamics

Michele Santarpia

First Edition - 04/05/2024

Come utilizzare questo materiale?
Innanzitutto, questi appunti non sono completi essendo nati per supportare il libro di appunti di Fabio Colella e
sono riferiti al corso di Space Flight Dynamics del 2023 tenuto dal professore Giancarmine Fasano. I primi tre
capitoli (quelli più soft, ovvero Introduction, Reference Frames e Two-Body Problem) sono incompleti dato che non
presentano tutte le informazioni ma potresti trovarne altre che non sono riportate nel Colella.
Inoltre, tutto il blocco appunti non è completo anche perché:
1. Non sono presenti immagini.
2. Non sono presenti tutte le formule.
3. Menziona poco GMAT.
Dal capitolo ’Orbital Maneuvers’ a ’Interplanetary Transfer’ sono presenti tutte le informazioni date dal professore
durante il corso. Il valore aggiunto di questo libro di appunti risiede nell’approfondimento di alcuni argomenti e
nella trattazione di altri non presenti nel testo di Colella.
Se vuoi utilizzare questi appunti, il mio consiglio personale è quello di studiare inizialmente dal Colella e magari
successivamente utilizzare questo libro di appunti per integrare e/o capire meglio certi argomenti. Un’alternativa
è quella di utilizzare questo libro di appunti ed i PDF ’SFD whiteboards’ e ’SFD slides’ per avere le immagini
e le equazioni mancanti in questo materiale. Infine, come alternativa, non posso non menzionare anche il blocco
appunti di Giuseppe Puleo e Clemente Tecchia (relativi anch’essi al corso del 2023) che personalmente non ho usato
ma sono convinto che siano altrettanto validi per affrontare questo esame. Il libro di appunti di Colella, Giuseppe
Puleo e Clemente Tecchia ed i PDF ’SFD whiteboards’ e ’SFD slides’ li trovi nella cartella alla quale puoi accedere
scansionando il QR code. Se ci sono degli errori, puoi segnalarli attraverso il Google Form presente nella medesima
cartella. Infine, se sei interessato, nel drive è presente anche l’estratto del mio libro di appunti relativo al corso del
secondo anno di Spacecraft Dynamics and Control pubblicato da PrincipiDi.
Contents
1 Reference Frames 2

2 Two-Body Problem (Keplerian Dynamics) 4

3 Orbital Maneuvers 12
3.1 Coplanar Maneuvers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.1.1 General Orbital Maneuver . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.1.2 Hohmann Transfer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.1.3 Hohmann-like Maneuvers for Elliptic Orbits . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.1.4 Bi-Elliptic Transfer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.1.5 Infinite Bi-Elliptic Transfer (also called Bi-Parabolic Transfer) . . . . . . . . . . . . . . . . . 18
3.1.6 Example: Tangential ∆v applied not at the perigee . . . . . . . . . . . . . . . . . . . . . . . . 19
3.1.7 Hohmann, Infinite Bi-Elliptic, Finite Bi-Elliptic comparison . . . . . . . . . . . . . . . . . . . 19
3.1.8 Bi-Elliptic Transfer Advantages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.1.9 Single-Impulse Strategy for Change of the Argument of Perigee Maneuver . . . . . . . . . . . 22
3.1.10 Two-Impulse Strategy for Change of the Argument of Perigee Maneuver . . . . . . . . . . . . 23
3.1.11 Example: Radial Impulse due to an Error in Attitude . . . . . . . . . . . . . . . . . . . . . . 24
3.2 Out-of-Plane Maneuvers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.2.1 Only Change of Inclination for Circular Orbits . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.2.2 Only Change of Inclination for Elliptic Orbits . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.2.3 Single Impulse Strategy for Change of RAAN and Inclination . . . . . . . . . . . . . . . . . . 26
3.2.4 Restricted Three-Impulse Plane Change Strategy for Change of RAAN and Inclination . . . 28
3.2.5 General Three-Impulse Plane Change for Change of RAAN and Inclination . . . . . . . . . . 30
3.2.6 Mixed Maneuvering Problem - Hohmann Transfer with Plane Change . . . . . . . . . . . . . 30
3.3 Launch Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.3.1 Initial Velocity due to the Earth Rotation Rate . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.4 Rendezvous Maneuvers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.4.1 GEO Repositioning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.4.2 Heliocentric Problem - Earth to Mars Transfer . . . . . . . . . . . . . . . . . . . . . . . . . . 43

4 Orbit Determination 44
4.1 1) On-Board Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.2 2) Ground-based Systems and 3) Space-based Systems . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.3 Introduction to IOD and POD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.4 Initial Orbit Determination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

5 Orbital Perturbations 52
5.1 Perturbations Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
5.2 Cowell’s Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
5.3 Relative Importance of Perturbations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
5.4 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
5.5 Special Perturbation Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
5.6 General Perturbation Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
5.7 1) Non-Sphericity Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
5.8 2) Drag . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
5.9 3) Third body effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
5.10 4) Solar Radiation Pressure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
5.11 Ground Track . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
5.12 Repetitive Orbits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99

6 Orbit Maintenance 104

6.1 Orbit Maintenance for LEO Satellites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
6.1.1 Drag Make-Up Maneuver . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
6.2 Orbit Maintenance for GEO Satellites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
6.2.1 Errors for GEO Orbits in terms of orbital parameters and effects on longitude and latitude . 108
 6.2.2 Sources of Errors (Perturbations) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
6.2.3 Maintenance Maneuvers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110

7 Relative Motion 114

7.1 Hill’s Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
7.2 Two-Impulse Rendezvous Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
7.3 Limits of Hill’s Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
7.4 Formation Flying . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130

8 Interplanetary Transfers 133

8.1 Earth to Mars Transfer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
8.1.1 Heliocentric Transfer Conic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
8.1.2 Departure Conic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
8.1.3 Arrival Conic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
8.2 Gravity Assists: Leading Side Fly-By and Trailing Side Fly-By . . . . . . . . . . . . . . . . . . . . . 140
8.3 Earth to Saturn Mission . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
Training Objectives
The course is aimed at introducing the methods of Space Flight Dynamics that are applied to real space systems.
Starting from the basic knowledge linked to two-body mechanics, several topics will be covered in depth, includ-
ing orbit perturbations analysis and propagation methods, orbital maneuvers, orbit maintenance approaches, and
interplanetary trajectories. Special emphasis will also be given to the study of relative dynamics in space and its
application to distributed space systems, and to autonomous rendezvous and docking in missions such as on orbit
servicing and active debris removal.

Syllabus Program
1. Orbital maneuvers and launch constraints [1 CFU].
2. Fundamentals of orbit determination and estimation, introduction to space surveillance [1 CFU].
3. Orbit perturbations analysis [3 CFU]:
• General and special perturbation techniques. Orbit propagation methods and tools.
• Perturbation effects on different classes of Earth orbits.
4. Orbit maintenance for LEO and GEO satellites [1 CFU].
5. Relative motion in space [2 CFU]:
• Hill’s equations and advanced mathematical models.
• Formation flying, on orbit monitoring, rendezvous and docking.
• Spaceborne collision avoidance strategies.
6. Fundamentals of interplanetary trajectories [1 CFU].
Space Flight Dynamics Michele Santarpia 1

Introduction
Orbital Dynamics is separated from Attitude Dynamics. Orbital Dynamics refers to the motion of the CoM of the
object while Attitude Dynamics refers to the motion around the CoM. In some problems, it is important to consider
the coupling between the Orbital Dynamics and the Attitude Dynamics.
In the 2-body problem, we consider a primary body with a mass M and a body with a mass m characterized by a
spherical symmetry in terms of mass distribution. What we study is the motion of the mass m around the primary
body M, where the m is much lower with respect to M.
Impulsive approximation means the concentration of the thrust in a single time instant. It can be acceptable
for orbital maneuvers if the thrust over weight (T/W) is significant which is true in the case of chemical propulsion;
the same is not true in case of the electrical propulsion. Most of the discussion will be based on the impulsive
approximation: so, in the course, we will talk about the impulsive maneuver or a sequence of impulsive maneuvers
applied to the spacecraft. There are different types of orbital maneuvers: in-plane (same orbital plane), out-of-plane
(changing in the orbital plane) and mixed ones.
In order to reach a given orbit, the launch site has to be properly selected; the optimal moment to launch is when
the launch site crosses the orbital plane due to the rotation of the Earth, in order to end up in the right plane and
in order to have the right phasing i.e. the correct angular position along the orbit: this means that the launch has
to happen in a certain launch window.
How happens the orbital transfer of the Soyuz to reach the ISS? Firstly, we have the rocket launch, then the insertion
orbit, Hohmann transfer, phasing orbit, bi-elliptic transfer and, finally, station orbit.
Passive Safety Ellipses is the natural dynamics in order to avoid collisions in formation flyings like TanDEM-X
or in the case of Soyuz-ISS.
3rd body effect is the effect due to the gravity forces of the Sun and the Moon in the case in which the study of
our problem is a satellite orbiting around the Earth.
Space Flight Dynamics Michele Santarpia 2

1 Reference Frames
Let us consider the motion of the Earth around the Sun: this orbit is called ecliptic. The rotation axis of the Earth
is not orthogonal to the ecliptic plane and it is inclined w.r.t. to the normal of a constant angle equal to 23.44deg.
The main consequence of this angle is the changing of the seasons on the Earth. Let us consider the intersection
between the ecliptic and equatorial planes: this is creating the so-called line of equinoxes which identifies the
two equinoxes i.e. autumn equinox (≈ 22/9) and spring equinox (≈ 21/3). By considering the normal line to the
equinox line, we obtain the winter (≈ 21/12) and summer solstices (≈ 21/6). In this way, it is possible to define
the Heliocentric Inertial reference frame i.e. a reference frame which is centered in the Sun whose directions
are fixed w.r.t. to the stars.
The directions will not be really constant w.r.t. to the fixed stars due to the phenomenon called precession of the
equinoxes: this is due to the precession motion of the Earth’s axis around the normal to the ecliptic plane. For this
reason, the 1st point of Ares will move. In order to fix the directions, it is possible to consider the direction from
the Sun to the 1st point of Ares at a specific date: generally, it is used the J2000 epoch (Jan 01 2000 12:00:00.000
TDB). The frame is treated as stationary at the J2000 epoch and is not rotating. Or we can consider the definition
of the reference frame at a current date (true of date definition). The true reference frames include nutation and
precession effects. It is important to consider also the nutation motion of the Earth’s rotation axis. So, for an
accurate definition of the reference frame, all these motions have to be taken into account: for this reason, in our
J2000 definition, it is important to consider these contributions.
Also for Earth-Centered Inertial (ECI) reference frame, the inertiality has the same approximation done
before: the directions are not fixed due to precession of the equinoxes. For this reason, it is used a true of date
approach considering the position of the equinoxes at a specific date which allows to have a not ambiguous definition.
In GMAT, the ECI frame is called MJ2000Eq.
We can have J2000 for the Earth where the Z-axis is in the direction of the vector normal to the mean equatorial
plane at Julian year 2000.0, pointing towards the Northern Hemisphere, the X-axis is in the direction of the vector
pointing from the center of the Earth to the mean vernal equinox at Julian year 2000.0 while the Y-axis is the
direction of the vector perpendicular to the x- and z-axes, forming a right-handed coordinate system. The origin is
in the center of the Earth.
It is possible to see how the right ascension and declination of the Sun vary during one year. The right ascension
of the Sun changes with a rate of change which is not constant since the trajectory is not a circle in this relative
problem and αs is not the anomaly since we are considering the projection of the Sun in the equatorial plane.
Typically, in our discussion, it is possible to neglect this detail and assuming a rate of change of the right ascension
which is practically constant and equal to:
360◦
α̇s ≈ = 0.98◦ /day
365days

Let us consider the Heliocentric axes in red while Geocentric axes in black: they have the x axis in common.
The Earth Centered Earth Fixed (ECEF) reference frame is not an inertial reference frame since it is
rotating and it follows the motion of the Earth (axes fixed w.r.t. the Earth). This reference frame is also called
International Terrestrial Reference Frame (ITRF).
Let us evaluate the Earth rotation rate with respect to the stars (not respect to the Sun) i.e. the sidereal rotation
rate, where the period of rotation is 23h56min. Additional four minutes are needed to complete the rotation w.r.t.
to the Sun due to an additional angle caused by the precession of the Sun.
In order to pass from one to another reference frame, we need to use rotation matrices. So, in order to pass from
ECI to ECEF, we have that:

rECEF = M ECI→ECEF rECI = M 3 (αGR )rECI

This relation is true for position vectors, but not for velocity vectors in which we have to consider also the term
related to the rotation of ECEF w.r.t. ECI frame:

v ECI = M ECEF →ECI v ECEF + ΩL × rECEF

Space Flight Dynamics Michele Santarpia 3

If there are accelerations, we need to consider also them since they are creating apparent forces in non-inertial
frames.
In case of Topocentric Horizon Reference Frame, the origin is not in the center of the Earth but at a specific
location on the surface of the Earth identified by the parallel and the meridian passing through the point identified
on the Earth. Let us consider the spherical Earth as assumption. The direction from the Earth’s center toward
the point identified on the Earth is the Local Vertical. So, the reference frame has the origin in a given position on
the Earth and with axes called SEU (South, East, Up) which is a right-handed reference frame. The site can be
identified through the longitude λ and latitude ϕ (subscript s stands for site).
Note: a point identified through a parallel and a meridian is described through the right ascension α and the
declination δ in the ECI frame, while through the longitude λ and the latitude ϕ in the ECEF frame.
Let us consider a point on the surface of the Earth: let us take the local horizontal and then the local vertical. Since
the Earth is not a sphere, it is possible to see how the local vertical does not pass through the center of the Earth.
This allows us to define two different definitions of the latitude. The blue angle represents the geodetic latitude,
while the green angle the geocentric latitude. There are some relations to pass from geocentric to geodetic latitude.
In the topocentric reference frame, it is considered the geodetic latitude.
If we want to pass from ECEF to the Topocentric Horizon Reference Frame, we need to consider two different types
of rotations: one rotation of type 3 and one rotation of type 2.

M ECEF →T opocentricHorizon = M 2 (π/2 − ϕs )M 3 (λs )

The matrix M 3 (λs ) is:

 
cos λs sin λs 0
− sin λs cos λs 0
0 0 1

Why do we need Topocentric Horizon Reference Frame? In general, it is used for ground sensors (like ground-based
radars) which provide measurements in this frame.
Topocentric Equatorial reference frame: let us take one point on the surface of the Earth, in this case the
axes are not referred to the local horizon and the axes are also not fixed w.r.t. the Earth. Let us consider ECI as
geocentric frame and not ECEF, and so the point is moving in the ECI frame. The idea of topocentric equatorial
is to take the origin in the point and the axes the same of the ECI. Generally, it is used for telescopes since they
provide their measurements directly along inertial axes (because the origin of the measurements is in the telescope
but the directions are the fixed stars of the ones of the inertial frame since the telescope recognizes the star field
and in the processing of the telescope there is a star tracker operation).
Space Flight Dynamics Michele Santarpia 4

2 Two-Body Problem (Keplerian Dynamics)

Two-body problem: the main assumptions that we have are:
1. There are only two bodies: let us consider to have a primary body M and a secondary body m. We
want to study the motion of the mass m w.r.t. the mass M. The two-body problem is a particular one of the
N-body problem. In the two-body problem, we are neglecting the presence of the other bodies (practically
speaking, the presence of the Sun if we consider an orbiting satellite around the Earth).

The N-body problem equation is:

N  ′ 
µ X ri ri
r̈ + r = µi − µi = Gmi
r3 i=1
ri′3 ri3

In case of the presence of the other N bodies, at the 2nd member we have their contributions: in particular,
we have the sum of all the differences given by the gravitational force due to the 3rd body mi acting on the
satellite m and the gravitational force due to the 3rd body mi acting on the primary body M.
(a) The term µi (r′i /ri′3 ) represents the specific (per unit of mass) gravitational force of mi on the orbiting
body m.
(b) The term µi (ri /ri3 ) represents the specific (per unit of mass) gravitational force of mi on the primary
body M.
Remember that µi =Gmi . If we neglect the presence of the other bodies, the 2nd member is equal to 0.
µ
r̈ + r=0 µ = GM
r3
Using the 2-body approximation, we are neglecting the difference between the gravity of the 3rd bodies on the
satellite and the gravity of the 3rd bodies on the primary! Remember this important approximation, since it
is fundamental to remark that we are neglecting the difference between the gravity forces and not the
gravity forces of the 3rd bodies. The assumption of 2-bodies problem is more and more valid as the satellite
is closer to the primary and so the error due to the 3rd body effects is becoming larger and larger going
from LEO to GEO.
2. There are only gravitational forces expressed by Newton’s law:
Mm
F =G
r2
The law of universal gravitation is true only if we have point masses or if we have spherical symmetry in the
distribution of mass. The satellite is of course a point mass w.r.t. the Earth but the Earth is not a point
mass from the point of view of the satellite since the Earth has a radius of about 6000km and the satellite is
orbiting at an altitude of about hundreds of km for a LEO. So, from the point of view of the law of universal
gravitation, we are assuming that the Earth has spherical symmetry. But of course the Earth is not a sphere:
for this reason, the are also some non-sphericity effects that have been to take into account. We have to
take in mind that a body can be spherical from the geometric point of view but with not a symmetry in terms
of distribution of mass. For example, the Moon is not far from a sphere from the shape point of view, but
the distribution of mass is much less spherical w.r.t. the one of the Earth, so the gravitational field is much
stranger and the non-sphericity effects are more significant.
3. The mass M is much larger than the one of the orbiting object m (M >> m): thanks to this
reasonable assumption, we can define an inertial frame with origin in the center of the body with mass M (for
example, the Earth): it is inertial because we can neglect any acceleration produced on the primary by the
orbiting body.
Kepler’s Problem In Kepler’s problem, we know the initial position r0 and the initial velocity v0 of the satellite
Space Flight Dynamics Michele Santarpia 5

and we would like to obtain the position r(t) and velocity v(t) in the inertial frame at any subsequent time t:
 µ
r̈ + 3 r = 0



 r


r(t0 ) = r0 → r(t), v(t) ∀t






ṙ) = v 0


Instead, in Lambert’s problem, as initial conditions we have the position vector in two time instants: Lambert’s
problem:
 µ
 r̈ + 3 r = 0



 r


 t1 , r(t1 ) = r1 → r(t), v(t) ∀t





t2 , r(t2 ) = r2


This problem is interesting if we have an orbital interception problem or orbital maneuvering problem.
1. For example, it is important in the case of a rendezvous problem where with a spacecraft we want to intercept
the station: for this reason, starting from a condition in which in a certain time instant t1 we are in a given
position r1 , we want to be in a given time instant t2 in another position in space r2 : we need to understand
what is the orbit/trajectory to reach r2 in t2 .
2. Another problem could be the one of orbital determination because, for example, we have on ground some
radars used to measure the positions of a satellite in two different time instants and we have to determine
which is its orbit.
In interplanetary transfers, we use massively the Lambert’s problem solutions.
The basic results of the two-body problem i.e. of the Keplerian dynamics are:
1. The (specific) angular momentum vector is constant (1st Kepler’s law):

h = r × v = constant

This is because in our problem we have a spherical mass in the origin, then the gravitational force is a central
one and so there is no torque: for this reason, the specific (per unit of mass) angular momentum vector
is constant.

In this course, we will always refer to specific quantities, so for this reason we will call the specific angu-
lar momentum vector simply as angular momentum vector.

The vector is constant in the inertial frame, both in intensity and direction. Since the direction of the
angular momentum vector is constant, this means that the motion is planar: for this reason, we can define
the orbital plane.
2. The mechanical energy is constant since the norm of the angular momentum vector is constant (2nd
Kepler’s law):

v2 µ
ε= − + c = constant
2 r
We have to remember that the potential energy is given by − µr + c (since we have a spherical mass) and it is
defined up to an arbitrary additive constant which is set to be equal to 0 at the infinite. As we move along an
orbit, the mechanical energy is not changing (and this is because we are only considering gravity which is a
conservative force, there is no reason why the energy has to change along the orbit). All the trajectories that
are solution of the 2-body problem are conic sections (we obtain them by the intersection of an infinite cone
with a plane):
Space Flight Dynamics Michele Santarpia 6

(a) In case of an ellipse the mechanical energy is lower than 0 (and eccentricity e<1, while for circular orbits
e=0):

v2 µ
ε= − <0
2 r
In this case, r cannot be infinite (the ellipse is a closed-trajectory) otherwise the energy will be larger
than 0. In this case, we cannot escape from the main body.
(b) In case of a parabola, the mechanical energy is 0 (and eccentricity e=1):

v2 µ
ε= − =0
2 r
This type of conic section has the minimum energy that enables us to escape the gravity of the central
body. We can reach the infinite but it means that we need to lose all the kinetic energy when we are at
infinite since ε is 0. The idea is that we arrive at infinite and as we reach it our velocity is 0: this is the
reason why the parabola has the minimum amount of energy to reach the infinite. We can define at any
given distance the escape velocity as:
r
2µ
v=
r

(c) In case of an hyperbola the mechanical energy is larger than 0 (and eccentricity e> 1):

v2 µ
ε= − >0
2 r
Using this type of trajectory, we can reach the infinite with some residual velocity since we have a residual
energy when r is infinite. At infinite, the velocity is called hyperbolic excess velocity and it will be:

v∞ 2
ε= >0
2

Parabola and Hyperbola are open-trajectories: if we consider them, the satellite is coming from infinite, then
it is passing close to the primary body and leaving it. So, initially and finally, there is only kinetic energy and,
in the middle, there is a continuous exchange of kinetic energy and potential energy. Practically, for us, the
meaning of reaching infinite is that we are so far from a body that we can consider it as not a primary body and
so we can connect the concept of infinite to the one of sphere of influence of a given body which is a volume
outside of which we will neglect the gravitational field of the primary and we consider other sources like the Sun in
our interplanetary mission/problems.
Definitions:
The eccentricity vector e is a constant vector in an inertial frame and it is important because the direction of the
eccentricity vector defines the closest point of approach between the satellite and the primary i.e. the periapsis:
v×h r
e= − = constant.
µ r
The furthest point, in the case of ellipse, is called apoapsis. In the case of parabola and hyperbola we do not have
apoapsis since we go at the infinite. In the plane of motion, if we consider the instantaneous position of the satellite
and the position vector associated to it, the angle between the instantaneous vector r and the eccentricity vector e
is called true anomaly ν(t). So, we can write the following relation:

r · e = re cos ν

We can write the equation of trajectory (valid for all the conic sections) that relates the norm of the position
vector r with true anomaly ν(t):
p
r = r(ν) =
1 + e cos ν(t)
Space Flight Dynamics Michele Santarpia 7

where p is called semilatus rectum which is constant and it is the norm of the position vector when ν = 90◦ . The
only time-variant parameter is the true anomaly ν(t).
Elliptic Trajectory: in case of ellipse, the central body will be in one of the two focal points and not in the
geometric center of the ellipse. The eccentricity can be defined as the ratio between the semi-focal distance c
and the semi-major axis a:
c
e=
a
As the two focal points become more and more closer, the semi-focal distance reduces up to becoming equal to 0
in case of a circular orbit. To define the shape and the dimensions of the ellipse (so, of the elliptic orbit), we need
two independent parameters: for example, e and a.
The semilatus rectum is linked (one to one) to the norm of the angular momentum vector:
√
h= µp

From the 2nd Kepler’s law, we have that the radius spans equal areas in equal time:

h = r2 ν̇ = constant

is is the tangential unit vector, ir is the radial unit vector, iw is the normal (to the orbit plane) unit vector
(where w stands for whiteboard).
From Poisson’s Law:
dir
= Ω × ir
dt
where Ω is the instantaneous angular velocity which expresses the rotation of the unit vector ir .
Considering the motion of the satellite, we can evaluate the angular velocity of ir : basically, ir is rotating since
the satellite is moving and the motion of the satellite is expressed by the true anomaly and remembering that the
motion is planar we have that Ω = ν̇iw .
Finally:

h = r × v = r2 ν̇iw

h is a constant vector in the inertial frame, so the radius and true anomaly are changing but the product expressed
in h is not changing. The angular momentum vector direction is along iw and it defies the normal to the orbital
plane.
Mechanical Energy Equations: from the energy equation, we have that:

v2 µ µ
ε= − =−
2 r 2a
The semi-major axis a is linked (one to one) to the mechanical energy ε of the orbit. If we have a maneuver or a
perturbation (external forces) that does not change the semi-major axis a, the mechanical energy ε will not change,
or in other terms if we want to keep the mechanical energy constant ε we need to avoid to change the semi-major
axis a.
Parabola Semi-Major Axis Definition: for a parabola, the semi-major axis goes to infinite and this implies
that for a parabola we will not need two parameters to define the parabola but just one (typically the radius at
periapsis rp ).
Hyperbola Semi-Major Axis Definition: in this case, what happens to the semi-major axis?

v2 µ µ v2 µ
ε= − =− >0→ε= ∞ =− > 0 → −a
2 r 2a 2 2a
There are two different conventional approaches:
Space Flight Dynamics Michele Santarpia 8

1. The 1st one consists in defining an energy equation with positive sign.
2. The 2nd one consists in using the same equation defining a negative semi-major axis.
The one that we use is the 2nd one, so we will have -a when we define the geometry of the hyperbola (but, of course,
it is always defining a distance which is positive).
Let us consider elliptic orbits: we have to relate the norm of the velocity v to the true anomaly ν i.e. we have to
find an equation v=v(ν).
Let us start from the mechanical energy:

v2 µ µ
ε= − =−
2 r 2a
By using the equation of trajectory:
p
r=
1 + e cos ν
and by using the following relation for the semi-major axis:
p
a=
1 − e2
we get the velocity v as function of the true anomaly ν (valid for elliptic orbits):
r p
µ
v= 1 + e2 + 2e cos ν = f (ν)
p

For circular orbit (e=0, r=p=a), the velocity is constant and equal to:
r
µ
v=
r

Perifocal Reference Frame: let us consider the ECI reference frame and the intersection of the orbit plane
with the equatorial plane XY which defies the line of nodes. AN is the ascending node while DN is the
descending node.
We can define:
1. The Node Vector n (it is not an unit vector) in the equatorial plane which points towards the AN.

n = k̂ × h ||n|| = h sin i

2. The Right Ascension of the Ascending Node (RAAN), indicated as Ω:

n · Iˆ = n cos Ω
n · Iˆ
Ω = arccos Ω ∈ [0◦ , 360◦ ]
n
The problem in the equation of RAAN is that the arccos is defined between 0 and 180 deg, and so in order to
solve the ambiguity we can check the sign of ny . If ny is positive, RAAN is between 0 and 180 deg otherwise
it is between 180 and 360 deg.
3. The Inclination i:
 
h · k̂
i = arccos i ∈ [0◦ , 180◦ ]
h

To define how the orbital plane is placed, we need to know inclination i and RAAN (the direction of h is given
by only two angles).
Space Flight Dynamics Michele Santarpia 9

4. The Argument of Periapsis (AoP), indicated as ω, which is the angle between the node vector and the
eccentricity vector:
 
n·e
n · e = ne cos ω → ω = arccos ω ∈ [0◦ , 180◦ ]
ne
To solve the ambiguity we need to check the sign of ez (so, if it is in pointing in Northern or Southern
hemisphere, i.e. above or below the equator).
Through the angles i, ω and Ω, we know how the orbit is placed in the ECI reference frame. This gives us also the
possibility to define the Perifocal Reference Frame P̂ Q̂Ŵ , which is placed in the origin of the planet where P̂
is in the direction of e, Ŵ is in the direction of h and Q̂ will be in the orbital plane. This reference frame is inertial
since the center is still the center of the Earth and in our assumptions the directions are fixed w.r.t. the inertial
axes because the eccentricity vector and the angular momentum vector are constant in the inertial frame. So, now
we can define the components of the position vector r and velocity vector v in the perifocal frame. The good thing
is that the perifocal frame has two axes in the orbital plane, so position vector r and velocity vector v have only
two components. Since the unit vectors are constant directions, the time derivative is zero. Finally, we obtain the
position vector r and the velocity vector v as function of the true anomaly ν: so, in this way, we obtain that the
only variable that is changing in time is the true anomaly:

r = r cos ν P̂ + r sin ν Q̂
r
d d µ
v = (r cos ν)P̂ + (r sin ν)Q̂ → v = [− sin ν P̂ + (e + cos ν)Q̂] = f (ν) ν = ν(t)
dt dt p

Flight Path Angle Φ: let us consider our trajectory, the velocity vector and the angle between the velocity vector
and the local horizontal (the normal to r) called flight path angle ϕ.
 
π
r · v = rv cos − ϕ = rv sin ϕ
2
If we substitute r(ν), v(ν) and so on, we can obtain the flight path angle as function of e and ν i.e. ϕ = ϕ(e, ν).
Do not remember the expressions of these relations, but remember that they exist and that they are related to e
and ν.
When we are moving from periapsis to apoapsis, the flight path angle will be positive since the distance is increasing
(ṙ > 0, ϕ > 0). On the other side, from apoapsis to periapsis the flight path angle will be negative since the distance
is decreasing (ṙ < 0, ϕ < 0).
In the case of the ellipse (conic section in general), we have a symmetry with respect to the line of axes which means
that if we consider the same symmetric angle on the two sides, the angle is the same but with opposite sign.
Orbital Period τ : It makes sense only for the ellipse (time to go from periapsis to periapsis, or AN to AN, or for
any type of considered reference point):
s r
a3 2π µ
τ = 2π τ= n=
µ n a3

n is the mean motion.

If we consider the points A and B on our elliptical orbit, the time to fly from A to B and from B to A is not the
same. It takes less time to fly from B to A. Instead, the time to fly from periapsis to apoapsis and vice versa is the
same (half of the period). So, we need to understand how to solve the variation of the true anomaly with time.
Scope of Two-Body Problem: the scope of our problem is to derive position r(t) and velocity v(t) at any time
instant t in the inertial frame. The nice point of orbital parameters is that we can reformulate the problem where,
instead of having six scalar as unknowns which are all time-variable, we have only one scalar time variant parameter
ν(t) and the others five are constant (a,e,i,Ω,ω).
The information content in r(t) and v(t) is the same that we can have by knowing the six orbital parameters. We
can always pass in an not ambiguous way through a fully analytical formulation from position r(t)and velocity v(t)
to orbital parameters a,e,i,Ω,ω, ν(t).
Space Flight Dynamics Michele Santarpia 10

Parabola Trajectory: in the parabola, the periapsis is the only point in which r and v are orthogonal (for the
ellipse there is also apoapsis).
Hyperbola Trajectory: we are interested only in one arm of the hyperbola from the orbital dynamics point of
view. We consider an arc of circle starting from the asymptote and ending in the focal point where it is located
the central body, and let make a vertical line in order to define the periapsis. The semi-major axis is negative from
a dynamic point of view but positive from a geometric point of view, which is the distance from the center of the
periapsis.
The angle δ between asymptotes defies the rotation angle of the velocity starting from the beginning to the end
of the hyperbola. In this geometry, we arriving from infinite with a velocity v ∞ IN aligned with the direction of
the asymptote and we will fly away to infinite with another velocity v ∞ OU T and the only effect of flying over the
hyperbola will be rotation of the direction of the velocity by δ where the two velocities are equal in norm (we cannot
change the velocity in norm because we cannot change energy). In fact, we have that:

v∞ 2 (v ∞ IN )2 (v ∞ OU T )2 µ
ε= = = =− >0
2 2 2 2a
In the expression of the radius at perigee we have that a<0 and since e is larger then 1, (1-e) will be lower than 0
and so rp is larger than 0:

rp = a(1 − e)

Also, p is larger than 0 for the same reasons.

It could be useful to define the distance between the central body and the asymptote called ∆. The two triangles
are equal since they both the hypotenuse equal to c and an angle in common: for this reason, ∆ is equal to b and
it is typically called B-factor and it is important because it will define the approach geometry w.r.t. to the planet
and so where is the asymptote w.r.t. the planet. We will use hyperbola trajectory to escape from the Earth. From
the solar system perspective, an object with hyperbolic trajectory w.r.t. to the Sun is the case for example of an
interstellar asteroid with a fast passage (object with high energy in a solar-centered perspective, it leaves the solar
system in short time).
Elliptic Trajectory: we can relate the true anomaly ν(t) and the eccentric anomaly E(t) analytically (equa-
tions not to remember). Through E(t) we can bridge to the mean anomaly M(t) in analytical way, instead from
M(t) to E(t) we cannot find an analytical solution but we can proceed only numerically. Finally, from E(t) to ν(t)
we can proceed analytically.
Keplerian propagation: Kepler’s problem is the orbital propagation problem. Initially, we have the initial
conditions r0 and v0 , from which we evaluate all the six orbital parameters through the equations we have found
before. Now, what we can do is to propagate the true anomaly ν(t). From the initial true anomaly ν0 , we can
compute the initial eccentric anomaly E0 and the initial mean anomaly M0 analytically.

ν 0 → E 0 → M0

For each time different from 0 (also a propagation for t lower than 0 i.e. back-propagation), we can evaluate the
mean anomaly M(t):

M (t) = M0 + n(t − t0 ) ∀t

From M(t) we pass to E(t) numerically, and then we obtain the true anomaly ν(t) analytically:

M (t) → E(t) → ν(t)

Then, by knowing all the orbital parameters a,e,i,Ω,ω, ν(t), we can get r(t) and v(t) in the inertial frame using
analytically relations:

a, e, i, Ω, ω, ν(t) → r(t), v(t)

Under the Keplerian assumptions, we can consider that it is an exact propagation even if it is not a fully
analytical problem because passing from M(t) to E(t) requires the use of numerical solutions but this does not
Space Flight Dynamics Michele Santarpia 11

introduce a significant numerical approximation and it requires a not significant computational time (it is, we can
say, an instantaneous calculation).
In general, to solve numerically differential equations, we need to select a given scheme and the selection of the
time-step ∆t becomes important. The accuracy of the numerical solution is impacted by the selection of the
time-step.
In the Keplerian problem, it is not important what is the time-step: we are propagating along an orbit, we
are dividing, from a propagation point of view, the orbit in a certain number of points and the time-step is not
playing an important role because it is a propagation without making approximations (exact propagation): the
real approximation is that we are Keplerian!. So, the reason why what we calculate is not true is because our
assumptions for the Keplerian problem are not true.
Space Flight Dynamics Michele Santarpia 12

3 Orbital Maneuvers
Orbital Maneuvers: used to change the orbital parameters.
1. The 1st category consists in coplanar maneuvers (not changing the plane of motion) where we can have a
∆a, ∆e, ∆ω or ∆ν.
2. The 2nd category consists in out of plane maneuvers where we will have ∆Ω or ∆i;
3. Finally, we can have also mixed maneuvers.
The full general orbital maneuvering problem can be considered as a Lambert’s problem, where for example at a
current time we are in a certain position and after a given interval of time we want to be in another position. This
is called space interception problem, where we want to find a conic section i.e. a trajectory that satisfies our
problem.
Impulsive maneuvering assumption (τman = 0): we assume to have an infinite acceleration in a ∆t equal to 0.
Z
∆v = adt

What we obtain is a new velocity in the same time instant and in the same position since the application of the
maneuver is instantaneous.
If we change instantaneously our velocity, we are changing instantaneously our orbit, so we will have a new set of
orbital parameters since we pass from r, vold to r, vnew .
Impulsive maneuvering is an approximation because the application of a chemical thrust requires a given time
interval, and it happens a phenomenon called gravity losses. Once we consider not instantaneous thrusts and
so a certain time interval of application, the final ∆v will be lower with respect to the one computed with the
impulsive approximation. With impulsive approximation we are not conservative and so also for this reason we
need to introduce some margins. The loss of ∆v due to a non impulsive approximation is a function of the burn
duration w.r.t. the orbital period.
Maneuvering cost: the total cost in terms of ∆v (called ∆v budget) is given by the sum of the norm of each
applied ∆v (we consider the norm since the ∆v is a vector):
N
X
∆vtot = |∆v i |
i=1

As we can see from the Tsiolkovsky equation, using propellant with high specific impulse means having a less cost
in mass of the propellant with the same ∆v, since the exponential increases by increasing the specific impulse:
 
− ∆v
mp = m0 1 − e g0 Isp

There are some maneuvers in which it is not so important the propellant cost but instead, for example, it is very
important to apply very small changes in the velocity which means small ∆vs , so very fine maneuvers like the ones
needed in close proximity operations (for example, formation flyings) and so it becomes important the minimum
impulse bit of the propulsion system, and in the case of formation flyings we are talking in the order of mm/s.

3.1 Coplanar Maneuvers

Coplanar maneuver: let us consider the change of radius for a circular orbit (remember that the Hohmann
transfer is not the only option!). Let us consider two coplanar circular orbits with different radius, and consider
the flight directions on the two orbits, they are counterclockwise and so the angular momentum vector is coming
towards us.
If we want to change the flight direction on the 2nd orbit, we need to spend much more propellant so we need
an higher ∆v, and in this case we use a different strategy to obtain this solution because we need to perform a
change of the orbital plane from a dynamic point of view and so the inclination of the orbit is different (where one
is supplementary to the other one) and also RAAN is different (we are inverting the nodes).
Space Flight Dynamics Michele Santarpia 13

In the coplanar problem the two angular momentum vectors have the same direction and the orbital plane is the
same. Of course, the norm of the angular momentum vector is different since it depends on the radius of the orbit:
√ √
h = µp = µr
In order to perform this maneuver, a single impulse strategy is not possible. A single impulse strategy is possible
when the final and initial orbits have some intersection points. So, we need at least two impulses and so two ∆vs .
We need a transfer trajectory that must have intersections with both the orbits (initial and final): for example,
the green elliptical trajectory could be a possible solution since it intersects both the orbits. Of course, also an
hyperbola or parabola might be good.
Let us call transfer trajectory the orbit in green:
1. The periapsis has to be less or equal than the 1st radius.
2. The apoapsis, when existing, has to be above the 2nd orbit and so higher than the 2nd radius.
These two conditions tell us that there are two intersections points.
(
rptr ≤ r1
ratr ≥ r2

where r1 and r2 are the radius of the two circular orbits.

Let us assume that we choice our transfer trajectory whose shape and dimensions are identified by two parameters
i.e. semilatus rectum ptr (or semi-major axis atr ) and eccentricity etr .
The radius at periapsis and apoapsis have to be:
 ptr
rp = ≤ r1
 tr 1 + etr



 ptr
ratr = ≥ r2


1 − etr
When we set these conditions, we can work in the semilatus rectum-eccentricity plane (p-e plane) and these two
inequalities individuate the area of the plane where our transfer has to be. Of course, a given transfer trajectory is
individuated by a point in this plane. In particular, these inequalities are linear and so we have two straight lines
for the periapsis and apoapsis conditions, which are saying us that we need to be over those lines. So, finally, from
the combination of these two conditions we have understood that our transfer trajectory needs to be in a given
area. Let us individuate on this plane the dotted line corresponding to eccentricity equal to 1. So, in this portion
of plane, that we have individuated through the two lines, we can have both ellipses, parabola and hyperbola. If
we select a trajectory with e < 1 we have an ellipse (below e=1), if we are on e=1 we are on a parabola, while in
e>1 we are on a hyperbola (above e=1). So, any of these trajectories could be useful for our scope.

3.1.1 General Orbital Maneuver

Let us consider a generic solution: for example, let us take a general ellipse and let us compute this maneuver.
Once we decide our transfer trajectory, we know that it is described by a combination of ptr and etr . We need to
evaluate the two ∆vs . As we are flying on the 1st trajectory with velocity vAc1 , the 1st ∆v will be applied in the
intersection point A which brings us on the transfer trajectory with a velocity vAtr . The application of the ∆v 1
changes both norm and direction.
∆v 1 = v Atr − v Ac1
The velocity in norm on the circular orbit is simply to derive:
r
µ
vAc1 = = vc1
r1
The scope is to derive ∆v1 that can be computed by the Carnot theorem applied to the triangle individuated by
the velocities:
∆v12 = vA
2
c1
2
+ vA tr
− 2vAc1 vAtr cos ϕAtr
Space Flight Dynamics Michele Santarpia 14

So, in this equation, the velocity on the circular orbit is known. Now, we need to derive, at least in norm, the
velocity vAtr . In order to compute the velocity on the transfer trajectory in the point A, vAtr , we can use the
energy equation on the transfer trajectory assuming that it is an ellipse:

µ v2 µ
εtr = − = −
2atr 2 r
This last equation can be written in the point A, in order to extract the velocity in A:
s 
2
vA

µ µ 2 1
− = tr
− → vAtr = µ −
2atr 2 r1 r1 atr

where:
ptr
atr =
1 − e2tr
Now, we need to know the angle ϕ which is the flight path angle on the transfer ellipse in point A because the
velocity on the circular has the direction of the local horizontal. This angle, as we have seen before, is a function
of the eccentricity of the transfer ellipse and of the true anomaly on the transfer ellipse:

ϕAtr = f (etr , νAtr )

So, we need the true anomaly that we can obtain through the equation of the trajectory:
p
r=
1 + e cos ν
Let us apply this equation on the point A of the transfer ellipse:
ptr
r1 = → νAtr
1 + etr cos νAtr
We have a cosine of this angle, if we invert the relation, we obtain an angle between 0 and 180 deg and so we get
two solutions and we need to use the one that corresponds to the right point (in this case, it is between 0 and 180
deg).
So, summarizing we have that:

νAtr → ϕAtr → ∆v1

Finally, we can say that in general the orbital maneuvering problem implies a change of norm and of direction. The
change in norm is obvious because we need to have a change in the energy since we need to have a change in the
semi-major axis.

∆v → ∆ε > 0 ∆a > 0

In the point B we can plot the arrival velocity. As we can see in the geometry we have drawn, the orbit in green
has an higher energy, so we need to slow down the velocity and rotate to arrive in the orbit in red. In general, we
need to have a change of norm and angle. The derivation is the same as before.
Finally, the total cost in ∆vtot is given by:

∆vtot = |∆v1 | + |∆v2 |

Maneuvering time: we can analyze the time needed to perform this maneuver. It is evident that if we fly on the
trajectory we have drawn, it will be faster. How to compute the time? We need to know νAtr , through which we
evaluate EAtr and then MAtr . When we make these passages we are using the eccentricity of the transfer ellipse.
On the 2nd we can do the same: νBtr , EBtr and then MBtr . So, we can evaluate ∆M = MBtr − MAtr . Since the
mean anomaly is a linear function of time we can get the maneuvering time τman :
MBtr − MAtr
τman =
ntr
Space Flight Dynamics Michele Santarpia 15

where ntr is the mean motion of the transfer ellipse:

r
µ
ntr =
a3tr

We can play this game with hyperbola which will be even faster. In this problem, we have two degrees of freedom
since we are playing with two parameters: we are setting p and e of the transfer trajectory and then we get the
time as consequence.
It is evident that flying on the previous general transfer orbit we will be faster than an Hohmann transfer. In fact,
in case of Hohmann, the maneuvering time is fixed once we have the initial and final orbits. Instead, in the general
case, if we increase the energy of the transfer orbit (i.e. the semi-major axis) we can obtain a lower maneuvering
time (even if it means to have a less efficient maneuver).

3.1.2 Hohmann Transfer

Hohmann transfer: it is a particular case (simplified version) of the general transfer we have just seen.
What is better? Hohmann or the general case we have drawn?
Strategy: in the Hohmann transfer, we apply two impulses but we are not applying any change of direction of the
velocity v, so we apply two tangential impulses. It is clear that if we use our propellant only to change energy not
to rotate, it will be more efficient because we are not spending energy to rotate but only to change the energy of
the orbit.
In the diagram we have drawn in the (p,e) plane, there is a representation of Hohmann which is the point in red
given by the intersection between the two curves since in this way the perigee of the transfer is equal to r1 but also
the apogee is equal to r2 , we are exactly on the boundaries/limit conditions:
(
rptr = r1
ratr = r2

The transfer trajectory is tangent to the initial and the final one. So, the idea is to jump from initial circular orbit
to the transfer ellipse, we need to apply a tangential ∆v1 in the direction of the velocity and then as we arrive at
the apogee (which as construction is on the final orbit), we will apply the 2nd impulse ∆v2 to circularize the orbit.
If we take the transfer ellipse, we have that:
r1 + r2
2atr = r1 + r2 → atr =
2
So, once we have our initial and final orbit we will have our transfer trajectory.
The Hohmann maneuvering time is the half of the period of the transfer ellipse:
s
a3tr
τman = π
µ

As we can see, the maneuvering time depends on the ellipse, which depends on the initial and final trajectory. So,
once we know initial and final orbits we cannot change the maneuvering time which is fixed.

∆v1 = vptr − vc1

The velocity at any distance from the center (and we are interested in the one at the perigee and apogee of the
transfer ellipse) can be written as before:
s 
v2

µ µ 2 1
ε=− = − →v= µ −
2a 2 r r a

The velocity depends on the instantaneous position r and on the semi-major axis a of the transfer ellipse, i.e.
practically speaking on the energy content of the orbit.
Space Flight Dynamics Michele Santarpia 16

In the example we have done, ∆v1 > 0.

If we do not give the 2nd ∆v, we will always stay on the transfer ellipse. So, we need to apply the 2nd:

∆v2 = vc2 − vatr

In the example we have done, to circularize the ellipse we will need ∆v2 > 0.
It is clear that vc2 is lower than vc1 (LEO to GEO, we are passing from 7km/s to 3km/s). But take in mind that
both ∆vs are positive, because for example we jump on the ellipse from 7 to 8km/s so higher energy orbit, then as
we move on the transfer ellipse we will naturally slow down just because we are converting kinetic into potential
and so we will arrive at the apogee at 2km/s where we will increase again to jump to 3km/s.
Even if our final velocity is lower, the final energy is higher because now we have higher semi-major axis (i.e. higher
radius in this case).
If we want to reduce the radius, the equations are exactly the same, now we need two negative and tangential ∆vs .
This type of solution is very efficient in terms of minimizing cost of ∆vtot . In particular, Hohmann transfer is the
most efficient two impulse strategy among the all two impulses strategies. The cost that we have to pay is in terms
of the maneuvering time τman that is fixed, once we know initial and final orbits (no degrees of freedom).
The question that we need to face now is: is the Hohmann transfer the most efficient also if we use a strategy with
more impulses? Only when the initial and the final orbits are very far from each other, we will find some solutions
are more efficient than Hohmann but we will pay in terms of time.

3.1.3 Hohmann-like Maneuvers for Elliptic Orbits

Hohmann-like maneuvers for elliptic orbits: let us consider Hohmann-like maneuvers for elliptic orbits. Using
this type of maneuver, we will change the semi-major axis and the eccentricity of the elliptic orbit. We start from
a1 , e1 , rp1 and ra1 . We can apply a tangential impulse at the perigee or apogee to change the semi-major axis and
the eccentricity without moving the perigee or apogee since we are changing only shape and dimensions. If we apply
this tangential ∆v at the perigee, we are raising the apogee of the orbit, instead if we apply this maneuver at the
apogee, we are raising the perigee. So, we can obtain a ∆ra or ∆rp . Looking at the equations of semi-major axis
and eccentricity linked with the radius at periapsis and radius at apoapsis, it is clear that if we want to obtain a
specific value of the semi-major axis, we can apply a tangential impulse and obtain a change in the perigee or the
apogee and we will have as consequence a change also in the eccentricity. If we want to obtain a specific value of
semi-major axis and also a specific value of eccentricity, we have to apply two tangential impulses (we will see this
in the next paragraph).
ra + rp ra − rp
a= e=
2 ra + rp

If we apply a tangential positive ∆v, we are increasing velocity and according to the energy equation we are
increasing the energy and the semi-major axis. When we give this positive tangential ∆v, the point in which
we apply the maneuver remains as perigee of the orbit: this is simply because at the perigee we have that
the position vector and the velocity vector are normal w.r.t. each other and if we apply a tangential ∆v, we are
changing only the norm, so they will still be orthogonal to each other. So, we jump to the red orbit where the final
result is that we have changed the semi-major axis and the eccentricity obtaining a2 and e2 (so, we have raised the
apogee: for this reason, there is a ∆ra ).
s   s  
2 1 2 1
∆v1 = vp2 − vp1 = µ − − µ −
rp1 a2 rp1 a1

where 2a2 = 2a1 + ∆ra . We can do the same operation at the apogee. So, finally, if we apply a ∆vp we have a ∆ra ,
instead if we apply a ∆va we have a ∆rp .
Change of both semi-major axis and eccentricity: let us consider a problem in which we want to change
both the semi-major axis and the eccentricity: we typically need two ∆vs in order to adjust the perigee and the
apogee and to obtain the desired values of semi-major axis and eccentricity. So, we start from the semi-major axis
a1 and the eccentricity e1 and we want to obtain the semi-major axis a2 and the eccentricity e2 . This problem can
Space Flight Dynamics Michele Santarpia 17

be formulate considering that we want to change apogee and perigee, so the problem is reformulating passing from
∆a and ∆e to ∆ra and a ∆rp . The problem is reformulate in this way because we know that:
ra + rp ra − rp
a= e=
2 ra + rp

Of course, remember that when we do this drawing in reality the geometric center of the red ellipse is not the same
of the black one, because what is the same is the focal point.
In this case, we need to raise both the apogee and the perigee. The inputs that we have are a1 , e1 , a2 and e2 . We
can derive the initial rp1 , ra1 and the final rp2 , ra2 where:
( (
rp1 = a1 (1 − e1 ) rp2 = a2 (1 − e2 )
ra1 = a1 (1 + e1 ) ra2 = a2 (1 + e2 )
Starting from the black ellipse, the idea is to apply a tangential input at the initial perigee and then we need to
apply another impulse at the final apogee.
s   s  
2 1 2 1
∆v1 = vptr − vp1 = µ − − µ −
rp1 atr rp1 a1
s   s  
2 1 2 1
∆v2 = va2 − vatr = µ − − µ −
ra2 a2 ra2 atr
We only need to write the expression of atr :
rp1 + ra2
atr =
2
If we want to change only the eccentricity, the problem is always the same: we have an initial and a final ellipse
and, for example, we can apply a single tangential impulse at perigee or apogee to obtain the desired value of the
eccentricity.
What we have obtained is generalization of Hohmann for elliptic orbits.
Linearized form of the equations: we have written for now exact equations and we can make some considerations
if we consider a linearized form of the equations: assume we have small changes/corrections, so the ∆ becomes
differential. So, a variation of velocity implies a variation of energy. Let us consider the energy equation and let us
apply the differential:
v2 µ µ µ 2a2
ε= − =− → dε = vdv = 2 da → da = vdv
2 r 2a 2a µ
The differential of µ/r is zero due to the impulsive approximation because we are not changing the position when
we apply the variation of velocity (dr=0). The relation we obtain says us that as we change the velocity we are
changing the semi-major axis.
What we have obtained are the linearized forms of the tangential ∆v at the perigee and at the apogee. We can
apply these equations if we want to derive ∆ra and ∆rp , but of course these equations are approximated w.r.t. the
exact equations. If the changes are small, we get the same results that we have from the exact equations.
Why we want to slightly apply small changes to the semi-major axis and eccentricity?
Typically, this will be useful within orbit maintenance operations where we need to adjust parameters and
not to really change the orbit. It could be apply because, for example, our initial injection has an error and we
have to correct it, or for perturbing effects that modify the desired orbit and so periodically we need to recover the
parameters that we want. For example, in GEO we may have a generation of the eccentricity due to the
Solar Radiation Pressure and we do not like this change in eccentricity since it affects the geostationarity of the
satellite (in order to have geostationary satellite, we need to have e=0).
Other small changes of the semi-major axis are related to the decay caused by Drag in LEO and so we need to
raise the semi-major axis periodically.
In LEO, in general, we like to have a precise value of eccentricity in order to have frozen orbits. This value of
eccentricity is not zero but it is a number in the order of 10−3 : it is related to the non-sphericity effects (we will
see later frozen orbits).
Space Flight Dynamics Michele Santarpia 18

3.1.4 Bi-Elliptic Transfer

Bi-Elliptic Transfer: three-∆vs strategy (tangential) to change of radius for circular orbits (in case of ISS): we
have initial and the final circular orbits. Three-∆vs means that we will need to use more than one transfer ellipse.
We apply a tangential ∆v1 in the initial point, where we increase the velocity and so the energy and semi-major
axis. Instead of arriving on the final orbit, we can decide to go beyond the final orbit: so, the apogee
of the 1st transfer ellipse in not on the final orbit! So, we generate the transfer ellipse called tr1 . At a given
moment, after half of the period on the transfer ellipse, we apply a 2nd impulse that is again tangential and applied
at the apogee: we can tune this 2nd impulse in order to be now with the opposite point on the final orbit, so we
jump over another transfer ellipse called tr2 . Finally, we can apply the final impulse to circularize the orbit, so with
the 3rd impulse we need to reduce the semi-major axis and so it will be negative and, in this way, we jump on the
final trajectory. This type of maneuver is characterized by the parameter rb i.e. the distance of the point B from
the planet. The bi-elliptic transfer is not a single transfer because there are infinite bi-elliptic transfers that
can be obtained by placing the B point where we like. The parameter rb is a degree of freedom of the
maneuver. The three impulses that we apply are:
1. ∆v1 > 0 to pass from c1 to tr1 .
2. ∆v2 > 0 to pass from tr1 to tr2 and it positive because we are always increasing the semi-major axis of the
orbit.
3. ∆v3 < 0 to pass from tr2 to c2 and it is negative because we have to slow down since we have too much energy.
We need to write atr1 and atr2 :
r1 + rb
atr1 =
2
rb + r2
atr2 =
2
Let us write the three ∆vs :
 s   r
 2 1 µ
∆v1 = vPtr1 − vc1 = µ − −


 >0



 r1 atr1 r1





 s   s  
2 1 2 1

∆v2 = vAtr2 − vAtr1 = µ − − µ − >0


 rb atr2 rb atr1





 r s  
µ 2 1


∆v = v − v = − µ − <0


 3
 c2 Ptr2
r2 r2 atr2

What is the reason why we perform this type of strategy?

It sounds pretty strange but, as we will see, sometimes it costs less than Hohmann.

3.1.5 Infinite Bi-Elliptic Transfer (also called Bi-Parabolic Transfer)

Infinite Bi-Elliptic Transfer: differently from Hohmann, the bi-elliptic transfer is not unique, but we can have
a number of infinite bi-elliptic transfers because we can move the B point where we like. We want to check what
happens to this maneuver by increasing rb . We increase ∆v1 to increase rb . Let us consider a limit case, where
rb goes to infinite and basically, bringing rb to infinite means that the 1st transfer ellipse becomes parabola. So,
we jump on this parabola, we arrive at infinite with 0 velocity (it will take infinite time), and also the 2nd transfer
ellipse will be a parabola (at infinite we jump on the 2nd parabola), so we will come from infinite and then, as we
arrive on the final orbit, we apply the ∆v3 to circularize the orbit. Let us consider the three ∆vs :

∆v1 = vPpar1 − vc1 > 0

∆v2 = v∞par2 − v∞par1 = 0 − 0 = 0

∆v3 = vc2 − vPpar2 < 0

Space Flight Dynamics Michele Santarpia 19

The velocity at the perigee of the parabola will be the escape velocity. Since we arrive at infinite with 0 velocity,
it takes no energy to change the parabola. So, the total cost will be:
∆vtot = ∆v1 + ∆v3 = f (r1 , r2 )
So, if we bring the 2nd point to infinite, the maneuver is like Hohmann i.e. it depends only on the initial and final
orbit. This is what we call an infinite bi-elliptic transfer or bi-parabolic transfer. This is interesting because
it is a limit case that we can compare with other strategies.

3.1.6 Example: Tangential ∆v applied not at the perigee

Example: tangential ∆v applied not at the perigee: if we apply a nominal ∆v at the perigee, we have only
a change of semi-major axis and eccentricity. Instead, what happens if we apply a tangential ∆v not at the perigee
but, for example, we command a ∆v, called ∆vreal , at 90deg of the true anomaly where r=p? Let us see how the
orbital parameters are changing due to this maneuver.
What happens to semi-major axis and eccentricity?
Let us consider the energy equation: of course, by increasing the velocity, we are increasing the energy and so the
semi-major axis. So, what we have is that: ∆v > 0 → ∆ε > 0 → ∆a > 0 but in this case the amount of ∆ε is
not the same as the one that we will obtain if we apply the maneuver at the perigee because in this case we are
applying the ∆v in a different point corresponding to a different value of −µ/r:
v2 µ µ
ε= − =−
2 r 2a
So, in the nominal case, the ∆v is applied at the perigee where we are faster, instead in this other case we are
giving the same ∆v when we are slower. So, we are increasing the energy but a little bit less w.r.t. to the nominal
one. The relation is not linear, the new kinetic energy is given by (v+∆v)2 /2 and v is the velocity where we give
the ∆v (v is now slower).
Let us consider the zoom in the point where we are applying the ∆v, we consider the initial velocity (in green), the
local horizontal (in black), r=p, here something does not change and it is the flight path angle because we give a
tangential ∆v. So, now we can compute our new angular momentum:
√
hnew = |r × v new | = rvnew cos ϕ → hnew = µpnew → pnew → pnew = anew (1 − e2new ) → enew
By knowing the semi-major axis (by the energy equation) and the semilatus rectum (evaluated through the formu-
lation above), we can get the eccentricity. So, we get a new value for semi-major axis and eccentricity.
What happens to inclination and RAAN?
This is an in-plane maneuver, so nothing change to the orbital plane, we do not expect any change.
What happens to true anomaly and AoP ?
We miss the ∆ω and ∆ν: we can apply the equation of the trajectory where we consider the time instant in which
we apply the maneuver: we have r=pold , so we apply the maneuver and we jump instantaneously on a new conic
section and we can evaluate who is νnew :
p pnew
r= → pold = → νnew → ∆ν
1 + e cos ν 1 + enew cos νnew
So, the TA will be not 90deg as before. If we have a ∆ν, this will be equal and opposite to ∆ω: this is because
if instantaneously by definition the TA is the angle w.r.t. to the perigee, if we were at 90deg, and now we are at
45deg, this means that the perigee has moved of 45deg.
∆ν = −∆ω

3.1.7 Hohmann, Infinite Bi-Elliptic, Finite Bi-Elliptic comparison

Hohmann, Infinite Bi-Elliptic, Finite Bi-Elliptic comparison: as we know, the total ∆v for Hohmann will
be a function of r1 and r2 :
∆vhoh = f (r1 , r2 )
Space Flight Dynamics Michele Santarpia 20

Then, the total ∆v of the infinite bi-elliptic it is again a function of r1 and r2 :

∆v∞−biell = f (r1 , r2 )
while the generic total ∆v for a bi-elliptic will be a function of r1 , r2 and rb (which is our tuning parameter):
∆vbiell = f (r1 , r2 , rb )
We can compare these different strategies by simply considering the analytical representation of the ∆vs and looking
at their trends. In particular, it is interesting to compare these strategies by considering as independent variable the
ratio between the final and the initial radius of the circular orbits given by R=r2 /r1 . Let us define also R∗ =rb /r1 .
When R∗ >R (not dimensional parameters), this means that rb > r2 , so the apogee of the first transfer ellipse is
beyond the 2nd circular orbit.
Modifying the equations by introducing R, we can consider this ratio:
∆vhoh
= f (R)
vc1
This is a not linear function,not monotonic and the maximum of the function is obtained when R=15.58. So,
this means that if the radius of the 2nd orbit is 15.58 times larger than the radius of the initial orbit, we obtain the
maximum cost in terms of ∆v for Hohmann. After R=15.58, the maneuver starts to become cheaper and cheaper
as we increase the radius of the 2nd orbit.
Let us consider the infinite bi-elliptic: if we plot the function ∆v∞−biell /vc1 , we can see how it comes from
higher values and it has an intersection with Hohmann. After this intersection, it will be below Hohmann and
we have an asymptotic trend. The intersection happens in R=11.94. This means that if the ratio is smaller than
11.94, Hohmann is better than the infinite bi-elliptic, if the ratio is higher of course the infinite bi-elliptic costs
less than Hohmann, so for very large transfers it is better/cheaper/more efficient to use the infinite bi-elliptic w.r.t.
Hohmann.
Let us consider on the same plot the finite/real bi-elliptic curves in which we are more interested in since they
are feasible. We have different curves depending on different R∗ . There is a region where the bi-elliptic will be
cheaper than Hohmann. After the 2nd intersection the curve will change the derivative and it will become more
expensive. This second intersection happens when R=R∗ . If R∗ =R, we have Hohmann since the 2nd point is on
the final orbit. The bi-elliptic can be cheaper only when the rb point is beyond the final orbit.
Final Conclusions:
1. If R ≤ 11.94 (LEO to GEO is around 6), Hohmann is the best solution. So, this means that when R ≤ 11.94,
Hohmann is not only the best solution in the two-impulses strategy category but also in the three-impulse
strategy category.
2. If R ≥ 11.94, the infinite bi-elliptic is better than Hohmann. In particular, we have two regions: when
11.94 ≤ R ≤ 15.58, it exists at least one bi-elliptic transfer that is better than Hohmann depending on R*
but the required R* may be too large.
3. If R ≥ 15.58, any bi-elliptic transfer with R∗ >R will be better than Hohmann.
Bi-elliptic Maneuvering Time: let us consider to have very large values of R (for example, larger than 12),
using a bi-elliptic we can save ∆v but we will pay in terms of maneuvering time. The maneuvering time for
a bi-elliptic will be given by the sum of the half of the period of the two transfer ellipses:
s s
a3tr1 a3tr2
τmanbi−ell = π +π
µ µ

where:
r1 + rb

atr1 =


 2

atr2 = rb + r2



2
Space Flight Dynamics Michele Santarpia 21

For Hohmann instead:

s
a3tr
τhoh = π
µ

The bi-elliptic is characterized by an higher maneuvering time and, of course, if we take rb larger the maneuvering
time will be characterized by an higher value.
Bi-Elliptic vs Interplanetary Transfer: for example, let us consider an interplanetary transfer in which, as
first approximation, we assume an heliocentric perspective, i.e. the Sun as primary body (generally we consider,
for example, for a mission from Earth to Mars, that we have three phases in which we have Earth, Sun and Mars
as primary bodies and the conic sections related to the Earth, Sun and Mars must be consistent, they must be
patched conics). Let us imagine to pass from a point to another one of the solar system: the Hohmann transfer
will take a long time (for example, Earth to Mars takes seven/eight months), of course bi-elliptic takes much more
time, so even if we save ∆v, the time will be too much high.

3.1.8 Bi-Elliptic Transfer Advantages

In general, we like to use the bi-elliptic transfer not to save ∆v but in order to have a degree of freedom for
the maneuver and this means that we want to make a fine tuning of the B-point to satisfy some velocity or
timing constraints. Not necessarily it must be rb beyond the final orbit: if it is below, ∆v3 will be again positive.
Let us understand what is the meaning related to the tuning of B point.
Let us imagine that we want to pass on the final orbit where there is a target which can be virtual or physical.
In case of a virtual target, we have some timing constraints where we want to arrive on the final orbit at a specific
position and time. Let us consider to have our target in red, and we are the green spacecraft. We want to arrive
on the final orbit in the black point but we also want to remove this difference in anomaly ∆ν w.r.t. to the target.
Let us make a comparison between an Hohmann and a bi-elliptic transfer:
1. If we use an Hohmann transfer, we do not solve the problem of phasing since it has a fixed maneuvering
time: so, after half of the period of the transfer ellipse we are on the orange position, but where is the target
on the final orbit when we arrive? The target will not be where we are, but it will also be characterized by
an higher ∆ν. In fact, we have that atrhoh is lower than r2 , so the period on the transfer ellipse τtrhoh is lower
than τ2 . So, half of the period on the ellipse is smaller than the half of the period on the circular orbit, so in
the time in which our satellite will fly half of the orbit on the Hohmann ellipse, the target will fly less than
half of the orbit, so the separation is increased and so we have an increasing of drift.
2. With a bi-elliptic transfer we can solve this problem because we have a free parameter rb to set, and so
we can impose a condition of timing. So, how to impose the condition to find the bi-elliptic that adjusts the
anomaly? Let us consider the maneuvering time of the bi-elliptic and the angle ∆ν as a positive quantity. To
impose this condition, we have to consider that in the time in which the green satellite flies the two half of
the transfer ellipses, the red satellite has to fly one orbit plus ∆ν:
s s s
a3tr1 a3tr2 r23
τmanbi−ell = π +π = (2π + ∆ν)
µ µ µ

If we check this condition, there is only one unknown which is rb and so we can compute rb which is contained
in atr1 and atr2 . So, finally, through a bi-elliptic transfer we can change the orbit and we can arrive in the
right time removing the difference in anomaly w.r.t. our target on the final orbit.
If we want to use an Hohmann strategy, since the two objects are on different orbits, we can wait until they are in
the correct relative position. So, if we use an Hohmann transfer, we need to wait and sometimes it could be a very
long time and also higher than the bi-elliptic maneuvering time. So, instead of waiting, we can use the bi-elliptic
strategy. Of course, closer are the two orbits, longer will be the waiting time since the angular velocity for the two
satellites will be very similar.
Bi-Elliptic Transfer used to minimize ∆v3 : another reason to use the bi-elliptic is because sometimes we want
to have a limited relative arrival velocity on the final orbit and close to the final one, so our driver is ∆v3 .
If we look at the expression of ∆v3 , it depends on rb , so we want to have a small value of ∆v3 . In this way, we
arrive with almost the correct speed and our final burn will be a small one for safety. Imposing ∆v3 , we can find
Space Flight Dynamics Michele Santarpia 22

rb . Of course, we cannot impose the timing since we can impose only one condition and so in this case we need to
wait for the correct relative position between the target satellite and the green satellite.

3.1.9 Single-Impulse Strategy for Change of the Argument of Perigee Maneuver

Let us introduce another co-planar maneuver characterized by only a change of the argument of perigee ∆ω! The
shape and dimensions must be the same, we want to rotate the ellipse in the plane of motion. The ∆v has to be
in the plane of motion (in-plane ∆v) but which type of ∆v do we need? If we change the norm of the velocity as
we know from the energy equation, we have a change in the semi-major axis but we do not want this variation. So,
for this reason, we cannot change the norm of the velocity! This means that we need a purely in-plane rotation
of the velocity vector. So, we cannot have tangential ∆v, so Hohmann-like strategies are not good. So, in order
to understand the maneuver, we have to see how many impulses we need, where we have to maneuver, and which
is the required ∆v.
Let us sketch the two lines of axes of the two trajectories and let us indicate the two perigee of the orbits. For the
symmetry of the problem, the two ellipses will have two intersections points and these two intersections points have
to lie on the bisector of the angle ∆ω between the line of axes.
Since we have these intersection points, it is possible to have this change in a single impulse (there is no maneuvering
time): as we reach the point, we apply the ∆v and instantaneously we jump on the other trajectory.
So, we have a 1-∆v maneuver with two possibilities, the point A close to the perigees and the point B close
to the apogees.
The questions are:
1. Why this type of maneuver is a pure rotation?
2. Which is the best option between point A and B (in terms of lower ∆v)?
Pure Rotation Maneuver: let us plot the velocity vectors in the point A and B for the two orbits, and so we
can sketch the two different ∆v s that we can give. Let us understand which are the involved true anomalies.
The norm of the velocity for a given trajectory, as we know, is a function of TA:
r p
µ
v= 1 + 2e cos ν + e2
p

Since we have the cos ν, remember that the cosine is an even function i.e. cos(α) = cos(−α), and the TA has a
symmetry w.r.t. the line of axes, if we consider over a generic ellipse the points in green and light blue characterized
by a value of the true anomaly which is equal and opposite, they have the same velocities.
For this reason, also in the case of our interest, since we are on two orbits with same eccentricity and semilatus
rectum and due to symmetry, vA1 and vA2 have the same norm. The same is true for vB1 and vB2 .
(
vA,1 = vA,2
vB,1 = vB,2

From a mathematical point of view this confirms that this maneuver is a pure rotation, so we are not changing the
norm.
Since it is a pure rotation, we need to understand the angle that we need to rotate. Let us zoom on A point and
let us consider the local horizontal (black line), the velocities vA1 and vA2 . Once we have the local horizontal, we
have the definition of two angles, i.e. the two flight path angles. Due to the symmetry, the flight path angles are
equal. Let us remember that the flight path angle is a function of the eccentricity e and true anomaly ν where the
dependence is in terms of the cos ν, so for this reason the two flight path angles are the same.

ϕ = ϕ(e, ν) → ϕA,1 = ϕA,2

So, the rotation angle is related to the flight path angles. The impulse that we need is:
 
θ
∆v = 2v sin θ = ϕA,1 + ϕA,2 = 2ϕA,1 = 2ϕA,2
2
Space Flight Dynamics Michele Santarpia 23

So, in norm ∆vA is:

∆vA = 2vA1 sin(ϕA1 ) = 2vA2 sin(ϕA2 )

We know everything i.e. the velocity (it is a function of TA), and we know the sine of the flight path angle (it is a
function of TA and eccentricity):

e sin ν ∆ω
sin ϕ = √ ν=
1 + e2 + 2e cos ν 2

Change of Perigee implies Change of TA: as we change the perigee, there is also a change in TA.

∆ω = −∆ν

This happens because, when we apply a single impulse, we are in the same position but the perigee has changed.
So, by definition, we have changed also the TA! But, conceptually, if we consider that our orbit is defined by five
constant parameters and one variable one, we are practically changing only one parameter of the orbit because only
one of the constant parameters has been modified.
Maneuvering in A or B is the same: the ∆vA and ∆vB that we found are exactly the same. In theory, we have
that sin(π + ∆ω/2) is negative but, since we are considering the norm of ∆v, a negative quantity has no physical
sense, so we consider the positive sign. So, in terms of ∆v we can maneuver in A or B with the same amount of
∆v. When we are in A we are faster, so the vectors have a larger norm but we have to rotate of a smaller angle,
instead, in point B we are slower but with a larger angle. Mathematically, the two effects compensate each other
and so it is the same.
Maneuvering cost dependencies: finally, we have that the maneuvering cost is given by the following relation.
r
µ ∆ω
∆v = ∆vA = ∆vB = 2e sin
p 2

1. The required ∆v is linearly related to the eccentricity, so the more eccentric is the orbit, the more expensive
is the change of perigee. If we have a near circular orbit, this change will be very cheap.
p
2. Then, we have also another term which is µ/p which has the dimensions of velocity (km/s). If the orbit is
highly elliptic, we cannot really talk about the average speed of the satellite, but intuitively from a physical
point of view this term can be considered as a quantity that will be relatively close to our linear velocity, so
it will be in the order of our inertial orbital velocity, so it will be more convenient for an high orbit to
change the perigee w.r.t. a lower orbit since we are faster (we need to rotate longer velocity vector).
3. The cost of the maneuver depends on how large is the change of the perigee ∆ω, so a larger change requires
a more expensive maneuver.
Example: let us consider a relatively high change of perigee:

∆ω = 50deg → sin(∆ω/2) = o(1)

a=∼ 11500km
e = 0.001 → o = (10−3 )

We get 5m/s to rotate the perigee, so the order of magnitude is of m/s.

3.1.10 Two-Impulse Strategy for Change of the Argument of Perigee Maneuver

The one-impulse strategy is not the only way to change the argument of perigee, i.e. to rotate the line of axes,
and it is also not the cheapest solution in terms of ∆vtot . Let us remember that the nice point of the one-impulse
strategy is that the maneuvering time is ideally zero (practically the maneuvering time is the burn time of our
thrusters). Let us consider a two-impulse strategy. It is characterized by:
1. Two Tangential Impulses: there is no rotation of the velocity vector.
2. Disadvantage: of course, the maneuvering time will be different from zero.
Space Flight Dynamics Michele Santarpia 24

3. Advantage: we can save some ∆vs (when we talk about tangential impulses we have the cheapest solution).
To jump from the green to the blue ellipse, we can leave the initial ellipse at a given position and we jump on a
transfer ellipse that will be tangent also to the final ellipse.
The transfer ellipse that we can build has the line of axes which is the bisector of the angle. We apply the
pure tangential ∆v, and we have to remember that when we apply a tangential ∆v on an ellipse in a point which
is not the perigee, we are rotating the line of axes.
So, we apply the ∆v1 > 0 and as we arrive in the opposite point i.e. when we are tangent to the second ellipse, we
slow down the satellite with a ∆v2 < 0 and instantaneously we jump on the blue trajectory.
Of course, there is a maneuvering time depending on the features of the transfer ellipse and on the true anomalies
where we maneuver.
With this strategy we can save ∆v and, in particular, we have that it is the half of the 1st strategy (50%).
r
∆v1−imp µ ∆ω
∆v2−imp = = e sin
2 p 2

3.1.11 Example: Radial Impulse due to an Error in Attitude

Let us consider that we are on an elliptic orbit, we want to apply a tangential ∆v at the perigee to raise the apogee
but we make a mistake in the attitude of the satellite, so the ∆v is applied in the nadir/radial direction. What is
the consequence on the orbit?
This is always a co-planar maneuver and all the in-plane parameters will change. One initial question could be: the
perigee will be always the point in which we apply the maneuver? No, the perigee changes! Whatever we do, our
final velocity will be not normal to the final position, so the flight path angle is not 0 anymore and so it cannot be
the perigee.
When we apply the ∆v, the velocity will change and this new velocity has a different norm. We know the velocity
at the perigee v p1 , then we apply the ∆v, and we can sketch v new . So, of this triangle we know everything: we
evaluate v new from Pythagoras, and from v new we can evaluate the new semi-major axis anew by energy equation:
2
vnew µ µ
ε= − =− → anew
2 rp1 2anew
Of course, since we are increasing the norm, the semi-major axis increases.
The angle between v p1 and v new is known and it is the flight path angle (the black line is now the local horizontal),
so our ϕnew is known (it is a function of the enew and the new true anomaly νnew ). Since we have ϕnew , we have
also hnew from which we have pnew .
 
∆v √
ϕnew = ϕnew (enew , νnew ) = arctg → hnew = rp1 vnew cos ϕnew → hnew = µpnew → pnew
vp1
p = a(1 − e2 ) → pnew = anew (1 − e2new ) → enew
We are changing, as we can see, all the parameters by playing this game.
To get νnew we can use the trajectory equation where r=rp1 :
p
r= → νnew → ωnew
1 + e cos ν
From the TA we can evaluate the new perigee, because we have to remember that if we are instantaneously changing
the TA, we have that we are also changing the AoP:
∆ωnew = −∆νnew → νnew
In practice, if we apply a radial ∆v, all the in-plane parameters change.
This situation can be interesting when there is an explosion because, for example, if there is a fragment with a
given velocity, the new velocity will be given by the composition between the velocity of the parent plus the ∆v
due to the explosion. There are some fragments that have a positive tangential, negative tangential, positive radial
etc. and each fragment has its own orbit, and each fragment will undergo its own perturbations.
Space Flight Dynamics Michele Santarpia 25

3.2 Out-of-Plane Maneuvers

We want to change the orbital plane. Our orbital plane is identified by the direction of the angular momentum
vector ĥ, while in terms of orbital parameters we have i and Ω.
When we said change of the orbital plane, we want only to have ∆Ω and/or ∆i and not other orbital parameters.
If we want to change only out-of-plane parameters, we should not change the instantaneous norm of our velocity
otherwise it changes ε and so the semi-major axis: in practice, we need to have a pure rotation of our velocity
vector.
Let us consider the geocentric inertial frame and our orbit:
1. Changing only the inclination means that we have to change the direction of ĥ keeping the same line of
nodes, so we need to rotate ĥ around the line of nodes (node vector n).
2. Changing only RAAN means moving the line of nodes, of course we are changing the direction of ĥ but
we need to keep the same inclination angle w.r.t. z axis, we have a cone around z, so we rotate ĥ on a cone
around z with semi-aperture i.
Again we have to maneuver and we have to answer three basic questions:
1. Where do we have to maneuver?
2. What has to be the direction of ∆v?
3. What has to be the norm of ∆v?

3.2.1 Only Change of Inclination for Circular Orbits

Let us consider circular orbits (e=0) and let us study the case of only change of inclination. Our 1st problem is:
where and how we should maneuver?
If we have two circular orbits with only a difference in inclination, we have two intersection points which
correspond to the nodes: so, where do we have to maneuver? We consider one of the two nodes, in which we
can consider a single-impulse strategy through which we can rotate the velocity vector instantaneously and, in
this way, we will be on the 2nd orbit.
In order to understand which is the required ∆v, let us consider our point of view along the direction of the node
vector and so we can see what happens from the POV of the ascending node. Let us consider the velocity of the
1st orbit and of the 2nd orbit where let us remember that the velocity 2 has a different direction but with the same
norm otherwise the orbits are not the same!
We can build a triangle in which we can identify the required ∆v (in norm) to change the inclination:


 ∆i = i2 − i1





 r

 µ
v = v1 = v2 =
 r



  
∆i


∆v = 2v sin


2
If we apply this ∆v at the AN, we rotate the velocity vector v around the line of nodes and we will instantaneously
jump on the 2nd orbit with the new inclination. Of course, the maneuver will be more costly if we want a larger
change of inclination and the cost goes linearly proportional with the orbital velocity.
These maneuvers are very expensive also with few deg of ∆i. Of course, the change of inclination will cost less if we
are on circular orbits at higher altitudes because we are slower (rotating a smaller vector is cheaper). For circular
orbits, if we are at the DN, it is the same in terms of cost.

3.2.2 Only Change of Inclination for Elliptic Orbits

If we remove the circular orbit assumption, in general, there will be a difference in maneuvering at the AN or DN
nodes. In the node, where we are slower will be better because we will pay less to rotate the vector.
Space Flight Dynamics Michele Santarpia 26

Let us plot the 2D geometry on the plane of motion where we can identify the nodes. Of course, when we are at
the AN, the velocity will be not normal to the position vector (we are not at the perigee) and we have a negative
flight path angle since we are moving towards the perigee.
In the expression of the ∆v we consider the projection of the velocity vector on the local horizontal since we need to
rotate only this normal component to the position vector around the line of nodes. For this reason, in the equation
we have v⊥ :
 
∆i
∆v = 2v⊥ sin
2

In general, there is a difference in maneuvering in the two nodes and maneuvering in one of the two nodes is better.
If there is a symmetry position in the nodes, i.e. in the case of perigee at 90 or 270 deg, there is no difference in
the maneuvering.

3.2.3 Single Impulse Strategy for Change of RAAN and Inclination

Let us consider the general case in which we want to have ∆Ω and ∆i and, for simplicity, let us come back to the
assumption of circular orbits.
Let us consider to plot our orbit on a sphere with radius of the orbit, so the orbits are arcs on the surface of the
sphere. Imagine that the rotation axis of the Earth is plotted. Let us consider two orbits with different RAAN and
i on this sphere: how they look like?
First of all, we have to understand which is the latitude that can be achieved by the orbit.
Let us take a point over the orbit, we consider the meridian (that starts from the North pole) that pass through
the point over the orbit; based on the construction, we have the equatorial plane, we get the geocentric latitude by
connecting the point on the orbit to the center of the Earth and considering the footprint of the meridian.
Maximum Latitude: which is the maximum latitude that we can get with a given orbit that has a given inclina-
tion? Let us consider a prograde orbit i.e. an orbit characterized by an inclination lower than 90◦ where in the
inertial motion we move towards east (not w.r.t. a ground observer) and let us consider the POV of the AN and
let us make a sketch in 2D where the orbit plane appears as a line (we see the trace of the orbital plane).
The maximum latitude that it can reach is the latitude of the point in red (A) because it is the pole of the orbit,
and it corresponds at the inclination.

max ϕ = i

Let us consider the geometry in 3D on the sphere. Each latitude identifies a parallel where the pole of the orbit is
identified as A in red, if we take the meridian, then the latitude of this point is equal to the inclination of the orbit.
Of course, the parallel has a latitude equal to the inclination and so the orbit will be tangent to the parallel.
Orbit with 90◦ of i is a polar orbit where the angular momentum vector is in the equatorial plane. The maximum
latitude is 90◦ , and this orbit conceptually touches the North pole.
If i is larger than 90◦ we have a retrograde orbit, the inertial motion of the satellite is in the west direction.
Which is the maximum latitude? We consider always the POV from the AN. The orbit pole is A, and the maximum
latitude is π − i. Let us see on the sphere what happens. Let us consider a parallel with latitude equal to π − i,
and then we can plot the orbit which will be tangent to the parallel and we can identify the A point.
Orbits with only ∆i: let us see how look like orbits on the sphere with difference in only inclination. Since they
have a difference in the inclination, it means that they will touch different parallels since the maximum latitude is
different. Let us plot the two parallels, and let us consider the two orbits with the same AN but different i.
Orbits with only ∆Ω: let us see how look like orbits on the sphere with difference in only the RAAN. They have
the same inclination and so they have to touch the same parallel because the maximum latitude is the same; let
us consider the parallel in black and let us plot our two orbits. Let us consider the angle on the equator and let
us connect with the Earth center: we can underline the ∆Ω = Ω2 − Ω1 . The orbits will be tangent to the same
parallel but in different points so the poles of the orbit will be different. These two orbits have two intersection
points! These intersection points, of course, are not coincident with the poles but they are close to them. So, if we
can see on the sketch, the orbit pole will be a little bit after the orbit pole on the 1st orbit, and a little bit before
Space Flight Dynamics Michele Santarpia 27

of the 2nd one. We have these two intersection points which are opposite w.r.t. Earth center, the other one will be
close to the orbit poles in the Southern hemisphere. So, now the question is:

How can we change only the RAAN?

Thanks to the two intersection points, we have two opportunities per orbit where we can apply the maneuver
using a single ∆v strategy. So, it is clear that we can rotate the velocity vector but around which direction? We
will take all these considerations in the general case.
Confusion between angles and arcs: before considering the general case, we can define some arcs: since the
orbits are on the sphere with radius equal to the radius of the orbits, this arc can be written as the product r · ∆Ω.
If we take the meridian passing through a point over the orbit, we can identify the latitude and we can evaluate the
arc on the sphere. Since we are working on this sphere, the radius does not have any importance for our calculations,
because all the arcs on the sphere will be given by the product between the angle and the radius, what we can do
is to make a confusion between angles and arcs (so, arcs as angles) and this means that the radius is 1. We will
remove r. So, we will work with angles: some angles will be angles, some other angles will be the arcs on the sphere.
Analysis of the General Case: let us consider the general case where ∆i and ∆Ω are different from zero. The
will touch different parallels and they will pass through different nodes. There will be always two intersection points
on a general position (not the nodes). Let us suppose to maneuver in one of the two points, we have to rotate the
velocity vector: we need to apply a pure rotation maneuver but around which line? In order to understand this, let
us consider the POV along the instantaneous radial direction corresponding to the intersection point.
These two vectors must have the same norm, we can build a triangle and the rotation angle θ which corresponds
to the angle between the orbital planes. We need to rotate the velocity vector around the instantaneous
radial direction r (in black, from Earth center to the intersection point).
Let us consider the general case in which we want to have ∆i and ∆Ω that correspond to a variation in the direction
of the angular momentum vector ∆ĥ. The two intersections points are opposite with a difference in anomaly of
180deg.
We need to answer to two questions:
1. Which is the required ∆v?
2. Maneuvering Positions: where along the orbit we need to apply the maneuver?
In order to answer these questions, we need to play with the spherical triangles in the image (remember the
relations). Let us remember that we use angles instead of arcs. We define the argument of latitude u = ω + ν
(sum of the AoP and the TA) which is the anomaly w.r.t. the ascending node (for circular orbit, we do not have
the argument of perigee of course, not defined).
To extract what we need, we apply the spherical trigonometry: remember that it can be applied not to any arc on
the surface of the sphere but only to maximum circle arcs which are obtained intersecting the sphere with some
planes that contain the center of the sphere. In particular, to evaluate the angle θ, we can apply the law of cosines
for angles.

cos θ = − cos i1 cos(π − i2 ) + sin i1 sin(π − i2 ) cos(Ω2 − Ω1 )

Thanks to the knowledge of θ, we can evaluate the required ∆v∆Ω∆i .

Now, we need to understand where to maneuver which means that we need to know u1 and u2 . To obtain them,
we can write the law of cosines for angles also to (π − i2 ) and to i1 :

cos(π − i2 ) = − cos i1 cos θ + sin i1 sin θ cos u1

cos i1 = − cos θ cos(π − i2 ) + sin θ sin(π − i2 ) cos u2

The case in which ∆i=0 and ∆Ω ̸= 0 is just a particular case of this general one that we can obtain by putting
i2 = i1 .
Examples from Vallado 6.5 and 6.6: as we can see from the two examples 6.5 and 6.6, the ∆v required to make
a change in both RAAN and i (≈ 3.61km/s) is less w.r.t. the one required to change only RAAN (≈ 3.69km/s)
because this problem depends on the value of θ. In this 2nd case, since θ is lower, the two orbital planes are
Space Flight Dynamics Michele Santarpia 28

closer and the intersection points are in different positions and so the maneuver is less expensive. So, the cost of
the maneuver depends on θ, and the value of θ has not a linear dependency with ∆Ω and ∆i (there is spherical
geometry involved).
The cost is pretty similar due to the fact that is pretty similar θ. The real difference is that we are rotating velocity
in an another point since u1 and u2 are different, and so we generate also a variation in inclination.
In the 1st case in which we want to change only RAAN, we have that u1 is slightly after the orbit pole, so
u1 > 90deg (90deg corresponds exactly to the orbit pole, it is exactly 90deg from the AN where we reach the
maximum latitude), u2 is lower than 90deg because the intersection point is slightly before than the orbit pole.
There is symmetry because u1 and u2 are both 13deg distant from 90deg.
Instead, in the 2nd case the shape of the orbit is different since the inclination is not the same, so they will touch
different parallels and the intersection point between the two orbits will be different w.r.t. the 1st case and this is
the reason why we have a different solution. In particular, both u1 and u2 are larger than 90deg, so the intersection
point is after the two orbit poles.
Due to plane-change, we have a large cost in terms of ∆v: so, the question is: why do we need to make a plane-
change? We want to try as much as possible to avoid plane change since they are expensive. So, we will try to be
launched directly on the orbital plane that we need but this is not always possible. If we cannot avoid plane-changes,
we may want to find other strategies to save ∆v. Maybe, we can make a compromise between maneuver cost and
time, instead of a single impulse we can make more impulses in order to save some ∆v but with more time.

3.2.4 Restricted Three-Impulse Plane Change Strategy for Change of RAAN and Inclination
We apply two tangential ∆vs and another one which is not tangential. The idea of this strategy is that we take
advantage from the fact that is cheaper to rotate a smaller velocity vector and so it can become cheaper to change
the plane. We will see that there are some conditions where the additional cost of using two transfer ellipses is
compensated by the advantage that we rotate the velocity vector at a lower velocity.
Let us consider to have two circular orbits, the initial (light blue) and the final one (red). We only want to change
our plane by rotating the orbital plane of an angle θ.
As we know from previous discussions, if we want to have a change of ∆Ω and ∆i, this implies a given value of
θ, and in the relation we have the argument of latitude u1 which is saying us where to rotate the velocity vector.
Now, we will talk only in terms of θ because any plane change implies a rotation θ, and so we have to imagine that
in the geometry that we have now in the sketch this is not the POV of the equator, and the segment that we see is
the segment that connects the two intersections points of the orbits (in general, these intersection points
are not on the equator; they will be on the equator if we want to change only the inclination but in the general
geometry the intersections points will be somewhere else along the sphere). So, just for graphical clarity, we are
considering the geometry from the POV of intersection points.
Maneuvering Strategy:
1. We apply a tangential ∆v1 to jump on a transfer ellipse (we have increased the radius at the apogee) where
the apogee will be at a relatively large distance from the planet, and after half of the orbit we will
be at the apogee. We are far and slow enough that the cost of the plane rotation is cheaper. So, applying
this ∆v on a circular orbit, we jump on an ellipse, we do not change the plane, and the point where we have
applied the ∆v becomes the perigee and we increase the height of the apogee.

∆v1 = vp,tr − vc1 > 0

r1 +r2
We write vptr without specifying 1 or 2 because they are equal in norm. Remember that rptr = 2
(Hohmann-like maneuver).
2. As we arrive at the apogee, we can rotate the velocity vector and so rotate the orbital plane, we have a pure
rotation ∆vθ (∆v2 is out-of-plane), and we will be over an ellipse with same shape and dimensions of the pre-
vious one but on the final plane. For the 2nd ∆v, it make no sense to consider a sign because it is not tangential.

In order to evaluate ∆v2 we have to consider that it is a pure rotation, and we can consider that rotat-
ing of θ the velocity along the ellipse is equivalent in terms of cost of rotating of θ the velocity along a circular
Space Flight Dynamics Michele Santarpia 29

orbit, so we can use the same expression as before.

 
θ
∆v2 = 2va,tr sin
2

3. After half of the orbit, we will be back at the initial point, now we can slow down the satellite because if we
do not nothing, we will continue over the green ellipse: so, we apply a tangential ∆v3 < 0 to arrive on the
final orbit.

∆v3 = vc1 − vp,tr = −∆v1 < 0

Finally, the total ∆v is:

3
X
∆vtot = |∆vi |
i=1

As we can see, w.r.t. the 1st strategy we are adding two ∆vs : this is an additional cost but sometimes we can
save a lot of ∆v on the 2nd impulse that can compensate for the additional cost derived from the two additional
impulses.
This strategy is called restricted three impulse plane change because we concentrate all the plane change
in the 2nd impulse ∆v2 .
In this case, we have a maneuvering time which is different from zero and equal to two half of the period on the
transfer ellipses (so, one entire period since the semi-major axis of the two transfer ellipses is the same) (of course,
having a maneuvering time is a disadvantage w.r.t. the one-impulse strategy):
s
a3tr
τman = 2π ̸= 0
µ

The possible advantage depends on θ: if θ is below a given threshold, this strategy is not convenient, whatever we
do with r2 this strategy will cost more.
The restricted three impulse plane change is not a single maneuver but we can have an infinite number of possible
maneuvers because we have one degree of freedom which is r2 . Larger is r2 , the longer will be the time for the
maneuver.
Results: let us focus on the diagram of ∆vtot /vc1 as a function of the ratio r2 /r1 .
1. The ratio r2 /r1 = 1 is related to the single-impulse strategy (it has in fact the behavior that we expect
from its formulation, with maximum in 180deg).
2. The curve r2 /r1 = ∞ is the limit case in which we are jumping not on an ellipse but on a parabola,
we arrive at infinite where we change the plane and then we come back on another parabola. Of course, what
is interesting is that at the infinite we arrive with zero velocity where rotating a zero velocity vector is free
(∆v to change the plane is 0), this is the reason why the infinite curve is horizontal (the limit case does not
depend on θ) and the ∆vtot related to this case is given by twice the ∆v required to jump on a parabola and
the other one to jump back on the circular case.
3. Until about θ = 40deg, the minimum ∆v is the one of the single impulse strategy.
4. As θ increases above 40deg, the minimum ∆v solution will be with an increasing value of r2 /r1 , so the larger
is the plane change, the more convenient is to increase r2 which means that it is cheaper to rotate the velocity
vector farther.
5. Above 60deg, the best solution is the ∞ case (pure theory).
Do not forget that with the restricted three-impulse plane change strategy we have one degree of freedom
i.e. the parameter r2 .
Space Flight Dynamics Michele Santarpia 30

3.2.5 General Three-Impulse Plane Change for Change of RAAN and Inclination
The restricted three impulse plane change is a particular case of the more general strategy called general three
impulse plane change.
• In this general case, we decide to apply a little bit of plane rotation also with ∆v1 and ∆v3 which
are not anymore pure tangential. Most of the rotation is applied in ∆v2 .
• The key message is that this general case could be a little bit better than the restricted one, we can again save
some ∆v. When we consider this strategy, it is better than the single impulse strategy above
22deg.
• This strategy has also additional degrees of freedom because we are not only deciding where it is r2 but
we are also deciding which fraction of plane change is applied in the ∆v1 , ∆v2 and ∆v3 , so there
are infinite strategies for more than one parameters (not only one).
• If we look at the plot for the general case, the numbers are expressing the ratio which is minimizing the ∆v:
for θ near to 22deg, it becomes convenient to apply the general strategy with r2 just a little bit larger than r1 .
• Increasing the value of θ, it becomes more convenient to increase r2 . As θ becomes in the order of 60deg, the
best general three impulse becomes the restricted one and so the two diagrams converge to same one.
• The larger is θ, the more convenient becomes to use a general three impulse plane change w.r.t. the single
impulse case.
• The three impulse plane change is a strategy which is different from all the other ones because it is not coplanar
but it has also some elements of mixed maneuvers (we are applying some impulses which are changing at the
same time norm and rotate the velocity vector).

3.2.6 Mixed Maneuvering Problem - Hohmann Transfer with Plane Change

The scope of this maneuver is to change radius and plane of motion. We assume an Hohmann-like perspective.
For example, we have a circular orbit that we want to change in another one with different radius and inclination
(different inclination means different plane). The most popular problem is the maneuver from an inclined LEO
orbit to a GEO stationary one. Of course, if the LEO orbit is equatorial, we obtain the case of a pure Hohmann
transfer.
Before proceeding, let us define the Geosynchronous and Geostationary Orbits.
Geosynchronous Orbit: it is characterized by the fact that the period of the orbit has to match the sidereal day
(rotational period) of the Earth given by 2π divided by the inertial / sidereal rotation rate of the Earth (i.e. rate
w.r.t. to the stars):
s
a3GEO 2π ∼
τGEO = τL → 2π = = 23h56min → a = aGEO ∼ = 42164km
µ ΩL

Geostationary Orbit: it is a particular case of a geosynchronous. In fact, a geostationary orbit is:

1. Geosynchronous.
2. Circular (e=0).
3. Equatorial (i=0).
The satellite is fixed w.r.t. a ground observer (sub-satellite point is fixed).
Let us consider different strategies to jump on a geostationary orbit. In the 1st strategy, we apply two Hohmann
tangential impulses plus a pure rotation impulse while in the 2nd, 3rd and 4th strategies we need to apply some
mixed impulses:
1. 3∆vs strategy: we can change the radius becoming geosynchronous and, then, when we are slower we can
change the orbital plane: so, first of all, we can use an Hohmann transfer jumping on a transfer ellipse, so
due to Hohmann we arrive on a circular orbit that is geosynchronous but there is an inclination, and then as
we arrive in one of the nodes/intersection points we can apply the 3rd impulse and instantaneously we will
Space Flight Dynamics Michele Santarpia 31

be on the final geostationary orbit. So, the 1st two impulses are Hohmann ones and the 3rd one is for a plane
change.


 ∆v1 = ∆v1hoh






∆v2 = ∆v2hoh



  

 ∆i
∆v3 = 2vGEO sin

2

2. 2∆vs strategy: in this case, we need to apply some mixed impulses, we cannot apply only tangential and
rotational. With the 1st ∆v (mixed ∆v, not tangential and not pure rotational), we jump on the ellipse to
GEO orbit and also remove the inclination (plane change at perigee), so we jump on an Hohmann ellipse
which is already equatorial, and then we apply the 2nd ∆v of Hohmann at apogee.
3. 2∆vs strategy: also in this case, we need to apply some mixed impulses, we cannot only apply only tangential
and rotational. the first impulse consists in a Hohmann without change of plane, and in the 2nd impulse we
circularize and change the plane at the apogee.
4. 2∆vs strategy: it is called Hohmann with split plane change since we divide the plane change in two
parts. In this case we need to apply some mixed impulses, we cannot only apply only tangential and rotational.
In both the impulses we are changing the plane. With the 1st ∆v we increase the norm (Hohmann) and we
change a part of the plane, and then with the 2nd impulse we complete with an Hohmann and change of
plane. So, we jump on an ellipse which is intermediate in terms of plane.
We need to understand which strategy is the best. For example, strategy 2 has no sense because we are changing
the plane when we have a large velocity (we have introduced it only as term of comparison).
Let us compute the ∆vs for the first and fourth strategy (the 2nd and 3rd strategies are particular cases of the 4th
one, so we study only the 4th strategy):
1. 1st strategy:

∆v1 = ∆vhoh1 = vptr − vLEO






∆v2 = ∆vhoh2 = vGEO − vatr



  
|∆i|



∆v3 = ∆v∆i = 2vGEO sin
 ∆i = −iLEO
2

2. 4th strategy: As we have seen, in this strategy the two impulses are modified to have some plane changes.
If we concentrate the plane change at the perigee we have the 2nd strategy, otherwise we have the 3rd one.

So, we have to correct our inclination through a ∆i. Let us consider to have a parameter called s which
represents the fraction of plane change that we are applying at perigee. Of course, if s is equal to
1 we have strategy 2 otherwise if s is equal to 0 we have strategy 3.

Maneuvering Cost: we have to compute the cost of the maneuver. To compute the cost, let us see what
happens at perigee: on the initial orbit we have the velocity vLEO and we want to increase the norm of the
velocity because we have to jump on a transfer ellipse. For this reason, we have to apply a ∆vhoh1 . With the
1st impulse, we are going to correct only a fraction of the inclination and so we have an intermediate angle
α1 = s∆i. So, our new velocity has to stay on this blue line and the norm has to be equal to the velocity at
perigee of the Hohmann transfer ellipse. So, we can find the ∆v1 used to go from v LEO to v ptr . In norm, we
have that:
s  
2 1
vptr = vLEO + ∆vhoh1 = µ −
rLEO atr
Space Flight Dynamics Michele Santarpia 32

We need to compute ∆v1 through the Carnot theorem as we did in the general transfer case. We have a
triangle of velocities and we can apply the theorem to the triangle:

∆v12 = vLEO
2
+ vp2tr − 2vLEO vptr cos α1 α1 = s∆i

We can compute ∆v2 at the same way, and we will be at the apogee but conceptually we can imagine to have
a similar situation, we are at the DN, so now we have v atr of the transfer ellipse, so now we need to complete
the plane change and we need to circularize the orbit (so we need to apply the Hohmann): so, from the point
of view of norm we need to apply the 2nd ∆vhoh2 , and so we can sketch the final green velocity v GEO that
we need to have which is on the equatorial plane, and we apply the rotation α2 = (1 − s)∆i. We can sketch
the ∆v2 i.e. the single ∆v that increases the norm and also change the rotation of the plane.
 r
µ

 vGEO = = vc2
rGEO








 s  
2 1
va,tr = µ −
rGEO atr










 2 2
∆v2 = vGEO + va2tr − 2vGEO vatr cos α2 α2 = (1 − s)∆i

Of course, we can have infinite values of ∆vtot depending on s:

∆vtot = ∆v1 + ∆v2 = f (s)

We are interesting in minimizing ∆vtot , and so we want to find that value of s which minimizes this function:
for this reason, we just apply the derivative.
∂∆vtot ∆v1 vGEO va,tr sin[(1 − s)∆i]
= 0 → sin(s∆i) = → sbest
∂s ∆v2 vLEO vp,tr
The equation is not solved analytically (both ∆v1 and ∆v2 contain are function of s) but numerically: we get
the best value of s (in terms of ∆v) in the order of 0.1. So, the best solution is to have a change of plane of
the 10% at perigee and 90% at apogee.

Of course, this discussion in general, it is not related only to LEO to GEO but it is for a problem in which
we want to change the radius (r1 to r2 ) and the inclination of an angle θ.
LEO to geostationary transfer results: in the LEO to GEO transfer, let us see the results for the different
strategies in order to get some sensitivity with numbers. The 2nd strategy is the worst since ∆vtot ∼ = 6.39km/s.
Instead, we retrieve that it is pretty efficient to concentrate most of the plane transfer at the apogee (4th strategy,
sbest =0.077, ∆vtot ∼= 4.22km/s). The 3rd strategy is not the best one but it is really close, in fact we have a
difference of just 20m/s w.r.t. the 4th strategy.
Why the strategy 1 (∆vtot ∼ = 5.38km/s) is so bad? We know that, through the 4th strategy, we apply in a single
impulse both the change of norm and of angle applying a diagonal ∆v 1 (the violet one in the mixed strategy).
Instead, if we split the application of ∆v 1 (where firstly we change the radius and then we change the plane), we
apply a logic in which we have ∆v hoh1 as 1st impulse and then ∆v α1 as 2nd impulse. So, we are separating the
∆v 1 : in this way, if we consider the norm of the sum of two vectors ∆v hoh1 and ∆v α1 , it is smaller then the sum
of the norms. This is the reason why the strategy 1 is not efficient (it is pure geometry as explanation).
The only problem of all these discussions is that it’s not applicable in the real world: as we want in mission analysis,
we desire to minimize the total ∆v computing the best value of s but we cannot apply this because this maneuver in
general are very expensive and we typically use solid propellant rockets, we need a lot of thrust in this case. In an
impulsive strategy, it is the only propulsion system that we can use. The problem with this type of system is that
once we start the combustion we cannot stop it (not possible to turn off). We play with the Tsiolkovsky equation:
− I∆v
mp = m0 [1 − e sp g ]

What comes out from this equation is that finally in our mission design problems we cannot really obtain the ∆v
we want, so our best solution cannot be apply in the real world.
Space Flight Dynamics Michele Santarpia 33

Let us consider our LEO to GEO: we need our two huge impulses, so we will have the 1st stage to give us the 1st
one and the 2nd stage for the 2nd one (two solid propellant rockets).
We cannot have a customized solid rocket (catalogue of available solid rockets). From mission analysis we have ∆v1
and ∆v2 we need, we have a limited selection of rockets, we can only look to that rockets which can give us a ∆v
which is larger or equal to the one that we need.
m0 = mstage1 + mstage2 + mpayload

mp is the mass of propellant of the stage 1 that we decide to buy.

From the equation we get a given ∆v, so this means that we cannot obtain exactly the ∆v1 , there are no degrees
of freedom. The problem of solid rockets is that we cannot turn off, for example if we have a ∆v higher than the
required one the idea could be to turn off before in order to have the one that we want. So, we cannot adapt the
rocket to our ∆v: this problem is called fixed ∆v transfer because once we make a selection of the solid thruster
we have a given ∆v. Since we have found from mission analysis the optimal maneuver but we cannot realize it, and
so we need to solve this problem. So, for example the ∆v could be larger w.r.t. the one we need. If the ∆v that we
can have is lower than the one we need (optimal one for example), we need to change the orbit, so in general the
problem is always that the ∆v is larger than the minimum we need.
Which are the strategies to obtain what we want?
1. Remove some propellant: 1st strategy is that we have to modify the rocket in particular we remove some
mass of propellant (off-loading propellant) in order to give the exactly amount of ∆v we need. This strategy
is not smart for some reasons: we are paying for a larger rocket and then we are removing some propellant
(not using the entire capability of the rocket) but the other problem is that if we modify we may need to
re-qualify the rocket.
2. Playing with m0 adding ballast: 2nd strategy we play with m0 , we can reduce the ∆v by increasing the
mass adding some mass in the two stages making them heavier (adding ballast weight). Where to add the
masses? Pay attention to this.
3. Changing the maneuvers (∆v1 , ∆v2 ): 3rd strategy is the smartest one! It does not require to modify
rocket or add mass. In this case, since it is a problem to adapt the rocket to the maneuver, we change the
perspective adapting the maneuver to the rocket that we find. So, we change ∆v1 and ∆v2 in order to match
with the rocket possibilities. What means changing the maneuver?
(a) The 1st idea consists in using s as a free parameter: this means that we can use a value of s which
is different from the best one. In this way, we decide to spend more by splitting the plane change in
a different way from the best solution. Of course, since we need to match two constraints i.e. ∆v1 and
∆v2 , we need to play with another free parameter (since if we use only s we cannot satisfy two conditions
at the same time). For example, we can add as free parameter the inclination of the initial orbit:
this means that we can work on the amount of ∆i that we have to apply (we are making the maneuver
more expensive by increasing the inclination). For example, for a LEO to GEO transfer, we can decide
to change the initial inclination of the LEO parking orbit.
(b) The other possibility is to change the maneuver meaning that we do not use the Hohmann transfer plus
plane change but we use non Hohmann transfer because it can be much better since we change the
type of in-plane impulses in order to spend exactly what we can spend (spend more).
(c) Finally, we can have a combination of these strategies.

3.3 Launch Constraints

Before talking of fixed ∆v transfer, we need to better understand the concept of Launch Window which is linked
with the idea of changing the parking orbit. In particular, the Launch Windows is one of the Launch constraint that
we have. The discussion about Launch Constraints is linked with the idea of plane change maneuvers.
Which are the constraints that are coming from the launch base?
As we know, plane change maneuvers are pretty expensive and we want to avoid them as much as possible, but the
need of applying plane changes depends on the location where we launch since it often does not allow us to directly
launch on the desired orbital plane.
Space Flight Dynamics Michele Santarpia 34

We have to make some considerations about:

1. Launch Direction, with the concept of angular sector.
2. Launch Site Location, with the relation between the orbit achievable inclinations and latitude of the LS.
3. Launch Window (i.e. the timing), which is important because it gives us the right moment in which we
need launch.
Let us take the geocentric inertial reference frame xyz and a LS in the Northern hemisphere which is
rotating with the Earth w.r.t. the inertial frame. Let us sketch the final desired orbit identified in green which is
not moving in this reference frame. If we want to minimize plane change maneuvers, we should try to launch
the satellite directly on the final orbital plane.
As we know, the launch requires some time (few minutes). We can make some assumptions to neglect the duration
of the launch but, in this discussion, we can simply ignore the time of launch. We consider that the satellite is
injected immediately on the final orbit.
Which is the right moment to launch?
As the Earth rotates, the launch site will have the possibility to cross the orbital plane (which is fixed). It is
intuitive that we should launch, if possible, when the LS is crossing the orbital plane in order to find our
self directly on the final orbital plane in order to avoid out-of-plane maneuvers. It is intuitive that the LS crosses
twice per day the orbital plane (two intersection points).
Launch Site Angular Sector: for each LS, there is an angular sector (see the figure 6.9) that gives us the interval
containing the possible launch directions, i.e. from each site we can launch from a given set of directions. The
angular sector is in eastward direction w.r.t. the LS. For example, in Kourou, the possible launch directions
are always above the ocean. Baikonour has a very large angular sector. The size of the angular sector is related
to safety considerations. We can see that each launch site has a minimum and maximum azimuth angle
which is giving us the allowed launch directions. Let us imagine that we are in a given LS, we consider the local
North direction and a given launch direction (given by the initial velocity get by the launch), we can derive
the azimuth angle which is clockwise from North direction. The interval of possible azimuth angles
defines the angular sector where we can launch our rocket.
Launch Options: what are the launch options? It is possible to launch directly on the final orbit, we will have
in general two opportunities per day to launch since there are two points in which the LS crosses the orbital plane.
Let us consider the spherical geometry. We can proceed in two possible ways:
1. The 1st one is to consider our final orbit and we consider the intersections of the orbital plane with the surface
of the Earth considered as a sphere. Let us take the parallel corresponding to our launch site. The projected
orbit over the Earth is in green and as we can see there are two intersection points on the parallel. But in
this geometry we are not considering the rotation of the Earth (it is in the inertial frame).
2. The 2nd perspective considers a sphere at the radius of the final orbit, and practically it is the same because,
as we know, we are working with angles and, for this reason, the radius has no impact/importance.
Let us consider X axis of the inertial reference frame. We can plot the meridians corresponding to the two intersection
points. The two points LS1 and LS2 allow us to identify the launch opportunities that we can have.
Let us consider the velocity of the satellite when it is in the point LS1, the angle in light blue is the azimuth angle
A1 i.e. the angle of the velocity w.r.t. the North (since the meridian represents the North). From this geometry,
the two intersections of the orbit with the parallel give us the inertial rotation of the two launch opportunities, and
for each of this opportunity we have a given direction of the velocity since each opportunity is associated to a given
azimuth angle (that’s why we have A1 and A2 ).
If we consider the meridians, the LSs are identified by the directions pˆ1 and pˆ2 . An important angle is the one
between the AN of the orbit and pˆ1 which is the angular separation between the AN and the footprint of the
meridian, indicated with δ, which identifies the 1st launch opportunity, and we have two of these angles.
Let us see this problem from the direction of z i.e. from the perspective of the horizontal/equatorial plane. There
is a symmetry for the angle between the DN and pˆ2 .
Space Flight Dynamics Michele Santarpia 35

We can convince of this symmetry looking at the drawing in which we plot the parallel in black, the orbit in green,
the two meridians in red and δ in purple, the 1st is δ after the AN and the other one is δ before the DN. δ contains
the info about the launch windows or the moment when to launch.
To compute δ, let us consider the spherical triangle in yellow, where we consider the arc of meridian ϕLS which is
the latitude of the LS (arc as an angle in this case). We can identify the argument of latitude u1 corresponding to
the 1st intersection of the LS with the orbit identified by LS1.
We need the cos δ and sin δ (both of them to discriminate the sign, which depends if we have a prograde or retrograde
orbit). Based on our input parameters i.e. the latitude of LS, the inclination of the orbit we want to reach and the
launch direction, we can evaluate δ. We can write the law of cosines for angles to A1 .
   
π π cos A1
cos A1 = − cos i cos + sin i sin cos δ = sin i cos δ → cos δ =
2 2 sin i

We can write also the law of cosine to i (important, remember this equation!):
   
π π
cos i = − cos A1 cos + sin A1 sin cos ϕLS = sin A1 cos ϕLS → cos i = sin A1 cos ϕLS
2 2

Mathematically, the sine of A1 is always lower or equal to 1 and higher or equal to -1 (−1 ≤ sin δ ≤ +1), so the
cosine of the inclination is always lower or equal than the cosine of the latitude of LS (cos i ≤ cos ϕLS ), and so the
inclination is always larger or equal than the latitude of LS (i > ϕLS ). As result, the minimum inclination that
we can directly achieve with the launch is always equal or larger to the latitude of LS. If we launch,
for example, from Cape Canaveral (we are at 28deg of latitude), we cannot obtain an orbit with inclination lower
than 28deg. This is the reason why we cannot fly directly on an equatorial parking orbit but the initial orbit will
be always inclined.
When the inclination is equal to the latitude of the LS?
This is possible if the sine is equal to 1 which means that the azimuth must be 90deg, so we need to launch with
an azimuth of 90deg i.e. to launch towards east along the parallel.
We can write the law of sines to A1 and i:
sin i sin A1
=
sin ϕLS sin δ
Finally, we obtain:
tan ϕLS
sin δ =
tan i

We can have three cases:

1. Two Launch Opportunities per Day: if we plot again our geometry with the Earth, the parallel and
the orbits when we have two intersections, in this case the inclination is always larger than latitude of LS; in
fact, when we consider the launch directions we can see how we are not launching towards east but in the 1st
launch opportunity we are launching in North-East direction while in the 2nd launch opportunity
in South-East direction, and so not exactly along the parallel. When we are in the case in which we are
launching along South-East, the consequence is not that we are achieving a lower inclination but that we are
launching the satellite in a descending phase of the orbit but still i > ϕLS . So, two intersections means
two opportunities per day to launch. For example, in Vandenberg it is possible to launch towards South and
we will be on the descending phase phase of the orbit.
2. One Launch Opportunity per Day: when the azimuth is equal to 90deg, the orbit is tangent to
the parallel and so the launch velocity is exactly along the parallel i.e. along east direction, remember
always that we are neglecting the duration of the launch and so as we launch we are instantaneously with
this velocity, and so immediately after the launch we are at the pole of the orbit and start the descending
phase. So, when we get the minimum achievable inclination, we have one intersection and so, from the point
of launch window, we have one opportunity per day to launch.
Space Flight Dynamics Michele Santarpia 36

3. No Launch Opportunities: if the inclination of the orbit is lower than the latitude of the launch
site, the launch is not feasible, and so there are not intersection points between the orbit and the parallel.
The idea of an inclination which is less than the latitude is not possible to obtain. This is an important
constraint and so latitude is playing an important role.
So, the minimum inclination is always larger or equal to latitude and it is equal only when 90deg is one of the
possible direction. From Kourou, we can see how we can achieve an orbit which is very close to be equatorial.
For Vandenberg, min i is 67.3deg and latitude is instead 34.669deg. To achieve min i of 67.3, we are launching
South-east because instantaneously we find on the descending phase of the orbit. If we launch South west we will
find on 107.9deg which is of course larger than 90. The key message is that latitude plus possible directions
will give us the constraints on the achievable inclinations.
When do we have to launch?
Launch Window: let us consider the two angles α1 and α2 which correspond to the two opportunities. To consider
the time to launch we have to consider the rotating Earth and so we have to consider which is the exact rotation
of the Earth at a given time and so to understand the time instant we need to go beyond the inertial frame and to
understand where is the Greenwich meridian in this geometry. Let us consider an instantaneous picture in a given
time, imaging that the Greenwich meridian is in a given position representing by the direction XGR . The position is
identified time by time by the αGR (t). The angle from Greenwich to the footprint of the meridian passing through
the LS is by the definition the longitude of the LS which is a constant w.r.t. to Greenwich. From this geometry:

α1 = Ω + δ = αGR (t) + λLS

The information on time is included in αGR (t), and this is an equation with one unknown. As we solve this equation
analytically, we get the time when to launch.
The 1st launch opportunity is represented by t1 : we can get it because from this equation we know Ω because we
know which is the orbit where we want to launch the satellite, we know δ because it has been computed and, of
course, we know also the longitude of the launch site. Of course, if i > ϕLS we have two launch opportunities per
day and in order to compute them for each day we can use that:

αGR (t1 ) = αGR (t0 ) + ΩL (t1 − t0 )

where αGR (t0 ) is the right ascension at a given reference time. We can imagine that t0 is, for example, midnight
UT (Universal Time) of the day we are considering, of course we can make this discussion every day because every
day we need to find the two launch opportunities.
We need to express the midnight of a given day in Julian Date and we get αGR (t0 ) (this angle contains the position
of the Earth w.r.t. to the stars at any given time).
So, finally, once we have defined t0 , we can come out with the equation which gives us the time instant of the 1st
launch opportunity for a given day corresponding to the day through which we have defined t0 :

Ω + δ − αGR (t0 ) − λLS

t1 = t0 +
ΩL

The 2nd launch opportunity can be derive by the following equation where the only unknown is t2 (it can be
obtained exploiting the symmetry of the problem):

α2 = Ω + π − δ = αGR (t2 ) + λLS → t2

So, we have that:

Ω + π − δ − αGR (t0 ) − λLS

t2 = t0 +
ΩL

Launch Window: when we talk about the launch window, we are referring to a time interval and this means that
there is some tolerance: where does it come from? If we make a mistake in the timing, what are the affected orbital
parameters?
Space Flight Dynamics Michele Santarpia 37

1. Timing error means that we will end up on a different orbit, we will have an error only in the ascending
node, it is clear that if we launch with the right direction we get the same inclination, so we are generating
the same orbit but in the inertial frame we are in a different position. So, if we can set a tolerance on the
ascending node, it is clear that from it we can obtain a tolerance in the time interval (from ∆Ω → ∆t).
So, the time window given by t1 + ∆t is substantially depending on our tolerance in the AN.
2. Launch direction error translates in an inclination error.
Launch Window for Parking Orbits: if we want to launch on a parking orbit (like the case of passing from
parking LEO to GEO or from parking LEO to inclined MEO like Galileo GPS), the launch window can be relatively
large. In fact, if we launch in different time instants of the launch window, we also need to change slightly the
direction of launch (depending on the specific time instant, since different time instant means different RAAN) so
that we get parking orbits which are also different in inclination. This is because the transfer to the final orbit is
characterized by a fixed ∆v. So, depending on the timing, we will adjust the parking orbit before doing the transfer
since the problem is characterized by certain constraints in terms of ∆v1 and ∆v2 which have to be of a certain
value (so, instead of changing rocket which is not possible, we adjust the launching maneuver and so the parking
orbit). So, ∆Ω depends on the ∆v budget, we fix a maximum ∆v for the initial acquisition, and so given a ∆Ω
we can evaluate from the equations the ∆v.
What happens if we launch in the wrong time instant and if we have to use a fixed 2∆vs transfer?
Summary: when we want to perform a plane change from a parking orbit to the final one, we have found that the
best strategy is using a two ∆vs , with s equal to 0.1. Of course, the problem is that the rocket is not able to deliver
exactly the values of ∆v1 and ∆v2 . We have two possibilities: adapt the rocket to the required ∆vs or adapt the
maneuvers to the rocket. The first option is not so smart since it requires the modification if the rocket. The second
option, instead, is the smartest. We select the rocket from the market which is able to deliver a ∆v in the order
of ∆v1 and ∆v2 . Then, we can work by adapting ∆v1 and ∆v2 to the ∆v of the rocket. Since ∆v1 = f (s, θ) and
∆v2 = f (s, θ), we can change the values of s and θ. If we launch in the wrong moment, the value of Ω is different
and this means that the value of θ and so the values of ∆v1 = f (s, θ) and ∆v2 = f (s, θ) will be different and not
equal to the one of the rocket: in order to have the right values of ∆v1 = f (s, θ) and ∆v2 = f (s, θ), we need to
change the inclination of the parking orbit in order to obtain a different value of ∆i and so finally the right value
of θ and so the right values of ∆v1 = f (s, θ) and ∆v2 = f (s, θ). Remember that in order to change the inclination
of the parking orbit, we need to change the launch direction.
Final considerations: as we know, in general, we have seen how there are two launch opportunities per day but
several times the 2nd attempt is done after different days for different reasons related to launch operations, safety,
weather forecast, avoid night launches. Sometimes, there are some procedures, but there are also some orbital
reasons! Even if we are able to restart immediately, we would not like to try in the same day: in all the discussions
we are doing, we are completely ignoring any timing constraints we can have, we are just considering to launch
a satellite on a given orbit, but when we arrive on an orbit on a given time we will be on a given anomaly and
sometimes we do not want to stay only on a given orbit but also in a given time with a given anomaly. Imagine
the space station problem: let us consider the inertial axes and plot a prograde orbit of the ISS, when we launch
on this orbit we have that launching at different times we will end up in different positions in a given time, but
our scope is to arrive where the ISS is. In the rendezvous problem, what is important is the ∆ν between ISS and
our spacecraft. We can correct this error, but we typically like to start closer and, in general, we have a tolerance
about a maximum anomaly separation w.r.t. our target. So, for each launch opportunity we have to check where
is the target to understand which will be our angular distance (it is a problem of synchronization). But the timing
constraints are also present often even if we do not have any target to reach. Timing constraints can be linked not
to the inertial orbit of the satellite but they can be linked to the ground track which means the motion w.r.t. the
ground observer.

3.3.1 Initial Velocity due to the Earth Rotation Rate

Let us consider the inertial reference frame, the Earth and a given a launch site in the Northern hemisphere. Let
us consider the vector that gives the position of LS w.r.t. the Earth center rLS and the Earth angular velocity ΩL .
There will be an initial velocity at the launch site v LS w.r.t. the inertial frame which is along the parallel:

v LS = ΩL × rLS
Space Flight Dynamics Michele Santarpia 38

This initial velocity v LS is given to the launch velocity of our spacecraft by the LS due the Earth rotation rate. As
the launch site is at higher latitudes, the initial velocity will be lower (maximum at the equator).
If we look at the problem in 2D, we get that:

|v LS | = ΩL RL cos ϕLS

Imagine that we are launching from an equatorial LS in the eastward direction and we want to get a zero inclination.
We obtain more or less an initial velocity of 465m/s which is already available from the beginning since we are
already rotating in the right direction:

|v LS |max = |v LS (ϕLS = 0)| ∼

= 465m/s

If we want to launch to an orbit with 180deg of inclination (equatorial westward orbit) we should give with our
propulsion system the orbital velocity and we should also compensate for this initial contribution. So, in summary,
the velocity of LS in the inertial frame gives a contribution, and this contribution is not negligible especially if we
are at the equator.

3.4 Rendezvous Maneuvers

In rendezvous maneuvers, we would like to approach a physical or virtual target. Virtual means that we want to
have a given anomaly at a given time. The rendezvous problem can be modeled assuming to have a virtual target
that we are trying to rendezvous with.
The rendezvous maneuver is characterized by different phases: during the 1st phase, we perform an absolute
navigation where we fly in order to arrive close enough to our target but at a given moment we will be so close
that we need to consider the relative motion with respect to the target and so below a given distance we will go
with the 2nd phase which is the relative navigation (final rendezvous, final operation) and in this problem we
will use special equations called Hill’s equations (for circular orbit).
Now, let us concentrate about absolute navigation, and in our exercise the idea is that we reduce the angular
separation to zero but, as we have said just now, let us remark that in the real world this is not a single maneuver!
Absolute Navigation for Rendezvous Problem
We have three possible cases of interest:
1. Coplanar Rendezvous with target on the same orbit: the target is on the same orbit, we only need to
remove a difference in anomaly.
2. Coplanar Rendezvous with target in the same plane but different orbit: we reach a target with a
different orbit in the same plane.
3. Non Coplanar Rendezvous: out-of-plane transfer + coplanar rendezvous discussions to merge.
Coplanar Rendezvous with target on the same orbit: let us consider a circular orbit (e=0). In the discussion,
we consider the target as passive and our scope is to maneuver the interceptor, and we have an initial anomaly
separation ∆ν. The scope is to nullify the angular separation. We can have two possible cases:
1. Trailing Interceptor (”intercettore in coda”) case where ∆ν > 0.
2. Leading Interceptor (”intercettore in testa”) case where ∆ν < 0.
Trailing Interceptor: let us analyze the 1st one in which the target is in front of the interceptor (i.e. the
interceptor is chasing the target, chasing means ”inseguendo”), we are following the target.
If we want to reach the target, we need to slow down. Let us consider that we give a negative tangential impulse
to the interceptor, that as we know this means that instantaneously we will reduce our velocity and so in the
subsequent instant, our separation from the target will increase because instead of moving towards the target we
are moving in the other direction. What happens is that we jump on a transfer ellipse, called phasing ellipse,
where the point in which we apply the impulse will be the apogee of the orbit:

∆v1 < 0 → ∆a < 0 ∆e ̸= 0

Space Flight Dynamics Michele Santarpia 39

The nice point is that aph < a, and so the period on the phasing ellipse is lower. If the period is lower, in the time
needed by the interceptor to complete one phasing ellipse, the target will not have enough time to complete one
orbit! So, the target, after one the period on the ellipse, will have a lower difference in anomaly. After one period,
the target will be closer to us, and if we play this game tuning the numbers in the right way, after one orbit we will
obtain the rendezvous with the target. We need to shape the transfer ellipse in the right way, so we need to find
the correct aph . We need that one entire period on the phasing ellipse is equal to the time needed by the target to
fly an orbit minus ∆ν. Once we have aph , we can compute ∆v1 which is an Hohmann-like ∆v. Of course, the total
∆v is given by two times ∆v1 in order to jump again on the circular orbit:

∆v2 = −∆v1 → ∆vtot = 2|∆v1 |

If ∆ν > 0, 1-∆ν/2π < 1, so aph < a and so ∆v1 < 0.

Leading Interceptor: the discussion is similar in which we give a positive ∆v. We jump over an higher ellipse
with higher apogee where the period is larger. Again, we can size the ellipse in order to have that in the time
needed by the interceptor to complete one orbit, the target has to fly one orbit plus an angle |∆ν|.
The problems with these type of maneuvers are:
1. Atmospheric Effects: in case of trailing interceptor, since there is the atmosphere around the planet, it is
an important challenge to face with. To remove ∆ν, we jump on a phasing ellipse which might be very close
to the planet. In the real life, as we enter in the atmosphere, we have atmospheric Drag (atmospheric density
gives us non-negligible effects) that will take energy away (dissipative force), and so instead of keeping energy
constant we will lose it, with the consequence of losing semi-major axis. So, we will come-back not to the
original orbit but we come back to a lower point (losing of altitude). We need to pay attention to the radius
at the perigee of the phasing ellipse that should not be below a given threshold given by the sum of the radius
of the planet and ∆h (which depends on the atmosphere).

rp,ph > Rplanet + ∆h

2. ∆vtot too large: in this case, we can trade time with cost. Instead of correcting in a single phasing ellipse,
we can wait more time making more ellipses. So, instead of spending one orbit on the phasing ellipse, we
will spend an integer number kint of orbits on the phasing ellipse, and the target will fly another number ktgt
orbits. We are splitting the maneuver in a number of orbits.
Let us consider that we have a ∆ν to remove. The interceptor flies kint orbits, the target will fly ktgt orbits. We
can get the semi-major axis once we decide the coefficients. In general, if the coefficients are equal, then we can
write an equation for the semi-major axis of the phasing ellipse and we are correcting a fraction equal to ∆ν/kint
for each orbit on the phasing ellipse. In addition, decreasing the total ∆v with this idea of making more orbits
on the phasing ellipse is also useful to reduce our problem with the perigee because, as we increase the number of
orbits kint , the phasing ellipse will be closer to the original trajectory avoiding the risk to fly in the atmosphere.
s s  23
a3ph a3

ktgt ∆ν
2πkint = (2πktgt − ∆ν) → aph = − a
µ µ kint 2πkint

If ktgt = kint :
  32
∆ν
aph = 1 − a
kint 2π

We are correcting ∆ν/kint per orbit on the phasing ellipse.

kint = ktgt ↑ |aph − a| ↓ ∆vtot ↓ τman ↑

Coplanar rendezvous from different orbits: the bi-elliptical transfer is conceptually a coplanar rendezvous
from different orbits, and the idea in this case is to solve numerically the equations for the bi-elliptical transfer. Of
course, if we consider some simpler transfer strategies we can proceed analytically. To keep this simple again, we
consider that the interceptor and the target are on two different circular orbits (e=0). We want to jump over the
final orbit to catch our target.
Space Flight Dynamics Michele Santarpia 40

Firstly, there is an initial angular separation ∆ν(t0 ) which is a function of time: in fact, if we do not do
anything, the angular separation will change since the interceptor and the target have different mean motion
along the orbits. To solve the problem, we can use a bi-elliptic transfer imposing a timing condition: to make
this simpler, let us assume to use Hohmann transfer. For Hohmann, as we know, the time of maneuver is fixed.
There are no tuning parameters (no degrees of freedom), so there are no way to adapt this maneuver to satisfy
timing requirements. So, we will end up on the transfer orbit but not in the position we want. Since we cannot
adapt the maneuver, we just need to wait that the angular separation with the target when we start the maneuver
is the correct one. There will be a special angular separation called ∆νhoh which guarantees that, if we start
the Hohmann transfer when we have this angular separation, we will catch up with the target at the end of the
maneuver. There are two problems:
1. How do we find ∆νhoh in order to satisfy our timing constraint?
2. Once we define this angular separation, how much time do we have to wait in order to get the ∆νhoh ?
The total time of the maneuver will be given by the waiting time plus the Hohmann maneuvering time:
τtot = τwaiting + τman

∆νhoh computation: let us compute ∆νhoh . We need to impose a timing constraint. In the time in which the
interceptor flies half of the Hohmann ellipse, the target should fly half of its orbit minus the Hohmann angular
separation ∆νhoh :
s s s  3 s s  3 
a3tr r23 1 r1 + r2 r23 1 r1
π = (π − ∆νhoh ) →π = (π − ∆νhoh ) → ∆νhoh = π +1
µ µ µ 2 µ 8 r2

Target Must Be in a Forward Position if the Initial Orbit is Lower than the Final: if the initial orbit is
lower than the final (r1 < r2 ), the semi-major axis is smaller than the radius of the final orbit, so the target takes
more time to fly (in the time in which the interceptor flies half of the orbit, the target will fly less than half of the
orbit), so we need to have the target in a forward position to have our match and this is the reason why we have
that ∆νhoh > 0.
Waiting Time: regarding the waiting time, let us write the anomaly of the target and of the interceptor, and let
us consider the difference. What is important for this problem is the relative angular velocity:
r
µ

νint (t) = νint (t0 ) + nint (t − t0 ) nint =

3



 r 1



 r
µ
ν tgt (t) = ν (t
tgt 0 ) + ntgt (t − t 0 ) n tgt = 3



 r2





∆ν(t) = ∆ν(0) + (ntgt − nint )(t − t0 )


Let us compute the time at which we acquire ∆νhoh by imposing that, when ∆ν(t) = ∆νhoh we have that t=thoh ,
so we can evaluate how much time we have to wait:
∆νhoh − ∆ν(t0 )
∆ν(thoh ) = ∆νhoh → thoh = t0 +
ntgt − nint
This is the 1st opportunity and the optimal separation is repeated after a certain amount of time: for this reason,
we can write +2kπ.
∆νhoh − ∆ν(t0 ) + 2kπ
thoh = t0 +
ntgt − nint
The dependency is on the initial angular separation ∆ν(t0 ): if it is close to the optimal one and it is changing
towards it, we have to wait a little time. The larger is the relative angular velocity, the less we will have
to wait and also the period through which this optimal separation will be repeated depends on this relative velocity.
In fact, what is interesting to evaluate is the ratio called synodic period, equal to:
2π
τsyn =
|ntgt − nint |
Space Flight Dynamics Michele Santarpia 41

It is the period of the relative motion (period to obtain the same angular separation). If two orbits have a significant
difference in semi-major axis, we will wait less time.
The waiting time will be given by:

τwaiting = thoh − t0

We can solve this problem with a bi-elliptical transfer or we can try to solve with other maneuvers (we have defined
infinite ways to jump from r1 to r2 ), so if we think numerically we can try to find some maneuvers to optimize
timing constraints.
Non-Coplanar Rendezvous: in case of a transfer from an inclined LEO to a geostationary orbit, as we have seen
before, the optimal transfer was a mixed maneuver with two-impulses and the best case was to have a little bit
of plane change at the perigee and most at apogee. In case of non-coplanar rendezvous, we have orbital transfers
constraints but also a timing one (they do not match each other). So, we want to pass from an inclined LEO to an
equatorial geostationary orbit and also catch up a target, so we want to arrive in a specific time instant.
Let us consider to have a passive GEO target that we want to reach from an inclined LEO orbit. In this problem,
we need to have the right synchronization. The problem is that, for example, in the same time in which a LEO
satellite flies 15 orbits, the GEO target flies only one. So, while a LEO satellite flies one orbit, the GEO target has
moved 1/15 of the orbit (i.e. 24deg of true anomaly). We can solve this problem with different strategies:
1. Waiting Strategy: the 1st strategy is the waiting one. The problem is that, practically speaking, we will
never have the situation in which we will come back in one of the nodes when the target is also at the correct
angular separation ∆νhoh ! In the case in which we do not have a timing constraint, there is a continuous
change of the relative positions: so, it is easy because when the satellites are in the right position, we start
the maneuver. In our case, instead, we cannot wait because when the GEO target is in the right position,
we cannot start the maneuver since we should stay in one of the nodes! So, the two events have to happen
together and this is the reason why the waiting strategy is not good.
2. GEO to LEO + Phasing Maneuver: the 2nd strategy consists in jumping to GEO (LEO to GEO standard)
and then we perform a phasing maneuver. This is not very efficient because we spend maneuver for jump and
phasing. So, the idea of waiting is better because we do not spend extra propellant but we just wait the right
time instant.
3. Phasing Ellipse from LEO: a possible 3rd idea (good one!) is to use a phasing ellipse from LEO. Let us
imagine to give a positive in-plane tangential impulse jumping on a phasing ellipse. This idea is nice because
we can try to design the phasing ellipse (instead of waiting that the two events happen together which could
require too much time). So, by changing the relative phasing, we can reduce the time of waiting. We can
impose that, after some time, the interceptor is at one of the nodes and the GEO target is exactly in ∆νhoh .
In this way, after an integer number of orbits on the phasing ellipse, the interceptor will be in one of the nodes
and the GEO satellite will be exactly in ∆νhoh . When we have the synchronization of the two phenomena,
then we will use the two impulse mixed maneuver (we will not come back to the original LEO). So, finally,
this 3rd strategy uses three impulses: through the 1st impulse we are increasing the energy of the orbit since
we are with an higher semi-major axis. Since we are on an higher energy orbit, the sum of the two following
impulses will be lower w.r.t. the 2nd strategy (the starting orbit is not a circular LEO orbit). So, the third
strategy is the smartest one.

Let us analyze deeper the 3rd strategy. Let us consider to have an initial anomaly for interceptor
νint (t0 ) and an initial anomaly for the target νtgt (t0 ). First of all, to start this process, we need to bring the
interceptor to one of the nodes (i.e. we need to propagate to one of the nodes): this will require a certain
∆tint , which is the time needed to fly from the initial anomaly to one of the node. The time needed to arrive
at the AN is:
2π − νint (t0 )
∆tint = ∆νint > π
nint
while the time needed to arrive at the DN is:
π − νint (t0 )
∆tint = ∆νint < π
nint
Space Flight Dynamics Michele Santarpia 42

Practically speaking, this is the time we need to wait to arrive at one of the node, it will be a short time
(some tens of minutes).

Let us consider the anomaly of the target νtgt (t0 + ∆tint ):

νtgt (t0 + ∆tint ) = νtgt (t0 ) + ntgt ∆tint
This anomaly corresponds to the real starting condition: so, the GEO target is at a certain anomaly and the
interceptor is at one of the nodes. Now, we want to impose our phasing condition: we need to impose that the
time needed by the interceptor to spend kint orbits on the phasing ellipse must be equal to the time needed
by the target to reach the desired ∆νhoh (since the interceptor is at the ascending node, ∆νhoh corresponds
exactly to the desired anomaly for the target). At the numerator we have the angle that the target has to
fly to reach the good angular separation, while at the denominator the mean motion of the target ntgt : the
fraction is representing the time needed by the target to fly that angle. By imposing this condition, we have
that the interceptor is flying an integer numbers of phasing ellipse (so, it will be again at the AN) and in the
meanwhile the target will have reached exactly the good separation in anomaly i.e. ∆νhoh :
s
a3ph ∆νhoh − νtgt (t0 + ∆tint )
2πkint = → aph → ∆v1ph
µ ntgt
From the above equation we get aph , from which we can compute ∆v1ph and then we can also compute ∆v2
and ∆v3 paying attention that the initial velocity in the Carnot theorem is not exactly the velocity on the
circular orbit. In this way, we will have a strategy where we arrive on the GEO orbit exactly catching up with
our target.
Example of a GEO target: when we fly to a geostationary orbit, we always have a GEO target which is virtual,
because when we are on a geostationary orbit, we do not want to stay at a generic anomaly but we want to stay
at a given longitude: if we imagine to be a telecommunication satellite that wants to provide services to Southern
Europe, we need to be at that specific longitude. So, finally we have always this problem and the scope of the game
is to reach a given specific longitude.

3.4.1 GEO Repositioning

Example 6.5 from Curtiss: GEO repositioning is modeled as a rendezvous problem, in which we want to change
the longitude for a geostationary satellite. Let us consider, for example, the case in which we want to move towards
west and we want to do this in three revolutions of the phasing orbit. Let us consider the ECEF axes: of course,
the geostationary satellite does not move along the circle in this frame. If we consider the satellite as red point in
a given position, we will have a given longitude of the satellite and we want to move towards west. Conceptually,
there is a virtual target at the longitude that we want to achieve. So, the satellite is our interceptor. From the
maneuvering POV, it is better to see the maneuver in the inertial axes. It is clear that the change of longitude that
we want to achieve is equal to the change of true anomaly.
∆λ = ∆ν
We give a positive impulse jumping on a higher phasing ellipse. We split the ∆ν using three ellipses, imposing a
timing constraint:


 ∆ν = −12deg





∆νorbit = −4deg/orbit





ktgt = kint = 3









aGEO ∼

= 42164km

s s 2
a3ph a3GEO

∆v 3
kint 2π = (ktgt 2π − ∆ν) → aph = 1 − aGEO ∼
= 42788km > aGEO
µ µ 6π
Space Flight Dynamics Michele Santarpia 43

The three ∆vs are:

 s   r
 2 1 µ ∼

∆v1 = µ − − = 11.26m/s
a a a

GEO ph GEO







∆v2 = −∆v1 = −11.26m/s




∆vtot = |∆v1 | + |∆v2 | ∼


= 22.52m/s

3.4.2 Heliocentric Problem - Earth to Mars Transfer

Example 6.10 from Curtiss: let us assume circular and coplanar orbits (this is an approximation since orbits of
Earth and Mars are not circular and coplanar around the Sun). We want to compute which has to be the relative
position between Earth and Mars so that we can rendezvous with Mars starting from the Earth using an Hohmann
transfer. So, we have to find:
• The time needed for the Hohmann transfer.
• The required ∆νhoh .

∼

8
rL = 1.496 · 10 km

rh ∼
= 2.279 · 108 km
3
µs = GMs = 1.327 · 1011 km


s2

rL + rh ∼
ahoh = = 1.80 · 108 km
2 s
a3hoh
τhoh = π
µs

In this case µ is the gravitational constant of the Sun. To fly from Earth to Mars orbit, we need about 258 days
(about 8.6 months). Regarding the ∆νhoh , we need to apply the timing constraint equation from which we can find
the optimal relative geometry between Earth and Mars:
s s
a3hoh a3h
π = (π − ∆νhoh ) → ∆νhoh ∼
= 44.33deg
µ µ

If we miss the right alignment how much time do we have to wait?

We need to evaluate the synodic period, and we have to wait more than two years (2.16 years) to have the same
relative position between Earth and Mars:
2π ∼
τsyn = = 777.9 days
|nL − nh |
where:
µs µs
r r
nL = nh =
a3L a3h

It is clear that when we consider the problem of how often we want to have mission to Mars, we can use a general
transfer strategy which is not necessarily the Hohmann one. We can plan this type of mission with a general timing
and we do not need to wait for this ideal geometry but, of course, if we are not in the optimal geometry we will pay
in terms of energy that we need for the transfer. When the relative angular position is the right one, we have the
best combination between time and energy/cost otherwise we will pay in terms of propellant that we need. When
we consider this example, the budget contains some approximation because we are considering just the heliocentric
part, we are forgetting that initially we are under the gravitational field of the Earth and finally of Mars. Of course,
due to the time of the maneuvers, some typical approach as bi-elliptical lose their sense in this context due to high
values of maneuvering time.
Space Flight Dynamics Michele Santarpia 44

4 Orbit Determination
The Orbit Determination and Estimation can be performed using:
1. On-Board Systems (only active satellites).
2. Ground-Based Sensors (for all Resident Space Objects).
3. Space-Based Sensors (for all Resident Space Objects).

4.1 1) On-Board Systems

An active object like an orbiting satellite or an aircraft uses on-board systems to retrieve the position and velocity.
Of course, for non-natural object we cannot get this information. The main on-board systems we have are GNSS
receivers which stands for Global Navigation Satellite System which includes GPS, Galileo, Glonass, Beidou,
and so on. GNSS receivers exploit the presence of the GNSS satellites which are on MEO orbits (semi-major axis
in the order 20000km and a period of 12h).
Principle of Operation for GNSS Receivers: let us imagine we are an atmospheric user like an aircraft and at
a given moment the antenna on the vehicle has a number of GNSS satellites in view. Each GNSS satellite transmits
a signal to the aircraft (unidirectional transmission from satellites to aircraft). The number of satellites in view
has to be at least four. The basic concept is that, using the receiving signals, the GNSS receiver can compute the
distances from the satellites and by knowing the positions of the satellites in ECEF we can get our position.
Practical Example: let us imagine that the satellite ’GNSS SAT1’ is in a known position in space given by:
 
XECEF,SAT i
 YECEF,SAT i 
ZECEF,SAT i

We measure the distance RSAT 1 between the aircraft and the satellite GNSS SAT 1: using a single satellite, what
we conclude is that the aircraft has to stay on the surface of a sphere centered in the position of ’GNSS SAT1’ and
with radius equal to RSAT 1 . Using more satellites, we have more spheres and so in general using at least three
satellite we can remove the ambiguity on the aircraft position. In fact, by the intersection of the three spheres we
obtain just one point in space which satisfies the three constraints. So, we get the position of the aircraft.
Why the minimum number has to be equal to four if it is enough to use three satellites to remove
the ambiguity?
The reason why we need at least four satellites is related to an additional unknown that is related to timing: the
basic problem is that, since we are not using a two-way communication, there can be a clock error. In fact, the
distance between the aircraft and the GNSS satellite is measured by computing how much time the signal takes to
arrive from the GNSS satellite to the aircraft. The problem is that, since we are not exchanging any information
with the GNSS satellite, the clock of the receiver will measure the time in a different way w.r.t. the clock of the
GNSS satellites. So, the distance measurement contains a clock error since the clock of transmitter and receiver
are not synchronized w.r.t. each other. In conclusion, the clock error of the GNSS receiver becomes the 4th
unknown: this is the reason why we need to have at least four satellites in view. For this reason, the constellations
are built so that every user has at least four satellites in view.
Measurement Accuracy: let us consider to have a LEO satellite, let us imagine that our antenna is on the top
and it is in view of our GNSS satellites. At any given time, we will get signals from these satellites and typically
we will have enough satellites to compute our position. If we compare with a ground user, the difference is that the
relative velocity between GNSS satellites and LEO ones will be much higher but this does not make the operation
impossible because we have enough satellites in view and also these satellites are typically in a good geometry:
the accuracy of our position measurement will depend on how many satellites we have in view but will also
depend on the geometry of the satellites.
Poor and Good Geometry Example: let us imagine that we receive information from a number of satellites
which are concentrated in a relative small solid angles, so the directions from the different satellites are very close
each other, typically this corresponds to a bad/poor geometry and this means that, given the same uncertainty
about the distance from each satellite, we obtain a worst uncertainty on our final position. The good/sparse
geometry is the one in which the directions from the different satellites are very far each other.
Space Flight Dynamics Michele Santarpia 45

Uncertainty Example: let us consider to receive a signal from the satellite GNSS1. We know that the aircraft
has to be over the surface of a sphere but this surface is characterized by an uncertainty. Instead of having a sphere,
we have a sort of ”thickness” of this sphere. If we imagine GNSS2 in light blue, we have a similar situation, and in
this geometry there will be a large region of uncertainty which is compatible with both the measurements: this
is the reason why we have a poor geometry. Instead, if we get the same uncertainty but with a sparse geometry,
the final uncertainty in the position will be much smaller. So, depending on the geometry and with the same
uncertainty in the distance measurement, we can obtain different types of uncertainties on the final position.
GNSS vs LEO satellites: if we are a LEO satellite, we have typically enough GNSS satellites in view with a good
geometry. When we have a good geometry of navigation satellites, we say that we have a good dilution of
precision and so a good accuracy in the position measurement, while a poor geometry will be characterized
by a bad dilution of precision. So, large dilution of precision means large uncertainty on the final position.
Satellite vs Ground User: the big advantage for a satellite w.r.t. to a ground user like a car is that the signals
do not travel through the atmosphere. The atmosphere is a problem since it generates disturbances on the signals
which make the positioning uncertainty worst (ionospheric and tropospheric delay). For a satellite, we have a meter
level positioning uncertainty using GNSS satellites without significant differences compared with ground users.
GNSS vs Geostationary Satellites: if we are on higher orbits, the problem is a little bit different. Geostationary
satellites do not get signals from GNSS satellites below them, simply because GNSS satellites are designed to
broadcast signals towards the Earth and not in all directions. We can get signals from the satellites on the other
side of the Earth because conceptually we are able to receive those signals. This situation is more challenging and
so we are less accurate. There are at least three problems:
1. SNR goes down: the 1st one is related to the lower power of signal we receive due to larger distance so for
GEO satellites the SNR goes down.
2. Atmospheric disturbances: the 2nd point is that the signals may travel through the atmosphere because
we get signals close to the Earth and so we can have atmospheric disturbances and the level of disturbance is
larger compared with ground users because instead of having nadir propagation we have a propagation along
a path through the atmosphere.
3. Bad satellites geometry: the 3rd effect is related to the geometry which is worst because the satellites
which are in view now are in a relatively small solid angle, so the geometry will not be so sparse as in the
previous case (poor geometry w.r.t. LEO, large dilution of precision) so the uncertainty of position estimate
on GEO will be worst (tens of meters of error so not meter level).
GNSS vs Lunar Orbits: GNSS signals can be beneficial also in Lunar orbits, and so we can make position
estimate also when we are travelling towards the Moon.

4.2 2) Ground-based Systems and 3) Space-based Systems

Other sources of information (that works for any space object) are ground-based systems/sensors. They are
traditional sources of information (ground-based radars or telescope or laser stations or radio) and also space-
based sensors which are more challenging in terms of dimensions (we have something in space that is needed to
get information about the rest of space objects like a space-based telescope used to observe the orbits of our interest
and it is not observing universe or Earth). These sources work for all space objects.
For ground-based System, first of all, in order to retrieve enough information, the atmosphere has to be transparent
to the signals.
Radio-Frequency System vs Radar/Laser: there is a fundamental difference between a radar/laser or a radio-
frequency system (that relies on a transponder). When we use reflected energy as in case of lasers and radars,
the SNR has a dependency on the fourth power of the distance:
1
SN R ∝
R4
Instead, for re-transmitted energy in case of radio-frequency system the question is different because it is not a
concept of reflected energy but it is more the idea of interrogation response: so, we transmit an interrogation
and then the system gives a response with an independent transmission.
Space Flight Dynamics Michele Santarpia 46

So, re-transmitted energy is the idea of transponder which is a one-way communication where the SNR
is proportional to the 2nd power of the distance (less challenging) and so it can be used if we work with a system
that cooperates with us:
1
SN R ∝
R2

Radio-Frequency System: we send a radio-frequency interrogation, there is a system called transponder on

board the satellite that will send a response after a well-defined amount of time/delay. Through the transponder,
the information will travel back after some delay.
Active Sensors (Radars and Lasers): active sensors are used to transmit a signal towards our resident space
object and we compute the time delay between transmission and reception of the signal. Of course, we need to
apply some corrections to the basic formula due to antenna pattern and atmospheric effects. With radars, we have
meters level of uncertainty while for lasers we can arrive to centimeter level.
Lasers Domain of Application: the domain of application of lasers (techniques called SLR i.e. Satellite Laser
Ranging) is more limited w.r.t. radars. To operate lasers, we need to have some cooperative systems on-board
our target satellite, and so if we consider a generic orbiting body, the surface bodies must have good property to
reflect laser energy. To make an orbiting body bright from the POV of laser (to get enough energy), we install
retro-reflectors on the surface of the orbiting body which reflect all the incoming energy.
Lasers Challenges:
1. The laser has a very narrow beam w.r.t. to radars.
2. High power lasers can be a problem for the optics of an optical payload (pay attention!).
Information retrieved by Lasers and Radars:
1. Lasers are used for ranging measurements.
2. Radars are used for range, range rate information (component of relative velocity) and angles. Specifically,
the range rate Ṙ is computed by measuring the doppler shift fD :
2 dR
fD = −
λ dt

Use of Topocentric Reference Frame: let us consider the Earth, a point identified by the intersection between
a meridian and a parallel, and the Topocentric Horizon Reference Frame. rtopo is the position vector of the satellite
in the Topocentric Horizon Reference Frame. We can project rtopo in the local horizontal plane. In this reference
frame, we refer the azimuth w.r.t. the North direction in clockwise (the North direction is in opposite w.r.t. South
direction). So, once we have projected in the local horizontal plane, we identify the azimuth and the elevation.
Uncertainty: the uncertainty in the measurement is typically linked to the aperture angle of the antenna θ:
λ
θ∝
D
where λ is the wavelength and D is the diameter. The equation about θ works also for telescopes, where D is the
optics diameter. For the optical sensors, it depends on the diffraction theory (instead of antenna theory). From
lasers, as output, we typically get only range with high accurate below 1m. From telescopes, as output, we only get
angles, i.e. right ascension and declination, produced w.r.t. topocentric equatorial.
Quality of data:
1. In case of troposphere, the quality of data depends on the weather (humidity, etc.).
2. Instead, for ionosphere, we typically have a delay of the signal that is proportional to the inverse square of
the frequency.
We need to compensate for this effects using information about weather measurements and models.
Quantity of data:
Space Flight Dynamics Michele Santarpia 47

1. Through ground sensors, we can get measurements only in some conditions, of course with a telescope we
will be always dependent on weather (cloud coverage). It is clear that the relative geometry between orbiting
objects and the ground sensors will give us how many observations we have.
2. Radar are all time all weather sensors.
Limits:
1. The limits in range exist only for radar.
2. For a telescope, we have problem for minimum apparent magnitude that is detectable.
3. We can have limits in azimuth and elevation which depends on the obstacles (we try to install telescopes on
mountains in order to have no obstacles in the FOV). There is also a minimum on the angle over the horizon
which degrades the observation quality due to the atmospheric effects. We typically calibrate the sensors in
terms of bias and noise.
[Discussion could be improved on the sensors]
Calibration: we need cooperative objects for calibration and some satellites whose positions are accurate very
well. For example, GNSS satellites are nice because, for their operations, the orbits are determined very accurately
and so we can use them since the accurate positions of GNSS is given periodically. If we have a telescope to be
calibrated, we can spot with it some GNSS satellites and we take very accurate ephemeris (very accurate orbital
information) which represents the benchmark.
Information we can deliver if we pass from sensors measurement to orbital state: if we know the orbital
parameters of an object, we can deliver a number of services, and one of the scope is, for example, to avoid a
collision risk and so if we have to maneuver, when to maneuver, or to know if some space systems is acting as a
thread for other systems.
The question now is: how do we pass from what sensors measure to the orbital state?

4.3 Introduction to IOD and POD

Initial Orbit Determination (IOD): there are some methods to get the orbital state of an object for the first
time and without any prior information using sensors measurements in order to understand which is the orbit.
Precise Orbit Determination (POD): we know some information in advance, we are tracking/following the
motion of the object, as we receive new measurements we inject them to improve the accuracy. It is a process that
receive three types of inputs:
1. Prior information about the object (previous estimate of orbit determination).
2. Dynamic model of the object.
3. New measurements from the sensors.
From the magic box, we get an updated orbit determination and so, for each object, we will update the orbital
state in terms of position and velocity in the inertial frame or orbital parameters. There are two main approaches:
1. Batch processing: we accumulate a number of measurements (temporal sequence) and then we use these
information all together: so, if we want our orbital reconstruction, it will be the one that fits in the best way
all the history of measurements (numerical fits).
2. Recursive processing: as we get measurements, we inject them in the estimation (we do not wait) to
improve it. The classic strategy by which we perform recursive processing is an algorithm called (extended)
Kalman Filter. When we consider these three inputs to the POD box, all of them are characterized by an
uncertainty. The scope of this algorithm is to have an updated estimate of the orbit determination that gives
the right weight to the different pieces of information that we want to combine. This algorithm will give an
higher weight to what is considered more accurate. This is the idea of Kalman Filter state estimation
and this is the same idea that we use for orbital determination. The way through which this algorithm works
is the reason why we need to calibrate our sensors because when we have new measurements we need to know
which is the uncertainty associated to these measurements in order to give the right weight.
Space Flight Dynamics Michele Santarpia 48

When we have new unknown objects (as result for example of fragmentation events), as we have some measurements
we need to make first of all an Initial Orbit Determination to understand which is the orbital state and where they
are coming from. Then, we will go through Precise Orbit Determination.
In the concept of precise orbit determination, we want to have the best estimation of the state and this is called
state estimation problem. There are different algorithms like filtering ones that are needed to combine all these
informations in input to the precise orbit determination block.
Dynamic Model: a dynamic model expresses the derivative of the orbital state of the object x, or generically
called state of the system, (which contains the position and velocity and that we want to estimate) as a vectorial
function of the orbital state and the input vector u:

ẋ = f (x, u)

So, our knowledge of the orbital dynamics will become a function of this type which explains how the state will
evolve in time. When we play with POD we are trying to be more accurate, so we need to go beyond the Keplerian
Mechanics, so the idea is that in the function f we include perturbations.

4.4 Initial Orbit Determination

Initial Orbit Determination methods: we can use different methods to have an initial estimate of the orbit
state and in this case we apply Keplerian mechanics. It is dependant on the measurements we have. Depending on
the method that we use, we can have different input information:
1. Site-Track Method: at a given time instant and w.r.t. the Topocentric Horizon Reference Frame, we know
the position vector rT OP O and the velocity vector vT OP O of an object. It is not a truly orbit determination
method since we need to just convert in the ECI frame the information that we take thorugh the sensors.
2. Gibbs Method: we know three position vectors to the object w.r.t. ECI frame r1 , r2 , r3 in three different
and consecutive time instants.
3. Lambert’s Method: another possibility is that we have two time instants t1 , t2 and two position vectors
r1 , r2 .
4. Gauss Method: we only have vectors of directions û1 , û2 , û3 in the topocentric equatorial reference frame
(telescope).
5. Numerical Method No.1: we have sensors that measure the distances at different time instants Ri (for
i=1,...,N).
6. Numerical Method No.2: we can have distance Ri and range rate Ṙi (for i=1,...,N).
1) Site-Track Method: we know the position vector r and the velocity vector v of an object in the Topocentric
Horizon Reference Frame at a given time instant. Let us consider an inertial frame and a given ground site (where
there is the sensor which is the source of our information). Let us consider the meridian passing through the GS.
From ECEF to topocentric horizon, we have a sequence of Euler’s angles 3 (λGS ) and 2 (π/2 − ϕdGS ). So, we know
position and velocity of a space object in the topocentric, and we want to know the orbital state: this is not a truly
orbit determination technique but just a conversion in the correct reference frame.

rECEF = rT OP O + rGS

We need to consider the transformation matrix MECEF →ECI in order to obtain rECI . The rotation matrix is related
to an angle αGR (t) of type 3.
In ECEF and TOPO, the velocity is the same since they have a constant geometry w.r.t. each other, there is no
angular velocity of the topocentric because the GS is fixed w.r.t. ECEF.

v ECEF = v T OP O

In ECI:

v ECI = v GS + v T OP O + ΩL × rT OP O v GS = ΩL × rGS
Space Flight Dynamics Michele Santarpia 49

But this is not a true orbital determination technique but it is called in Vallado ”site-track”. The sensor itself is
not giving us directly position and velocity in the topocentric frame. A direct measure is, for example, Ṙ in doppler
radar.
2) Gibbs Method: we use three position vectors (not distances) in the inertial reference frame (ECI) in three
subsequent time instants. Whatever sensors we have, we will not have directly the vectors in the ECI reference
frame but we need to convert the measurements. For example, in case of radars, we have to pass from topo to
ECEF and from ECEF to ECI.
We can estimate which is the orbit passing through these three position vectors (single solution). Gibbs Method
uses the equations under the Keplerian assumptions and it is a fully analytical method (no approximations), so we
have a direct procedure that, starting from the position vectors, allows us to obtain the orbital state.
In particular, the idea is that, starting from the positions vectors, we can compute the velocity vi (where i=1,2,3)
and so if we have these three vectors, Gibbs gives a set of equations that allow us to compute which is the velocity
in one of the three positions. So, if we know the velocity in one of the three positions, we know the orbit and the
problem is solved.
Remember that the method does not use time information, so we do not need to give accurate time to the positions
vectors.
Uncertainties: Gibbs method is analytical but each input information is characterized by an uncertainty σri
(where for each component of the position vector, we have an uncertainty). Remember that Gibbs method is
exact but under Keplerian assumptions, so another source of error is related to the fact that in reality these
assumptions are not true. As we will see, timing is important to understand how large is the error we make.
Initially, we apply these methods without paying attention to perturbations because the other sources of error are
larger.
Let us consider two scenarios (which have the same position uncertainties):
1. Three position vectors very close to each other (short arc).
2. Three position vectors largely spaced (long arc).
What is the best scenario?
In order to understand this, we need to understand how the (sensor) uncertainty propagates, through the Gibbs
method, to the orbit determination uncertainty.
The uncertainty can be seen as a sphere: when the observations are closer in space, they will generate a velocity
error which is larger. If the observations are relative to a longer arc of the orbit, the uncertainty will propagate to
a smaller final orbit reconstruction error. So, the propagation of the uncertainty will depend on the geometry of
the observations. Finally, we can say that there is an final error due to input uncertainties.
Gibbs Method Variant: another possibility is that the observations/the three position vectors are close to each
other and so they correspond to a short arc of the orbit. While the traditional Gibbs method suffers of some
numerical computations, the variant Herrick-Gibbs works better (a sort of adaptive Gibbs method).
Deeper Analysis of Gibbs Method: let us say something more only about Gibbs Method (other methods are
only described in summary later)!
In the Gibbs case, we have nine scalars in input (given by three components for each position vector ri ) and we
want to compute six scalars in output (given by r and v in one of the three positions).
This method applies the basic results of the Keplerian motion. We can exploit the the 1st result of the Keplerian
motion i.e. the fact that the motion happens on a plane.
Let us define the following unit vectors:
 r
 û1 = 1


 ||r 1 ||

 r × r3
ĉ23 = 2


||r2 × r3 ||
Space Flight Dynamics Michele Santarpia 50

The unit vector ĉ23 is normal to the orbit plane.

Mathematically, the planarity can be expressed in two ways:
1. Planar motion means that when we consider the dot product between û1 and ĉ23 (prodotto scalare), it will
be equal to zero:

û1 · ĉ23 = 0

2. Another way to express planarity is based on the following concept: if three vectors belong to the same plane,
one of the three can be expressed as a linear combination of the other two. The vectors are not linearly
independent.

r2 = c1 r1 + c3 r3

Then, we use the eccentricity vector e definition:

 
v×h r r
e= − →v×h=µ e+
µ r r
Let us premultiply vectorially by h:
   
r h×r
h×v×h=h×µ e+ → h × (v × h) = µ h × e +
r r

We use double vectorial product called BAC-CAD rule:

A × (B × C) = (A · C)B − (A · B)C
   
Ŵ × r µ Ŵ × r
h × (v × h) = (h · h)v − (h · v)h = h2 v → h2 v = µ hŴ × eP̂ + h →v= eQ̂ +
r h r
If we solve the equation of the velocity in one of the three time instants, we have solved our problem. At the second
member, we have something that we already know in any of the three time instants i.e. r and its norm r. If we are
able to determine the elements at the 2nd member i.e. norm of the angular momentum vector h, the eccentricity e,
the directions Q̂ and Ŵ (remember that these are constant elements), starting from r1 , r2 and r3 , we can determine
the velocity. The key idea is to use all the other Keplerian mechanics equations we know:
 √

 h = µp




 p
r=

 1 + e cos ν




p = a(1 − e2 )


Finally, we obtain that the velocity v is given by:

r  
µ D×r
v= +S
ND r
First of all, we need to find the constants of the motion as function of r1 , r2 and r3 :



 N = r1 (r2 × r3 ) + r2 (r3 × r1 ) + r3 (r1 × r2 )




D = r1 × r2 + r2 × r3 + r3 × r1






S = r (r − r ) + r (r − r ) + r (r − r )
1 2 3 2 3 1 3 1 2

Then, for example, we write the the equation at the time t1 to find v 1 . The important relation is the final one, all
the passages to derive it have not done by the professor. Finally, we obtain an equation for the velocity vector v
Space Flight Dynamics Michele Santarpia 51

where we need to remember that it depends on N, D and S which are function of the three position vectors r1 , r2 ,
r3 . So, as we can see, there are no approximations (no development in series), it is an analytical equation.
3) Lambert’s Problem: we have in input two position vectors r1 and r2 in ECI which give us six scalars plus two
time instants t1 and t2 . It consists of a two-body problem plus boundary conditions:
 µ
 r̈ + 3 r = 0



 r


r(t1 ) = r1 → r, v






r(t2 ) = r2


We can have two type of solutions:

1. In the first one, we have an orbit where ∆ν < 180deg.
2. In the second one, ∆ν > 180deg.
Lambert’s problem can be considered as orbit determination technique or as an orbital maneuvering tech-
nique depending on our scope.
4) Gauss Method: we work in topocentric equatorial reference frame by knowing three unit vectors of directions
û1 , û2 , û3 (not distance, but only angles) in three different time instants, obtained through a telescope. Each unit
vector is characterized by two angles, the right ascension α and declination δ, which individuate the direction of an
object for each time instant. Again we have six scalars in input and we can compute the orbital state given by r
and v.

(t1 t2 t3 ) → [(α1 , δ1 ) (α2 , δ2 ) (α3 , δ3 )] → r, v

When we do not have the distances in input, we have less information and so the final error is larger (additional
uncertainties).
Different applications need an orbital reconstruction with different accuracy! Practically, a good initial
orbit determination (and so a good uncertainty we can accept) could be the one that allows us to plan the acquisitions
of the other sensors in the future to refine the estimation. If the error is so large that we do not know how to point the
telescope in order to get measurements, this could be a problem. Sensor Tasking means planning the acquisitions
of the sensors (possible benefit of this method). Improving orbit determination is important because lower is the
uncertainty, the more efficient will be the collision avoidance planning maneuver.
5) and 6) Numerical Techniques: we can use range R and range rate Ṙ information that we obtain through
radars and lasers. The solution will be a numerical one and the orbit is the best fit:
1. Numerical Method No.1: we have sensors that measure the distances at different time instants Ri (for
i=1,...,N).
2. Numerical Method No.2: we can have distance Ri and range rate Ṙi (for i=1,...,N).
Space Flight Dynamics Michele Santarpia 52

5 Orbital Perturbations
Related to orbital perturbations, our objective is to answer the following questions:
1. Which are the main perturbations that we have to consider?
2. Which are the force models that we have to take into account?.
3. Which are the effects on the orbit (related to maintenance)?
As we know, the basic assumptions of Keplerian mechanics that we have considered are:
1. Only two bodies.
2. Only gravitational forces: this means that we are considering that our central body has a spherical mass
distribution (the spherical gravity equation come from the universal gravitational Newton’s law) since we are
not taking into account the non-sphericity effects.
3. The mass of the central body M is much larger than the mass of the orbiting body m (M>>m):
this assumption will be kept since it is reasonable (for example, the mass of the Earth is much larger w.r.t.
the one of the satellite).
The Kepler’s problem is expressed by:
 µ
 r̈ + 3 r = 0



 r


r(t0 ) = r0 → r(t), v(t)






ṙ(t0 ) = v 0


What are the problems with the assumptions of Keplerian dynamics?

1. Only gravity not true: we have SRP and aerodynamic forces. Furthermore, we have non-sphericity: the
mass distribution is not spherical since the mass is not uniformly distributed inside the planet.
2. Not only two bodies: we do not have only two bodies but we need to take into account the effects of the
gravitational attraction of the other bodies (for example, Sun and Moon for Earth orbits).
Short vs Long Time Scales: we talk about orbital perturbations because we will have in general a small change
in the behavior of our orbiting body, so the Keplerian approximations will give still a solution which is reasonably
accurate at least in short time-scales.
If we imagine to propagate the Keplerian orbit and if we check which is the real solution, they are not completely
different: of course, there will be a difference between them and the spatial separation (the propagation error)
will increase over time.
So, the reason for taking perturbations into account is connected to the timescale of our analysis. If we are
interested in a shorter timescale, the Keplerian approximations/solutions still fit our need, so the Keplerian
solution is good. Of course, if we are interested in more accurate solution or longer timescale, we need to consider
some additional effects which modify the Keplerian solution.

5.1 Perturbations Effects

1) Non-Sphericity Effects: first of all, when we consider the perturbations we have to take into account that our
primary body is not a sphere. If the density is not uniform, we will have that the gravitational force which operates
on the satellite is not directed to the center of the reference frame and this means that there will be a torque (so,
we do not have a central force). So, the orbital plane is changing:

dh
̸= 0
dt
Space Flight Dynamics Michele Santarpia 53

In Keplerian assumptions, the mechanical energy is constant along the orbit: in this case, whatever is the distribution
of the mass and so whatever is the gravitational field of the central body, since the gravity is always a conservative
force, on an average perspective we expect to have not a change in energy:
dε
=0
dt

2) Third Body Effects: if we do not consider the third body effects, we are neglecting the difference between
the third bodies gravity attractions on the orbiting body and on the primary: so, the 3rd bodies will attract both
the primary and the orbiting mass, and the gravity on the orbiting body (since it is closer) will be larger. This
perturbing force is not central and so there is a torque which affects the orbital plane:
dh
̸= 0
dt
Again, since it is gravity force perturbation, the mechanical energy is not affected by the 3rd bodies (the gravity is
conservative so it cannot take or add energy from/to our orbiting mass).

dε
=0
dt

3) Drag: in LEO orbits we have very rarefied conditions, and so from a fluid dynamics POV the regime is called
free molecular flow. The free path of the molecules is much larger than the dimensions of the objects, so every
molecule exchanges momentum and energy with the body, so we cannot expect the same behavior we have in
the continuum domain. Of course, satellite are odd-shaped bodies (“corpi tozzi”). Due to the particular fluid
dynamics domain and due to the shape of the satellite, finally the only aerodynamic force is Drag (we have no lift).
If we consider the velocity of the satellite flying in the atmosphere, there will be an aerodynamic Drag opposite to
velocity. The velocity that we have to consider of course is not the inertial one but the one w.r.t. the atmosphere.
The velocity w.r.t. the atmosphere depends on the assumptions that we make on the atmospheric model. In general,
it can be seen as the difference between the inertial velocity of the satellite minus the velocity of the atmosphere
(how the atmosphere moves in the inertial frame):

v Sat−Atm = v − v Atm

Who is v Atm ? Let us consider two possible scenarios:

1. Earth rotates but the atmosphere does not rotate.
2. Rotating atmosphere which keeps a fixed geometry in the Earth-fixed coordinates.
In reality, none of the two options is true: there is a very particular wind effects at hundreds of kilometers of
altitude. This introduces some complexities which introduce some uncertainties.
So, in general, we can model the motion of the atmosphere but, of course, we have to consider that there are some
local spatial variations (vertical motion, horizontal motion if we change position etc.) and temporal effects.
Finally, we can say that the real uncertainty does not come from velocity uncertainties but from the atmospheric
density model.
The 1st effect is that Drag (non conservative force) takes energy away (dissipative), so the mechanical energy ε
will decrease and so the semi-major axis a will decay:
dε
<0
dt

While 3rd body and non-sphericity does not depend on the configuration of the satellite, instead Drag depends
on the section of the satellite exposed to the atmosphere and so it is related to the exposed area: what is
important is the ratio Area/Mass:

Aatm
m
Space Flight Dynamics Michele Santarpia 54

So, two different objects will be subjected to two different Drag. This will give us an additional uncertainty. But
from space surveillance side this is good because by monitoring the behavior of the object under the effect of Drag,
we can estimate which is the area over mass ratio.
4) Solar Radiation Pressure (SRP): the incoming radiation will have a given direction and it interacts with
the satellite. The electromagnetic radiation will exchange momentum with our surfaces, there is some absorbed
radiation, and some out-coming radiation. This exchange of momentum will imply a mechanical pressure
(”Sun is pushing the satellite”) which produces an overall force and a total torque. As in the Drag case, we have a
strong dependency on the Area/Mass ratio (where in this case, it is important the area exposed to the Sun):

As
m
By nature this is a non conservative perturbation, so there is a change of energy and it is not necessarily negative
(for example, higher than zero in case of solar sail).

dε
̸= 0
dt

5.2 Cowell’s Formulation

We need to change the equation of motion where at the 2nd member we have a perturbing acceleration ap (specific
force i.e. per unit of mass).
 µ
r̈ + r3 r = ap






r(t0 ) = r0 → r(t), v(t)






ṙ(t0 ) = v 0


We can model also the thrust that we give to our satellite as perturbation: but what type of thrust?
1. If we think about impulsive maneuvers, this looks like not so reasonable, we concentrate the variation of
the velocity at a given time and then we are Keplerian again. The idea of impulsive maneuvers is that we
evolve naturally from point to point and then in some points we change velocity.
2. When we use low thrust (so, finite duration of the burns), since they cannot be modeled as impulsive
maneuvers, we can model the low thrust as a perturbation (so it becomes the fifth type of perturbations).
So, in ap , we need to consider the low thrust control action.

5.3 Relative Importance of Perturbations

The relative importance of perturbations depend on the orbit:
1. Drag is strongly affected by the altitude and the dependency is exponential. It makes sense only in LEO,
and above a given altitude (700/800km) it goes down and so it is smaller than other perturbations.
2. SRP does not depend on altitude (constant) but it is true that we pay more attention to SRP when we are
in higher orbits (MEO and GEO) due to the change of the relative importance in magnitude with the other
perturbations (it becomes comparable).
3. 3rd body effects increases with altitude because the delta gravity w.r.t. to the Sun / Moon between Earth
and satellites increases. We will take into account in MEO/GEO principally.
4. Non-sphericity is a deviation from spherical gravity, so we are adding a variation to the force due to spherical
gravity. Gravity of the primary decreases by increasing the orbit altitude, and so the absolute effect decreases.
In relative terms, we need to consider the non-sphericity effects in all the orbits (LEO, MEO, GEO). In all
these orbits there is the generation of some strange phenomena: for this reason, we cannot consider that this
perturbation gives a negligible effect.
Space Flight Dynamics Michele Santarpia 55

When we play with non-sphericity and 3rd bodies effects, we may have some resonance phenomena. The
perturbations effects might be relatively small but the relative geometry between the satellite and the primary body
causes a periodicity in the effect of the perturbation which can activate some resonance. With resonance, even a
small perturbation will show some non small effects, because we have a periodical perturbing force that always
accumulates due to resonance.
Finally, we can say that in LEO, the main perturbations are Drag and non-sphericity.

5.4 Methods
They can be divided in:
1. Special Perturbation Techniques: Ordinary Differential Equations (ODE) are solved numerically, and it
is the most accurate method in reasonable time scales and we use this approach when we want to be more
accurate then Keplerian. Numerical solver to use, we discretize the time using a time-step, we compute an
approximated solution, we will propagate forward our motion and with some techniques we are able to com-
pute the solution. If we have a long propagation (long time), the problem is the computational time.

Imagine we want to understand how the collision risk evolves in the next 100 years due to the all existing
objects and predicting also the launch of a given number of mega constellations. So, we need to propagate
thousands of objects for a very long time: the computational time will explode! So, again, a purely brute
force technique for very long time propagation is not good. But it is also a problem of approximations that
we enter, because we have some round-off and truncation errors because, as we propagate numerically we
make some approximations in our computer, and this is the reason why for very long propagation intervals,
the classic numerical techniques/schemes are not the best ones so we need to use other approaches in this cases.

A problem is that these techniques are special which means that if we consider to run a propagation on
GMAT (where we give our initial state, the parameters of the spacecraft and the forces and so on), we will
obtain the result which is the position r(t) and velocity v(t) for all the time interval but, of course, this is
an approach where we do not get general conclusions because if the orbit is not useful in its behavior and if
we have to change it, we have to change the initial conditions. The only approach that we can use to solve
is a sort of trial and error which is not good at design level because we have to understand which are the
degrees of freedom with which we can play and so we are not able to get general conclusions about the effects
of the forces. So, these techniques are very powerful but not for all the problems.
2. General Perturbation Techniques: since in the past there were computational limits, it was not possible
to apply special techniques. The general perturbation techniques are still important since we can retrieve
general ideas about the effects of a single perturbation. They are very powerful at the design level. Instead of
spending propellant in a not smart way to fight against the Drag, for example, we can design in such a way
the orbit taking into account the perturbations.
3. Semi-Analytical Techniques: powerful for very long propagation. They tend to be full analytical, they are
analytical thanks to approximations and so we need to accept some large errors. They use analytical tools
so that if we have a very long propagation time, what we can do is to write some equations good for long
timescales. The key concept is the interval of time in which we make our numerical computations. While
in special techniques, we want to have a small time interval in order to reduce the approximations when
we propagate forward, in these cases we use some developments so that our equations can be solved with a
very long ∆t like for example 1day. So, we use numerical tools to propagate directly for the next day, so we
propagate numerically with long time steps and then we use analytical developments to pass from this long
timescales to the behavior in a short timescale, so we use the numerical power of our computer to make the
longer jump, and then we have some theory that given what happens with long timescales, we can know what
happens inside each interval. This becomes very powerful because now we can save orders of magnitude of
computational time and we will not suffer of the same errors that we have before.
The numerical methods can be classified in different ways:
1. Single Step vs Multi Step propagator.
2. Fixed Time Step or Variable Time Step size.
3. Single Integration or Double Integration.
Space Flight Dynamics Michele Santarpia 56

Let us analyze deeper all these different methods:

1. Single Step vs Multi Step: most of the methods are single-step which means that we are propagating our
solution in a single-step, so we are propagating, for example, the solution from t1 to t2 directly and we use the
values of the derivatives computed at intermediate points. Other methods, like the Adam-Bashforth-Moulton
(ABM), are multi-step which means that the propagation happens using two-step (we have a prediction
correction solving scheme).
2. Fixed vs Variable Step Methods (step to solve the orbital dynamics): when do we use fixed or variable?
It depends on the orbit.
(a) A constant time-step makes sense for near or circular orbit where the constant time-step will correspond
to a constant anomaly difference, and so it makes no sense to use a variable one.
(b) If the orbit is an ellipse, we need to use a variable one because it makes more sense to have a constant step
in the angle instead of having a constant step in time, in this way we sample the orbit with a constant
angular sectors.
3. Single Integration and Double Integration Methods: single integration methods solve numerically 1st
order equations while double integration methods solve the 2nd order ones.

As we know our problem is expressed at the 2nd order:

 µ
 r̈ + 3 r = ap



 r


r(t0 ) = r0 → r, v






ṙ(t0 ) = v 0


But equivalently we can pose our problem using two 1st order differential equations:
 µ
 v̇ + 3 r = ap



 r




ṙ = v








 r(t0 ) = r0





v(t0 ) = v 0


From an analytical POV it is the same, but from a numerical POV we may have some methods which solve
the 2nd set of equations. Instead of having one vectorial 2nd order equation, we have two vectorial 1st order
equations so the maximum order of the derivative is the 1st one.
Runge-Kutta: it is a classic example of a single step method with an adaptive time step. For example,
Runge-Kutta 8-9 are methods characterized by an adaptive time step i.e. the step by which we propagate
forward our solution is not constant and it changes in an adaptive way. These techniques are based on development
in series and 8-9 means that we propagate using formulations of eight or ninth order. Then, we make a comparison
between the computations of the derivatives. When the difference is below a given tolerance, we consider the time-
step to be acceptable, if the difference is above, we assume that the time-step is too large and so we will reduce
it.
Sampling Theorem (Nyquist-Shannon): minimum frequency to have in order to sample a signal in order to keep
the original information content. Signal in the image at least two points per period (theoretical minimum).

fs ≥ 2f

where fs is the sampling frequency while f is the frequency of the generic function f(t).
Space Flight Dynamics Michele Santarpia 57

When we propagate numerically with a given time-step, we are sampling the perturbing effects i.e. the perturbing
accelerations ap , as we know we are trying to solve numerically the problem:
 µ
r̈ + 3 r = ap



 r


r(t0 ) = r0 → r, v






ṙ(t0 ) = v 0


To solve it, we use a given time-step. What we are sampling are the perturbing accelerations/forces, because a
perturbing force will have a mathematical expression which depends on the orbital state.
If we write our approximated equation only at some time-steps, we will write our perturbation only at these time-
steps. So, we are modeling the perturbation at these time-steps. Every will depend on the temporal (or spatial if
we consider the anomaly) frequency. A given perturbation will have a certain frequency (i.e. it is characterized by a
certain frequency content). If we consider the Fourier transform, for example, we will obtain the frequency content
of the function which represents the perturbing force, and so if we have a function f(t), by transforming we obtain
F(f) and we see what are the frequencies contained in this function. In the case of a sine function we have just one
frequency.
When we sample our orbit (when we propagate forward), we are sampling the function. So, the time-step must be
small enough to sample in an adequate/proper way the function. Higher is the (temporal/spatial) frequency of the
perturbing force, the smaller the time-step has to be.
Regularization: sometimes, we have fixed step but not in time. For example, we can have a constant step in
anomaly. This idea of using a constant step not in time is called regularization. So, the time is variable since
we can write an equation like the following one where we have a coefficient c, r which is the norm of the position
vector and s is the sampling variable (i.e. the constant step):

dt = crn ds

If we have that:



 n=2

1
c = p



µa(1 − e2 )

We have that ds=dν, so the constant step is in true anomaly:

r2 ds
dt = p
µa(1 − e2 )

So, we have that:

√ p r2 ds r2 ds
h= µp = µa(1 − e2 ) → dt = →h=
h dt
By the 2nd Kepler’s Law:

h = r2 ν̇

Equating, we have that:

ds ds
r2 ν̇ = r2 → ν̇ =
dt dt

Of course, if the orbit is circular it will be also a constant in time.

Space Flight Dynamics Michele Santarpia 58

5.5 Special Perturbation Techniques

Let us analyze the special perturbation technique called Encke’s method.
Encke’s Method: special perturbation technique used before the development of modern computers. It contains an
idea useful for general perturbation techniques: when we have a perturbed trajectory, we can consider a Reference
Keplerian Trajectory and we can work on the difference between the reference and the perturbed orbit because,
if we have a short time-scale, the difference between the Keplerian propagation and the perturbed one should be
small.
Imagine to have a primary and an orbiting body. Our propagation problem has an initial position vector r0 and
initial velocity v 0 . We consider the Keplerian solution ρ(t) and the perturbed one r(t) which are characterized by
the same initial conditions.
Perturbed trajectory is in red, while Keplerian in light blue. We know ρ(t) and ρ̇(t), and we want to know r(t), ṙ(t).
At a given time we have r(t) and ρ(t). Let us consider the term in red i.e.:
ρ3
 
1− 3
r
The main idea of this method is to write an approximated equation for the term in red by defining an auxiliary
variable q: in this way, the term in red can be seen as qF(q) where F(q) can be computed just one time and put
in a table (for solving numerically without computer was good!).
ρ3 ∼
 
1 − 3 = qF (q)
r
The key approximation made to solve the problem through this method was that δr should be relatively small.
So, this method could work for a limited amount of time. So, instead of computing the perturbed solution, we can
propagate Keplerian which is easy and then we can compute numerically the δ w.r.t. the Keplerian solution. At a
given moment, it becomes too large δr and so this method also contains a sort of adaptive logic: when δr becomes
too large, we change the Keplerian solution that we use and we consider the one with initial conditions equal to the
current position and velocity, so we consider a new Keplerian orbit kep2 (the violet one) and we start again. So,
with a given periodicity, we change our Keplerian solution (re-initialization of the Keplerian solution) and
we compute the difference. The Keplerian solution is called osculating orbit, and this gives us the idea that it is
the Keplerian solution which fits the initial position and velocity that we have (this is the Keplerian solution that
instantaneously corresponds to the real motion i.e. has the same behavior of the perturbed trajectory and then in
time there will be a difference). The operation of changing the osculating orbit is called rectification. This idea
of osculating orbit is used in the general perturbation framework.

5.6 General Perturbation Techniques

General Perturbation Techniques: as we know, the orbital state is described by six scalars (position and
velocity vectors) which change in time. We compare these six scalars to the orbital parameters that in Keplerian
dynamics only one of them changes in time.
We will start from our 2-body problem, we want to know position and velocity in every instant of time and we
start from the initial conditions from which we can pass to the six orbital parameters: the 6th parameter can be
one among the true anomaly ν(t), eccentric anomaly E(t), mean anomaly M(t), mean anomaly at time t0 i.e. M0 ,
or we can also have the time of the last passage at perigee tp because from it we can evaluate M(t) (of course, at
the instant tp we can write that M0 is zero since we are at the perigee):

M (t) = M0 + n(t − t0 ) → M (t) = n(t − tp )

Under the perturbing forces, the motion will be a perturbed Keplerian one where the orbit is like a Keplerian orbit
that slowly changes its characteristics. For this reason, we can still use orbital parameters but the difference w.r.t.
before is that now the Keplerian orbit has to change in time and so all the parameters are function of time.
Regarding the time-dependency, there are two time scales:
1. There is a slow motion, corresponding to a long time scale, related to the first five parameters (they change
slowly).
Space Flight Dynamics Michele Santarpia 59

2. There is a short time scale at which the 6th parameter (for example, the true anomaly) changes faster.
The derivative dν/dt evaluated in the perturbed condition will be slightly different from the one in the Keplerian
assumption.
The idea of considering the time-variation of the orbital parameters is the idea of a set of methods which are under
the name of ”Variation of Parameters (VOP) methods/equations”.
When we consider the perturbing scenario, the sixth parameter can be the true anomaly but at the same way we
can use again the mean anomaly M(t) where now:
   
dM dM
̸= =n
dt pert dt kepl

These derivatives are different. We can play the same game with E(t). The sixth parameter can be also in this case
M0 but w.r.t. the case of Keplerian where it is constant, in this case we have M0 (t) because this corresponds to the
idea that we have a Keplerian solution that changes, so all the orbital states were described by six parameters with
the last M0 (t), or we can consider the Keplerian passage at the perigee tp (t) which is a function of time.
We want to write all the derivatives.
We say osculating or instantaneous orbital parameters because they correspond instantaneously to the
Keplerian orbit that approximates our motion. The derivatives depend on the current (state of the) orbital
parameters and on the perturbing force ap . If we write these six equations, they are called Gauss VOP
equations. This is a very general tool, whatever is the perturbation (non-sphericity, 3rd body, thrust, SRP, Drag
etc.) the equation will work.
We can also give a look to another set of equations called Lagrange VOP equations but they are not applicable
to any type of perturbation but they are powerful in obtaining approximated analytical results.
When we say that position and velocity correspond to the six orbital parameters, as we know in Keplerian we have
analytical equations and now in the idea of instantaneous Keplerian orbit/solution we can use exactly the
same relations! So, instantaneously everything is valid.
Gauss equations: let us introduce a new reference frame R̂, Ŝ, Ŵ (rotating frame, not inertial, directions are not
fixed). Let us consider the:
1. Instantaneous Keplerian orbit and so Instantaneous Orbital Plane. characterized by an Instanta-
neous Normal Direction Ŵ .
2. Instantaneous RAAN and Inclination.
3. Instantaneous Position of the Satellite and the Instantaneous Radial Direction R̂.
4. Instantaneous Direction in the Plane Ŝ which is orthogonal to R̂ and Ŵ .
5. Instantaneous Radial Direction R̂ which has an angle w.r.t. line of nodes ω + ν = u.
To write the transformation matrix from ECI to R̂, Ŝ, Ŵ , we can use the information about the instantaneous
position and velocity vectors of the satellite in the inertial frame in order to directly write the matrix. If we need
the rotation matrix from an old reference to a new one, the matrix will be made by the components of the new
unit vectors along the old unit vectors (each vector is a row vector). So, if we know instantaneously position and
velocity, we can evaluate the instantaneous rotation matrix from ECI to R̂, Ŝ, Ŵ .
The reason why we have defined this frame is that now we consider the components of perturbing accelerations
along this frame and not ECI.
Then, we now can write the Gauss equations. R̂, Ŝ, Ŵ is similar to ORF (directions follow the same logic, but
center in Earth).
Ω is the instantaneous angular velocity vector that describes the rotation of the radial direction.
What is the logic by which we determine the VOP equations? In Keplerian mechanics, both the mechanical
energy ε and the angular momentum vector h are constant. In general, due to perturbations, they are not constant.
Space Flight Dynamics Michele Santarpia 60

In order to derive the VOP equations, we can work on the fact that the mechanical energy and the angular
momentum vector are changing in time:
dε dh
̸= 0 ̸= 0
dt dt
From the mechanical energy variation equation, we can derive the semi-major axis VOP equation:
• Mechanical Energy Variation: we can write that the elementary variation of the mechanical energy will
be equal to the elementary work done by the perturbing force, so it will be given by the acceleration times
the elementary displacement of the orbiting object ds.

dε = ap · ds

The time derivative of the mechanical energy dε/dt will be equal to the instantaneous power expressed by the
product between the perturbing force and the velocity (force times velocity is a power, so acceleration times
velocity is power per unit mass):

dε dr
= ν̇ar + rν̇as
dt dν
Only in-plane components of the perturbing force ar and as can affect the mechanical energy. The opposite
is not true because it is possible that there are in-plane components but as we compute the sum the total is
zero.
dε
ap = aw ŵ → =0
dt

• 1) Semi-Major Axis VOP Equation: under the assumptions of osculating Keplerian parameters, we can
use the usual Keplerian formulations about the mechanical energy, where a is the osculating semi-major
axis. After we have obtained the final expression for the semi-major axis da/dt, we can see that, since
the mechanical energy variation does not depend on the normal component of the perturbation, also the
semi-major axis does not depend on the normal component of the perturbation.
da
ap = aw ŵ → =0
dt
If the perturbation is in plane, it may be a change in semi-major axis (not always true that there is the
implication as for energy).
From the angular momentum vector variation equation, we can derive the other five VOP equations:
• Angular Momentum Variation: the derivative of the angular momentum vector dh/dt will exist if the
perturbing force generates a torque on our satellite. If the perturbation is radial, it does not affect the angular
momentum vector.
1. In particular, the tangential component will have an effect and this will be a change of norm of the
angular momentum vector.
2. Instead, the normal component of the perturbation will generate a change of the angular momentum
along S and so a change of the angular momentum direction, so this means that the normal perturbation
will generate a change of the orbital plane.
By considering the derivative of h, and considering the Poisson equation to make the derivative of the unit
vector, we can obtain an equation in which we have Ω∗ (instantaneous angular velocity of the unit vector
W). Whatever happens instantaneously, the angular momentum vector will always be normal to the radial,
so based on this the angular velocity vector has to be along the radial direction.
dh
= ras Ŵ − raw Ŝ
dt
This is the key equation that we need to exploit to get the VOP equations of the other orbital parameters.
Space Flight Dynamics Michele Santarpia 61

• 2) Eccentricity Variation: the change of eccentricity can be calculated based on the change of semi-major
axis and norm of angular momentum vector. When we substitute dh/dt, we have to consider only the norm
which is given by ras . So, in the equation of the variation of the eccentricity we can see that dh/dt only
depends on as , while da/dt only depends on ar and as , so in the final expression we have no contribution
deriving from aw . So, if we have a normal perturbation, it does not change the eccentricity of the orbit but
the opposite implication is not always true also in this case.
• 3) Inclination Variation: since aw causes a change of direction of the angular momentum vector h, this
di
means that it will change the plane of motion, impacting the RAAN and the inclination i. Let us derive dt
dΩ
and dt . The key point of these discussions is that we can solve everything by using our analytical tools.

During the passages to derive the final equation, let us remember that:
1. The 1st derivative dh/dt is a vectorial one, while dh/dt is only the norm.
2. The projection of k̂ over the orbital plane is along n̂⊥ .
3. Let us consider the POV along n̂ from the AN and then a POV2 along Ŵ .
Finally, we obtain that:
di r cos u
= √ aw
dt na 1 − e2
2

1. The inclination can only be changed by a normal perturbing force aw and any in-plane perturbations
does not change the inclination.
2. This equation has singularities (this type of singularity exists also in the case of the derivative of the
semi-major axis; in all these cases there might be some singularities) when the eccentricity e=1.

When we have mathematical singularities or when we have conditions which approach singularities,
we have some numerical challenges because the numbers will explode. Typically, the solution consists
in changing the orbital parameters. These equations are with the classical orbital parameters but we can
combine the orbital parameters in different ways, so that we avoid some singularities.

These singularities do not represent real physical problems: it’s a matter of how we model
our orbits, we just need to use other elements like, for example, the equinoctial elements which is
another set of six parameters where we combine the traditional elements in a different way and we write
equations which have not the same singularities. The idea is similar to the case of the attitude kine-
matic equations where we know that by using trigonometry we have singularities, and we can change the
selection of the Euler’s sequence to avoid singularities or by using quaternions.
3. For the inclination, we have a dependency on the cosine of the argument of latitude, and so the
effect of a normal perturbation depends on where we are on the orbit w.r.t. the AN, so the maximum
effect on the inclination by aw is coming when the perturbations are applied close to the nodes of the
orbit. When instead it is applied at the orbital poles (let us consider the highest point of the orbit), aw
does not have any effect on the i.
4. Link between our-of-plane maneuvers and normal perturbation: there is a strong link between
these considerations and the out-of-plane maneuvers. Orbital maneuvers, orbital perturbations and rela-
tive motion (three topics of the course) are not independent. When we talk about the impulsive orbital
maneuvers, we say that if we want to change the inclination we need to rotate the velocity vector at
the nodes, while if we do this at the orbital poles it will become mostly a change of RAAN. Now, we
have not an impulse, but a perturbing acceleration of the orbit and it contains similar concepts compared
with out-of-plane maneuvers.

Low thrust to change the inclination: as we know, the thrust can be considered as a perturba-
tion particularly in the case of low thrust since it has a similar effect of the natural perturbations. Let
us consider, for example, that aw is given by our thrust (electric thruster): practically, let us imagine to
orient our satellite thruster along W in order to change the inclination. If we want to change more the
inclination, we can apply the thrust for more time: but this not sounds completely reasonable. Because
Space Flight Dynamics Michele Santarpia 62

if we need to change our inclination, and we apply a constant thrust along the orbit, on average we
are not changing the inclination due to the cos u. The effect of thrust is the one depicted in the figure.

So, how do we change the inclination?

(a) The 1st option could be the one in which we apply the thrust only in a fraction of the orbit, so we
can apply the thrust for the 1st quarter of the orbit and as we fly towards 90deg the effect on
the inclination goes to zero, and then to avoid having the opposite effect we thrust again when we
are in the final quarter of the orbit, so this means that we use our thruster only for half of the
orbit.
(b) The 2nd option is to change the direction in which we apply our thrust, and so for example in the
1st quarter we have a positive aw , and then we change our attitude and so we push in the opposite
direction. So, if we have a failure and the thruster cannot be turned off the effect is an oscillation of
the orbital plane.
• 4) RAAN Variation:

dΩ r sin u
= √ aw
dt na2 1 − e2 sin i
In the final expression we can see that:
1. Singularities: we have singularities when e=1 (parabola) and when the orbit becomes equatorial
(i=0deg). It makes sense because, when the orbit is equatorial, the RAAN is not defined anymore
(in this case, we use not RAAN as orbital parameter but we change it with another one to describe the
orbital plane). In the real world, the orbit will never be exactly equatorial, there will be always some
residual inclination (we will see this from non-sphericity effects). When we have an orbit which is nearly
equatorial orbit (geostationary with some perturbations), the change in RAAN becomes more and more
a change in anomaly (contribution to the angular motion of the satellite).
2. In this case we have the dependency on the sine of the argument of latitude sin u, so the discussion
is similar to before but opposite because if we have a normal perturbation aw the effect on RAAN will
be present when we are at the orbital poles and it is zero when we are at the nodes.
3. Low thrust to change the RAAN: the same discussion is for the constant perturbation like the
constant thrust, if we apply a constant thrust the RAAN again oscillates with zero average (net effect
is zero), so also in this case if we want to apply the electric thruster to change RAAN we need to apply
the thrusting in some portions of the orbit or adjust the sign of thrust in different portions of the orbit.
If the oscillations of an orbital parameter is within our tolerance (so, it is not too large) and if we have a
constant normal perturbation, we will not fight against them and we accept them.
• 5) AoP Variation: in the final expression, we have that the variation of the AoP is a function of the all the
three components of the perturbing force:
dω
= f (ar , as , aw )
dt
Note: on Colella, the starting expression that is considered is wrong. The correct expression is:

R̂ · n
cos u = = cos(ω + ν)
||n||

• 6) Variation of the Initial Mean Anomaly: the sixth derivative is expressed by using dM0 /dt. In
Keplerian perspective, the initial mean anomaly M0 is a constant. As we know, we can relate the initial mean
anomaly M0 to the mean anomaly M(t):

M (t) = M0 + n(t − t0 )
Space Flight Dynamics Michele Santarpia 63

In the perturbation analysis, we can use the same expression to get the instantaneous mean anomaly M(t)
where M0 =M0 (t) (it is changing over time):

dM0
= f (ar , as )
dt
Substantially, in case of perturbations, we are changing the mean motion (angular velocity) of the satellite
which is given by the sum of the mean motion n plus an additional term related to M0 (t).

The derivative dM0 /dt is only a function of ar and as perturbations (only in-plane components), and it
makes sense because, if we have out-of-plane perturbations, we do not alter the motion along the orbit w.r.t.
the perigee.

Note: in the slides, we have the forces F instead of accelerations.

The derivative which changes faster is the anomaly but we do not see it in the last equation because we
have M0 .

ar as aw
Mechanical Energy Yes Yes No
Semi-Major Axis Yes Yes No
Angular Momentum No Yes Yes
Eccentricity Yes Yes No
Inclination No No Yes
RAAN No No Yes
AoP Yes Yes Yes
M0 Yes Yes No

5.7 1) Non-Sphericity Effects

It is hard to use the Gauss VOP equations to get immediately analytical results regarding perturbations, especially
when we consider gravity: for this reason, we like to use another set of equations called Lagrange planetary (or
VOP) equations where, at the 2nd member instead of having the components of the perturbations, we have some
partial derivatives of a scalar function R, called perturbation potential, w.r.t. the orbital parameters.
While Gauss equations are general, Lagrange are not general because they contain only what regards conser-
vative perturbations (so, they do not include Drag or SRP).
In general, a conservative force can be expressed as the gradient of a scalar function called potential: in this
case, we can express the perturbation acceleration as the gradient of the perturbation potential R which
will have some dependencies. For example, two effects can be modeled using the Lagrange VOP equations (defining
a potential perturbation):
1. Non sphericity of the primary where the potential R contains the deviations from the spherical gravity.
2. 3rd body effects.
Let us consider a body with a generic shape and density distribution. A given point in space is subjected to a specific
gravity force a: this is still a gravity force that can be expressed using a (gravitational) potential function U
which is made by two terms: the spherical term called two-body potential U2−body (which is the term related to
a sphere with spherical symmetry), and the perturbation term given by the perturbation potential R.
We can write the potential also in Cartesian coordinates.
If we are able to model the perturbation potential R for any mass/body, then we can compute the perturbation for
any body and so we can understand what happens to our orbit.
Non-sphericity: let us understand the dependencies for R. When we consider non-sphericity, the idea is that when
we write the two terms U2body and R, of course U2body only depends by the distance (due to spherical symmetry)
and when we consider some other effects R depends also by the latitude and longitude. For this reason, R is a
Space Flight Dynamics Michele Santarpia 64

function of three independent scalars r, latitude and longitude (w.r.t. equatorial plane of the object).

U2−body = f (r) R = f (r, Φ, λ)

We have to use two angles which are body fixed and this is the reason why we use latitude and longitude since
the rotation of the planet will influence our orbit because we experience a variable gravitational field
depending on which is our position w.r.t. the planet (we do not use two angles like the one related to the inertial
frame because if the mass rotates and we are in a given point, the gravity changes due to the rotation of the Earth).
If we consider the osculating (instantaneous) orbital parameters due to the perturbations, they are given by
the sum of three terms:
1. Secular Effects, secular variation of the orbital parameter which accumulates in time.
2. Long Periodic Effects: then we have some long periodic oscillations (long period means much more than
the period of the orbit).
3. Short Periodic Effects: we will have short periodic oscillations (short period means in the order of the
orbital period).
Not necessarily we have that these three terms are different from zero, sometimes we may have only periodic
phenomena and it depends on the type of orbit and perturbation we consider. These periodic oscillations are
activated by conservative perturbations like non-sphericity or 3rd body.
Not always we are interested in the entire evolution of the osculating parameters and so not necessarily we
are interested in the short periodic effects, so for a number of practical scope we may be interested only in the
secular or only in secular plus long periodic effects. Instead of evaluating the entire set of instantaneous parameters,
we can have a solution that does not contain these oscillations and these solutions are more easy to obtain and so
we can adopt some analytical ideas.
How do we isolate what we are interested in from what we are not interested in?
1. If we are not interested in the short periodic oscillations/phenomena, for example, we take our
Lagrange equation and we apply an average over one orbital period and so instead of working on instantaneous
orbital parameters, we work through this average over one orbital period. From a practical POV, we can apply
an average over 360deg of mean anomaly: so, in this way, we take only the effects whose average is different
from zero, this will remove all the short periodic phenomena.
2. If we are interested in removing the long periodic oscillations, we apply a 2nd average which might
be on a much longer period and the value of this long period depends on the type of perturbation on which
we are working on. For example, in case of 3rd body effects of the Sun, maybe to remove long periodic
phenomena, we can compute an average over one year since the motion of the Sun is periodic in one year.
So, in general, we have this idea of two average operations (one over orbital period and the other on a longer
time scale), once we perform these operations, we can derive some equations which are much simpler that have an
analytical solution.
After applying these operations, we will have an analytical solution for our orbital parameters given some pertur-
bations and the orbital parameters that we consider if we remove all the periodic phenomena will be called mean
(or secular) orbital parameters (there will be no oscillations) and so we extract some approximated orbital
parameters (they do not contain some effects).
There are different theories to obtain these secular parameters: in GMAT for example, we can have Brouwer-
MeanShort which are the orbital parameters without short periodic oscillations based on Brouwer theory.
Let us consider the problem of non sphericity, our body has a generic distribution of mass, we are interested in
the gravity effect in a generic point, in general we can imagine that our body is a sum of infinite elementary
masses. We can consider the relative position between the elementary mass and the point P. The elementary
gravitational potential is dU(x,y,z) and it is the potential of a point mass which is in fact given as we know by
GM/r.
The Laplacian of the gravitational potential is proportional to the local density:

∇2 U = 4πGρ(x, y, z)
Space Flight Dynamics Michele Santarpia 65

1. The local density will be not zero if we are inside the body.
2. Outside the body (which is our case of interest, because we want to evaluate the gravitational field around
our body), the local density is zero and so the Laplacian is zero.
The gravitational potential can be written as a development in series and, in particular, as a development in
series of spherical harmonics. So, it will be given by two terms: the 1st one is the two-body one and then we
have the perturbation part. ϕgc is the geocentric latitude (elevation angle over the equator). While U of the two
body depends only on r, the perturbation potential instead depends on r, latitude and longitude. Pnm are Legendre
polynomials where in the case in which we write Pnm (sin ϕgc ) we mean that the polynomial is a function of the sine
of latitude.
Theoretically, the perturbation potential R is a sum of infinite terms/spherical harmonics (this is similar
to Fourier where we have the sum of infinite harmonics represented by n and m). Inside the sum in n, we have
a sum in m where the order m goes from 1 to n. Conceptually, each harmonic represents a contribution to the
gravitational field.
1. Zonal Harmonics: if we consider the sum in n, we find the harmonics of degree n and order m=0 which
have a special characteristic: if we check the expression there is no dependency on longitude. So, if we have
a contribution to the gravitational potential that does not depend on the longitude, these contributions are
not altering the axial symmetry of the primary body and so they represent deviations from the sphere which
happens at the same way at any longitude! These harmonics are called zonal harmonics and so they depend
only on the distance from the center and latitude.

The zonal harmonics are modeling variations w.r.t. spherical gravity that happen at the same way
at any longitudes and they only depend on latitude and so, for example, they are modeling the fact that we
have an excess of mass at the equator (oblateness of the Earth, the distribution of mass is not spheri-
cal) which is an effect axial-symmetric. Not only the oblateness of the Earth, but also higher order effects
where, for example, we can model that there is different mass between the Northern and Southern
hemisphere (variations which happen at the same way at different longitudes).
2. Sectorial Harmonics: then, we have other infinite harmonics which will be characterized by a given degree
n and order m equal to n (harmonics 2,2 or 3,3 in terms of coefficients). These harmonics depend on longitude
and so there is no axial symmetry of the body and so when we consider the body there is non uniform
distribution of mass at the different longitudes. So, these harmonics model the tri-axiality of the planet
(sectorial harmonics and they model the deviations from spherical gravity due to the mass that is added
or subtracted in bands of longitudes and this is something which is removing the axial symmetry).
3. Tesseral Harmonics: then, we have infinite harmonics with degree n from 2 to infinite, and of order m
different from n (3,2 4,3 etc.), and these harmonics depend on distance, latitude and longitude and so they
model again the tri-axiality of the Earth and we are modeling the fact that the mass is not distributed
uniformly. These harmonics are called tesseral harmonics because from a physical POV because these
harmonics are adding and removing mass in given patches of the Earth.

U = U2−body + ∞ harmonics(n = 2 → ∞, m = 0) + ∞ harmonics(n = 2 → ∞, m = n) + ∞ harmonics(n = 2 → ∞, m ̸= n)

If we use enough of these terms (so, enough degree and order), we can model a mass distribution that is really non
homogeneous.
By increasing the degree of the harmonics n, we are considering deviations with increasing spatial frequency
and so we are able to model effects which happen at higher frequency.
If we consider the sectorial harmonic (2,2), when we look at the planet from the rotation axis what we see is not a
circle but an ellipse (at higher order of analysis the Earth is not axial symmetric).
When we increase degree and order like (3,3), (4,4) etc. we are modeling variations with increasing spatial frequency.
More terms we introduce, more complex is the explanation we can give.
Finally, we can see how there are the tesseral harmonics characterized by order different from 0 and from the degree
and these harmonics are modeling deviations which happen in specific intervals of latitudes and longitudes. When
we increase the degree and order, we are modeling the addition or subtraction of mass at specific places.
Space Flight Dynamics Michele Santarpia 66

The key idea is that if we can use enough harmonics (sum enough contributions), we may model a central body
with any shape or any mass distribution.
As we want to inject more harmonics (with higher spatial frequency), we will want to model effects that happen
at higher frequencies and this will bring us to a shorter time-step in the numerical perturbation topic in order to
satisfy the sampling theorem.
In the expression of U, we have some dimensional terms like the radius at the equator but this term in our problem
of interest will be of order 1. The Legendre polynomial is computed in the sin of latitude and again it is of order
1, and then we have some coefficients are non dimensional. Each zonal harmonic is characterized by a coefficient
Jn , while Cnm and Snm are the coefficients for the sectorial and the tesseral harmonics. These coefficients are non
dimensional and they give us the order of magnitude of these harmonics.
If we consider the zonal harmonic of degree n, Jn will give the order of magnitude of this term w.r.t. the spherical
potential. For example, J2 (which models oblateness) is in the order of 10−3 for the Earth and it means that the
zonal harmonic (2,0) is 1/1000 smaller than the two-body potential (the spherical gravity). So, in relative terms,
this term describes how much is not spherical or how oblate is the Earth. What distinguishes one planet from
another one, the non-sphericity of the body is contained in the coefficients: each body has its own coefficients.
J33 is smaller than 10−6 .
The largest effect is given by J2 and this is the harmonic (2,0) i.e. the ring of mass around the equator which
describes the oblateness of the Earth. All the other terms are at least three orders of magnitudes smaller. If we
consider which is the effect on the gravitational field of the ellipticity of the equator described by J22 (due
to the triaxility, if we see from the top it is more an ellipse than a circle), it is three orders of magnitudes smaller
than the oblateness.
So, if we want to introduce in our explanation the non-sphericity, the 1st solution can be the one in which we take
into account only J2 , so we are not Keplerian and the next level of approximation is to consider J2 . If the accuracy
has to be increased, than we will consider the other harmonics in our potential in order to consider the other smaller
effects.
If we consider a body with a shape that deviates more from a sphere and the density is highly non homogeneous, the
coefficients will be larger. If we have a very non symmetrical shape, we need a relatively large number of harmonics
to describe this gravitational field and there will be more harmonics with the same order of magnitude. For example
the Moon has a more irregular gravitational field which is due to the mass distribution inside.
How to derive the Earth and Asteroid Gravitational Field Model?: When we encounter an asteroid, how
we can derive the coefficients? How do we know this number? How do we know the coefficients that model the
gravitational field of the Earth? Remembering that the coefficients are modeling the shape and mass distribution
of the Earth, so how is the Earth gravity?
1. Earth: if we have a satellite on its orbit and we follow its motion, we can invert. In fact, as we know:
(a) Using the gravity model, we can enter in the theory of perturbations and then we can derive the
effects on the orbit.
(b) In this case, we can proceed in the opposite direction by measuring the effects on the orbit, using the
theory of perturbations and then retrieving the gravity model.
Of course, this is true if there is only non sphericity and so we can maybe select an orbit where we only have
non sphericity, but as we fly higher (to avoid Drag) the effect of gravity reduces since it is related to the
square of the distance (less sensitive we are), we need to fly at very low Earth orbit to have much sensitivity
(the lower is better) but, of course, we have Drag and so we need to remove the other perturbations (for Drag
we have to remove it and we use electric propulsion like ion engine and we compensate the Drag). By using
the orbital effects, we describe the satellite geodesy (geodesia satellitare) (see the GRACE mission with
two satellites or GOCE mission with one satellite characterized by a very accurate accelerometer in order
to measure the perturbations, and both make use of electric propulsion to compensate for Drag, we can see
from the model higher gravity where we have the mountains, we can derive up to 200 harmonics).
2. Asteroid: when we approach an asteroid, we can try to enter in the orbit of the asteroid and typically in
these missions we have an initial prediction of the orbit with some tolerance and as we approach the celestial
body we learn in real time the gravitational field and we can make possible adjustment.
Space Flight Dynamics Michele Santarpia 67

For space mission POV, we do not use 200 harmonics but to design our mission we use much less harmonics and
also to control our orbit. The real scope of gravimetry is not to improve mission design but it is to describe the
Earth system because the distribution of mass and the motion of mass contain the motion of glaciers (ice melting,
climate change, etc. stuff).

Effects on the orbit due to First Zonal Harmonic J2 The largest effect introduced by non sphericity of the
Earth is given by the 1st zonal harmonic J2 which introduces the oblateness. How can we model the effect of
J2 on our satellite? We can approach the problem analytically. We come back to the Lagrange VOP equations
and we replace R with the J2 harmonic. We enter in these equations and we apply some average operations: in
fact, once we inject J2 we will have secular, long and short period variations of all the parameters but we may be
interested only in the effects what accumulate in time (secular) or secular plus long period oscillations.
Let us consider only the secular effects (effects which accumulate in time): we obtain the secular orbital param-
eters which means orbital parameters without oscillations.
1. J2 does not have any effects on a,e,i (constant).
2. As result of J2, there is a secular precession of the line of nodes Ω̇, a secular precession of the line of axes ω̇
and a change of the angular rate w.r.t. the perigee Ṁ .
3. If J2 =0, Ṁ = n (n=mean motion).
4. As a general result, only the even zonal harmonics give secular effects. In general, secular effects on
orbital parameters are only due to even zonal harmonics (J2 ,J4 ,J6 etc.). All the other harmonics give only
short and long period oscillations.
5. Due to J2, the orbital plane is moving: if we have a prograde orbit, the line nodes has a retrograde precession.
The same is for the perigee which is not fixed anymore.
If we want to fight against these effects, we will need to periodically bring again the orbital parameters to the values
that we like (if we want to have our Keplerian orbit where the nodes and perigee are fixed).
It is not so smart fighting against J2 but it is better to design the orbit taking these effects into account. Instead
of spending propellant, if we want to nullify some effects, we can design the orbit so that the variation of the
parameters is zero. But the question is: do we really need that the nodes are fixed? (no!)
As we can see, while the satellite moves along its orbit relatively fast, the rate of precession of perigee and AN is
10−3 of Ṁ , so the line of nodes and axes change slower than the motion of the satellite along the orbit.
Anomalistic and Nodal Period: since ω̇ is not constant, we can define two types of orbit period:
1. Motion w.r.t. the perigee: one orbit period is relative to the motion of the satellite w.r.t. the perigee
(time to fly from perigee to perigee) called anomalistic period; the angular rate of the motion w.r.t. the
perigee is ν̇ and on average it becomes Ṁ . So the anomalistic period is given by 2π/Ṁ . So, in this case, what
is doing the perigee is not considered.
2. Motion w.r.t. the line of nodes: however, when we consider the overall angular motion in the orbital
plane there are two contributions: the 1st one is the motion of the satellite w.r.t. the perigee and the other
one is the motion of the perigee itself w.r.t. the line of nodes. So, if we consider the total angular rate in the
orbital plane, it is given on average by Ṁ + ω̇ (it is a matter of relative motion), and as we know these two
rates differ by three order of magnitude. So, this will be the angular rate of the satellite w.r.t. the line of
nodes. So, we can define the period to fly from node to node called nodal period. If ω̇ > 0, the nodal period
is lower w.r.t. the anomalistic one, which means that it takes less time to fly from node to node because we
have that the perigee is moving together with the satellite, so the satellite is faster w.r.t. the nodes.
Nodal period considered at design level: when we take into account these perturbations (typically, we consider
J2 at design level), we consider the nodal period because it is related to the repetition properties of the orbit
(which is related to the meaning of how much time we need to fly again over the same spot).
Molniya orbits: they are high eccentric orbits used when we are interested in communication satellites at
high latitudes, since there are some challenges in using geostationary satellites at those latitudes.
First of all, let us consider the following two problems:
Space Flight Dynamics Michele Santarpia 68

1. Geostationary Communication Satellites are challenging at high latitudes: for geostationary satel-
lites, the sub-satellite point will be constant and at the equator it will be on the vertical of the same point.
These satellites have challenges in communicating at high latitudes, there are some latitudes limits which are
given by the tangent directions to the Earth that start from the satellite (so, it happens when the satellite is
exactly on the local horizon of the Earth), so there are polar regions which are not in view of the satellite.
RL ∼
The angle in the figure is arccos aGEO = 82deg. But as we know in order to communicate, we cannot use a
satellite which is on the local horizon but we need a minimum elevation angle w.r.t. the local horizon,
and so for this reason we have another maximum latitude (given by light blue), and if we consider now the
local horizon we have that the satellite is at a given elevation, and so in this way we can find the maximum
latitude at which we are able to communicate. This is a problem of atmospheric absorption, because, if
we consider the electromagnetic waves traveling through the atmosphere, when we are at very high latitudes
the path that the signal has to fly through the atmosphere is very long, so the absorption is too much and so
the communications do not work anymore (the problem here is in the link budget).
2. Plane change has a high cost: Russians have had always the problem of launch sites: the inclination that
can be obtained directly with a launch is larger or equal to the latitude of the LS, and since the LSs of the
Russians are all at 50deg of latitudes, the correction of 50deg of inclination is too much. So, other solutions
are needed: we need something which works like a geostationary satellite but only for high latitudes.
To solve this problem of high-latitude communications, we can use the so-called Molniya orbits which are char-
acterized by a high eccentricity (e=0.73) and a critical inclination (icrit =63.43◦ ). Let us give a look to the
orbit in the geocentric inertial reference frame. The perigee is in the Southern hemisphere (ω = 270◦ ) (i.e. it is
coincident with the Southern pole of the orbit), the orbit is prograde since i < 90◦ , a ∼
= 26335km, τ ∼= 12h. Being
high eccentric, we will pass most of the orbit in the Northern hemisphere (most of the time we are around the apogee
where we have a slower angular rate) and so we are with a clear view (good coverage) of the higher latitudes. This
concept is not too smart if the perigee moves: so, we need to keep the perigee in the Southern hemisphere in order
to keep the desired geometry fixed in time, that’s why we need the critical inclination. If we consider J2 , there
are two values of the inclination used to have a fixed perigee otherwise the perigee drifts (i=63.43◦ or i=116.57◦ ).
The line of nodes will move in a negative sense but we will keep the perigee fixed and so it will be always in the
Southern pole of the orbit and the apogee in the Northern hemisphere.
Precession of line of nodes:
 2
3 ΩLE
Ω̇ = − J2 Ṁ cos i = f (a, e, i) p = a(1 − e2 )
2 p

Ṁ is a function of (a,e,i), mainly on a and slightly on e and i:

Ṁ = f (a, e, i)

The ascending node is not fixed but it is moving: in particular, if we have a prograde orbit (from west to east in
inertial motion), the node will move in a retrograde way (from east to west w.r.t. inertial frame).
1. A polar orbit is the only one where the orbital plane behaves like in the Keplerian theory since Ω̇ = 0: a
physical interpretation of this is related to the fact that J2 is the representation of a ring of mass around the
equator (oblateness). If we consider a polar orbit and the Earth with this ring of mass, the gravitational field
will still have a symmetry w.r.t. a polar orbit: so, there is not a net torque due to the ring of mass and this
means that the angular momentum vector in a secular perspective does not change and so we have no change
of the line of nodes.
2. For the other orbits, since there is more mass at equator due to J2 , there will be a non-central gravity which
generates a net torque: for this reason, the direction of the angular momentum vector (i.e. the normal to the
orbital plane) changes moving along a cone with semi-aperture equal to i, and so the line of nodes rotates.
The direction of motion will depend on the inclination of the orbit.

Sun Synchronous Orbits Whatever is our orbit, we have a precession of the line of nodes. If we want a fixed
line of nodes, the only possibility is to have a polar orbit otherwise we have to accept this effect and, as we can
see in the propagation, the line of nodes moves relatively fast. Generally, this is something that we do not want
to counteract also because we need a lot of propellant but this is something that we like to exploit. If we cannot
stop the precession of the line of nodes, maybe we can synchronize the motion of the line of nodes with
Space Flight Dynamics Michele Santarpia 69

the motion of the Sun in order to obtain an orbit that keeps mostly a constant relative geometry w.r.t.
the Sun. This leads to the concept of Sun Synchronous Orbits. Let us consider the apparent motion of the Sun
in ECI True of Date: the ascending node of the apparent motion of the Sun is on the X axis, the Sun will move
over the ecliptic; as the Sun is moving, we can project it on the equatorial plane of the Earth, we can define the
right ascension of the Sun and the declination. The motion in right ascension is not constant, but at a 1st order of
approximation we can assume that it is constant and equal to 0.98deg/day.

α̇s ∼
= 0.98deg/day

In order to synchronize the motion of the line of nodes with the motion of the Sun in the equatorial plane, we
need to impose the equality between Ω̇ and α̇s . If two angular rates are equal, then the difference between the two
angles is constant:

Ω̇ = α̇s → Ω − αs = constant

If we want an orbit where the precession of the line of nodes has the same rate of the Sun, we are establishing a
constraint about a,e and i:

Ω̇ = Ω̇(a, e, i)

So, we cannot select them as three free parameters but there is a condition that connects them. When we consider
the Sun projection on the equatorial plane, the line over the equator is moving with a rate constant equal to
1deg/sec. In green, we can plot the line of nodes of the satellite orbit (plot not in scale). As we know, the Sun is
moving and also going up and down w.r.t. the equatorial plane but we consider just the projection in the equatorial
plane and so it is simply rotating as time passes. The line of nodes is also moving, and we obtain that the angular
rate, so the variation of the line of nodes, happens with the same rate of the motion of the Sun, and if this is
true we will keep a constant angle between the satellite and the Sun. This is what is called Sun-synchronous orbit
which is plot as a retrograde orbit not by case but because we need a positive precession of the line of nodes and
so the inclination must be higher than 90deg. In this situation, the geometry of the line of nodes w.r.t. the Sun
will be constant. Strictly speaking, we cannot say that the geometry of the orbital plane w.r.t. the Sun is constant
because the Sun not only rotates but also goes up and down w.r.t. the equatorial plane but neglecting the effects
of changing the declination, the geometry of the orbital plane w.r.t. the Sun is constant.
Why do we like the idea of SSOs?
1. Important for Remote Sensing orbits for illumination conditions that we have when we observe some regions.
2. For the Electric Power Subsystem.
3. Significant simplification for the Thermal Control Subsystem.
Since the angle given by the difference of the RAAN and right ascension of the Sun is constant, it is used to identify
the relative constant geometry between the Sun and the orbital plane (the line of nodes). This relative geometry
can be given if we have a single number which is the Local Time at the Ascending Node (LTAN). Let us
remark that this relative constant geometry does not change as time passes!
Let us call ŝ the Sun projection on the equatorial plane. If we are in one day during the spring, the Sun will be,
for example, in the direction depicted in the figure. If the difference between the angles is zero, the line of nodes
direction is equal to the Sun direction, and in particular, since the difference between the two angles is zero, we will
have the AN in the positive direction of ŝ.
Let us plot the SSO corresponding to Ω − αs = 0:
1. In this geometry, it is possible to understand that when the satellite is at the AN, the local time is equal to
12PM. This means that the LTAN=12PM.
2. When we are at the DN, the local time is 12AM. This means that LTDN=12AM.
For any type of orbit, instantaneously, depending on the position of the satellite w.r.t. the Sun, we have a given
local time. In case of SSO, something special happens: the conditions in terms of local time that we have at any
latitude do not change in time. Since the line of nodes and the sun direction rotate together, when we arrive at the
AN it is always 12PM while at DN is always 12AM. Since these two numbers are constant, we can call this orbit
12PM 12AM SSO orbit. This idea of constant local time is true for any latitude of the orbit. If we fix a given
Space Flight Dynamics Michele Santarpia 70

latitude, we will pass over that latitude twice per day (ascending phase and descending phase) and the local time
in the ascending phase and descending phase will be always the same.
As we have seen, the precession of the line of nodes is activated by the torque due to the oblateness of the Earth,
so we do not have a central force anymore. If we consider the secular effects (so, on average), there is a net effect
which causes the rotation of the line of nodes.
The relative geometry between the Sun projection and the ascending node will be kept constant: this means that
if we neglect the variations in declination of the Sun, the relative geometry of the orbital plane w.r.t. the Sun will
be constant also because remember that under the perspective of secular J2 effects the inclination is not changing
(it does not change in time due to the oblataness of the Earth as the eccentricity and the semi-major axis).
Let us consider the following scenarios:
1. 12PM 12AM SSO during Spring: let us consider the situation in which the difference between Ω and αs
is zero, which corresponds to a LTAN equal to 12pm.

Ω − αs = 0 → LT AN = 12P M

(a) Let us consider a POV from the vertical axis Z.

(b) Let us plot the Earth in order to understand which are the different local time and the normal to the
direction Ŝ. Of course, the central point represents the North Pole.
(c) Let us plot the orbit: since we are considering a 12PM/12AM, the AN will be where it is 12PM while
the DN will be the point where it is 12AM.
After we have put the AN and DN, we have to see how the orbit looks like when we look from the Z axis. We
need to remember that it is a retrograde orbit and so we do not see a line when we look from the top (line
only if it is a polar orbit), and so we will see the projection of the orbit which is an elongated ellipse. We will
have a given local time at the AN and a given local time at the DN, and in general we have a given local time
at any latitude of the ascending and descending phase. What is the local time when we are at the Northern
pole of the orbit (it is in red)? When we pass at the Northern pole of the orbit, the local time is 6AM. In the
other pole of the orbit (Southern hemisphere), the local time is 6PM. In order to obtain other local time, we
can consider the line of the y axis, and then we can take the corresponding intersections with the orbit which
correspond to 3AM (during night) and 3PM (during day). This type of geometry is kept constant as the time
passes during the year. In the case sketched in the figure, the geometry is referred to a 12PM 12AM during
the half of spring.
2. 12PM 12AM SSO at summer solstice: if we consider another date what happens to 12PM 12AM SSO?
Let us consider, for example, the summer solstice which is along the positive y direction (projection). We can
build the Earth and the orbit, place the AN and DN and build the orbit. As we can see, as the time passes
the geometry rotates with the Sun and we keep the same local time.

What is the local time when we pass at a latitude of 45deg in the Northern hemisphere (Italy)? We
will pass on that latitude twice, during the ascending and descending phase (so, we pass one time when it is
morning closer to noon and the other one when it is night closer to midnight).

This orbit gives us good illumination conditions because we will fly over this latitude close to noon, so
during the year the illumination conditions will change because the declination of the Sun changes at the
same local time but they will remain similar (advantage).

How do we can compute which is the difference in local time between the AN and the latitude corresponding
to 45deg? If we plot a tridimensional geometry and a retrograde orbit, if we have a point at latitude of 45deg,
we can sketch the parallel and identify the point and we can plot the meridian (same construction as in the
case of launch constraints), we use arcs as the angles, we can consider the triangle. We can compute δ which
represents the angular distance on the equator which will give directly the time difference w.r.t the ascending
node, and if we consider that 24h corresponds to 360deg, the time difference can be evaluated as:
time dif f erence δ δ
= → time − dif f erence = 24h
24h 360deg 360deg
Space Flight Dynamics Michele Santarpia 71

We are converting an angle in a difference in local time. If we have an inclination of 97/98deg, we get a time
difference of 15min, so we can see how it is highly non linear.
3. 6AM 6PM SSO at half of spring: let us consider the geometry in the case in which we are at half of
spring (remember to do exercise making the geometry in any moment of the year!).

Ω − αs = −90◦ → LT AN = 6AM

Let us give a look from the z axis. When we are at the Northern pole of the orbit it is midnight (12AM),
while at the Southern pole is noon (12PM).
4. 6AM 6PM SSO at winter solstice: if we have a retrograde geometry it means that when we move
along the orbit we have a retrograde local time, so the direction is from 6AM to 5AM and so on.
5. 9AM 9PM SSO at spring equinox: first of all, we place the Sun along x (where in this case it is the real
direction and not only the projection since declination is zero). This case corresponds to Ω − αs = −45deg.
6. 6AM 6PM (good for EPS):

6AM 6P M → Ω − αs = −90◦

In this type of SSO orbit, the Sun is approximately normal to the orbital plane. The direction of the Sun is
important because we like to have the solar panels in the direction of the Sun.

In general, to obtain this result, we can use some mechanisms and so we can point our solar panels to
follow the apparent motion of the Sun (or the unit vector to the Sun ŝ) but mechanisms present complexity
(we tend to minimize as much as possible the use of mechanisms in orbit). The consideration about power
problem becomes important if we have some systems that have an high power consumption: this happens
for radar mission (SAR, active payloads). In general, from the POV of power, all the SSOs have a basic
advantage: in fact, to follow the Sun we have to compensate only for the motion of the satellite along the
orbit. Instead, to follow the Sun for a generic orbit, we should consider the motion of the satellite along the
orbit and the motion of the orbital plane and so the variation of the relative geometry w.r.t. the Sun. Since
for SSO there is no change in relative geometry, at most we have to compensate for the motion of the satellite
along the orbit which means using a single pointing mechanism (so, compensation for a single angle).

For a 6AM 6PM, we can completely avoid the need of a pointing mechanism and the idea is to
try to keep the solar panels within the orbital plane since the orbital plane is approximately normal to the
Sun direction. Let us see this in a 2D geometry from the POV of the AN. We can consider the Summer
and Winter solstices which are interesting because they represent the two extremes of the Sun declination:
(a) Summer Solstice: first of all, let us consider what happens at the summer solstice considering a 2D
geometry: we are looking at the orbit from the ascending node, so what we will see is the trace of the
orbital plane (we will see a line) where we can see the poles (points in green). At the summer solstice
the Sun has the maximum declination (23.44deg). We can sketch the Sun direction through the vector
Ŝ, and the shadow of the Earth. The angle between the orbital plane and the vertical direction is about
of 7/8/9deg. In this case:
i. The Sun is almost normal to the orbital plane.
ii. We have no portions of the orbit that are in eclipse conditions, the satellite is always illuminated
by the Sun (so, we never have full eclipse conditions). If we consider the Northern hemisphere, we
are always in the illuminated side and when we consider the Southern hemisphere the satellite has
a given altitude and, for this reason, we do not end in the shadows of the Earth (good for power
considerations, we will not have to rely only on batteries).
(b) Winter Solstice: at the winter solstice the situation is similar but a little bit different, we always
consider a 2D-dimensional geometry from the AN and, as we know, the Sun has the maximum negative
declination. Due to the geometry of the orbital plane, at the winter solstice we have the maximum
angle between the direction of the Sun and the normal to the orbital plane, from the POV of
power we have the most challenging condition because we gather (raccogliamo) less energy and also,
since we are on a LEO orbit, there will be some parts of the orbit which is in eclipse: for this reason,
winter solstice=worst case while summer solstice=best case. So, what we can do is to take this
Space Flight Dynamics Michele Santarpia 72

worst case condition into account by taking some margins. We can design SP with some margins (larger
than the minimum) in order to account for this worst case condition and so through these margins we
can still avoid the need for mechanisms.
Some questions:
1. Why do we not use 6AM 6PM SSO for any type of remote sensing missions?

If we have a 6AM 6PM orbit this is good for radar but it is not good if we have to use optical sensors
because if we consider the local time discussion this is an orbit where we will pass/fly at European latitudes
(mid-latitudes) before Sunrise and after Sunset and for an optical payload this is very bad due to the il-
lumination conditions. So, for radars missions, in general, we use a 6AM 6PM SSO orbit while, in
case of optical missions, it is very likely to use an orbit closer to 12AM 12PM SSO or the inverted
one 12PM 12AM SSO. If we have both payloads on board, we will have some intermediate solutions: for
example, EnviSat (equipped with SAR and optical sensors) which uses a 10PM 10AM SSO orbit and once we
have this orbit we need some pointing mechanisms for the SP in order to compensate for the motion of the
satellite along the orbit. EnviSat is famous for active debris removal since it is now not operative and so it is
a debris.
2. What is the type of pointing mechanism for a 12PM 12AM SSO?

If we consider our side-view (POV from minus the normal to the orbital plane), we can see that we al-
ways have eclipse condition (the shadow changes during the year) i.e. also a portion of the orbit in eclipse.
If the red is the solar panels, when we rotate if we keep a three axis stabilized geometry, we will lose the
orientation and so the SP are not pointed towards the Sun and so we need to have some pointing mechanisms.
3. What about 6PM 6AM and 6AM 6PM SSOs? Some properties/advantages?

Let us consider a 3D plot and let us consider the Sun projection Ŝ. Let us plot in green the 6AM 6PM
and in blue 6PM 6AM. The line of nodes is the same. The orbital plane is not the same because for the 1st
Ω − αs = −90deg while for the 2nd Ω − αs = +90deg, they need to have the same inclination and to have
the same inclination they will be on different orbital planes. These two orbital planes forms an angle between
them (it is a V-shape) of about 15deg.

Are they both useful for power?

If we plot again from the POV of AN the situation at winter and summer solstice, and we can see that
we reach similar conclusions from the power POV (same properties and benefits) but what we can conclude
about the eclipse and the orientation of the normal to the orbital plane towards the Sun is likely different.
(a) CosmoSkyMed satellites are on 6AM 6PM.
(b) TerraSarX satellites are on 6PM 6AM.
Due to the geometry of the two orbits, for the 6AM 6PM the summer solstice is the best case and the winter
is the worst case, while for 6PM 6AM it is exactly the opposite situation, so there is not eclipse at the winter
solstice and instead the worst case is at the summer solstice.
Regarding thermal control, since for SSO we have a constant relative geometry, this will simplify in general the
design of the thermal control subsystem because we have some more constraints in the relative direction of the Sun
w.r.t. to the satellite.
Inclinations of our interest: once we set the constraint about the precession rate equal to the rate of the right
ascension of the Sun, we have a constraint which links three parameters i.e. a, e and i that cannot be independent
anymore:

Ω̇ = Ω̇(a, e, i) = α̇s

If we focus our attention on circular orbits (e=0), the condition of Sun synchronicity generates a constraint which
links the semi-major axis a and the inclination i: so, for each semi-major axis a there is only one inclination i
that makes the orbit SS. We can build a plot where we have the altitude h on one axis and the inclination i on y
axis, and it is possible to see how the inclination of our interest will be slightly above 90deg, and in particular for
Space Flight Dynamics Michele Santarpia 73

LEO orbits we are interested in what happens between 400km and 800km which correspond to inclinations of our
interest approximately of 97/98deg.
SSO disadvantages:
1. Due to the constant relative geometry w.r.t. to the Sun, we have some disadvantages which concern some
additional perturbations. In particular, as we keep a constant geometry w.r.t. the Sun, some perturbations
due to the Sun will be emphasized. The SRP and the 3rd body effects due to the Sun may be emphasized
simply because we are experimented the same relative geometry; so, for generic orbits we might have some
effects that create a compensation, while with the same relative geometry, even if some effects are small, there
may be a resonance and so the effects may be amplified (amplifications of perturbations). In particular, 3rd
body effects will tend to slightly change the inclination, but this small change in the inclination ∆i
generates a small change in the precession rate due to J2 i.e. ∆Ω̇J2 :

∆i → ∆Ω̇J2

This will mean that, if we change a little bit our inclination, we can lose the condition of Sun synchronicity,
so what we typically do is to periodically compensate for the perturbations of the inclination due to the 3rd
body effects of the Sun because we want to keep our Sun synchronicity.
2. Another disadvantage is related to the coverage: if we have an inclination of 98deg, the maximum latitude
we can cover is about of maxϕ=82deg, so we do not cover the poles but we cover a really wide range of
latitudes. Of course, if we are only interested in small range of latitudes, this may become a disadvantage: in
fact, if we consider lower inclination orbits, we will spend the entire time in a narrower range of latitudes and
so as a consequence our revisit time performance will be better.

Frozen Orbits They are related to non-sphericity effects and, in particular, to the zonal harmonics. The concept
of frozen orbit is not an alternative to Sun-synchronous (we may have an orbit that is both frozen and Sun-
synchronous). The idea of frozen orbit is that we want to keep a frozen argument of perigee ω and also
nullifies the long periodic oscillations of the eccentricity e.
If we consider only J2, we can keep a fixed perigee just selecting a critical inclination but, for remote sensing mission,
it is typical not so nice for two problems:
1. The 1st problem is that if we consider a critical inclination orbit, it will be non SS.
2. The 2nd problem is that we have an inclination of about 60deg, and if we are interested in a wider range of
latitudes, this inclination becomes a disadvantage.
Why do we like to have frozen orbits for remote sensing?
The rationale for frozen orbits (keeping frozen perigee) is connected to the fact that we like to have a constant
altitude profile along the orbit and so we want to pass at given latitudes always at the same local altitude: this
is typical important for remote sensing, for example, to compare subsequent measurements.
Which are the constraints?
Let us consider to be on a circular orbit around a spherical Earth: the local altitude will be constant. Actually,
this constant local altitude is not possible and there are three elements to consider:
1. Earth is not spherical: the Earth is not a sphere (but an ellipsoid) and so even if we have a perfect circular
orbit there is variation of 20km in the local altitude between the equator and the poles. So, instead of
requesting for a constant local altitude, we can just request a constant altitude profile which means
that, as we fly orbit after orbit, we see a repeat in the altitude profile.
2. Long periodic effects for eccentricity to consider: since we are including non sphericity, we are not in
Keplerian dynamics anymore: as we have said before, when we include non-sphericity, from an osculating-
parameter point of view (i.e. from an instantaneous perspective), all the parameters will oscillate and so the
eccentricity also will oscillate, and so if we propagate an orbit (like in GMAT) only with non sphericity effects
(gravity perturbation), even if we give an initial eccentricity which is 0, there will be an oscillation, and the
eccentricity cannot be negative, so there will be a behavior with an oscillation without the possibility to be
negative and so finally the average eccentricity of the orbit will never be zero. So, for this reason, since
Space Flight Dynamics Michele Santarpia 74

the average of the eccentricity cannot be zero, we can see how we do not have a circular orbit (since we
are looking from an instantaneous perspective) and so they cannot exist due to gravity perturbations.
3. Secular and long periodic effects for AoP to consider: since there is always an average eccentricity, we
need to pay attention to where is the perigee and this will impact the altitude profile that we will encounter
during our orbit.
So, a constant altitude profile means that, orbit after orbit, when we pass above a given latitude, we pass at the
same local altitude.
How do we obtain this phenomenon?
The concept of frozen orbits is based on the fact that we want to try to have orbits where:
1.
dω
=0
dt sec+lp

and what we are considering now are also the secular effects and the long periodic effects (so, we want to add
something more than before).
2. Then, we want to have orbits for which the long periodic variation of the eccentricity is zero:

de
=0
dt lp

So, in order to have a constant altitude profile, we need to have a fixed perigee and a fixed value of the eccentricity.
In order to realize an orbit like this, we want to consider not only secular but also long periodic perturbations. If
we have J2 , there is one way to nullify the movement of perigee: using a critical inclination.
Is there a way to nullify the rate of change of the AoP without using the critical inclination?
We have to go beyond J2 , and in particular we can use J2 and J3 together and we can play with secular and long
periodic effects (J2 generates secular and long periodic effect, J3 generates only long periodic effects). Remember
that only even zonal harmonics can give secular effect (so, J3 cannot give secular effects). J2 models the oblataness
while J3 the asymmetry of the Earth (the fact that we have more mass in one hemisphere).

J2 ∼ 10−3 J3 ∼ 10−6

Let’s write the secular and long periodic variation of perigee due to J2 and J3, and also the variation of the
eccentricity (only long periodic, only due to J3).
If we want to nullify the long periodic oscillations for the eccentricity, we have several possibilities:
1. The 1st and the 2nd possibilities have no interest because otherwise we lose the benefits of SSO. In case of
the 1st one, we have the sin i = 0 which means inclination equal to 0 while for the 2nd one we need to have a
critical inclination.
2. The 3rd condition is about the argument of perigee, i.e. we can impose cos ω = 0 and we can find two possible
solutions: ω = 90◦ or ω = 270◦ .
After selecting ω = 90◦ or ω = 270◦ , we impose θ = 0deg to nullify the secular and long periodic oscillations
for the AoP and we find the required eccentricity. So, finally, if we want an orbit where, from the secular and
long periodic perspective, both the perigee and the eccentricity do not change, we need to select a perigee equal
to ω = 90◦ or ω = 270◦ and, then, we need to select an eccentricity which is in the order of 10−3 (given by
the expression that we have found imposing θ = 0). In this way, we can select the inclination based on other
considerations like, for example, Sun synchronicity, and then we can have an orbit which is also frozen. So, finally,
with good approximations, the perigee will not move so much, so we will limit the maintenance of our perigee
(perigee correction maneuvers).
Numerical approach POV: if we approach the problem from a numerical POV, we can use a similar method
with more zonal harmonics: it is possible to take into account higher order effects and so the concept of frozen
orbits may be applied with higher degree zonal harmonics (J4, J5, J6, etc.). The required eccentricity will change
Space Flight Dynamics Michele Santarpia 75

just a little bit. If we consider the CosmoSkyMed mission, the eccentricity is 0.00118. So, typically, for remote
sensing missions these numbers of eccentricity come from frozen orbits equations.
How do we really obtain SS + frozen orbits?
The problem is that, in theory, if we change the eccentricity we lose Sun synchronicity. Let us consider the logical
flow of what we are doing:
1. We want a SS frozen orbit for remote sensing.
2. We start as design parameter from e=0.
3. We know that SS generates a link between the semi-major axis and the inclination, so once we select the
semi-major axis (altitude) we have a given inclination.
4. As we want to add the frozen condition, we say that the argument of perigee must be 90◦ or 270◦ while the
eccentricity must be the one derived from the equation.
5. Apparently, we violate SS because now eccentricity e is changed and Ω̇ is changed which is not equal anymore
to α̇s . Actually this is pure theory: the nice point is that the angular rate Ω̇ is slightly sensitive to eccentricity
and so it is negligible for near circular orbit. If we have a δe, what happens to δ Ω̇? We consider a 1st order
budget where:

∂ Ω̇
δ Ω̇ ∼
= δe
∂e e=enom

This is our 1st order development. If we are considering near circular orbits, we can consider that our enom
is close to zero so:

∂ Ω̇ ∂ Ω̇
δ Ω̇ ∼
= δe = δe
∂e e=enom ∂e e=0

If we compute the derivative ∂∂eΩ̇ |e=0 , it is equal to 0. So, for small eccentricities, the nodal precession is really
not sensitive to the change of e. So, the key message is that we can adjust the eccentricity to have frozen
orbits without really losing the Sun synchronicity, that’s why it becomes easy to have a SS frozen orbit.

Effects on the orbit of the 1st Sectorial Harmonic J22 In the development of the gravitational potential,
the 1st sectorial harmonic J22 takes into account the tri-axiality of the Earth and this term models the fact that if
we look at the Earth from the top we do not see a circle (i.e. we do not see an axial-symmetry in the distribution
of mass) but we have an ellipse (so more mass along a given direction).
J22 effect is in the order of 10−6 (pretty small), so why do we pay attention?
This is the case in which, even if the perturbing force is pretty small, in some cases, there might be some resonance
which will make visible some effects. If we look from the z-axis (North pole), we have an ellipse. Let us sketch the
Greenwich meridian which forms an angle of 15deg w.r.t. the direction of the Earth semi-major axis. The Earth is
a little bit elongated with the elongation which happens at 15deg west of Greenwich. Imagine to be on a orbit and
to be in a given position instantaneously, imagine to have an equatorial orbit so the satellite is in the orbital plane
(so, we do not consider the projection) and consider that it is subjected to the gravitational field.
From the POV of the satellite there are two portions of the Earth / two half of the Earth in terms of masses, and
we can consider the gravitational attractions of these two portions: what happens is that the two masses are the
same for the symmetry but the light blue one is closer to the satellite due to the geometry, so it is intuitive that,
due to this geometry, we will have a sort of net force that pulls the satellite towards the blue region, so there will
be a perturbation due to the non-sphericity (due to the ellipticity of the equator) that pulls the satellite
in that direction.
This concept of non central force is pretty small as effect and this will generate some additional oscillations that
will not give significant effects.
This effect becomes significant if we have resonance: it happens when the orbit is such that the satellite is
always subjected to the same perturbation, so it is subjected continuously to the same type of gravitational pull
due to this perturbation. So, the idea is that if the satellite moves in a generic geometry, the satellite will be pulled
Space Flight Dynamics Michele Santarpia 76

in a given direction but as the satellite is in another position w.r.t. the Earth, it will be pulled in another way so
the net effect will not be significant.
There are some satellites that instead will always have the same type of gravitational pull due to this perturbation
and this will activate a resonance: this is especially true for geostationary satellites because this type of satellites
for their nature stay in a fixed longitude, so they will always experiment the same type of gravitational pull. So,
even if the forcing function is pretty small, then the effect may become pretty visible.

Instability of Geostationary Satellites Let us consider the Earth as an ellipsoid due to J22 and let us imagine
that our satellite is at a given longitude λ. The satellite sees an asymmetry in the gravitational field, because there
will be a larger gravitational pull from the light blue half compared with the yellow half of the Earth.
The circle in red DOES NOT represent the orbit of the satellite (it is used only to sketch different longitudes that
the geostationary satellite can have)! In fact, the plot that we are considering is not in the inertial coordinates but
in ECEF: when we look at a geostationary satellite in ECEF, it is not moving (it stays at a fixed position).
Once we have a given satellite at a longitude λ, we can consider the gravitational pull from the two parts: J22 will
generate on this satellite, that nominally is geostationary, a net force that pulls the satellite towards the longer axis
of the ellipse because there is more gravitational attraction. If the satellite is in another position, we can imagine
to have another net force. The result of this situation is that there are long periodic oscillations with a period
of hundreds of days (in GMAT by propagating form 1200days we look at a very significant drift but it looks like
a drift due to the long time). These oscillations does not have any dumping and once they are activated they will
never stop (because gravity is conservative, we cannot dissipate any energy). This means that the longitude starts
to change due to this effect.
Concerning the discussion about the net force, there are only four longitudes where there is no net force
which correspond to the points of the semi-major and semi-minor axes because, even if there is an ellipticity, the
gravitational field will look perfectly symmetric and so there will be no net pull in any direction. So, there are four
equilibrium points: only at four longitudes the geostationary satellites will stay nicely quite, otherwise there will
be the net force that will tend to change the longitude. Of these four points, two of them are stable and the other
two are unstable. Since there is no dumping, we cannot have asymptotic but only simple stability or instability.
It can be intuitive that the stable points are the ones on the semi-major axis (longest direction) because the net
force pulls in that direction, instead counter-intuitively those two points are the unstable ones. The S points i.e.
the ones along the semi-minor axis (one corresponding to India seen in simulation) are the stable ones and so they
are not the ones towards which we have the net force.
How do we explain this phenomenon?
What we have to remember is that we are in a rotating frame (ECEF, non inertial frame!), and so the final effect
on the change of longitude will be opposite to the direction of net force!
This concept is linked to our discussion about the rendezvous, where we have an INT and TGT. The TGT is a
rotating object and so if we consider our motion w.r.t. the TGT, this is the motion w.r.t. the rotating frame that
is fixed with the TGT. We have seen that in order to approach the TGT, we do not need to push the INT towards
the TGT but we need to give a ∆V in the opposite direction because, in this way, we generate with this negative
∆V an ellipse with smaller semi-major axis and, finally, in the rotating frame of the TGT, if we push the satellite
in the opposite direction we will finally move towards the TGT.
We have to consider a similar perspective: here, we have a rotating frame, we consider our motion in longitude, so
our motion w.r.t. a rotating target which is the nominal geostationary satellite.
The idea is that if the net force is in that direction, this corresponds to a ∆V in that direction, so if the satellite
is pulled in this direction with a sort of ∆V due to J22, finally it will start to drift in positive longitudes and this
explains why the equilibrium points are not the ones toward which we have the net force but the opposite ones so
in the points from which the net force is pulling the satellite.
[remember that the circle in red is not the orbit that a satellite flies on but the satellite is a point on this circle
because this is the Earth fixed geometry]
Non-Sphericity Effects - Summary:
Space Flight Dynamics Michele Santarpia 77

• Effects on all orbital parameters and if we run a propagation in GMAT with non-sphericity we will see
oscillations in all the orbital parameters. Not all these effects are secular.
• Secular effects (effects which accumulate in time) are only due to even zonal harmonics and will appear on
Ω̇, ω̇ and Ṁ (rate of mean anomaly).
• J2 secular effects: 1) precession of the perigee which gives us the difference between anomalistic and nodal
period but also leads to the concept of critical inclinations (how to nullify secular effects? linked to Molnyna
orbits) 2) precession of the line of nodes which leads to the idea of SSO.
• Secular J2 and long periodic effects J3 which allow us to define the concept of frozen orbits.
• J22 which leads to instability of geostationary satellites.

Differential J2 Effects Let us study the differential J2 effects to analyze:

1. J2 from an orbital maneuvering point of view.
2. Fragmentation event.

J2 from an Orbital Maneuvering Point of View We can use this perturbation in order to maneuver the
satellite and change the plane. We know that J2 generates a precession of the line of nodes, so we can work with
the idea of differential J2 effects. Sometimes, in fact, we are interested in generating a δ Ω̇ (a difference in the
precession between two satellites) by changing the parameters from which the perturbation depends. We know that
Ω̇=Ω̇(a,e,i): so, this means that, if we are able to change a little bit semi-major axis, eccentricity or inclination, we
can change a little bit the precession of the line of nodes δ Ω̇. This is interesting when we have to change our plane
but we want to save propellant.
Let us imagine that we are trying to launch our satellite on a SSO with a given local time at the AN: to launch
a satellite on a final orbit, we have to launch at the right time! If we make a mistake on timing but we launch in
the right direction, what changes is the AN position. Let us imagine to have a late launch with some delay, which
means to launch some time after the right moment, which will be the final effect? The final effect is that we are
able to launch on an orbit where basically the nodes have moved (red orbit): this depends basically on the fact that
the Earth has rotated a little bit more and so we have launched in another inertial position.
We would like to be on the blue orbit! So, there is an error ∆Ω at the time t0 given by the difference between Ω1,0
and Ω2,0 . How do we can correct?
If the orbits have the same inclination and radius, there is only an error in RAAN, the two orbits will exhibit
the same precession. So, we are on the wrong orbital plane (red one) but the correct orbital plane (blue one) will
drift and also our plane. So, the problem is to bring our satellite again on the blue orbit. There are two possible
strategies that we can use:
1. Common strategy: the 1st idea to correct ∆Ω is to use the orbital maneuvering strategy where we have
two intersection points that we can use to rotate the velocity vector to jump on the correct orbit. There is
an angle θ between the two orbital planes, we need to rotate of this angle around the radial direction.
2. Differential J2 perturbation: the 2nd option consists in trying to generate a differential J2 perturbation
in order to have that, naturally due to J2, the red plane will move towards the blue plane.
Since Ω̇ = Ω̇(a, e, i), we can consider to give a δa, δe or δi. Let us write the variation in δ Ω̇ at 1st order:
     
∼ ∂ Ω̇ ∂ Ω̇ ∂ Ω̇
δ Ω̇ = δa+ δe+ δi
∂a ∂e ∂i
Let us compute the partial derivative in the nominal conditions (initial orbit) in order to understand the
sensitivity of δ Ω̇ w.r.t. the eccentricity, inclination and semi-major axis:
(a) Sensitivity on eccentricity: if the orbit is near circular:
 
∂ Ω̇
=0
∂e e=0
This basically means that the nodal precession is not sensitive to eccentricity.
Space Flight Dynamics Michele Santarpia 78

(b) Sensitivity on inclination: let us compute the partial derivative with respect to the inclination:
 
∂ Ω̇
= f (sin i)
∂i

It depends on the sin i. Since SSO are quasi polar orbits, the sin i is around its maximum value and so
the sensitivity is good.
(c) Sensitivity on semi-major axis: also for semi-major axis there is a good sensitivity.
So, we can change δa or δi to have δ Ω̇:
(a) Analysis on inclination: let us consider that we want to have δ Ω̇ < 0 because we want to move
towards the other orbit. We can apply a negative δi and this will slow down the nodal precession. Since
we slow down the red orbit, the blue orbit will approach the red one. Of course, changing the inclination
means apply a ∆V = 2v sin(θ/2) = 2v sin(δi/2) which could appear not so smart since it is an expensive
maneuver but, since δi << i, it makes sense, so this is a small plane change that wants to activate a
differential J2 effect, so the maneuver will be a small one but as always we need to wait some time. If
we need to correct a ∆Ω, this will require some time because ∆Ω = δ Ω̇∆t and, of course, smaller is the
change of inclination, the smaller will be δ Ω̇, larger will be ∆t. When we reach the correct orbit, we will
need to bring again the inclination to the one of SSO.
The idea that we have used consists in exploiting the differential J2 effects as a way to maneuver our
satellite. Instead of maneuvering by ourself, we change the orbital perturbations in the direction that we
like: this is smart but we will always to trade off between cost and time.
(b) Analysis on semi-major axis: why do we not use δa? In theory we can since the sensitivity is good,
but we need to pay some attention because, since we are in LEO, we can have problems with the Drag
and so the recovering can require more energy. We need to pay attention to the direction towards with
we change δa (side effects of this maneuver).
The nice idea to change the inclination is good also because we are not changing significantly the other perturbations
(and so Drag). If we use δi, it is true that we have a change of δ Ω̇ but also a variation of the rate of change of
mean anomaly δ Ṁ since Ṁ = Ṁ (a, e, i), where these three dependencies are not of the same type because the
one from the semi-major axis is very strong (Keplerian dependence), while the dependence from the eccentricity
and inclination is only related to J2. So, in general, the sensitivity of the rate of change of mean anomaly to the
inclination will be 1000 smaller than the Keplerian one (but present!). So, a change of the inclination means a
change in the orbital period. Sometimes, we need to pay attention to the phasing of the satellite and so we need to
pay attention also to δ ω̇ (another side effects of δi).

Fragmentation (or Break-Up) Event The differential J2 effect explains what happens to a cloud of debris
generated after a fragmentation event. Break-up events are violent fragmentation phenomena which can be of
two types: explosions or collisions. So, if we imagine our satellite flying over a given orbit, we can imagine two
catastrophic scenario:
1. Explosion of the satellite.
2. Collision with another impactor that is coming.
Immediately after the break-up, instead of a single object (instant before), we have a certain number of fragments
(instant later) where each fragment will have a given area and mass. From the orbital point of view, the break
up phenomenon can be seen as a phenomenon where we are giving a different ∆v to each fragment (we are in the
same position with a velocity v before, and then in the same position and instant we have a certain ∆v), they are
spread in every direction (there is a sort of isotropy), so a generic ∆v will introduce a ∆ in the orbital parameters.
There will be a ∆vi for each fragment and so for each fragment there will be a ∆a,∆e, ∆i, etc. The orbits of the
fragments are so modified orbits of the parent (original orbit + ∆v due to the break up event).
The cloud of fragments, due to the difference in the orbital parameters (∆a, ∆e, ∆i, ∆Ω, etc.), will not be on the
same orbit. So, for example, they will be characterized by a difference in the mean anomaly rate ∆Ṁ . For this
reason, we start from a dense localized cloud of debris but then the cloud will be dispersed along the orbit. At
a given moment, we will not be able to identify a limit in the mean anomaly at which we can find the fragments
Space Flight Dynamics Michele Santarpia 79

because they are spread all along the orbit, primarily due to Keplerian effects (so, a little bit of change of semi-major
axis will spread all the fragments in along-track).
Then, we will also have differential nodal precession because we have ∆i and ∆Ω. Different fragments will have
some difference in the inclination which will remain the same. Regarding the variation of AN, initially, all the
fragments will have similar AN because the orbital planes are all close each other. Then, because of ∆i, in time we
will have a variation in Ω: so, finally, after some time there will not be an orbital plane on which all the fragments
more or less will be since the nodes will spread for all the 360deg becoming a belt of fragments.
Let us look what happens to the fragments on the Ω, i plane: each point on this plane defines the orbital plane. The
point in red is the parent that generates the fragments. As we have the fragmentation event, all the fragments will
be located around the parent and then, because of J2, the difference in inclination will remain the same while the
difference in RAAN will increase. In the diagram, t0 represents the fragmentation event instant, t1 the instant after
the event, while t2 represents a time instant much larger than t1 at which the fragments will be spread in RAAN
covering all the values from 0 to 360 degrees (the fragments will have the same interval in terms of inclination).
So, in summary, initially we have a cloud, then this cloud will quickly spread along the orbit because of Keplerian
effects, then the fragments will be spread all around because of the differential J2 effects. This makes evident that,
if some time passes after the fragmentation event, recovering which was the origin of the objects may become more
and more challenging (also because we have some uncertainties due to also Drag).
Qualitative discussion about formation flyings: when we consider satellites that fly together in a formation
(satellites that will fly relatively close each other), it is intuitive that they are flying on very similar orbits, so their
orbital parameters are very close each other, they will have only small differences in the orbital parameters and
this explains why they move close and slowly with respect to each other. When we are in a formation and we
select a different inclination, we have to expect that the different members of the formation will be subjected
to different J2 effects. This differential might be relatively small but the nodal separation will increase. So, in
formation flyings, if we have a ∆i, this will lead to a ∆Ω̇ due to J2. This means that, even if initially there was not
∆Ω (the nodes were close), the nodes will tend to depart (allontanarsi) from each other: so, J2 tends to destroy our
formation because if the orbital planes are characterized by different rates the separation will increase. Typically,
with maintenance we will try to remove differential J2. Another idea is avoiding to have a difference in inclination,
so we try to have the geometry we want with the same inclination (we become smarter in the design of formation).

5.8 2) Drag
Aerodynamic Drag: it is the 2nd main perturbation. The considerations about non sphericity are not applicable
in this case. Satellites do not have an aerodynamic shaped body (odd-shaped bodies). Only in VLEO we tend to
have an aerodynamic shape where we really try to minimize the Drag effects. The satellite will move with a given
velocity v. Due to the interaction with the residual atmosphere, we will have the Drag force D that will push the
satellite in the opposite direction of the satellite velocity w.r.t. the atmosphere. We have to play with this force
that counteract the motion of the satellite.
1
D = − ρv 2 CD Av̂
2
We are in Free Molecular Flow (FMF) (highly rarefied region): in LEO, the effects on the orbit are significant
and decrease with h. For all the orbit stuff, we are interested in the specific force (perturbing acceleration) due
to Drag:
1 2 CD Aatm
apDrag = − ρvsa v̂sa
2 m
1. The Local Atmospheric Density ρ is completely different from the one we have in troposphere or strato-
sphere but it is related to the residual atmosphere.
2. The velocity is the velocity of the satellite w.r.t. the atmosphere (in aeronautical terminology it is the
airspeed) given by the difference between the inertial velocity of the satellite v and the inertial velocity
of the atmosphere vatm , so we need to make some assumptions about the motion of the atmosphere in the
inertial frame:

v sa = v − v atm
Space Flight Dynamics Michele Santarpia 80

3. The Drag coefficient is challenging to predict with significant high accuracy: in general, for all the satellites
(given our configuration and the type of fluid dynamics we are using), it is a number between 2 and 3.
4. The area that we should consider is the area exposed to the atmosphere Aatm (cross section of the
satellite with respect to the atmosphere).
This expression contains a lot of uncertainties: once we consider the cross section, we are introducing
something which depends on the attitude (at the same time, the attitude depends on the configuration we have).
Let us see the satellite from the top: for the case in which we do not have any elongated surface, if we change
the attitude, the cross section does not change significantly. Instead, if we consider a satellite with some elongated
surfaces, the situation is different: in fact, if we look at the satellite in the local horizontal, the exposed area
(cross section) will be completely different (so, the change of attitude amplifies the Drag). This dependency on
the attitude and on the surface interaction adds complexity and uncertainties. Our capability to predict this
phenomena is limited. Of course, if we know our satellite, we will have a decent estimate of the cross section.
Instead, if we have a debris object we do not know initially the area, the mass and the Drag coefficient (we do not
know the materials or in general we have uncertainties). Furthermore, the debris object is tumbling and so the
cross section will continuously change. This will introduce a very strong uncertainty in our analysis.
Modeling the Drag for satellites is complex and this generates a significant uncertainty. The uncertainty on Drag
cannot be separated from the uncertainty on density because many times in our analysis we do not have a
clear way to separate what depends on variation of density and what depends on the change of ballistic coefficient
CD A/m (the two parameters are unobservable in an independent way). In general, the uncertainty of Drag cannot
be removed and we need to accept it and take into account in our analysis that’s why we typically assume margins
or we assume a stochastic perspective and not a deterministic one when we talk about Drag.
When we try to estimate the orbit of an object under the effect of Drag, we may have two main sources of
uncertainties:
1. One is on the object itself (what is the real ballistic coefficient and how it changes in time?).
2. The other one is on the atmospheric density.
When we use the two uncertainties together, if we have a given effect on Drag, it is statistically possible that the
effect is double or much smaller, so the uncertainty is of the order of 100%. Typically, the density plays for the 50%
and an uncertainty on the object adds another 50% (typical weights of the uncertainties). Finally, we will need to
take into account the uncertainty in the analysis. So, when we make a prediction with Drag in LEO and we predict
that an object is somewhere with a given velocity, there is a volume of uncertainty or to be more technical there
is a covariance associated to this prediction.
If we consider the prediction of an orbit, it should be intuitive that our uncertainty due to Drag is not applied in
the same way in all directions: when we try to predict forward the motion of the satellite, most of our uncertainty
will affect the along-track coordinate (our anomaly, we are uncertain in terms of where we will be in terms of
anomaly) while the other components of uncertainty will increase much more slowly. This discussion explains why
we are not able to estimate the risk of collision with a lot of time in advance because, if we have a debris in LEO,
the uncertainty in the prediction will be large and so there is a limit in the analysis we can do when we consider
one month of prediction for example and everything will become more clear as we approach the event and this
explains why the collision avoidance maneuvers in space are done not with much time in advance to avoid being
over conservative.
For the atmospheric density there are many models: there is a significant temporal and spatial variability
because the density depends on the interaction between the upper atmosphere with the Sun and with the magnetic
field of the Earth, so the main actor is the Sun or the solar wind. The solar radiation interacts with the atmosphere
and this will change completely the atmospheric density. When we are in Sun conditions, so when the temperature
increases, at all the altitudes the atmospheric density will increase (the atmosphere inflates under the effect of
the Sun). When the Sun activity is very intense, we will have a further increase of atmospheric density (cycle of 11
years of the Sun, where on average we have solar maximum and solar minimum but there are in this period the solar
storms with significant increase of atmospheric density and this links the Drag with space weather). When we have
extreme solar events, this is dangerous for satellites electronic components. From the orbital dynamics POV, for
LEO satellites there is a very significant effect of space weather and it will influence the atmospheric density
and so Drag. A very famous event that explains the impact of space weather on Drag is an event happened on
2022: solar storm knocks 40 SpaceX satellites out of orbit. Kp index interesting for solar activity (and so Drag).
Space Flight Dynamics Michele Santarpia 81

Main effects of Drag on orbital parameters:

1. Orbital decay phenomenon: since Drag is a dissipative force, the energy of the orbit is impacted (losing
of energy) and so the semi-major axis will decrease:

dε da
<0→ <0
dt dt
Our capability to fight against the orbital decay will determine our life in orbit. As we finish our propellant,
there are no ways to fight the orbital decay. From a mission POV, this adds cost because in order to keep
our orbit, we need to spend propellant in order to periodically correct the semi-major axis. From a space
environment POV, this is a positive phenomenon, since the Drag will continuously cleans low orbits and
especially the VLEO from debris and any type of object.
2. Circularization of the orbit: we have also an effect on orbital eccentricity and it accumulates in time as
effect. Drag will continuously tend to decrease the eccentricity of the orbit.
de
<0
dt
If we imagine our planet, a qualitative representation of the atmosphere and an elliptic orbit with one part
of the orbit which is well within the atmosphere. If we have an orbit like this, the effects of Drag will change
dramatically as we move along the orbit. There are two aspects to be considered:
(a) At perigee we have the maximum density (since density will decrease with altitude) and also the
maximum velocity which means that Drag is maximized when we are at the perigee.
(b) At the apogee, due to similar considerations, Drag is minimized.
Drag is something which gives us a negative ∆v: Drag is pushing our satellite against its velocity. At 1st
order of approximation, we can imagine to concentrate the effect of Drag only at perigee. The idea is that
because of Drag we will have a ∆vDrag that acts at perigee (of course, it is not true that it acts only at perigee
but it is an easy way to understand what happens). If we apply a tangential ∆v at perigee, one point is not
affected i.e. the perigee itself and all the effect will be on the opposite points and so we will keep the same
perigee bringing the apogee down. If this phenomenon goes on, of course we will have that the semi-major
axis is decreasing but also the orbit is becoming more and more circular (eccentricity is going down since the
difference between perigee and apogee is reducing). This is an extreme simplification but it is real that when
we have an ellipse the Drag operating at the perigee will be the maximum w.r.t. the one at the apogee, and
so considering the cross effects i.e. perturbing the satellite at the perigee the effect is on the opposite point,
what will happen is that both the apogee and the perigee will decay but the apogee will decay faster than
the perigee. If we plot the radius at perigee and apogee in time for an orbit subjected to Drag, what happens
is that the perigee will decay with a given rate lower with the one of at the apogee. It is evident that this
difference in rate will reduce the eccentricity. Sometimes, in planetary applications when we have a planet
with an atmosphere we can try to use it to circularize our elliptic orbit. This will be something which gives
us some help in Martian missions.
3. Non linear decreasing of the semi-major axis: the linear behavior for the radius of apogee and perigee
makes sense only when we consider a relatively short time scale, so the linear decrease is an approximation
while in reality there is an increasing negative derivative both for perigee and apogee. For example, let us
check the semi-major axis in time: it will decrease and in the long time the decreasing is not linear (the rate
of decrease will be larger and larger until we will typically re-enter the atmosphere). The non linearity is
related to the increasing of density as we decrease the altitude and so we are entering in lower layers of the
atmosphere and also due another phenomenon: when the eccentricity is 0, the linear velocity on a circular
orbit as we know is:
r
µ
v=
r

So, as the satellite decreases, the semi-major axis decreases but also the velocity is becoming higher and higher
and so the Drag is increasing. So, finally the non linearity in the decreasing of the semi-major axis is due to
the increasing both of density and velocity.
Space Flight Dynamics Michele Santarpia 82

4. Drag paradox: Drag is a force which is pushing the satellite back and differently from what happens for
an aircraft (where due to Drag the velocity is decreasing), in space what happens is exactly the opposite (it
sounds strange) because due to Drag we are decaying and due to this phenomenon we are becoming faster,
so Drag is making us faster. This concept is called Drag paradox: it makes sense since we are playing with
energy which is given by:

v2 µ µ
ε= − =−
2 r 2a
The potential energy is decreasing more by reducing the semi-major axis and the kinetic energy is
increasing less and finally the total balance is given an energy which is reducing.
5. Change of phasing: as an interesting consequence of Drag paradox, becoming faster causes a change
of phasing: as we know, we have some tolerance on the nominal semi-major axis which means that we will
let decay the orbit for some time but, as we reach the minimum semi-major axis, we will try to increase again
the semi-major axis (Hohmann maneuvers or something similar). Typically, for many times, this problem
of correcting the semi-major axis is not the only problem we have but the other one is that we need to pay
attention to the timing / phasing along the orbit. Let us consider in blue our nominal orbit: due to
Drag,
p we will decay becoming faster both in linear and angular terms, in fact if we consider the mean motion
n= µ/a3 , due to the Drag paradox the mean motion n will increase (and the orbital period reduces). So,
decaying we will be on a lower orbit in red but also we will be at a different anomaly: so this means that,
w.r.t. the nominal satellite in blue, in reality, due to Drag the real satellite will be on another orbit
but also at a different anomaly.

So, what is the relative motion when we consider these two objects?

It’s like that the real satellite is going down and then moving forward. So, when we have to ma-
neuver, we have not to correct just the semi-major axis (to correct a ∆a) but also the anomaly (to correct a
∆ν).

Why do we are interested to correct also the difference in anomaly?

Let us start considering, for example, to apply an Hohmann maneuver (in yellow) to the real satellite. When
we arrive on the nominal orbit, where is the nominal original satellite? As we know, the time to fly half of
the transfer ellipse (so, the time of the Hohmann maneuver) is:
s
a3tr
τhoh = π
µ

From the geometry, anom > atr it means that:

s
a3nom
π > τhoh
µ

So, this means that in the time in which we perform the Hohmann maneuver, the nominal satellite will fly
less than half of its orbit. For this reason, not only we are not correcting but the error also increases (there
is an additional angle).

How do we adjust timing and so remove the anomaly?

We do not use Hohmann, but we need more degree of freedom. For this reason, we use a bi-elliptical
transfer using two transfer ellipses, so we have to select the B-point, and we set the B-point in order that
in the time in which we fly these two transfer ellipses, the nominal will be exactly where the real is. So, we
can try to adjust the B-point in order to correct the timing. In the real world, we do not use a bi-elliptic like
the previous one (i.e. we do not go back to the nominal) because, typically, we like to minimize the number
of maneuvers and so the objective is instead of going back to the original orbit, we like to go on an higher orbit.

As a general policy (politica), if the nominal orbit is the blue one and we are decreasing on the red orbit,
Space Flight Dynamics Michele Santarpia 83

we will like to directly jump on a orbit that is higher than the nominal i.e. a green orbit in order to take more
time to apply a new correction because actually we like to minimize the number of ∆vs and also we like to
minimize the number of maneuvers because when we maneuver we do not do anything else, so we do not use
the payload and so we are just correcting the orbit, we like to minimize the numbers even if they are bigger
as maneuvers. We can use the bi-elliptic to correct the anomaly, typically we like to go back not to the initial
point but to an higher orbit and also we like to do something smarter about the ∆ν.
Considerations about phasing: let us imagine to have our planet, the nominal orbit and the nominal satellite.
As we know, we do not like to be back on the nominal orbit with a ∆ν because we like to pay attention to the
relative phasing: but why? which is the difference between the two satellites (blue and red)? If we have two
satellites on the same orbit with different anomaly, why should this make a difference for us? From a mission
POV why do we want a given timing? For a GEO satellite, for example, there was a consideration about
longitude because if you are in a different position we are at a different longitude. For a LEO satellite instead?
The real difference between these two satellites is not in the inertial orbit but in ECEF, so in terms of ground
track (motion w.r.t. ground). The two satellites have a different ground track, if we imagine that we fly over the
orbit and the Earth is below, since they cross the equator at different times, the point under the satellites will be
different because in the time corresponding to ∆ν the Earth has rotated, so these two satellites do not pass over
the same spot of the Earth. Sometimes we like to keep a given ground track because, if we want to fly over Naples,
we do not want to have this error in anomaly, because, if we have this error in anomaly, as we arrive at the correct
latitude, the Earth below is in a different position.

The key concept here is that: many times for our missions we need to pay attention to the ground track and
finally the real driver for orbit maintenance will not be ∆a and so how much we are allowed to decay BUT the real
driver is how much is the tolerance about the ground track (±15km). This tolerance will impact ∆a driving the
orbit maintenance strategy. This type of idea is typically applied for high performance satellites where we really
want to have a tight ground track control. Some other satellites have simpler strategies where we just really control
the range of semi-major axis and we are not controlling the phasing and so the motion w.r.t. the ground track.
Comment about ground stations: sometimes we use stations at high latitudes and if we use high inclination
orbits such as SSO, the parallel is so small that finally the change of the ground track will not have a strong impact.
Instead of course if we have stations at low latitudes, it will have an impact.
Formulation passages about da/dt: let us use the Gauss VOP equations where all the derivatives are
function of the specific force ap . We need to write the components in the RSW radial tangential and normal frame.
We have to consider the atmospheric motion for which we have two main models: we can use the fixed atmosphere
model (vatm = 0, so the velocity of the satellite w.r.t. the atmosphere vSA is equal to the inertial one v) and the
rotating atmosphere model (where for fixed or rotating we mean w.r.t. the inertial frame).
Let us consider the fixed atmosphere. The perturbation due to the atmosphere is along the inertial velocity
direction v:
1 CD A
apDrag = − ρv 2 v̂
2 m
It is an in-plane perturbation since it is aligned with the inertial velocity. So, if we have a given position of our
satellite and a given velocity, apDrag will be opposite to v and, in particular, it can be split in two components: we
can take into account the flight path angle ϕ in order to obtain the three components in the rotating frame. Then,
if we want to check analytically what happens, we have to write the velocity along the orbit, and then sin ϕ and
cos ϕ along the orbit. Then, we have to substitute these expressions in the components of the acceleration and then
we can go in the Gauss VOP equations. Finally, the effects are:
3
1 + e2 + 2e cos ν 2

da CD A √
= −ρ µa <0
dt m 1 − e2

[We do not need to remember the expression by memory but keep in mind it is non linear, it is linear only as an
A
approximation in short time scales, and keep attention to the dependence given by the ratio m ].

This is the instantaneous variation in terms of the osculating parameter (in fact, we have the true anomaly).
This expression is always lower than 0 and the semi-major axis decays more if the ballistic coefficient is larger. What
A
is important is the ratio m , if we have an object which is small but with a large mass (like a very dense CubeSat),
Space Flight Dynamics Michele Santarpia 84

this will decay much less than an object with a very wide surface and with less mass. da/dt is not linear and it
is evident that, as time passes, the density will increase and the effect will be the escalating (crescente) one. If
we have some eccentricity, this will generate some periodic variations and so there will be some oscillations. If the
eccentricity is zero or very small, we can neglect all the other terms and so da
dt will be given by (there is no reason
why the effect should oscillate):

da CD A √
= −ρ µa
dt m
Non linear: density is not constant but also the semi-major axis is not constant. The non linearity is underlined due
to the escalation effect we have. There are some scenarios where we let this effect to happen only for a relatively
short time, so if time is short we can consider linearity (constant in time): for orbit maintenance we do this.
Satellite in formations (as RODiO): let us imagine to have two satellites over the same orbit: one is small and
massive (CubeSat) and the other one is larger, so they have two different ballistic coefficients. Furthermore, we
want that these two satellites have to fly together and at a given distance on the same orbit. It is clear that they
will be subjected to different Drag (differential Drag) and we need to do something otherwise they will fly away
from each other. They are decaying at a different rate and we want to understand which is the best configuration.
Is better to have the small satellite forward or back w.r.t. the larger one? We would like that, naturally,
they do not approach each other but we want that they will fly away from each other in order to avoid the collision.
So, what is the relative geometry which minimizes the risk of collision? Imaging that they are both decaying and
moving forward, the larger satellite (1) tends to decay faster, so it will go down and it will tend to approach the
other satellite (2). So, in the geometry in which the small satellite is forward, the will fly towards each other.
Instead, the best relative geometry is the one in which we have the larger satellite in front because, since it will
drift forward and decay faster, it will drift away from the other satellite.
If we want to use the derivative of the semi-major axis to make some analytical prediction, we need some model
of density: for example, if we can express the dependence of the density from the semi-major axis, we will have a
differential non linear equation to be solved, we integrate the equation and predict the orbital decay in this
way: these types of predictions are simplified since they use simplified models of the density but they are useful for
the operating life to have some order of magnitude regarding the entire orbital decay. Remember that there are
uncertainties and whenever we make an estimation of this type we need to have large margins.
We can extract also the derivative for the change of eccentricity with the same approach. Mathematically,
we can demonstrate that is negative and the structure is similar in terms of density, ballistic coefficients and other
terms like short periodic term with cos ν but the other term is the secular one where there is an oscillating term
and another one that does not oscillate with anomaly.
We have no out-of-plane effects since the atmosphere is fixed.
Analytical exponential model of the atmosphere: h0 is a reference altitude at which we have a reference
density ρ0 . All the complexity of the phenomenon is modeled by the scale height H of the atmosphere since
there are some models which relates the scale height to the local temperature T, the gradient of the temperature
∂T /∂h where h is the radial direction i.e. how the temperature changes in vertical, and by the local gravity g(h).
Due to the changes of the variables, we will have a complex dependence. The density increases with temperature
due to the inflate of the atmosphere. So, with this idea, if we have different orientations of the orbit w.r.t. the Sun,
there will be different conditions.
For the synchronous orbits, we can have two extreme cases:
1. Assuming to have the Sun projection in the spring, we can have the 12PM 12AM orbit (line of node aligned
with the direction of the Sun) where, if we look from a side perspective of view the geometry of the orbit,
one portion of the orbit is in eclipse and so, of course, the satellite will encounter a lower density which
means different Drag but of course this is in general an order of approximation used to have a very accurate
prediction.
2. If we have a 6AM 6PM orbit, the situation is different and now the satellite will be always in illumination
condition except for some periods of the year (close to the solstices and, for example in 6AM 6PM orbits, we
have some eclipses only when we are close to the winter solstice). So, the key message is that: even if the
satellites on these two orbits are at the same altitude, they will encounter a different Drag, so we need to take
these temperatures effects into account.
Space Flight Dynamics Michele Santarpia 85

Through the exponential model, we can understand how the density changes by order of magnitudes by changing
the altitude. This type of model is made of different intervals, which means that the equation has to be used with
different value of the scale height H. For each interval we have a different value of h0 and ρ0 . If we want to have
some approximated predictions we can go with the exponential model, if we want to have some more accurate
predictions we can use the JacchiaRoberts, etc. models. There are some models which are more accurate than
others but we cannot really solve the uncertainties of this kind of propagation. If we use a complex atmospheric
model the propagation will be much slower (practical problem) which means to have many computations to make
in a single step of propagation.
The temperature affects the density and so as we can see from the plots the relative effects will be very significant
as we increase the altitude.
As we can see from the table, there are order of magnitudes from 200 to 500 km. In relative terms, the dependence
from the solar illumination conditions (night/day, min/max) is higher at higher altitudes (there is a
reduction of one order of magnitude passing from night to day at 500 km, or from solar min to max): the effect of
inflation of the atmosphere is more evident at the external layers.
If we have a complex model, it has a number of parameters used to describe the atmosphere. If we want to be
more accurate, we can have some local measurements that give at a given space and time the coefficients for the
model to be used: if we have in theory thousands of satellites with Drag sensors around the Earth, we can use them
to estimate the local density which can be used for our model (in situ sensors which gives in situ measurements
and we are in this way more accurate). In the Drag equation, we have the density and the ballistic coefficient, so
the only way to have a good estimate of the density is to have a very accurate ballistic coefficient and so we can
separate the effects (if we have an uncertainty on both the density and the ballistic coefficient of the satellite, we
cannot separate what depends on the density and what depends on the satellite).
Rotating atmosphere model: it means fixed in ECEF but rotating w.r.t. inertial frame. We know that the
Earth rotates with an angular rate ΩL , the atmosphere will follow the surface of the Earth and so it will have a
fixed position in ECEF, so whenever we consider the atmosphere, if we take a generic point over a certain meridian
and parallel, there will be a velocity v atm along the local parallel given by ΩL × r where r is the position vector
w.r.t. the center of the Earth. The model can be applied at any altitude. The higher is the altitude, the faster is
the atmosphere in linear term (in angular term it moves constantly):
 
π
|v atm | = ΩL r sin − ϕ = ΩL r cos ϕ
2
To understand the effect of the rotating atmosphere, we need to plot again our orbit and we consider a sphere at
the height of the orbit. We start from the axes, we build our model of the Earth at the height of the orbit, we
can take the equator and the North pole, we can build the parallel corresponding to the maximum latitude that is
reached by the orbit. For graphical clarity, let us plot a retrograde orbit (but it does not depend by the fact that
it is retrograde), we can identify the inclination, the parallel corresponding to the highest reachable latitude has
a latitude equal to ϕmax = π − i. Let us draw the parallel corresponding to a generic latitude, a point and the
meridian passing through this point. For simplicity, let us assume to have a circular orbit (e=0) and this means
that we really have the entire orbit and not the projection on the sphere. If we have a circular orbit, the inertial
velocity of the satellite will be a velocity along the orbit, let us consider the velocity of the atmosphere.
To estimate Drag, we can write the equation of the perturbing acceleration ap where now we have the unit vector
v̂sa :
1 2 CD A
apDrag = − ρvsa v̂sa
2 m
Let us consider the geometry at the parallel of our interest with latitude ϕ. The inertial velocity will be v.
Drag has two components:
1. One along the direction of velocity v.
2. One component which is normal to the orbital plane.
The perturbing acceleration apDrag is made of three components:
1. The radial component ar is equal to 0 since it is related to the sine of the flight path angle (it exists in case
of elliptic orbits). Since we are assuming to have a circular orbit (e=0), the radial component is zero.
Space Flight Dynamics Michele Santarpia 86

2. The tangential component as is related to the cosine of the flight path angle. In case of a circular orbit
and absence of the motion of the atmosphere, the tangential component is the only one different from zero.
3. The normal component aw exists if the atmosphere is rotating.
Let us give a look to POV1. As we can see, we have an out-of-plane phenomenon. For the satellite in the
ascending phase, the atmosphere is pushing the satellite from our left. Being pushed by the atmosphere is a concept
similar to the cross wind for an aircraft: as we reach the pole of the orbit (highest point), the velocity is exactly
along the parallel, so we are not pushed anymore laterally (there is no cross wind), then we start our descending
phase of the orbit and now, for the rotation of the Earth, the satellite is pushed from our right. So, it is clear
that this out-of plane effect is an oscillatory effect/perturbation by nature. We need to evaluate the normal
perturbation aw (which is small): there will be effects on dΩ/dt and di/dt (the orbital plane is changing).
Key point is that the effects of aw of the inclination and RAAN are small and with an oscillatory nature:
1. In relative terms, small means that v is relatively much larger than vatm at our heights of interest.
2. Mathematically, we can convince of this oscillatory nature considering the Gauss VOP equations of di/dt
and dΩ/dt. In fact, the derivative di/dt is a function of aw and cos u, while the derivative dΩ/dt is instead
a function of sin u and aw . Also, the perturbation aw is a function of cos u because it changes sign passing
from the ascending to descending phase. This perturbation has its maximum at the equator because the
atmosphere is pushing at the maximum way. Finally, there is an oscillation of the inclination and RAAN
which is at twice the orbital frequency with a dependence like cos(2u+ϕ).
Since the effects of aw is small, this explains why we can predict our orbit using a fixed atmosphere model (there
is no real effect).
What is important for the rotating atmosphere is in terms of minimization of Drag because the main effect of
the rotating atmosphere is not the out-of-plane effect but the increasing of the cross section which depends on our
satellite. If we have a satellite with some elongated surfaces and if we imagine to keep a three-axes stabilized
profile, the idea is that we have to compute the cross section due to the fact that we have v sa (we have changed
the direction of the velocity). Now, we have a different cross section and so we have a larger area A which means
a larger Drag. We do not like to ignore this type of effect but we want to take it into account: how do we solve
this problem? Of course, by increasing Drag we decay faster and so this means spending more for maintenance, so
we need to have an approach similar to aircraft when are typically close to landing, so we need to point our nose
towards the wind or in space-terms we need to align the longitudinal axis with the direction of vsa , so
we need to rotate our satellite in order to minimize the cross section through a yaw steering maneuver (this
geometry is in the local horizontal plane, we are looking from the radial direction what is happening, so we are
rotating along the local vertical of the satellite to point towards the wind).
Let us consider the angle between the velocity and the parallel which is not the inclination because we are not at
the equator. Let us consider to look from the z direction (POV 1). What we see are different circles representing
the different latitudes. We can sketch the orbit as seen from the top (orbit pole in yellow). When we are a generic
parallel, the angle is iϕ and at the poles iϕ becomes 0 (or 180) because we become tangent.
Assuming we know iϕ , we need to understand which angle we need to rotate (red angle) i.e. -Ψ (minus if we consider
that it is positive in the other direction). ORF to BRF (321 sequence) Ψ, θ, ϕ, typical selection for body is forward,
right, down.
To solve this problem we have to compute the components of the vector vsa .
In the expression of the normal component, we have -vcos iϕ and +ΩL rcosϕ but the two terms are in the same
direction (retrograde orbit, the cos iϕ is negative).
To complete the discussion we need to express iϕ : it depends on i and latitude ϕ. To do this, let us come back to
the original tridimensional geometry and let us consider the meridian passing through the pole of the orbit. We
consider the triangle. That angle is 90deg because is is the pole of the orbit. If we consider the POV2 from the
AN, and we can represent how the orbit looks like in order to represent the arc of the orbit corresponding to the
pole. We can apply now the sine law to compute iϕ . So, after we have find the expression for the cosine of iϕ , we
can substitute the expression of iϕ in the 1st equation of tan[180deg-(iϕ -Ψ)] in order to find the final expression
for the tan Ψ which gives us the yaw angle required to keep the satellite pointed towards the velocity w.r.t. the
atmosphere and as we expected it has an oscillatory nature depending on the cosine of the argument of perigee (at
Space Flight Dynamics Michele Santarpia 87

the poles is zero, at the equator is maximum). The amplitude is of few degrees, so it is not a very significant
attitude maneuver (just a small variation w.r.t. the zero attitude but this maneuver will make a difference for
our satellite if we are in LEO and with an elongated shape). In general, this is an analytical approximation that
works only for a perfectly circular orbit and in general we will work numerically and we have some slightly little
adaptations of this strategy.
This maneuver is used to minimize Drag, we are aligned our satellite w.r.t. vsa which is the same direction of
vECEF and if we consider how we compute vsa , we are considering the atmosphere fixed in ECEF, the velocity
w.r.t. the atmosphere is equal to the velocity w.r.t. ECEF. So, vsa = vECEF . So, we are aligning the velocity w.r.t.
ECEF, so our velocity w.r.t. the ground. This is not only important for Drag but also because if we have a SAR
antenna, aligning the SAR antenna w.r.t. vECEF (velocity in ECEF is the velocity w.r.t. a ground observer) this
will generate a zero-doppler effect (this will impact the type of data we can have from the radar antenna).
In general, if we want to align the antenna w.r.t. ECEF, we do not only require a yaw rotation. The orbit is not
perfectly circular, and so to align our satellite w.r.t. ECEF we also need an additional rotation which depends on
the ellipticity of the orbit. If we imagine our elliptic orbit, and we imagine a generic position, if we have a zero
attitude, the antenna is not aligned with the velocity. If we want fully align the antenna with vECEF , we also need
an additional rotation which is a pitch rotation (we need to pith the satellite up to align the antenna with the
velocity) i.e. additional pitch steering (radar satellite performs also pitch steering, we can move the beam of the
antenna electronically called zero or full-doppler steering). From the point of view of Drag, it is important only
yaw.

5.9 3) Third body effects

3rd body effects: they consist in gravitational perturbations due to other bodies (not gravity of the primary).
1. We consider the differential gravity, not the entire gravity of the 3rd bodies but the difference between the
gravitational attraction on the orbiting object and the gravitational attraction on the primary one.
2. Since it is differential gravity force, it consists in a conservative perturbation and so we do not expect any
consequences on average on the semi-major axis (i.e. on the orbital energy). The discussion is similar to the
non sphericity effects and, as in that case, we can find a 3rd body perturbation potential that can be
used in the Lagrange equations.
3. The difference with non sphericity is the order of magnitude of the perturbation since it increases for higher
orbits (larger are the effects because the differential gravity is larger).
4. Instead, in LEO the effect is pretty small and we take them into account only when there is resonance.
There will be resonance when we have some conditions which are repetitive in time: for example, SSO are
characterized by a constant relative geometry and, for this reason, we will feel always the same 3rd body
perturbation of the Sun (same type of force) which can activate the resonance.
5. For Earth orbits, we may have two main sources of 3rd body effects i.e. Sun and Moon.
Let us study the phenomenon:
1. To model it, we can imagine to have our usual geometry with the satellite in orbit described by the orbital
parameters. We can use orbital parameters also for the 3rd body: these orbital parameters will be available if
we have the ephemeris of the 3rd body which basically means that we know the 3rd body current state.
So, we can imagine that the 3rd body has a given trajectory w.r.t. our reference frame and we can model
in this way the motion of the 3rd body by using the orbital parameters (index p because it refers to the
perturbing body).
2. We have to work with differential gravity, and so what is important is the instantaneous relative
position vector between the satellite and the 3rd body.
3. We can write the perturbing force due to the 3rd body apT B .
4. Whatever is the 3rd body, we are in conditions where the 3rd body is at much larger distance from the
primary compared with the satellite. We will use this approximation which leads to some linearization
without significant errors.
Space Flight Dynamics Michele Santarpia 88

5. Phenomenon time-scales: this phenomenon happens on two main time scales. Whatever is the 3rd body,
the period of the orbit of the 3rd body will be much higher than the period of the satellite: this gives us the
perspective that in most of the time, in one single orbit of the satellite we can consider the 3rd body at a
constant position, so the two time scales will be:
(a) The short time scale related to the single orbit of the satellite (and there will be some short periodic
effects due to the change of the 3rd body perturbation due to the fact that in reality the 3rd body moves
during one orbit of the satellite).
(b) The long time scale related to the period of the 3rd body motion and so the period of these effects will
be the 3rd body one (for the Sun, for example it is one year).
6. After the derivation of ap , we need to write the three components in order to enter in the Gauss VOP,
otherwise if we are able to write the components of the perturbing specific force as partial derivatives of the
perturbation potential we can also use in the Lagrange equations.
7. Let us concentrate on aw , i.e. on the out-of-plane perturbation of the 3rd body because, even if the
absolute value of the perturbation is relatively small, there are some consequences on the orbit.
8. In order to develop the equation we need to compute the dot product ir · iw which depends on the geometry
of the problem: we know instantaneously the orbital parameters and the ephemeris of the 3rd body, so we
can write the equation.
9. From ECI we can jump to ir is iw with Ω, i, u or to the perturbing body reference frame through another set
of angles. Once we know the rotation matrices, we know how they are oriented w.r.t. each other and make
the computation. Once we substitute, we can work with VOP we can try to understand what happens to the
change of inclination and the change of RAAN.
Let us see the results:
1. Typically, we are not interested in osculating phenomena but in long periodic or secular effects: so,
we use a time averaging process where the key idea is that, if we want to remove the short periodic effects,
we compute the average over one orbit of the satellite (since the short periodic effects are characterized by
the satellite period). There are two steps of time averaging which work with short and long time.
2. Finally, the result is that we will observe short periodic phenomena on all the orbital parameters:
the oscillations behavior qualitatively is similar to non sphericity (since it is gravity from other sources).
3. (a) The semi-major axis will undergo only short periodic variations.
(b) For inclination and Ω there can be secular or long periodic phenomena and they will become
important for some types of orbits: it may become important in case of resonance due to a repetitive
nature of our perturbation.
To be practical, which are the orbits for which we have resonance? Let us see two examples:
1. Sun Synchronous Orbits:
(a) We are typically in LEO and so the perturbation is pretty small but we have some long period variation
of inclination due to the Sun.
(b) Instead, from the point view of Moon, a SSO is not different from any other types of orbits, there is no
reason why there is a resonance due to the Moon for a SSO because the geometry w.r.t. the Moon is not
repetitive.
(c) The variation of inclination due to the Sun for SSO in long time is related to k which is a
coefficient that is of about -0.047deg/year. The effect will still look small because 0.047deg/year is just
a little bit of variation of inclination as maximum value. So, why are we really interested in a variation
of 0.047deg/year of inclination? The reason why is that we know that we need a special inclination to
have the SSO, we do not want to lose the Sun-synchronicity. So, the 3rd body effects of the Sun generate
a very small change in i, where now when we consider J2, this generates a change of ∆Ω̇ so the nodes
are not moving exactly as the Sun and we are changing the local time at the AN. To be practical, this
requires some corrections, so we will like to fight against the Sun and we will like to remove these small
Space Flight Dynamics Michele Santarpia 89

changes of inclination. If we need to make corrections, there are not happening with very high frequencies
so we will have a very relatively long time after which we will adjust the inclination.
The phenomenon is small but it depends on the local time at the AN. In fact, if we define the difference
between the two angles, this means defining the LTAN:
(a) There are some orbits that, even if with resonance, the effects will be zero like 12PM 12AM since
the difference between αs and RAAN is zero. As we know, the Sun is not in the orbital plane but
approximately it is, so it is reasonable that the 3rd body perturbation of the Sun will be mostly in
the orbital plane and so the out-of-plane component is negligible, and even if we have resonance is still
negligible. So, there is no effect on inclination simply because there is no really out-of-plane action of
the Sun (Sun is almost in the orbital plane, aw ∼= 0, so also the differential gravity, if everything is in the
orbital plane, will be in the orbital plane).
(b) Also when the difference is 90deg (6AM 6PM, Sun is 90deg forward w.r.t. the AN), there are no effects.
It is a little strange that the long periodic effects are zero. Let us consider the problem from the AN
POV, we consider the orbit, the Sun projection, let us take the two poles of the orbit, one portion of the
orbit is in the Northern, the other one is in the Southern.

Let us try to understand which is the direction of the 3rd body perturbation of the Sun. For exam-
ple, if we are in one pole, which is the direction in which the satellite is pushing (i.e. the 3rd body effect
direction)? Let us consider the differential gravity, what we have to consider is the difference between
the position of the satellite and the Earth center:
i. When we are in point A, the Earth center is closer to the Sun w.r.t. the satellite, so the gravity of
the Sun on the satellite is less than the gravity of the Sun on the Earth, so since the 3rd body effect
is the difference between the two gravity forces, the 2nd term is higher, so the 3rd body perturbation
is in the direction in the figure.
ii. When we are in point B, it is the opposite.
It looks like that we are stretching the orbit in the same direction and intuitively we think that the effect
is maximum because it seems that when we are in Northern hemisphere we push in a direction and in
Southern in another one. But the net effect is zero: we have to remember that in the Gauss VOP
there is a dependency from the cos u. The reason why the net effect is zero is that if we divide the orbit
in four quarters, there will be a sort of compensation. So, for example, if we consider a 1st arc there
will be an effect in a given direction, but then the cosine will change the sign and so in the other arc we
have the same effect (aw is the same) but with opposite sign and so the two effects will compensate each
other. So, in the Northern hemisphere the two effects are balanced and the same in the Southern.
(c) The maximum effect is when the difference of angles is of 45deg (LTAN 9AM 9PM).
2. The other case of interest is for geostationary orbit:
(a) In general, for geostationary, 3rd body effects are more significant. We need to consider both the
effects of Sun and Moon.
(b) It can be demonstrated that if the orbit is geosynchronous (r=42160km, n=ΩL ) and the inclination is
very small (so almost geostationary), we can obtain a type of general approximated equation which gives
us the long periodic effects on the inclination of a quasi geostationary satellite due to Sun and Moon.
(c) For the Moon, the gravitational constant is many order of magnitude smaller w.r.t. Sun but the distance
of the Earth from the Moon is smaller. So, there is a compensation between Sun and Moon, and so they
have similar effects on the geostationary satellites.
(d) For the Moon, the inclination changes in an interval from 18 to 28 deg and there is a long periodic motion
of the Moon (it does not have the same inclination w.r.t. the Earth centered inertial frame)
(e) Also Ωd changes. When the inclination is minimum or maximum, Ωd is zero.
(f) The Sun and Moon effects have the same dependency and order of magnitude but, on average, the effect
of the Moon is larger w.r.t. the one of the Sun (0.5 vs 0.25 more or less). If we combine the two effects
together, the final effect depends on sin Ω and k which is more or less 0.9deg/year [remember this number
Space Flight Dynamics Michele Santarpia 90

and dependency]. This is something useful to remember, and we do not like this effect because we want
our inclination to stay down and so we will fight against this effect with maintenance.
In general, the equation contains the right ascension of the Moon. The motion of the Moon in the Earth centric
reference frame happens on an orbit where the right ascension changes and the inclination has a long period variation
which is in the order of 10deg, so the plane of the Moon orbit changes the inclination in time, so in general when
we consider 3rd body effects we should take into account the current inclination and the current right ascension
and to have an idea of the basic dependencies and orders of magnitude we consider the extreme cases of minimum
inclination and maximum one where in these two cases the right ascension is zero and so we can write the equation
without the right ascension of the Moon:
 
di ∼ deg
= 0.478 · sin Ω id = idmin = 18deg
dt LP ,d year
 
di ∼ deg
= 0.674 · sin Ω id = idmax = 28deg
dt LP ,d year

The effect changes and it is larger when the inclination is higher. We have to remember the order of magnitude.
Finally, the dependency of the long periodic change of inclination due to Sun and Moon has a similar dependence
where we have a coefficient k which changes in time where in the order of magnitude average is 0.9, so we can write
that the effect is:
 
di ∼ deg
= 0.9 · sin Ω
dt Sun+M oon,GEO year

This equation states that, depending on the right ascension of the satellite, Sun and Moon will affect differently the
inclination. Let us imagine to have a right ascension between 0 and 180 degrees which means that the sine of the
RAAN will be positive: so, starting from an initial value, the inclination of the satellite will increase. If we have a
right ascension between 180 and 360 degrees, (the maximum effect happens at RAAN equal to 270) the inclination
will decrease. So, there is a dependency on the current position of the nodes i.e. on the current right ascension: we
will take this into account for orbit maintenance. If we change the position of the nodes, we change how Sun and
Moon affect the inclination, so by changing right ascension, indirectly we can control the inclination.
So, from a theoretical point of view, for inclination we do not have secular effects but long periodic ones but from
an engineering POV if we have a long periodic effects that take a very long time, for us it is like a secular effect
because if we cannot accept this, we will fight against and if the time is pretty long we will not really follow the
oscillation, but we will just counteract the effect. If we remember the instability of geostationary satellites for
non-sphericity, theoretically speaking it is a long periodic effect but from the orbit maintenance POV it’s like a
secular one because we will not wait for the long oscillation because we will keep the oscillation before it will reach
an amplitude that we cannot accept. So, finally, very long period long periodic effects are like secular from
the orbit maintenance POV.

5.10 4) Solar Radiation Pressure

Solar Radiation Pressure:
1. There are some similarities with Drag, the 1st one is that it is a non conservative perturbation, so the
energy will be not constant and so the semi-major axis will be affected, and unlike (a differenza di) Drag,
this is not necessarily a dissipative force because SRP can increase the energy and not only take energy
away from the orbit (trying to get energy as in the concept of solar sails, using the pressure of the Sun as a
propulsion strategy). So, in general it is non conservative but non necessarily dissipative where the dissipative
depends on the relative geometry of the orbit w.r.t. the Sun and also it depends on how the surface exposed
to the Sun changes during the orbit.
2. The other similarity with Drag is that the force depends on the satellite configuration (different satellite
will be affected by a different type of perturbation from SRP), and it comes from the surface interaction
between radiation and the surfaces of the satellite. The larger are the surfaces compared with the mass,
higher will be the effect and so again there is a ratio of As /m similar to Drag but this is not the same because
A is the area exposed to the Sun, so it is important the relative geometry of our satellite w.r.t. the Sun
direction.
Space Flight Dynamics Michele Santarpia 91

3. Since it is configuration dependent, this will add a number of complexity aspects: the different satellites
will undergo different effects where it is not only a problem of geometry, but also a problem of reflectivity of
the surfaces: the parallel with Drag can be done in terms of different Drag coefficient for different surfaces. In
fact, also in this case we can introduce a reflectivity coefficient that plays the same type of role, so different
types of surfaces will interact in different ways with the Sun and so the complexity is not only for the shape of
the satellite. In general, finally, what happens to our satellite will be the integral of a mechanical pressure
which is local at each exposed surface of the satellite and we need to have the integral of all the local pressure
that will lead to a total force and a total torque. In general, in mission analysis we do not really reach this
level of depth of prediction, and we will accept some simplified models that are still good enough to have
a good prediction of the effects and we are not integrating numerically the local pressure but we are
having a more direct integrating view that contains approximations. So, there are uncertainties due to the
complexity in terms of interaction between Sun and the surfaces and also because as in the case of
Drag the Sun itself is complex, the Sun activity is variable in time and we have a solar cycle, the amount of
radiation will change and then we will have local temporal phenomena that will change the solar pressure, and
if we have a solar storm we will have an increase of radiation at a given point which will imply an increasing
of perturbation. So, this will give us challenges which will be relevant for high are over mass (effect is small
for small ratio). Since the effect is dependent on the ratio, we can try to use this effect to characterize our
space object.
Exchange of momentum: in Drag we have an exchange of momentum between the molecules of the atmo-
sphere and the surfaces of the satellite, while in case of SRP we do not have molecules (incoming particles) but
photons (incoming radiation) that hit our surfaces but again we have an exchange of momentum. So, we use the ra-
diation for power purposes and it has a thermal impact but it has also a mechanical impact. The amount of photons
per unit of time and surface is dependant on the distance from the Sun. At 1AU, there are 8 · 1017 photons/(cm2 · s).
The energy E is equal to hf, where in average the frequency f is equal to 556nm. The Solar Flux SF (energy
per unit time and surface) on average will be 1367W/m2 . Besides power purposes, the SRP carries an incoming
momentum per unit of time and surface. The link between the incoming momentum h and the solar flux SF
is through the theory of relativity:

E = mc2
E SF N
E 2 = m2 c4 + p2 c2 → E 2 = p2 c2 → E = pc → p = → = PSR = 4.57 · 10−6 2
c c m
If we have a quiet particle means that it is without momentum p, so we can evaluate the energy associated to the
particle only because it has a mass. When we divide the solar flux by the speed of light, we obtain a momentum
per unit time and surface (it is a sort of momentum flux i.e. flusso di quantità di moto) and it is called Solar
Pressure PSR which represents the incoming momentum per unit time and surface. Any object at 1AU from the
Sun is subjected to the radiation of the Sun that will bring an incoming momentum per unit time and surface equal
to 4.57 · 10−6 m
N
2 . This momentum will interact with the satellite and there are different types of interaction

mechanisms.
Interaction mechanisms between incoming momentum and surface: let us consider a surface and the
direction of the incoming radiation i.e. of the incoming momentum P in . We have to consider the 3rd law of
momentum (principle of action and reaction): the surface will produce a change of momentum of the radiation
∆P = pout − pin , and so -∆P will be the effect on the surface (this is the general idea we have to follow).
1. Transparent: the radiation is not affected by the surface, so the outcoming momentum will be exactly equal
to the incoming one, no interaction with the surface and so there will be no mechanical pressure on this
surface. In this case, ∆P =0, the radiation is not affected and so there is no mechanical pressure on our
surface. This situation does not happen in space because our satellite is made by material surfaces which are
not transparent. It is a limit case.
2. Absorption: We can imagine that the entire radiation is absorbed by the surface, so there is no outcoming
momentum and so there is no outcoming radiation, so P out is equal to 0, so ∆P =P out − P in = −P in . So,
all the momentum of the radiation has been transferred to the surface of the satellite, and so the mechanical
pressure will be -∆P and so equal to the incoming momentum per unit time and surface P in . This is another
limit case.
3. Complete reflection: in this case all the radiation is reflected back and so pout = −pin , so the change of
momentum on the radiation is ∆p = −2pin , while the change of momentum on the surface is the opposite
Space Flight Dynamics Michele Santarpia 92

−∆p = +2pin (mechanical pressure). The effect ranges from 0 to 2, it can be null to twice.
1. The specific perturbing force acting on the satellite due to solar radiation pressure can be expressed considering
the direction to the Sun from the satellite Ŝ.
2. We need a coefficient which describes the type of phenomenon called reflectivity coefficient Cr (same role
of Drag coefficient): full absorption means Cr equal to 1 while full reflection Cr equal to 2 (it will amplify
the incoming radiation).
3. In the formula we have the direction from the satellite to the Sun. What we expect is that our perturbation
will be in the direction from Sun to the satellite and so there will be a minus sign.
As rSATs
apsat ∼
= −PSR cR
mSAT ||rSATs ||
This is the typical approximated equation we consider to describe the effect of the SRP. This expression
is intrinsically approximated: we are considering directly an entire effect and we do not have the concept of
integrating local pressures, and so from a theoretical POV it contains errors (it is not true that the perturbation
is given by this equation). There are several approximations we are considering when we apply this in the
real life:
(a) We typically use an average value of the incoming solar pressure PSR but instantaneously it changes.
(b) There is a complex interaction between the exposed surfaces of the satellite and the Sun, and we
are including all these complexity in a scalar number Cr and our capability to estimate this number is
approximated.
(c) Approximation also in the area exposed to the Sun, because it depends on the shape and attitude,
and our capability to predict the attitude may be limited.
(d) Also it is not true that the force is exactly in the direction from the Sun, and to understand why this is
approximated we need to consider that in general our radiation will not be normal to our surfaces and
since it is not normal the different types of surface interaction will not only change the norm of the final
pressure but also the direction of the force.
Let’s consider a flat plane and an incoming radiation which is not normal to the surface. We want to understand
which is the direction of the force, depending on the type of mechanism.
1. Absorption: there is no outcoming radiation which means P out equal to 0, so ∆P =P out − P in = −P in . So,
the net force on the surface -∆P is in the direction of the Sun. We are pushing exactly from the direction of
the Sun.
2. Reflection: the change of momentum of the radiation is not in the direction of the incoming radiation, and
so since the change of momentum is in another direction, the net force will be itself in another direction. In
particular, when we consider this non normal geometry, we have to think that there are two mechanisms for
reflection:
(a) Specular reflection: the incoming radiation with its momentum will be fully reflected keeping the
same angle with the normal to the surface. Checking this geometry, we can find the net force which is
in the direction of the normal to the surface (so, not in the direction of the Sun).
(b) Diffuse reflection: the reflection will be uniformly distributed around the normal to the flat surface.
The reflection happens in several directions, so the energy is spread in different directions with a given
dependence, the largest amount of energy is along the normal to the surface and then there is less energy
as we move in other directions, so finally what we can built is a sort of eggle which describes how the
energy is reflected. So, we have an incoming wave and then most of the reflection is along the normal and
then we have also some energy in other directions. When we integrate all the reflected momentum (pout
is a combination of all the contributions), it is reasonable that the change of momentum will be a vector
that has a direction which is intermediate between the incoming and the normal to the surface. We can
consider simply pout as just the vertical. What is happening in the diffuse reflection is that the overall
change of momentum will be in an intermediate condition between what is happening in the absorption
case and what in happening in the specular reflection case. This means that actually the net force will
be something which is really in between. So, the net force is not along the direction of the Sun and not
along the normal.
Space Flight Dynamics Michele Santarpia 93

The real phenomenon will be a combination of absorption, diffuse and specular reflection. So, a portion is
reflected in diffuse way, a portion in specular, and so on. We are summing different contributions to the net force.
The final net force will be in a direction general in between the normal and incoming. Based on this, when we
integrate all the contributions on each surface where each contribution is not in the direction of the Sun and in
general each contribution is different, now it is clear the reason why the formulation is approximated.
Only for the spherical satellite, whatever is the type of interaction, the final force will be in the direction of
the Sun. Let us consider the incoming radiation on the surface of the satellite. We can consider a combination of
absorption, diffuse and specular reflection. For specular, we should consider all the normals for each point, all the
specular reflection due to the symmetry gives us that when we sum the total force will be in the same direction.
The same is for diffuse reflection. So, for a spherical satellite the net force is in the direction of the Sun. The
amount of the net force will depend on the combination of absorption, diffuse and specular reflection.
Generation of eccentricity for a GEO satellite due to SRP: let us consider a geostationary satellite
(plane of whiteboard is the equatorial plane, view from North pole), let us consider the Sun direction and let us
consider that the Sun is in the orbital plane as 1st level of approximation (that’s why the out-of-plane effects are
small because it is pushing in the orbital plane). When we consider the satellite on the orbit, it is pushed by the
Sun which is like a positive ∆v. Considering an impulsive approximation and so the Sun is concentrating the
impulse in the A point, we are generating an elliptic orbit, where we are raising the other point B. When we are
in B, instead, the Sun is pushing again in the same direction but now with a negative ∆v and so we are lowering
the perigee of the orbit (esagerazione nell’immagine). So, since we have 1st a positive ∆v and then a negative
one, it is reasonable to think that the semi-major axis is the same, and what is changing is the eccentricity. So,
we obtain a ∆e > 0. So, we are giving energy and then we taking energy away, so from the semi-major axis POV,
there is a balance. So, w.r.t. Drag we have an opposite behavior because with Drag we have the circularization
of the orbit and it reduces the eccentricity while the SRP makes the orbit eccentric for a GEO satellite.
If we look at eccentricity versus time, and we put 365days we can see that it becomes a long periodic effect
with a period equal to one year. In the picture we have an instantaneous moment, where the Sun is pushing
from the right and it is making the orbit eccentric, but of course the Sun in that geometry during the year is
moving along the orbit and so it will push from all the different directions. So, there will be a balance of the
different effects. When the Sun is on the opposite direction (violet), the Sun in the other position will remove
the eccentricity that was created before. So, the eccentricity is not diverging and the orbit is not becoming really
elliptic. But of course, imagine that we need a circular orbit because the object has to be geostationary, we reach
a maximum eccentricity of 0.006 and the question is: can we accept this value? If not, we need to
spend propellant before we reach the maximum, and how we understand which is the maximum eccentricity? This
depends on the maximum acceptable variation of the GT!
What is the effect of the Sun due to the SRP on SSOs?
Let us imagine to have a satellite with a very large area over mass ratio (As /m), in a SSO and let us study two
possible cases of interest (12PM 12AM and 6AM 6PM orbits):
• Secular effect of generation of eccentricity for a SSO satellite: In case of 12PM 12AM, there is a
secular generation of eccentricity. Let us imagine to start from a near circular orbit (classic for remote
sensing). This is similar to the case of generation of eccentricity for a geostationary satellite but in this case
the effect is secular and not long periodic.

Let us consider the side-view of 12PM 12AM where the Sun is almost in the orbital plane. Let us
consider the Earth, the orbit, the equator, and the Sun direction. The geometry is not in the horizontal
plane (equatorial) but almost in a polar plane, and it is clear that there is a generation of eccentricity where
conceptually is similar to GEO. There is not a periodicity of one year as before now we have a relative
geometry which is the same. Conceptually, this effect is really secular, we have always this geometry, they
will move together and so it will always push in the same direction, it makes sense that in this case there is a
generation of eccentricity which continuously increases.
• In case of 6AM 6PM, the Sun solar pressure is mostly out of the orbit plane (since the Sun is normal to
the orbital plane). Let us consider the orbit and let us consider the Sun projection. The Sun is pushing from
the side. If we consider the side view, we have the equator, the Earth axis, the orbital plane in green, and
more or less the Sun will be normal to the orbital plane (mostly out-of-plane effects we have to expect). The
out-of-plane perturbation aw , if the area exposed to the Sun and the configuration parameters do not
Space Flight Dynamics Michele Santarpia 94

change, will be constant and, if it constant, it will generate just short periodic oscillations of the orbital
plane. Finally, there will be no effects that accumulate in time. If we want that some effects accumulate,
we have to change the attitude so that we push the satellite only in some portions of the orbit (as we can
remember, we have the sin and cos of the argument of the latitude, so we need to activate the solar pressure
only in some portions of the orbit, so we need to change the sign in order to have a net effect that is not zero).
In general, in LEO, SRP is not the largest effect.
Altitude at which the Drag becomes lower than SRP: the specific perturbation force ap due to SRP is
constant with height while the one due to Drag decreases with the height. So, at a given altitude SRP is
higher than Drag and it becomes more important. Let us try to understand on average what we are talking
about and so at which value of the altitude the SRP and Drag are becoming comparable. Let us compare the
norm and so the ratio between the two perturbing forces. To have some average ideas, let us assume that the Drag
coefficient is 2 and reflectivity is 1 (on average). In order to simplify, let us consider that the area exposed to Drag
is equal to the one exposed to the Sun. We are interested in understanding when the ratio is equal to 1. The density
and velocity are function of the height. The value of the height at which the ratio becomes 1 is about 800km. We
need to check case by case, but the idea is that below 800km Drag will tend to be larger. We know that Drag is
exponential and, as we reduce the height, Drag will increase dramatically compared to SRP. Above 800km SRP
becomes dominant.

5.11 Ground Track

Ground Track: it is the motion of a satellite with respect to a ground observer (not w.r.t. the inertial reference
frame), so the motion of the satellite in the Earth-centered Earth-fixed reference frame (motion in ECEF). At any
given time instant, we can define an instantaneous sub-satellite point, so if the satellite is at a given position
we can consider which is the Earth location that has the same latitude and longitude of the satellite. By definition,
the GT is the locus of all the sub-satellite points. If we have an orbit, we have numerically position and
velocity vectors for each time instant in ECI i.e. rECI (t) and v ECI (t) (propagation in the inertial frame). We have
to jump to the position in ECEF using a rotation matrix of type 3 which depends on αGR (t) because we know
where is the Earth at any given time w.r.t. to the stars and we have our rECI (t) from the propagation:

rECEF (t) = M 3 [αGR (t)] rECI (t)

So, we obtain the position vector in ECEF: in order to plot our GT in a Matlab script, we need to pass to latitude
and longitude and for these transformations we use analytical relations that will depend on the model of the
Earth (WGS84 ellipsoid model). Typically we use the geodetic latitude.
We need to understand how the two different motions of this phenomenon interact, i.e. the motion of the satellite
in its orbital plane and the motion of the Earth w.r.t. the line of nodes: it is useful to have some analytical
stuff because we want to synchronize these two motions in order to built repetitive orbits i.e. those orbits for
which after some time the GT will repeat itself exactly at the same way (it is a problem of relative motion). We
can study this problem by splitting it in two different motions and finally making combination of them to retrieve
the complete motion. We can try to understand independently how they contribute to the GT:
1. The 1st motion that we consider is the rotation of the Earth w.r.t. the line of nodes.
2. The 2nd one is the motion of the satellite along its orbit within its orbital plane.
So, the idea is that the motion of the line of nodes of the satellite is included at point 1 and not at point 2. So,
when we are at point 2 the orbital plane is not moving but we are just considering the motion of the satellite along
its orbit.
An important aspect is to decide how many perturbations we want to take into account: we do not use Keplerian
dynamics since it is a too simple description. What we do in this discussion is to consider only J2 secular effects.
So, in this discussion we will take into account only Ṁ = Ṁ (a, e, i), ω̇ = ω̇(a, e, i) and Ω̇ = Ω̇(a, e, i). We are
considering the motion of the line of nodes and this is the reason why we need to pay attention to the rotation of
the Earth w.r.t. the line nodes.
1. Let us start from the 1st motion: the satellite is not moving, we want to understand which is the im-
pact/effect on the GT due to the rotation of the Earth w.r.t. the line of nodes. If we only consider
the rotation of the Earth w.r.t. the line of nodes, it affects only the longitude and so we can write that
Space Flight Dynamics Michele Santarpia 95

the variation of the latitude ∆ϕ(t) is equal to 0. So, we have our line of nodes which has a precession rate
(depending on a,e and i) and the Earth rotates towards east with a certain angular velocity: if there is no
motion of the satellite, the effect is only on longitude. We want to consider the ∆λ(t) that we have due to
only the motion of the Earth w.r.t. the line of nodes (so the variation in longitude that happens in time).
The longitude is positive towards east, and if we imagine our satellite, for example, fixed at its AN and with
an initial longitude, since the Earth rotates, the effect of the Earth is with a negative term because the Earth
rotates towards east (so, for this reason the Earth is contributing to the longitude with -ΩL ), so the satellite
will drift in longitude towards west, but we have also to consider the line of nodes precession and so there is
also a contribution due to the Ω̇. The relative velocity of the Earth w.r.t. the line of nodes ΩL − Ω̇ is called
nodal velocity of the Earth and it is generating this change in longitude.

Since we can define this angular velocity of the Earth w.r.t. the line of nodes, we can have the defini-
tion of the nodal day.

Let us see all the different definitions of the day that we can have:
(a) Sidereal Day: as we consider ΩL and so the inertial velocity, this angular rate is the one of the Earth
w.r.t. the fixed stars. So, the sidereal day is the time needed by the Earth to complete a rotation with
respect to the fixed stars. Let us consider the view from the top so from the North pole and let us
imagine to have a given meridian, and we have the fixed stars, after a sidereal day we will have exactly
the same configuration w.r.t. the stars (the Earth will move but the stars are at an infinite distance, so
it is only an inertial motion).
(b) Mean Solar Day: time needed by the Earth to complete a rotation with respect to the Sun. In this
case we have ΩL − α̇s , so there is the relative angular velocity of the Earth w.r.t. the Sun. α̇s
is the rate of change of the right ascension of the Sun. The Sidereal Day and the Solar Day are
different: in fact, let us imagine to have the Earth and the Sun, let us consider the starting time of one
day t0 and in red we have two meridians from the North pole, when we wait for the sidereal day the
Earth is in the same exact configuration w.r.t the fixed stars, but to complete the rotation w.r.t. the
Sun we need some more time simply because the Earth is moving along its orbit, on average we need
four minutes more to complete the rotation w.r.t. the Sun.
(c) Nodal Day: we can now define a day which depends on the satellite, which is the time needed by the
Earth to complete a rotation w.r.t. the line of nodes of a given satellite. In this case, we have (ΩL − Ω̇)
i.e. the angular rate of the Earth w.r.t. the line of nodes. How long is the nodal day? Is it longer
or shorter than the sidereal and solar? It depends on the sign of Ω̇:
i. If Ω̇ > α̇s , the line of nodes rotates faster than the Sun, so the Earth will need more time to reach
the line of nodes, and so in this case the nodal day will be larger than the solar day, and so the
Earth will need more time to recover the position w.r.t. the ascending node.
ii. If Ω̇ = α̇s , they are the same and this is the case of SSO.
iii. If Ω̇ < α̇s , the nodal day is lower w.r.t. the solar one (lower can also mean negative rate because
we know that, if the orbit is prograde, the nodes precess with a negative velocity). The nodal day
becomes shorter than the sidereal.
We like to introduce the nodal day because we need to consider the motion of the Earth w.r.t. the nodes to
define the repetition properties of the orbit. We will work with the nodal day through the idea that it
is close to the solar day and equal only for SSO.
2. Let us see the 2nd motion i.e. the motion of the satellite in its orbital plane and remember that the
orbital plane is not moving. We are considering secular J2 effects, so the angular velocity of the satellite in
the orbital plane is Ṁ + ω̇ because we are taking into account that, in general, the perigee is moving: this al-
lows us to define the nodal period of the satellite τN which is the time needed to fly from the AN to the AN.

We can now plot our geometry in order to understand the effect on the GT using spherical trigonome-
try: let us consider e=0, prograde orbit, we take a generic point along the orbit, we take the meridian, we
are considering the geocentric latitude, this geometry is inertial (fixed) because we are not considering the
rotation of the Earth or the motion of the line of nodes, in our point we can define the argument of latitude
Space Flight Dynamics Michele Santarpia 96

u i.e. the arc of the orbit, we will have our generic latitude, and our variation of longitude: the effect on
longitude given by the motion of the satellite is represented, for a given latitude, by an arc indicated as
λ′ (so, λ′ is the variation in longitude only due to the motion of the satellite without considering the motion
of the Earth since we are splitting the problem in two). We need to understand who are the relations for the
latitude and longitude. We can use the spherical trigonometry. For latitude, we apply the law of sines. For
longitude, we use the law of cosines. Since λ′ varies from 0deg to 360deg, we need both the cosine and sine to
have a non ambiguous definition. If at time t0 = 0 we are at the AN, we can write that u = (Ṁ + ω̇)t. So, at
any given time we are able to get the GT due only to the motion of our satellite.
When we consider the combination of the first and second motion, which is the final longitude and
latitude?

The latitude is just the one derived from the law of sines. The longitude at a given time will be equal to the
longitude at the AN at the time t0 plus longitude due to the satellite motion in the orbital plane and then minus
the contribution due the relative motion rotation of the Earth w.r.t. the line of nodes:

λ(t) = λAN (t0 ) + λ′ (t) − (ΩL − Ω̇)(t − t0 )

Latitude Longitude Map

Let us consider a LEO prograde orbit: we want to build a diagram called latitude longitude map (as in GMAT).
We assume that, initially at time t0 , the satellite is at the AN and that the AN is on the Greenwich meridian. So,
at time t0 , we assume that:

t0 → u = 0 λAN = 0

Which is the shape of GT?

We need to understand which are the boundaries: the boundaries are expressed in latitude. The maximum latitude
that the satellite reach is equal to the maximum inclination (+i) while the minimum latitude is the equal to minus
inclination (-i).
If we are in LEO, the trend looks like a sinusoidal one: let us imagine that we are flying towards east, increasing
the latitude up to the maximum one, then we will start the descending phase, we will cross the equator, then we
will reach the maximum negative latitude, and then we will come back at the equator. What we should expect
will be something like in the figure. Let us draw some auxiliary points in red to draw the orbit. Of course, we
will come back in the same point because the Earth has rotated. Since we are in LEO, about 90 minutes has
passed, the Earth has rotated of a few tens of degrees in longitude (it should be 24 degrees i.e. 1/15 of 360 degrees
since we have 15 orbits in a day). This variation corresponds to a difference in longitude between two subsequent
passages at the AN. This variation in longitude is due to the rotation of the Earth w.r.t. the line of nodes
during the entire orbital period, and it is called westward shift/displacement S. This shift S is an angle (so,
it is non dimensional) expressed in radians and it is given by the product between an angular rate times a period,
where the angular rate is the angular rate of the Earth w.r.t. the node ΩL − Ω̇ and the time that has passed
is the period of the satellite. Since we are considering J2, the period is the nodal period. So, after one orbit, we
have a displacement of the GT in longitude equal to S:
2π 2π
S = (ΩL − Ω̇)τN → S = (ΩL − Ω̇) →S=
Ṁ + ω̇ Ṁ +ω̇
ΩL −Ω̇

In the final expression of S, we have the ratio between two angular velocities (Ṁ + ω̇)/(ΩL − Ω̇): at the numerator,
we have the angular rate of the satellite in the orbital plane (Ṁ + ω̇) and at the denominator we have the
angular rate of the Earth w.r.t. the line of nodes (ΩL − Ω̇): these are the two angular velocities that have
to be synchronized in order to build a repetitive orbit.

Q is a non dimensional number called repetition factor of the orbit. To have a repetitive orbit we have to
set a specific number of this factor Q:

Ṁ + ω̇
Q=
ΩL − Ω̇
Space Flight Dynamics Michele Santarpia 97

S can be expressed as:

2π
S=
Q

How do we take the three red points to sketch the ground track?
We take displacement in longitude S that we have in one orbit, so the 3rd point in red on the right is corresponding
to a drift that we have in half of the orbit (so, the distance is the half of the one we have in one orbit) and this is
the reason why the DN is not at 180deg of longitude but it is 180deg minus a westward drift equal to the half of
the drift that we have in an entire orbit.
We may convert the displacement in angular terms S in a linear distance Slinear : if we want the displacement at
the equator, we can multiply by the equatorial radius RLE .

Slinear = (ΩL − Ω̇)τN RLE

Of course, once we return at the equator, we start the 2nd orbit and the GT is exactly the same but displaced in
longitude. We can compute the longitude displacement/drift at any latitude. For example, we may be interested at
the drift at 45deg of latitude: in longitude terms (i.e. in angle terms) we have the same value while in linear
terms we have to take into account the radius of the parallel which is different from the radius of the equator.
For this reason, we have to consider the cosine of latitude: so, as the latitude increases the distance between the
ground tracks will be shorter.
We have drawn a LEO GT. It can be understood for two main reasons:
1. We have plot the westward drift in one orbit as few tens of degrees. An orbit which is not LEO, like a GPS
satellite (12h), will take 12h to come back at the equator and so when it comes back the drift will be half of
the equator.
2. If we increase the altitude, the shape will change and it becomes less and less sinusoidal.
Considerations about Map scale: when we have this latitude longitude map, the scale of the map is variable
and so we need to pay attention to which are the consequences that we take into account from this map. In fact, we
have a rectangular map because if we consider two points at the equator they have a given difference of longitude
which corresponds to a linear distance. Of course, if we take two points with the same difference in longitude but
a different latitude, the scale is completely different and the linear distance could be very short. When we consider
the limit case at the poles, the lines at ± 90deg are not true lines but points (because they are the poles). So, since
the scale of the map is variable, we have not to consider this map as standard map with uniform scale (error!) since
it leads to wrong conclusions.
Let us consider a retrograde LEO orbit, we can plot the longitude latitude map, we plot the equator in the
center, the Greenwich meridian in the vertical. The boundaries are given by: the maximum inclination of 180◦ -i
and the minimum one of -(180◦ -i). We can plot the GT starting from the AN at zero longitude. In the inertial
motion, we are moving towards west. The westward drift between consecutive ground tracks S (ground
track distance between consecutive orbits) is always the same and it does not depend from the fact that the orbit is
retrograde or prograde. In linear terms, at the equator, we have that Slinear ≈ 2700km for a LEO orbit. We take
the entire equator and we divide it in more or less 15 parts since we fly more or less 15 orbits per day.
Why if we change the altitude, we have the deformation of the GT?
So, let us consider different types of prograde orbits, with different radius but with same inclination. Let us consider
to be at the AN of the orbit. We want to understand which is the local inclination of the GT: this angle does not
depend only on the inclination of the satellite but also on the orbital period.
1. LEO orbit
2. MEO orbit
3. GEO orbit
Let us draw the inertial velocity of the satellite which has an inclination i, but the local inclination of the GT
is not equal to i. In fact, it is a problem of relative motion since we need to understand the motion of the satellite
Space Flight Dynamics Michele Santarpia 98

w.r.t. the ground observer at the equator which is moving with the Earth, so we need to consider a relative velocity:

v GT = V − v L

v GT = V LEO − v L

v GT = V M EO − v L

v GT = V GEO − v L


So, the local inclination is not equal to i, but it is pretty close to it.
 r r r
µL ∼ µL µL
vLEO = = 7km/s vM EO = vGEO =


 r
LEO r
M EO r GEO



v
LLEO = ΩL rLEO cos ϕ vLM EO = ΩL rM EO cos ϕ vLGEO = ΩL rGEO cos ϕ

where we have to consider rLEO , rM EO or rGEO i.e. the radius of the satellite orbit (because for the relative
motion we need to consider what happens to the altitude of the satellite, because the ECEF point that corresponds
to the satellite position moves in a way proportional to the distance [1h05min, part 1, immagine supplementare])
and we have at a generic latitude cos ϕ which of course at the equator is equal 1.
1. For LEO satellite, vL is relatively small, so the deviation of the ground velocity from the inertial velocity is
relatively small.
2. In case of MEO satellite, considering the same inclination, the inertial velocity is smaller while the velocity
of the Earth will be larger since the radius is larger. When we sum the vectors, the velocity of GT will become
almost vertical, in fact the GT for navigation satellites is really like this one (we really have vertical tangent
at the equator).
3. For GEO satellites, the velocity of GT will be more inclined. This explains why for GEO satellites when we
are at the AN, even if the orbit is prograde (i.e. in the inertial motion the satellite is moving eastward), in the
relative motion the GT is moving towards west, that’s why we are starting that ’eight’ pattern (it is like an 8).

In this geometry, after one day everything repeats because we are again at the node and at the same spot,
but we pass again over the same spot also after half of the orbit (for this concept, let us consider also the
other sketch where we are at the AN at a certain meridian, then we can understand that the satellite to pass
from the AN to DN we need 12h and in 12h also the meridian has rotated and it is in the same position of
the DN i.e. it is also in the opposite point and so we pass over the same spot after 12 and 24h).
With this type of explanation we can also understand what happens to non circular orbit. When the orbit is not
circular we will have more degrees of freedom: the 1st one is that the inertial velocity will change, and its link with
the equator will depend on where we place the perigee, if the perigee is at the AN the inertial velocity is high for
example.
Analytical explanation: we can try to understand this type of dependence also analytically. The dependence of
longitude on time as we have seen is:

λ(t) = λAN (t0 ) − (ΩL − Ω̇)(t − t0 ) + λ′ (t)

tan λ′ (t) = tan u(t) cos i

To understand which is the inclination of the GT, we need to understand the derivatives of longitude and latitude.
The variation of latitude contains the inclination and does not depend on the rotation of the Earth. What is
interesting is the variation of the longitude:

λ̇ = λ̇′ − (ΩL − Ω̇)

ϕ̇ = f (i)

This term λ̇′ depends on the satellite, that can be computed by applying the time derivative to the equation of the
Space Flight Dynamics Michele Santarpia 99

tan.

u = (Ṁ + ω̇)(t − t0 )
λ̇′ (Ṁ + ω̇) cos i
=
cos2 λ′ cos2 u
2 ′
cos λ
λ̇′ = (Ṁ + ω̇) cos i
cos2 u
cos u
cos λ′ =
cos ϕ
(Ṁ + ω̇) cos i
λ̇′ =
cos2 ϕ
At equator:

λ̇′ = (Ṁ + ω̇) cos i

Finally at the equator the instantaneous derivative of inclination:

λ = (Ṁ + ω̇) cos i − (ΩL − Ω̇)

r
µ
n=
a3

This of course depends on the rotation of the Earth but it contains the inertial angular motion of the satellite
(Ṁ + ω̇), so it is clear that as the orbit increases in height this number becomes smaller and smaller, so the 1st
term becomes smaller and smaller while the 2nd is constant so at a given time becomes negative. So, even if the
inclination is positive, instead of moving towards east we move towards west. This was the analytical explanation
of the discussion with vectors seen before.

5.12 Repetitive Orbits

Repetitive orbit: it means that after a given amount of time the ground track will repeat exactly in the same way
(repetitive ground track). For this repetitive orbit, we can introduce the repetition period which is the amount
of time after which we have the same ground track.
In this course, we talk about only about the repetition of the GT. In general, from a mission perspective point of
view, we refer to revisit time which is related to the time needed to observe again a given spot of the Earth. The
concepts of repetition period of the GT and the revisit time are different because the revisit time involves the
payload on-board the satellite (sensor). In fact, to revisit a given area we do not necessarily need to fly again
with the same GT because the sensor has a given FOV and it has a certain pointability. Instead, if we consider the
repetition of the GT, it is only related to the orbital motion of our satellite.
In case of repetitive orbits, our objective is to synchronize two motions, i.e. the rotation of the Earth w.r.t. the
line of nodes and the motion of the satellite within its orbital plane. Since we are considering J2 effects, Q
is a function of a,e and i. The dependency on the semi-major axis is related to both keplerian and J2, instead the
dependency from the eccentricity and inclination is only related to J2. For this reason, these three dependencies
are not of the same importance: the one from the semi-major axis is stronger than the ones from eccentricity and
inclination.
Repetitive orbits definition: an orbit is repetitive if and only if Q is a rationale number. Q can be computed for
any orbit: in general, it is real but if we have an orbit for which Q can be written as a rationale number, the orbit
will be repetitive. So, rationale means that Q can be written as the fraction of two integer numbers prime to
each other.
If Q is a rationale number, why the orbit is repetitive?
This concept is linked to the angular velocities of the two motions. The repetition factor can be seen equivalently
as the ratio between the nodal day and the nodal period. The repetition factor is a number that tells us
how many orbits the satellite flies per day. So, it is between 14 and 16 for LEO and as the altitude increases
Space Flight Dynamics Michele Santarpia 100

Q decreases (2 for a MEO satellite, 1 for Geosynchronous). Let us consider that the satellite at an initial time t0
is at the AN and we are on the Greenwich meridian.
(
u=0
λAN = 0

The repetition factor Q can be written as the ratio between R and N. This means that:

DN N = RτN

This equation tells us that in the time in which our satellite flies exactly R orbits, the Earth rotates of
exactly N days. Imagine that we are at the initial time and we wait a time equal to DN N (or equivalently RτN ),
after this time the satellite has completed exactly R orbits and it is again at the AN, the Earth has completed
exactly N rotations, so the AN is again corresponding to the Greenwich meridian. If this is true, after this time,
everything will repeat exactly as it was at zero time. So, this is the reason why if we can find a rationale expression
for Q, the GT will be repetitive and the repetition time will be equal to DN N = RτN . So, there is a perfect
synchronization: the satellite completes R orbits and the Earth has completed a given number of days DN , so
everything will repeat.
If we want to have a repetitive orbit (i.e. repetitive GT), we have to set our orbit in order to have these two numbers
as integer and prime each other. Set our orbit means we need to play with a,e and i.
We can choice the repetition time as we like but there will be a cost to pay (let us remember that we cannot change
the 15 orbits per day), if we improve in some ways our repetition time, as we reduce the repetition time we will pay
in other terms.
Repetition Factor Q=15 and Q=15.5: let us imagine to select an orbit in order to have Q=15 (this is rationale
and it is integer) which means that R=15 and N=1. Let us consider the ground track of this orbit and we look
only at the passages at the equator and only the ascending passages at the equator. We draw the equator as a line
which corresponds to a given longitude. Let us consider the 1st passage (orbit 1) which starts at the extreme of the
line. The 2nd passage has a westward drift:

Slinear = SRLE

We selected our orbit so that we fly exactly 15 orbits per day, so this means that in the time in which we complete
exactly 15 orbits the Earth completes exactly 1 rotation. When we complete the 15th orbit, the situation starts
exactly in the same way of the initial instant. The B point is coincident to A. We start with the 16th orbit which is
actually the same passage of the 1st orbit. The repetition time here is 1 nodal day or equivalently 15 nodal periods.
We have obtained an orbit called one-day repeater. After exactly one nodal day, we fly again over the same
spot. The interesting point of this orbit is that it takes only one day to pass over the same point/spot. The cost
we are paying is that the distance between two consecutive passages is pretty large that will be a problem
in terms of resolution for remote sensing (2700km). For this reason, we are in general more interested in having a
more denser pattern of the GT. For example, if we are interested in the position in light blue, what we can do?
(since we are flying too far from that position). We can design another orbit and we can accept that the repetition
time is not 1 day. For example, we can select Q=15.5=31/2. We fly 15.5 orbits per day. Let us draw the pattern
now. Since Q is larger, S will be lower but it will be of the same order and whatever we do it will be always of
this order because we fly in 90min. We do 15.5 orbits in 1 day, so we complete the orbit 15 and when we start the
orbit 16 we are not in the same point A, what happens is that in one day we fly exactly 15.5 orbits (so, 15 orbits
plus the half of the orbit 16). When we are at the half of the 16th orbit, the Earth has completed one
rotation but we are not at the ascending node, to come back at the AN we need another half of the
orbit and so the orbit 17 will have a displacement of S/2. Now, the passages of the 2nd day are not the
same of the 1st day and they are exactly in the half during the 1st day passages. Remember that we will always
drift of S between consecutive orbits. So, we complete exactly 31 orbits in exactly 2 days and so we will come back
in B=A where it starts the orbit 32 which is equal to 1. Now, in the 3rd day, the GT will repeat. So, we have
created a change in the pattern of the GTs so that we have a denser coverage but we will pay the cost that
we are taking 2 days to fly again over the same spot.
The distance between two consecutive tracks is S which is always the same. From the coverage POV, we like
the fact that there is another distance that is the half of the previous one (it is the half for the two-day repeater)
Space Flight Dynamics Michele Santarpia 101

called s i.e. the distance between adjacent tracks where these two tracks are in two different days, but at the
end of the repetition period, if we consider which is the minimum distance, it is equal to s.
S
s=
2

Repetition Factor Q=15.67: let us change the repetition factor just a little bit by changing the altitude of the
orbit where now Q=15.6=47/3. In the time in which the Earth completes three rotations, the satellite flies 47
orbits. CSK has Q=237/16. We fly 15.6 orbits per day, so 15 orbits + 2/3 of one orbit. Now, the idea is that when
the 1st day ends, we are at 2/3 of our orbit (the number 16). If the distance is S, we can take a distance which is
2/3 of S and so we will need another 1/3 of S and so after another 3rd of the orbit to come back at the AN. Every
passage of the 2nd day will be displaced of 1/3 of S w.r.t. the passages of the 1st day. When we are in orbit 47 we
are exactly at a distance equal to S w.r.t. the point A=B and in fact this is an orbit where we fly exactly 47 orbits
in exactly 3 days, so after 47 orbits everything will repeat and so we have that 48=1. The nice idea is that since
we are using a different repetition period / time, we have a clear compromise between the density of the pattern
(coverage) and the how much time we need to fly again over the same spot (basic trade-off).
The distance between adjacent tracks will be s=S/3. So, in general, s=S/N where N is the numbers of the
nodal days (repetition period) that can be also written as:

S 2π 2π 2π
s= = = R =
N QN N N R

In this case of interest, R=47. If we have more orbits, we will sample the equator in smaller steps / segments and
the number of segments is equal to the number of orbits. So, if we take 300 orbits to repeat, we will divide the
equator in 300 parts.
As we have a different latitude, the longitude discussion will be the same, the linear distance is reduced due to the
cosine of latitude at a different parallel.
Before talking about Drag, let us consider again the nominal discussion without Drag and let us say something on
the design of the orbit and how we work about the compromise.
As we know:
DN R
Q= =
τN N
The distance between consecutive tracks:
2π
S=
Q
The distance between adjacent tracks:
S 2π
s= =
N R
where for s, as we can see, we have degrees of freedom where we can make our selection.
In order to have a repetitive orbit, Q must be a rationale number, this means that we have a discrete number of
possibilities to have a repetitive orbit.
On x-axis, we have the orbital period or altitude. The orbits for which this diagram is applicable are SSO
circular ones (if the orbit is not SS, the result will contain some little approximations in the link between the
period and the semi-major axis or the altitude). If e is small, the dependence is negligible. Let us remember that
if we have e=0, the Sun-synchronicity creates a one-to-one link between semi-major axis and inclination,
so for each inclination there is an semi-major axis that makes the orbit SS. (there is a plot which links inclination
and semi-major axis):

Ω̇ = Ω̇(a, e, i) = α̇s
Space Flight Dynamics Michele Santarpia 102

When we consider a discussion of dynamics secular J2, the orbital period does not depend only on a but it depends
also on inclination and eccentricity.
2π
τN = Ṁ = Ṁ (a, e, i) ω̇ = ω̇(a, e, i)
Ṁ + ω̇

where here Ṁ = Ṁ (a, e, i) the dependence on a is the strongest (Keplerian dependence), the other two are much
smaller (less sensitivity, they are just variations related to J2).
So, mostly, the nodal period will depend on a but also a little bit on the other quantities. So, τN = τN (a, e, i).
While in Keplerian we pass from τ to a while from secular J2 this is not possible because there is an impact related
to e and i. But we can still say that there is a one-to-one link of the nodal period with the semi-major
axis if we add that the inclination is the one of SSO and the eccentricity is zero. This is the reason why on
the plot we can related the altitude and the nodal period.
On the y-axis (not continuous axis, are integer numbers) we have the repetition time, this orbit makes sense only
for repetitive orbits. Each point on the diagram is a repetitive orbit. Each point represents a value of Q that
will correspond to a given altitude and N. If for example we start with N=1, we have only few points which will
correspond to a given repetition factor, where the repetition factor is the axis on the top. As we know Q depends
on a,e and i but the orbit is circular and SSO so it depends only on a. For example, N=1, Q=15, R=15, h=560km.
What we have are all the possible repetitive orbits that we can use up to 20 days of repeat cycle, we have a discrete
number of possibilities because we have a discrete number of rationale numbers up to 20 at the denominator.
Increasing the N, increases the number of possibilities.
As we can see, we can change the repetition properties even if we have a small change of altitude: since the
repetition properties come out a synchronization between two velocities or periods, for example by changing a little
bit the period of the orbit the synchronization changes completely. For example, for N=17, we have some points
which are pretty close to 15.5 (31/2) where in those points the pattern of the GT is completely different because
it repeats after 17 days and not 2 days, but the orbits are pretty similar in terms of altitude. So, the value of Q
is pretty close but the repetition properties are completely different. So, with fine adjustments of the altitude we
can obtain the repetition properties that we want. This is something we like from a design perspective. Imagine
that we have as requirement h=500km for the design, of course if a payload works well at 500km it will work well
also if we change a little bit the altitude (like 505). So, we will move in a range around 500 and we can decide the
repetition properties having different solutions. So, we can tailor the ground track to the pattern that we want and
we will still around 500km.
SS repetitive Frozen Orbit Design Flow: design of an orbit means that we need to choice the orbital parameters.
Let us assume that we want a SS repetitive frozen orbit for remote sensing.
1. Generally, the starting orbital parameter is the altitude h (and so the semi-major axis a) which comes
from payload needs (remote sensing, optical payload, altitude linked to resolution).
2. Let us consider to start from a circular orbit i.e. eccentricity equal to 0 (e=0).
3. Since we have eccentricity equal to 0 and we want to have a SSO, there is a one to one link between the
semi-major axis and the inclination. For this reason, we have also the inclination i.
4. At this stage, we consider the GT and, in particular, the repetition properties: so, we consider that we have
initially a value of Q called Qold = (Ṁ + ω̇)/(ΩL − Ω̇) (this value can be evaluated since Q=Q(a,e,i) and we
have selected before a,e and i). Let us imagine, for example, that Q=15.6. We look at the diagram Q-R-N,
and we see that we do not have a point corresponding to 15.6. For this reason, we have to adjust R and
N in order to have a new value of Qnew which is on the diagram and closer to Qold (change a little bit the
orbit to obtain the repetition properties we want). So, we have a new value of semi-major axis anew and we
can recompute the inclination inew . So, we start from a SSO at a given altitude, and now we have a SSO
repetitive at another altitude closer to the previous one.
5. Typically, we like to have also a frozen orbit where to obtain this condition we need that the eccentricity
must be a specific value e=o(10−3 ), and we will set the perigee typically at 90deg or 270deg. So, the orbit is
not anymore circular and, for this reason, we are affecting the Sun-synchronicity but, finally, from a practical
POV, Ω̇ is not changing and the orbit is still SS.
Space Flight Dynamics Michele Santarpia 103

6. To complete the orbit design, we have to decide the RAAN (typically, what we really impose typically is the
LTAN). The value of Ω depends on the launch date: in fact, to get a specific SSO, we need to have a certain
geometry of the orbital plane w.r.t. the Sun. The position of the Sun, described by αs and δs , depends on
the launch date: if we want to have a Ω − αs = 0, we will impose to have the RAAN equal to αs .
So, we have designed our nominal orbit in the inertial space with five orbital parameters. When we design an orbit
there is one parameter which is not designed and, if we consider the flow until this point, we are not interested in
the anomaly of the satellite or in the timing. It could be important if we have some timing requirements.
Let us imagine that we select just for simplicity Q=15 and let us consider the pattern on the equator (or a generic
parallel). If we consider two satellites that have two different true anomalies, the GT pattern/structure does not
change (the distance between tracks is always the same) but the GT itself is different. If we consider SAT A and
SAT B on the same orbit and spaced in anomaly, the black GT is of SAT A, the light blue GT is of SAT B, there
is a spacing/offset between the passages, and so we are sampling the parallel in different points because we are
at the AN in different time instants (the starting point is different) since the satellites cross the equator in
different times.
So, if we are interested in a GT not only in terms of repetition properties but also in terms of longitude at which
we have to pass, we need to pay attention to timing: when we launch it is similar to a rendezvous problem because
our scope is not only to fly over the green orbit but we want to be also in the right position and so we have to pay
attention to the anomaly (phasing considerations).
This fact that we can have different ground tracks with some separation can be exploited when we have a constel-
lation of satellites over the same orbit because, for example, if we use more satellites, finally with this configuration
we can have a denser pattern in one day because we are using more satellites. If we have a very large value of R, it
means that we are really giving more emphasis to full coverage compared with revisit time (more denser pattern).
Space Flight Dynamics Michele Santarpia 104

6 Orbit Maintenance
6.1 Orbit Maintenance for LEO Satellites
What we have seen for a LEO satellite is pure theory: all the stuff that we have described happens in the ideal
world where we design the orbit and nothing happens to the orbit but we have perturbations (not J2 because
we have considered J2) that breaks the repetitive pattern. We are in LEO so we have Drag which affects the
Ground Track.
Let us consider our pattern on the equator considering the ideal world in which we have only J2 in black. The
basic problem is that Drag changes the GT, in fact we decay (losing of energy) and so semi-major axis decreases
(escalating effect). If we consider a short time scale, the decay of semi-major axis can be approximated as a linear
decay, then we have the Drag paradox so the mean motion and the angular velocity go up (Keplerian perspective
but as we know the main effect is Keplerian, it is the same if we consider J2 where Ṁ + ω̇ also goes up). The period
of the satellite (nodal period if we consider also J2) reduces and so since we are faster, compared with the ideal
orbit we design, we reach the equator in advance and so instead of taking 90min for example we take 89min. If we
are arriving before, the Earth has no enough time to complete the nominal rotation, so because of this phenomenon,
when we arrive at the equator the GT will not be the nominal one and, for this reason, we will have a displacement
(in red), and this displacement is an eastward drift w.r.t. the nominal passage. This effect will accumulate
in time and so the red pattern will have an error towards east that will increase. At a given moment, since we will
have a tolerance, we violate this tolerance and so as we violate we need to maneuver. When we maneuver we will
compensate for the effect of Drag because we need to raise again the orbit, but this is a problem where the phasing
is important. So, if our tolerance is on the GT, the objective will be to control the GT and as a consequence we
will control the semi-major axis.
When we are on our nominal orbit, if the semi-major axis is not our nominal one but has a small change, the
GT changes completely and so what is a degree of freedom at design level becomes a challenging aspect in the
operations because if we want to keep some repetition properties we need to pay attention to the semi-major axis,
and so we need to have a fine control of the semi-major axis and the main driver for orbital maintenance is in the
revisit properties of the orbit.
There are different types of strategies. We want to counteract Drag for orbital decay. If we do not do anything we
decay and the mission ends, the semi-major axis has an escalating effects. Let us see the three different strategies
that depend on the different types of requirements:
1. Requirement on ∆a: the simple case is when we do not really pay attention to timing or to the GT itself
but our driver is ∆a and we have a tolerance on ∆a. We start from a nominal value of a, we decay linearly
considering small time scales, we have the tolerance and once we arrive at the boundary, we control raising the
orbit (Hohmann maneuver or something like that), we go back not to the original a, but to the other extreme,
in order to increase the time between maneuvers (time between subsequent maneuvers). Each maneuver will
require a ∆vman . We estimate the orbital decay da/dt, which is used to evaluate the time between maneuvers,
we evaluate ∆vman for each maneuver, we have the lifetime, and so we can understand which is the numbers
of maneuvers τlif e /τman = Nman , and so by ∆vtot = Nman ∆vman .
2. Requirements concerning GT: our tolerance is on GT and so we need to pay attention to the real GT
vs the nominal one. We have our nominal GT and so the nominal passages at the equator, and we consider
the passage nN OM and we have a tolerance in km or in longitude ∆λtolerance , so we want to stay within this
interval. So, of course, there will be a certain ∆a, but the phasing will become important. So, we need to
keep our distance from the nominal satellite within some intervals (there is a sort of virtual target idea).
3. Orbital Tube: sometimes the control may become very precise and this case is the one of orbital ’tube’. As
for 2, we have a tolerance on the motion w.r.t. the Earth, but now instead of asking 20km of tolerance, we
can say that we have to stay on the nominal orbit w.r.t the Earth with an error of few hundreds meters.
The tolerance is a tube now around the nominal orbit which is pretty small (nominal orbit in ECEF!! not
inertial! in general, the concept of tolerance is in ECEF, we have a given desired motion w.r.t. the ground and
w.r.t. the nominal trajectory we have a tube). For example, the maximum tolerance of 500m can come from
payload requirements like radar or optical (since for radars differential interferometry we use subsequent
observations which need very closer observations geometry).
Space Flight Dynamics Michele Santarpia 105

6.1.1 Drag Make-Up Maneuver

It is a maneuver for compensation of the effect of Drag. Let us draw the nominal passages in black, and due to
decay the ground track will not be the nominal and so it will change. The real passages are in red and they have
an eastward displacement / shift w.r.t. the nominal. At a given moment, we will violate some tolerance on the
ground track, and so we will need to maneuver. Because of Drag, da/dt < 0, as a consequence dn/dt > 0 due to
the Drag paradox, so the satellite becomes faster both in linear and angular terms, so dτN /dt < 0, we pass at the
equator sooner than the nominal, and this is the reason why we have the eastward shift. The real dynamics is the
one depicted in red (decay and drift towards east). So, there will be also a difference in anomaly ∆ν which is the
source of error, and we have a ∆t ∼ = ∆ν/n and so we arrive some seconds in advance and the error in longitude
is ∆λ = (ΩL − Ω̇)∆t (error in GT), we arrive sooner and the Earth has not rotated of the same amount. This
phenomenon is not constant because the angular velocity increases and so the angular separation will increase more
and more. Which are the functional dependencies? Of course:
da CD A √
= −ρ µa
dt m
In limited time scale, we can assume a linear decay of the semi-major axis (da/dt is constant). The mean motion:
r
µ
n=
a3
r
3 µ 3n
dn = − 3
da = − da
2 a 2a

da negative, dn is positive since we become faster.

r
dn 3 µ 3 n da
=− da = − ≈ constant
dt 2 a3 2 a dt

So, also n versus t is non linear, but again we can assume in short time scales that dn/dt is constant and so n varies
linearly going up.
∆ν is the drift in anomaly (angular separation between nominal and real satellite) and it is linked to the error in
the GT; to have ∆ν, we have to consider the integral in time of the effect of the differential angular velocity.
So, we have that:
Z t
∆ν(t) = dndτ
0

So, the idea is that we have two objects with a different angular speed, so the angular separation will be the integral
of the differential speed. So, the key message is that since the difference in angular rate is increasing linearly, the
anomaly separation between nominal and real will increase quadratically (it is linear only if the difference in
angular rate is constant in time).
We have a maximum tolerance, so we have a maximum tolerance in anomaly ∆ν.
We can follow the phenomenon using four diagrams:
1. Up Left: we plot the semi-major axis as function of time where we plot the nominal semi-major axis
and the boundaries.
2. Up Right: we plot ∆S as function of time that gives us the error we have in the GT. The real passage n
in red and the nominal in black n, we call ∆S the distance in the ground track crossings between the
previous nominal n-1 and the current real. So, in absence of errors so without Drag ∆S will be equal
to S nominal. Because of Drag, we have a reduction of ∆S and we can put the boundaries.
3. Bottom Left: we have a view of the geometry w.r.t. the nominal.
4. Bottom Right: we plot the current ground track passage and the boundaries that we have due to the
maximum error in longitude that we can accept.
What happens in time?
Space Flight Dynamics Michele Santarpia 106

We are decaying, so we can update the diagrams. ∆S is decreasing in a quadratic way because it contains ∆ν. At
a given moment we reach the maximum tolerance, let us call the point with A (maximum allowed eastward error).
At this moment, we have to maneuver, let us see where this point is in each diagram, which will correspond to the
minimum semi-major axis, to the boundary for ∆S.
Which maneuver we have to apply?
We need to correct semi-major axis and the phasing: this looks like a rendezvous from different orbits. We need to
bring the point A to the nominal, so we need to drift backward in anomaly and we will back again to the nominal
and so once we correct the phasing the GT is the one at the starting point. This is typically what we do not to.
Generally, we want to simplify the maneuvers and maximize the time between the maneuvers. Instead of
going back to the original point, we bring the orbit to a larger a (i.e. to point B). We raise the semi-major axis.
What we do in terms of anomaly (important for phasing)?
We do not correct the phasing with the maneuver! The idea is that we jump to an higher orbit keeping the
same phasing w.r.t. the nominal (same separation in anomaly). In the diagram right bottom, we have the relative
phasing between the satellites, since we are not changing the relative phasing, when we maneuver we are still there
so A=B. Also in terms of S it is exactly the same.
Why do we not correct the phasing?
Because after we maneuver, we exploit the natural relative dynamics between the real and the nominal.
The real is on a higher orbit w.r.t. the nominal, so it will be slower in terms of mean motion and so from the B
point it starts to go back, starts to drift back, so the error in GT starts to be correcting itself, so if we look
at the bottom right the ground track starts correcting, the error going from left to right, in terms of ∆S we have
change the sign of the drift so we start increasing quadratically in the other sense. So, it is drifting back but this is
not a phenomenon where we just drift back because the real satellite is subjected to Drag, so it will start to decay
from B point. As it decays, on one hand it approaches the nominal satellite but on the other hand its mean motion
becomes closer and closer to the one of the nominal satellite.
Why we have a shape like this?
Since we are drifting back but then we are decaying, so the differential velocity w.r.t. the nominal will go to zero,
and when we are on the same orbit they will have the same mean motion and this is way the tangent is
vertical because the relative speed is zero.
So, as we arrive in C we will continue to drift back because we are still higher, so this means that instantaneously we
will pass through S nominal, while in terms of semi-major axis we are decaying. So, since we are drifting backwards,
we are generating a westward error, so the GT will continue to drift from C to left (error with opposite sign). The
risk now is that we violate the error on the other side but this does not happens because this movement of the GT
will be slower and slower because the angular velocity of the real satellite becomes closer and closer to the nominal.
We arrive at the point D where we are instantaneously at the same value of semi-major axis (so, there is
no relative motion in the ground track, so the GT is not moving for an instant, the next instant due to Drag
to the real, the real starts decaying and so higher speed). From the POV of the GT this will be the point where we
have the maximum westward error. So, this is the maximum of ∆S, maximum error in the other direction. After
this instant, instantaneously we are at the same rate of the nominal satellite but we decay and so we start to drift
exactly in the opposite way, so we become faster, and we go below and forward, and there will be a qualitative
trend as in the figure. So, after the vertical tangent, from the GT POV, we are drifting with the GT in the opposite
direction, we will pass again with the same anomaly and at a given moment we will reach the point E=A where
we are again at our minimum semi-major axis and maximum eastward shift, and from the POV of ∆S we have
reached again the point E at the boundary. This is again the time to maneuver.
Finally, once we have ∆λ, with this strategy we can connect it to ∆a. The general maintenance strategy will
be the one in which we have boundaries on the semi-major axis but that are driven by a tolerance on
the GT.
How to pass / fly from A to B without changing the phasing?
Conceptually, this is the typically maneuver where we use a bi-elliptical transfer because we change
radius and then we have some timing constraint (anomaly constraint). In the ideal case in which the
situation is perfectly symmetrical, we can see that the situation is simpler: in order to understand this, we need to
Space Flight Dynamics Michele Santarpia 107

make another plot. Let us plot the nominal and the other two orbits of the boundaries. We can depict the nominal
satellite, the real satellite on the decayed situation and where we want to put the satellite i.e. on the higher orbit.
How do we increase of a quantity equal to 2∆a by keeping the same ∆ν w.r.t. the nominal? 1st
idea: using of Hohmann transfer, we have the initial and final orbit, so nothing can be decided, so we apply the 1st
impulse jumping on the final orbit and then second impulse.
When we arrive on the final orbit, the nominal satellite where is?
So, now, for us is a sort of rendezvous problem because what is important it is the angular separation w.r.t.
the nominal. The time of the maneuver τhoh is half of the period of the ellipse:
s
a3hoh
τnomhoh = π
µ
r1 + r2 anom − ∆a + anom + ∆a
ahoh = = = anom
2 2
So, since ahoh = anom , the time in which the real satellite flies the Hohmann ellipse, the nominal satellite has flown
exactly half of its orbit, so this is a case in which, because of symmetry, even if in general with Hohmann we cannot
have timing as degree of freedom, in this ideal case of perfect circular orbits and perfect symmetry, using
Hohmann we are able to adjust the semi-major axis keeping the same phasing. But in the real life it is
not exactly the Hohmann transfer but we can imagine that is pretty close to Hohmann. So, finally, we have reached
the B point where we have to remember that the B point is the one w.r.t. the nominal! In fact, finally, we are w.r.t.
the nominal at higher semi-major axis of ∆a and forward of an angle equal to ∆ν.
Considerations about Budgets: when we make this analysis, the final objective is to understand which is the
time between maneuvers, the total ∆v, the time of the maneuver (because when we maneuver we cannot use
the payload). Imagine that from the analysis we find that we have to maneuver every two days, and we will spend
a given amount of ∆v. When we fly in the real life or when we launch our satellite, this will become a prediction.
Since Drag is affected by uncertainties, also our maneuvering history will be affected by uncertainties,
so when we do these calculations in advance this gives us some budgets but in the real life the frequency of the
maneuver will not be constant and if the Sun is more active than usual we will maneuver more frequently (when
we consider other perturbations like non sphericity or 3rd body, it is really deterministic because our capability to
predict the effects is pretty accurate so we will really maneuver with that timing but when we consider the Drag we
will have some uncertainties, so on average hopefully the budget will be good but what really happens is affected
by uncertainties).
Correction of Inclination: since we are in LEO, with a certain frequency (every few days), we will maneuver in
order to correct the in-plane effect of Drags (Drag make-up maneuver) but then finally with a lower frequency
(few maneuvers per year) we need to apply also a correction of inclination: this is because since we have a small
∆i due to 3rd body perturbation and SRP, this will influence ∆Ω̇J2 and this is a problem because as we
know the precession of the line of nodes is connected to SSO.
Regarding to this, when we perform the Drag make-up maneuver we are correcting the GT, so from the POV of GT
if we have some loss of Sun-synchronicity, with this maneuver we are also correcting the loss of Sun-synchronicity.
When we consider the distance S between consecutive GTs, S = (ΩL − Ω̇)τN , when we think about our GT
correction, what we are doing is to adjust continuously S because the GT moves, we have an error and then we
correct since we have a given tolerance and so we will always be around the nominal S which is the one without
Drag. So, we are correcting S through our in-plane maneuver, if there is some error in inclination and so some
small loss of Sun-synchronicity, what is affected is the quantity Ω̇! So, this means that the error in inclination
is also affecting S. So, Drag impacts in a dramatic way the nodal period but the error in inclination
instead affects Ω̇. Well, in our strategy we are directly working in correcting S to correct the GT (it is like a global
correction for GT), if we have a slight error in inclination, even if we do not correct the inclination, our GT control
strategy will compensate for this effect of the inclination error. So, we are both correcting the two errors i.e. due to
Drag and the loss of Sun-synchronicity. So, from the POV of the GT control we may avoid to correct the
inclination (through our in plane maneuvers we will compensate for some out of plane effects, we are interested in
correct the GT). Anyway, it is still common that we really correct the inclination with low periodicity
(few maneuvers per year). Let us plot the inclination versus time, we can sketch the nominal inclination, depending
on our orbit we will have some trends of the inclination, and in short time we have a linear trend, we remove all
the oscillations and we are interested in only secular or long periodic effects. Typically, we will define a tolerance
Space Flight Dynamics Michele Santarpia 108

and as we reach the tolerance, we will then correct, typically here there is always the idea of minimizing the
numbers of maneuvers, and so instead to come back to the nominal, we will bring the inclination directly to the
upper limit. In green, we have the maneuver, in red the natural evolution. If we never correct for the inclination,
the GT will be okay but the LTAN has an error that increases because we are not compensating but we are just
adjusting otherwise the GT.

6.2 Orbit Maintenance for GEO Satellites

Different points to talk about:
1. Errors for GEO Orbits (∆a, ∆e, ∆i).
2. Sources of Errors: which are the perturbations?
3. Maintenance: how do we correct?

6.2.1 Errors for GEO Orbits in terms of orbital parameters and effects on longitude and latitude
1) Errors for GEO Orbits: for geostationary satellite, the tolerance is on the latitude/longitude box that we
can accept. A geostationary satellite which has 0 latitude and it has a specific nominal longitude. We define a
box for the tolerance. If we are not in the nominal longitude is not a problem as long as we are able to keep the
requirement that comes from the mission and in general we have also some tolerance w.r.t. other geostationary
satellites (we do not have to approach too much the other geostationary satellites). Since our tolerance is on latitude
and longitude, we have to understand which is the link between errors in orbital parameters and errors in latitude
and longitude (so, how errors in orbital parameters affect the ground track). For a geostationary satellite we have
a specific semi-major axis (period equal to sidereal day), eccentricity is zero and also inclination zero.
1. Error in Semi-Major Axis ∆a: what happens if we have a ∆a? We have a circular equatorial orbit with
wrong semi-major axis, so different period, so the orbit is not anymore geosynchronous and so not geosta-
tionary. Imagine to have the Earth and the geostationary orbit, and as we know the geostationary satellite
rotates with the same rotation rate of the Earth. With a certain ∆a, we have a longer period since we are
higher, we will cover a small angle, so while the nominal flies an angle, the real one will fly a smaller one and
we generate an error in ∆ν which is an error in longitude (because the nominal will always keep the same
longitude since it is our reference geostationary satellite), so this means that ∆ν = ∆λ.

How do we compute this ∆ν?

Let us work with Keplerian mechanics because we are trying to understand at 1st order of approximation the
effects of the error. We differentiate to obtain dn. In particular for the geostationary satellite:
3 nGEO
∆n ∼
=− ∆a
2 aGEO

Since the ∆ν = ∆λ, this means that ∆n = ∆λ̇ (in longitude we will drift forward or backward, and under
these approximations this is a linear phenomenon in time, we have an error that increases linearly). Finally,
we have to remember that:

∆a → ∆λ̇ Linear drift in longitude

For the GEO case we have that:

∆λ̇ ∼
= −0.013◦ /day · ∆a

where ∆a is in km. If we have 1 degree of drift in longitude, it is not small because the GEO belt is dense,
and so we can have problems of proximity conditions with other satellites. So, 1 deg of drift in longitude is
not negligible!
2. Error in Eccentricity ∆e: we have a ∆e, so the eccentricity is different from 0 (for example, due to SRP
that was active for some time and we have not compensated).

When the perturbations are small (errors are small), we can consider in an independent way the
Space Flight Dynamics Michele Santarpia 109

errors and then we have a superimposition of the effects which is reasonable to understand.

What happens to GT?

Let us consider the Earth, the nominal orbit, and we consider also the eccentric orbit and we assume that
the line of axes is vertical. Let us consider the nominal (=geostationary satellite) and the real which are
together at the initial condition (the real one is at the perigee). After some time, the nominal will be
in a certain position given by an angle equal to nGEO t = ΩL t (for the geostationary satellite, if we have a
given point on the Earth, the sub satellite point on the Earth will be at the same position i.e. same longitude).

What is instead the motion of the real satellite?

Let us consider the energy equation. Remember that nominal and real have the same semi-major axis,
because the error is only in eccentricity. So, the two orbits have the same period and same energy.
When the real is at its perigee the distance from the Earth is different, so ε is the same, the radius is smaller,
so the velocity is higher. The real will be faster than the nominal. So, the eccentricity for this reason will
generate an error in GT and as before the ∆ν is a ∆λ, because if we are moving faster, we will be drifting
eastward (we are forward with the real). At a given moment, at the intersection point, they have the
same velocity and there will be the maximum separation in longitude. After B, the real will be slower
than the nominal and so in the plot of λ we start to drift back. At the point C, they will be again together,
because they have the same period. In D, there will have again the maximum angular separation because
instantaneously in D they will have the same speed and after D the real will be faster. So, finally, if we have
an error in eccentricity, there will be an oscillation in longitude at the orbital frequency.

Can we accept this oscillation in ∆λ?

Of course, it depends on the value of eccentricity.

Let us understand this dynamics in analytically term. ∆ν is equal to ∆λ because the plane is perfectly
equatorial (no plane error), we can write that ∆λ = ∆ν = ν(t) − nGEO t where for the real satellite nGEO = n,
so nGEO t = M (t) i.e. equal to the mean anomaly of the real satellite at time t.

∆λ = ∆ν = ν(t) − nGEO t = ν(t) − M (t)

So, ∆λ = ν −M . We have to introduce the eccentricity in the formulation. We consider a linear approximation
(that makes sense if e is small) for ν. From the final formulation (general, not only for GEO), we can
understand which is the error in longitude. So, as we can see, the longitude has an oscillation at orbital
frequency (sin M ), with amplitude equal to 2e. If we plot the error in longitude due to the eccentricity
as a function of M (or time), we have an oscillation, where one interval is 2π and the amplitude is 2e. So, now,
we can understand that starting from a requirement in terms of ∆λ, we will define which eccentricity can be
accepted because if we are within the tolerance of the longitude, the eccentricity is acceptable otherwise at a
given moment we may need to correct for the eccentricity.
3. Error in Inclination ∆i: we have an orbit, it is no more equatorial. Let us take a generic point on the inclined
orbit, the meridian which passes through the point, due to the error in i, we have an error in latitude, from the
construction we did when we have seen the GT, in general we know that sin ϕ = sin i sin u. sin u = sin(nGEO t)
since it is a perfect circular orbit without any error in a. We have a sinusoidal oscillation of the latitude
with period equal to the one of the satellite and amplitude equal to the inclination that we
have not corrected. Of course, if we have an error only in i, there is not any error in the satellite
angular velocity but there will be an error in longitude because this correct angular velocity is not in the
correct plane. The change of longitude is due to the projection on the equator of the angular motion of the
satellite. We want to extract the error in longitude as final scope. Finally, we find that, when there is
an inclination, the error in longitude depends on the square of inclination, this means that it is
much smaller than the variation of latitude because the inclination is a small angle. So, finally, we have a
variation in latitude at orbital frequency and an oscillation in longitude at 2 times the orbital
frequency with amplitude which is much smaller w.r.t. the latitude one. What is happening is
the mathematical representation of something which is like an ’8’. This is because if we look at the latitude
longitude plane, we are saying that in latitude we have an oscillation between +i and -i while in longitude
Space Flight Dynamics Michele Santarpia 110

between i2 /4 and -i2 /4. We can sketch these points. Finally, so, the variation in longitude is smaller than the
one in latitude.
When we combine these three errors what happens?
1. Errors in Semi-Major Axis and Inclination i.e. ∆a + ∆i: we can adopt a superimposition of the effects.
Because of ∆a, the GT will drift while ∆i creates an ’8’, so we obtain a ’walking 8’ which will move in
longitude in one direction on in the other one depending on the error that we have in semi-major axis.
2. Errors in Semi-Major Axis and Eccentricity i.e. ∆a + ∆e: due to semi-major axis error, there is a
secular motion i.e. an increasing of the longitude error over time, instead due to eccentricity error, there will
be an oscillation of the longitude: so, finally, there will be an oscillation (due to eccentricity) that will
drift in a given direction (due to semi-major axis) i.e. the center of the oscillation will drift in
one direction or in the other.
3. Errors in Eccentricity and Inclination i.e. ∆e + ∆i: due to the eccentricity error, we have an oscillation
in longitude (pure longitude), while due to the inclination error we have that the GT oscillate with an ’8’
shape, so we have to combine an ’8’ with an oscillation in longitude, so it will be an inclined ’8’. We will
obtain an inclined ’8’ only if the perigee is at 0deg or 180deg. In fact, when we combine the ’8’ with the
eccentricity, we have to take in mind a key concept: the ’8’ is phased w.r.t. the nodes while the oscillation
of ∆λ due to eccentricity is phased w.r.t. the perigee. By deciding the perigee, we are deciding which is the
phase difference between these two motions. The ’8’ is always relative to the AN/DN because we are the
center of the ’8’ is the AN/DN, the oscillation due to eccentricity is always centered in the perigee. So, in
general, the two motions are phased w.r.t. different points. When the perigee is at 0deg, it means that the
perigee is on the AN. When the perigee is at 0deg or 180deg, the two motions are phased w.r.t. the same
point because in these two cases the two points are coincident. So, we obtain an inclined ’8’. The two motions
are 0deg in the same moment because at the same time we are at the center of the ’8’ and we are also at the
center of the oscillation due to the eccentricity. If we set the perigee somewhere else, we will have some other
results.
If you combine all three effects, there will be a drifting inclined eight pattern.

6.2.2 Sources of Errors (Perturbations)

2) Sources of Errors: In general, some sources of error depend on perturbations, some other depend on our
limited control performance (due to the fact that we do not have the possibility to nullify all the errors).
1. SRP: it depends strongly by the ratio area over mass, but in general it will generate a long periodic
oscillation of the eccentricity that produces an oscillation in longitude.
2. 3rd body (Sun and Moon): we will have two main effects.
(a) A variation of inclination di/dt, so the orbit will not be perfectly equatorial. Because of i, we have
an oscillation in latitude plus longitude (’8’ shape).
(b) A variation of the ascending node dΩ/dt, so the nodes are impacted. Regarding the RAAN, if we
have an orbit that is almost equatorial, a change of RAAN corresponds to a change of angle which is
almost in the orbital plane, so a drift in RAAN will generate a drift in longitude.
3. Non sphericity: we have principally J2 (oblateness) and J22 (tri-axiliality). Due to J2, we have again a
variation in the RAAN dΩ/dt and so again a drift in longitude. J22, instead, causes an instability of
the geosatellite, this is very long period and very large oscillation in amplitude, and this can be considered
from an engineering POV as a ”drift” in longitude.
4. Control Errors: means both in orbit injection and then during the operations, and they may imply errors
in eccentricity and inclination because there will be always a residual error, but we can also have an
error in semi-major axis. We will never have the semi-major axis equal to the correct one that prevents
from any drift. If we have a ∆a, will generate a drift in longitude.

6.2.3 Maintenance Maneuvers

3) Maintenance Maneuvers: our basic strategies of maintenance are connected to the GT, the two main objec-
tives are the control the latitude and the longitude; the maneuvers to control the longitude are called East-West
Space Flight Dynamics Michele Santarpia 111

station keeping and the ones to control the latitude are called North-South station keeping.
Among the East-West station keeping maneuvers, we find:
1. Phasing maneuvers (∆λ ↔ ∆ν).
2. Maneuvers to adjust the semi-major axis and/or the eccentricity (we can correct only the semi-major
axis using an Hohmann-like maneuver, or we can correct both semi-major axis and eccentricity by using a
2-∆vs Hohmann-like maneuver).
3. Maneuvers to correct semi-major axis, eccentricity plus phasing this means that we want to change
the orbit and we have also timing constraints. The bi-elliptic can be a strategy that gives us enough degrees
of freedom to correct the orbit.
To maintain the latitude, we know that only the inclination is our problem, so we need to remove this undesired
inclination. The inclination is introduced by third body effects due to Sun and Moon, because even if we have a
perfect orbit, they disturb our geostationary satellite. The maintenance of latitude is done through North-South
station keeping maneuvers. We want to keep the satellite in some bands of latitude and this is linked with the
inclination, we cannot let to increase too much the inclination. Let us imagine that we have a given tolerance on
the inclination σi which corresponds to the maximum latitude that we can have ϕmax . Let us imagine to have
initially not zero inclination and RAAN equal to 270deg:

i = σi Ω = 270◦

Let us plot in ECI frame. When we propagate a satellite like this, the RAAN will change but in a time scale
which is longer w.r.t. the one in which the small inclination will change. So, we can consider that the
RAAN is constant and so for this reason we will have that di/dt will be equal to -k. So, the inclination will decrease
linearly. In the time in which this small inclination σi will go to zero, since the time scale is so short, we can assume
that Ω is constant. So, in this condition, naturally the error in inclination is corrected. The orbit in this way tends
to become equatorial. The instant after the orbit became equatorial, we will have that the motion will continue
and, since the inclination cannot become negative, the result will be that we have an inversion (natural) of the
nodes (so, what was the AN becomes immediately the DN and vice versa). So, what changes instantaneously is
not the inclination but the RAAN, instead of 270deg the RAAN now becomes 90deg, so now in the equation we
have that di/dt is equal to +k. So, we will have the same linear trend but in the opposite direction. After the
same amount of time, we reach the maximum value. As we go up, we do not let this phenomenon evolve because
otherwise the inclination will increase more and more becoming not acceptable and we have to maneuver in the A
point, there are two possible ideas:
1. Restore the zero inclination: let us look the situation at the AN. An instant after the restoring of the
inclination to zero, it will re-start to grow again.
2. Inversion of node: another strategy consists in inverting the nodes, giving a ∆v normal to the equator,
where the new v will be symmetric w.r.t. the equator. We are converting again the AN in DN. We bring
instantaneously the RAAN from 90deg to 270deg, having again di/dt=-k. The point B will be again the point
in which we have to maneuver.
Which is the best strategy?
Strategy no.1 requires a smaller ∆v, but the strategy no.2 will require less maneuvers because the time between
maneuvers is the double w.r.t. the strategy no.1.
Which is the cheapest one?
Under linear approximations, they have the same cost. Typically, we like more strategy no.2 due to the less
maneuvers and also the time between maneuvers is two times the one of the first strategy. The only maneuver for
strategy no.2 is the one in red in the Ω vs time plot.
We can make some computations.
Computations of ∆vs :
1. 1st strategy:
 
σi ∼
∆v1st = 2v sin = vσi
2
Space Flight Dynamics Michele Santarpia 112

2. 2nd strategy:

∆v2nd = 2v sin σi ∼
= 2vσi

Time between maneuvers: the rate of variation of inclination is k, so the time between maneuvers for the
1st strategy:
σi
τm =
k
For the 2nd strategy will be:
σi
τm = 2
k
Once we have a given lifetime for the satellite lT (for example, equal to 10 years), the number of maneuvers and
the total ∆vtot for the 1st strategy will be equal to:
lT lT k lT k
Nman = = ∆vtot1st = ∆v1st Nman = vσi = vlT k
τm σi σi
2nd strategy:
lT lT k lT k
Nman = = ∆vtot2nd = ∆v2nd Nman = 2vσi = vlT k
τm 2σi 2σi
The total ∆v is the same. Under these linear approximations, the total ∆v does not depend on the tolerance σi .
Larger is the tolerance, larger will be the time between maneuvers, and larger will be the cost of each maneuver.
What does happen when we finish the propellant? What happens if we do not do anything, so if we
reach the B point and we do not maneuver?
So, the AN is at 90deg and the inclination will increase due to 3rd body effects of Sun and Moon. Let us understand
what happens in the long term. In the time scale useful for maintenance, we have said that the Ω is almost constant.
When we reach the threshold for the inclination, it continues to grow if we do nothing. We do not have the possibility
to maintain the satellite (we have no propellant) or we are analyzing a debris in geostationary orbit (like a solar
panel who detouched from an existing satellite, there is no control). The inclination will increase linearly, and so the
orbit goes towards 90deg but in the long time this is not what happens because there are non-sphericity effects
due to J2 (not negligible) that gives Ω̇.
If we have an orbit with an increasing inclination, what is the effect of J2?
If the orbit is prograde, and so Ω̇ < 0, so there is a retrograde precession of the line of nodes, as we consider
a longer time scale we cannot neglect this effect. Finally, the effect of J2 and the effects of 3rd bodies will
go together. Since Ω is no more 90deg, it starts to go down and as it goes down the growth of inclination will
not be linear since we have the sine of the RAAN (the rate of growth of the inclination is reducing with time). As
time passes, the RAAN goes to zero and so di/dt=0 (maximum), after this instant the RAAN will continue to go
down and since the sine of RAAN is negative, the inclination starts to decrease and so we have a sort of symmetric
behavior also for the right ascension. So, we will have a combined behavior (there is the coupling of the two effects).
Actually, there is also an effect of inclination on RAAN due to the cos i, that’s why the trend of RAAN is not
exactly linear. The maximum of inclination is of about 15deg. Finally, if we leave a debris on a GEO orbit, it
is not true that it will go to high inclination orbits, but it will reach a maximum inclination of 15deg and then the
trend will be inverted. We have a periodic behavior.
If we combine these types of behavior, if we consider another plane i-Ω, if the orbit is modified, we have a point that
moves in this plane. As we arrive at -90deg, in a short time we will then move again to +90deg. So, the evolution of
our orbit is something like in the figure. The evolution has a periodic behavior with a period of 55 years.
Depending on the time in which the evolution starts, we will find our space object in different positions. Most of
the debris in GEO are not so old. So, to arrive at the maximum of 15deg, it will take 27 years! The average range
of the debris in GEO is in the order of 15/20 years, so they are in the depicted portion of the curve (so, there is a
large density in a relatively narrow interval of RAAN and i).
An interesting practical consequence: let us consider an orbit in ECI with a certain Ω tilde with i of 10deg. So,
this explains why if we have a telescope and we are pointing towards the GEO belt and we want to maximize the
Space Flight Dynamics Michele Santarpia 113

number of objects that cross the FOV of the telescope, we have to point in these two directions (pointing directions
towards AN and DN), so there exist two couples of (α, δ) where we maximize the number of objects. These points
in which we have higher density are called pinch points. This is good for surveillance because we survey in these
regions (remember that for surveillance is important to have a given periodicity of the new measurements to have
a good orbit determination accuracy otherwise the uncertainty will increase).
Space Flight Dynamics Michele Santarpia 114

7 Relative Motion
Relative Motion: motion of a spacecraft with respect to another one, or the motion of a space object
with respect to another one. When we consider the relative motion problems, we are still in the perspective to have
a primary body and these two objects are orbiting around the primary. Relative motion does not mean that
we are considering the gravity interaction between these two objects: if we consider the space station
and an approaching spacecraft, we can make our analysis neglecting the cross effect due to gravity.
How to solve the relative motion problem?
When we consider that this gravity interaction is negligible, the problem can be faced in the following and well-known
way: we consider the two spacecraft on their orbits (red and blue), we are interested in their relative motion which
is represented by the position vector from 1 to 2 that we can call rrel , and our problem consists in understanding
what happens to rrel in time i.e. we are interested to understand rrel (t) and the relative velocity ṙrel (t) at any
given time. Then, we have to use two times GMAT, i.e. we have to propagate numerically the solution, because we
write twice a perturbed two-body problem for the two spacecraft. We can follow numerically (through a
satellite propagator) or analytically, and then we can find the solutions of the two problems, computing the relative
motion in terms of rrel (t) as the difference between the two position vectors and the same for ṙrel (t). This type of
solution gives us the relative position and the relative velocity in the ECI inertial frame. Of course, we can use the
rotation matrices to rotate the relative position. For the relative velocity, since we introduce a rotating frame, there
are some other terms that we need to pay attention. A reference frame that we like is the orbiting reference frame.
With this type of approach, we do not get general conclusions because we are considering a very special case
perspective. In order to retrieve some general tools, let us start by considering the mental experiment reported
on Colella.

7.1 Hill’s Equations

How to get general conclusions?
We need some tools that are of intuitive or of immediate usage i.e. some relative motions tools in order to have
a perspective so that we directly get the motion of one object w.r.t. the other one. Our reference frame
of interest is centered in one of the two objects. We want to understand, w.r.t. this other object, which is the
dynamics. These relative motion tools can be used for different types of application scenarios:
1. Rendezvous and Docking (similar to the problem of the passage of the ball between the two guys because
we need to understand how to fly towards another object in a specific direction, because we do not want to
hit that object but we want to rendezvous and dock in a given direction).
2. Debris Removal (which is a sort of rendezvous and docking with something which does not collaborate due
to the tumbling motion).
3. In-Orbit Servicing (IOS) (which is a scenario similar to rendezvous and docking or to debris removal,
depending on our assumptions that we make because it depends if the client collaborates or not) and In-Orbit
Assembly (IOA, it is new perspective of IOS).
4. Formation Flyings (in which two or more satellites that fly together, we do not want to approach up to the
final rendezvous but we need to avoid that the satellites fly away from each other).
All these scenarios require the understanding of the relative motion but the time scale is different because,
for example, for the rendezvous and docking we need something which is effective for a relative short time.
Typically, rendezvous and docking operations are short (if it is too long, there are many challenges that come out).
For formation flyings, the two spacecraft have to fly together for years and so the problem will be to maintain
the formation (maintenance of the relative geometry), because as we establish a nominal relative geometry, this
will not be kept, if we do not anything the satellites will fly away from each other.
To introduce these relative motion tools, first of all, we need to define a proper reference frame. We consider to
have two objects even if generally a formation flying can be characterized also by more than two objects. From a
mathematical modeling POV, we can now consider that we have only two objects, and what we learn can be scale
to the case in which we have more than two (n-objects). Let us start from the inertial reference frame, we plot only
the orbit of the reference (i.e. the target). We have two resident space objects (to talk in general, not always
spacecraft, but when we talk about of maneuver it is of course a spacecraft), where one is the reference (where it
Space Flight Dynamics Michele Santarpia 115

means that we study the motion of the other w.r.t. this 1st object) and then we have the ’other’. These two objects
can have different names:
1. Typically, in rendezvous and docking operations or in IOS or in debris removal, the reference is
called target (it is passive, it does not maneuver) and the ’other’ is called chaser (the one that maneuver).
2. In formation flyings terminology, we typically use other terminology like leader and follower or chief
and deputy (we are interested in a number of deputies flying around the chief).
In black we have the orbit of the target, and we consider the instantaneous position of the target and the one of
the chaser. We are interested in rrel which is by definition the difference between the position vector of the chaser
and the one of the target. We are interested in this quantity in a reference frame centered in the target CoM. We
introduce the Hill’s reference frame: we use the reference frame RSW as done for perturbations, but centered
in the target where R̂ = xH is the instantaneous radial, Ŵ = zH is normal to the instantaneous orbital plane of
the target, and Ŝ = yH is the tangential one. This reference is familiar because a similar one is used in attitude
determination and control for satellites, and it is called orbiting reference frame (so, the Hill’s one is just a different
convention of the ORF).
 r
 R̂ = T
||r T ||







r × vT

Ŵ = T


 ||r T × v T ||





Ŝ = Ŵ × R̂


We can compute the matrix from ECI to HRF:

 rT 
||r T ||
 
 
 (rT ×vT )×rT 
M ECI→HRF =  ||(rT ×vT )×rT || 
 
 
r T ×v T
||r T ×v T ||

Instantaneously, as we have position and velocity of the target, we can always pass from ECI to HRF. So, finally,
what we want to do is to study the motion of chaser w.r.t. the target in the Hill’s reference frame of the
target. Our solution will involve in a very limited sense quantities which are relevant to absolute astrodynamics
but only to relative astrodynamics. When the target is on a circular orbit, the 2nd axis is the one of velocity; if it
is not circular, it is close to the direction of velocity (only perigee and apogee will be coincident).
Let us see the derivation of the Hill’s equations, which give us a straightforward solution to our relative motion
problems. The derivation is important to understand how we apply the assumptions but it is very important that
we do not forget the assumptions because we need to have a sensitivity about which is the range of validity of what
we do and which is the error and when it is acceptable or not. We are going to derive a set of equations called Hill’s
equations (or Clohessy-Wiltshire equations). These equations are very powerful because they are simple, and
they are based on these three assumptions.
Assumptions:
1. Keplerian Motion.
2. Close Proximity i.e. (rrel /rT ) << 1.
3. Target on Circular Orbit.
1st approximation: Keplerian motion i.e. absence of orbital perturbations. Due to this assumption, we
might be accurate in some timescale and then we might become much less accurate as time passes due to the effects
of the perturbations.
We are going to use the other assumptions during the derivation.
Space Flight Dynamics Michele Santarpia 116

We want to study rrel in the Hill’s reference frame and so we are interested in a vector that will have three
components and we want the components in the HRF:
 
rrel HRF = xR̂ + y Ŝ + z Ŵ

We are also interested in the velocity and acceleration in this HRF.

 
ṙrel HRF = ẋR̂ + ẏ Ŝ + ż Ŵ
 
r̈rel HRF = ẍR̂ + ÿ Ŝ + z̈ Ŵ

The unit vectors are constant because we are considering the derivatives in the Hill’s Reference Frame and so they
are constant w.r.t. this reference frame, so we have just the derivatives of the components. The scope of the game
is that we need to write some differential equations where we have x, ẋ, ẍ, y, ẏ, ÿ, etc. As we write differential
equations with these components, we can try to solve these equations. Once we have solved them, we have the
relative motion directly in the HRF.
In order to reach this objective, we need to start from what we know i.e. from the absolute motion problem (Kepler’s
problem) and so we write the equations of motion for the chaser and target. For the chaser, we will place the thrust
since we want to keep the possibility to maneuver through the chaser:
µ F thrust
r̈c + r = athrust athrust =
rc3 c mc
The target is instead a passive object. The complete system is:
µ

r̈c + rc3 rc = athrust



r̈ + µ
T 3
rT
rT = 0

We take the difference between the two equations since we are interested in the relative motion.
µ µ
r̈rel = r̈C − r̈T = − r − r + athrust
rc3 c rT3 T

On Colella there is an error since it is not F thrust but it is athrust .

Let us consider those comments for the derivation:
• The target is passive, we know how it moves, so we want in the equations just the position, the motion of the
target and the relative motion, so we want that rc disappears.
• 2nd approximation: close proximity assumption i.e. rrel /rT << 1. We want to apply this approxima-
tion because, in the usual relative motion problems in which we are interested in, the distance between the
target and the chaser is much smaller than the distance of the target from the Earth center and the distance of
the chaser from the Earth center. This approximation can be applied at the denominator (as done for the 3rd
bodies). If we are in LEO orbit, and we are at 10km from a target, we have that rrel /rT = 10/6000 = O(10−3 ),
so the number is very small. The error due to this assumption will depend on the target orbit, since it depends
on the ratio.
• We put r−3 −3 3
T outside. Error on Colella: it is rT when we put outside and not rT .
r T ·r rel r T ·r rel
• The term 3 rµ3 rT2 rrel is similar to the term 3 rµ3 rT2 rT because in the 1st one we are multiplying the
T T
µ r T ·r rel
term 3 3
rT rT2 for rrel while in the 2nd case for rT . Since rrel is smaller then rT and since we are summing
r T ·r rel
the terms, we can consider that the term 3 rµ3 rT2 rrel is negligible w.r.t. the other one.
T

• rT is not a relative quantity but it is an absolute quantity and we are not writing this vector w.r.t. HRF but
it is the position vector w.r.t. the Earth center using R̂.
• The final equation that we obtain is simple but it has a problem: we are interested in the motion w.r.t. HRF
but the acceleration that we have at the 1st member is the inertial relative acceleration because we have
written the difference between the two inertial accelerations, so r̈rel does not have as components ẍ, ÿ, z̈. We
Space Flight Dynamics Michele Santarpia 117

need to make another conversion, and this is pure kinematic: we need to convert the acceleration in a given
frame to the acceleration in another frame (’equazione dei moti relativi’, Poisson law). So, we can write that:
 
ṙrel = ṙrel HRF + ΩH × rrel

where ΩH is the angular velocity vector of the Hill’s rotating frame.

• We need to understand who is the angular velocity vector of the Hill’s rotating frame. This angular velocity
vector exists because the target moves along the orbit and so the HRF rotates, so it is the angular velocity
vector of RSW and it depends on the orbital motion, we are assuming no perturbations (we are Keplerian),
so we have a fixed orbital plane, we just change our true anomaly, so the only reason why the HRF changes
orientation is due to the change of true anomaly. The change of true anomaly happens with a rate which is
ν̇, so for this reason ΩH = ν̇ Ŵ . Of course, Ŵ does not change because we are in Keplerian assumptions. Ω˙H
is the angular acceleration.
• Finally, we get the equation in [r̈rel ]HRF = ẍR̂ + ÿ Ŝ + z̈ Ŵ .
• Up to now, we have done two assumptions, so the target is on a general orbit which means that the orbit
can be circular, can have eccentricity etc. and this is why we have the angular acceleration because ν̇(t) for a
general target is not constant (we are faster at the perigee in angular terms etc.). Of course, what is constant
is rT2 ν̇T = hT . We can apply the last assumption and let us restrict our attention to the circular orbits. 3rd
assumption: target on a circular orbit. We assume that:

athrust = ax R̂ + ay Ŝ + az Ŵ

• We obtain the Hill’s equations (HCW equations) by projecting along the RSW reference frame
obtaining three scalar differential equations. These are of second order and linear with constant
coefficients differential equations. All the equations are linear because there are no products between x,y,z,ẋ,
etc. and the coefficients are constant because we have numbers and then we have the mean motion of the
target n which is constant due to the circular orbit. Since we have linear constant-coefficient differential
equations we can solve them analytically.
• The 3rd equation looks nicely since it is an harmonic oscillator with no damping. The z motion is
independent/decoupled from x and y (z motion is the cross-track direction, i.e. angular momentum motion,
i.e out-of-plane of the target motion, i.e. lateral), so it is independent from the in-plane motion. So, if we
are playing outside of the station, and we want that the ball comes back to us, we can throw it laterally,
and whatever we do it will always come back to us because the solution of the 3rd equation will be the
one in the sketch (sinusoidal). So, if we throw on the right our ball giving an initial velocity and after
45minutes (considering a LEO orbit) we can catch again the ball. If we are launching the CubeSat,
never launching it laterally otherwise it will come back exactly with the same speed that we have used to
launch and from the same direction. So, not launch it laterally!
So, in summary, in the relative motion problem, we are interested in the relative position of the chaser w.r.t. the
target evaluated in the HRF. In the three scalar differential Hill’s equations, the only absolute astrodynamics entity
is the mean motion n of the target. If we consider zero thrust, we obtain just a natural evolution.
In this case, the motion along z is represented by the equation of a simple harmonic oscillator without forcing
function: it will be an oscillation at orbital frequency where the period of the oscillation is 2π/n (equal
to one orbital period). In general, the phase of the oscillation will depend on the initial conditions which
are represented in terms of the initial position cross-track z0 and the initial cross-track velocity ż0 . If we
consider some lateral initial conditions like a lateral initial velocity (like for the satellite deployment), whatever
we do it will be always an harmonic oscillator, so there will be no divergence which means that z comes back
to 0 after some time, so the satellite will come back to the target.
Let us consider directly all the three analytical solutions in case of zero forcing functions:

ax = ay = az = 0

So, we have only some initial conditions as source of the relative motion (we do not anything, it is only
natural orbital dynamics). In this case of zero forcing functions, the equations admit analytical solutions and we
Space Flight Dynamics Michele Santarpia 118

need to give initial conditions in terms of the initial relative position and of the initial relative velocity:


 x0 ( (

ẋ
0 x(t) z0
→ → z(t)


 y0 y(t) ż0
ẏ0


The 1st four initial conditions will determine what happen in terms of x(t) and y(t) (in-plane problem), while
the last two conditions will impact z(t) (simple harmonic oscillator motion). We can solve the linear differential
equations applying the Laplace transform obtaining an algebraic problem and we can solve algebraically
in the Laplace domain computing X(s), Y(s) and Z(s) and then we perform the inverse transform coming back in
the time domain. After this (in Vallado is shown), we can write the solutions (remember that they are with zero
forcing functions / external forces). So, under our assumptions, if we leave the chaser at an initial relative position
and with an initial relative velocity, it will evolve in time in a way that is explained by these equations. We need
to understand which is the type of trajectory that we will get.
1. Motion along z i.e. cross-track motion: let us start from the simplest one i.e. the z(t) which is a simple
harmonic oscillator with an oscillation at orbital frequency. The mean of this solution is zero (for z(t) motion,
we have zero average) and it cannot diverge in time (bounded motion). Whatever we do in terms of initial
conditions, we will have a sinusoidal function. If we change the initial conditions, for the same amplitude we
can have for example the red curve or the light-blue one. In the functions we have sketched, it is changing only
the phase, but in general changing the initial conditions, it may change also the amplitude of the oscillations.

z(t) = Az cos(nt + ϕz )

The amplitude of the cross-track oscillation is:

s  2
ż0
Az = z02 +
n

The phase of the oscillation is:

 ż0 
ϕz = arctan n
z0

From the POV of the ball game, if we are at the center of the space station and we are launching laterally,
what we are doing is to give an initial cross-track velocity ż0 , and we are generating exactly the sine function.
Of course, there are other situations in which, for example, there is a lateral separation instantaneously, and
we leave the object just with a lateral initial relative position, this object will start to move towards the space
station, as we leave the object, it will start moving, after one quarter of the orbit, it will hit the target. The
key message is: the z component is independent, so if we play with the z component we will not activate the
other components, we will remain on the z component, whatever we do we cannot diverge and we will pass
through the target.

Just a little comment about perturbations: when we introduce perturbations, every will depend on the po-
sition along the orbit in which we give this cross-track motion. When we introduce Drag instead there will be
an impact of the ballistic coefficient of our object. When we try to inject perturbations in this discussion, what
is important is not the perturbation itself but the differential perturbation between the chaser and the target.
If the chaser and the target are subjected to the same perturbation, we will not see any effect.
2. Motion along x i.e. radial motion: x(t) is the radial (i.e. vertical) component. The amplitude contains
the in-plane initial relative velocities and the initial radial position. From this, we can see how the
motion along x and the motion along y are coupled. The 1st and 2nd terms are oscillations at orbital
frequency while the 3rd one is the radial offset (constant quantity) that depends on the initial conditions.
We can nullify the offset if we select the initial conditions in a proper way. There is no term that diverges
(bounded motion again as for z). So, under the Hill’s assumptions, we will never fly away radially from the
target.
Space Flight Dynamics Michele Santarpia 119

3. Motion along y i.e. along-track motion: let us analyze y(t), which has four terms. The 1st and 2nd
terms are oscillations with orbital frequency of the target and they depend on initial conditions, the
3rd term is a (linear) drift term (it is a term not bounded and we have some numbers which multiply time),
the 4th term is the along-track offset. If the initial conditions are such that the 3rd term is not nullified,
the relative motion will not be bounded and so it will diverge. As it diverges, we will reach a distance where
the Hill’s equations are not valid anymore (remember that the Hill’s equations are true when we are in close
proximity). When we take conclusions from these results, we need to pay attention to the moment when the
applicability of the Hill’s equations itself becomes tricky. We can plot it where in black we have the average
trend and around it we have an oscillation. This trend will be the combination of all the elements. If we check
only the oscillatory part, the period will always be 2π/n, but now we have the motion will diverge and there
are not boundaries, so at a given moment we will fly away from the target.
The relative motion is the composition of the three components.
What is the in-plane relative trajectory i.e. the motion of the chaser w.r.t. the target in the target
orbital plane xy?
The explanation that we have done before does not really describe what happens to the motion of the chaser along
x and y, it is not clear what is the real trajectory in the xy plane (target orbital plane), so we like to understand
which is the trajectory of the chaser within the target orbital plane (the so-called in-plane relative trajectory).
To understand this, let us give a look at the equations x(t) and y(t).
• Let us give a look only at the oscillatory terms. If we check them, we can see that the amplitude in x(t) for
the sine term, i.e. ẋ0 /n, is the same (but multiplied by two) that we find in y(t) for the cosine term. The
same is for the term which multiplies instead the cosine in x(t) and the sine in y(t). So, it is clear that there
is a specific phase relation between x and y and the amplitude of y is twice the amplitude of
x. There is a so-called phase quadrature between these two oscillations, so they have a difference of
90deg. So, the oscillator part of y(t) has a phase difference of 90deg and an amplitude which is twice w.r.t.
the radial.
• Let us now focus the attention on the radial offset and the along track drift terms. When we check the radial
offset and along-track drift terms, we can see that there is a link between them. So, the along-track drift
(i.e. the divergence term) is equal to:
3
along-track Drift = − nRadial Offsett
2
So, the radial offset and along-track drift terms are not independent. If there is not a radial offset,
we will not have an along-track drift. If there is no along-track drift, the solution/motion will be bounded, so
it will not diverge, the relative motion will be always contained in a given volume because the components,
and in particular y(t), will not diverge.
xc is the radial offset, yc0 is the along-track offset, ẏc t is the along-track drift.
3
ẏc t = −(6nx0 + 3ẏ0 )t = − nxc t
2
We have rewritten the components but in a more compact way, and now it is clear that there are oscillations where
the along-track is twice the amplitude of the radial and they are in phase quadrature, in fact there is sine for y(t)
and cosine for x(t). To better understand the implications, let us neglect the drift. Later, we will inject it. So, now,
this means that we have set the coefficient (6nx0 + 3ẏ0 ) equal to zero. If we neglect the drift, we have that:

x(t) − xc = Ax cos(nt + ϕx )
y(t) − yc = −2Ax sin(nt + ϕx )

where since there is no drift we can ignore the corresponding term and so we have yc . We can bring the amplitude
at the 1st member and then we square and sum the two equations. The equation that we obtain is an ellipse
where the center is (xc , yc ) and the semi-major axis is 2Ax (along y) while the semi-minor is Ax (along x):

(x − xc )2 (y − yc )2
+ = 1 = cos2 (nt + ϕx ) + sin2 (nt + ϕx )
A2x (2Ax )2
Space Flight Dynamics Michele Santarpia 120

The center depends on the initial conditions, and this is not a generic ellipse but it is called 2 by 1 ellipse
which has a precise eccentricity. The most interesting aspect of this discussion is that we have not made any
assumption about (special) initial conditions. So, we are saying that if the drift is zero or it is relatively
small, whatever are the initial conditions, the in-plane trajectory is always a 2 by 1 ellipse.
If the along-track drift is different from zero, what is the shape of the in-plane relative motion?
Let us consider the in-plane motion given by the 2 by 1 ellipse and let us so draw the orbital plane of the target
xy. Let us consider the ellipse where the chaser will fly on with a counter rotation w.r.t. the orbital motion of
the target. The description of the motion through the 2 by 1 ellipse is in general an instantaneous picture of
the motion, because in reality we have to consider that if we have a radial offset there will be an along-track
drift! So, the shape of the ellipse cannot be kept when xc is not zero (our motion is not centered in radial direction
around zero). So, we will have an along-track drift which is opposite to the radial offset: this means that if we have
a positive radial offset, the motion/ellipse will drift in the negative direction; if the offset is negative, the chaser
will drift in the positive direction. So, the idea is that the chaser instantaneously without drift will move on this
ellipse but due to the along-track drift this ellipse is moving backwards (the ellipse will move along the negative y
direction). the drift can only be in the horizontal (i.e. y) direction. These ellipses cannot move vertically, because
there is no divergence in the radial coordinates. Since in general there will be a drift plus the ellipse, the reality is
that the 2 by 1 ellipse may be not visible in the final solution. The final trajectory coming from the sum of these
two will not look like an ellipse, so the ellipse will be embedded in the solution, we will see that there is a drift and
while there is a drift there is the ellipse. So, we will obtain some strange shapes.
Instead if the along-track drift is zero, what is the shape of the in-plane relative trajectory?
When xc = 0, now the equation with the ellipses is exact because the drift is zero. Let us see the trajectory we
obtain. In this case, the chaser will move along this ellipse and according to Hill’s solutions it will continue to
fly over this 2 by 1 ellipse in time without drift. If we set to have an initial along-track velocity ẏ0 and an initial
separation vertical x0 that satisfy the relation ẏ0 = −2nx0 , there will be no radial offset and so no along-track drift
(there will be no divergence) and we will see the ellipse in the solution.
What is the solution without drift if we consider the tridimensionality of the motion?
When we consider the tridimensionality of the motion, we are taking z into account. We want to understand which
is the type of motion that we have to expect. Remember that z is independent, it is a simple oscillation. Of course,
depending on how we give the initial conditions, we will have a given phase relation. Since it is independent, it
will never modify the xy projection of the relative trajectory. Let us consider the entire solution. Let us consider
the HRF xyz and the ellipse. We can re-plot the same ellipse along z. We have to consider so the cross-track
oscillation, so the motion will be the composition of this in-plane motion given by the 2 by 1 ellipse
plus a cross-track oscillation. As a consequence of this, whatever are the initial conditions, the motion will
happen on the surface of this elliptic cylinder because the projection in the xy plane has to be the 2 by
1 ellipse. So, in general, the solution without drift happens on the surface of this elliptic cylinder with a given
amplitude.
Instead what is the solution with drift if we consider the tridimensionality of the motion?
In case in which we have a radial offset, instantaneously the motion will happen on the surface of a
cylinder but then this cylinder will drift backward or forward along y. So, if we have this radial offset,
this cylinder might not be visible in the final solution (motion that it is not bounded).
What happens if we give some specific initial conditions like x0 , ẏ0 , y0 , ẋ0 (which are the results)?
We need to check what terms are not zero, and then we will try to give some physical explanations of the reasons
why we have some results. We give the initial conditions 1 by 1, and this is a linear modeling of the dynamics, so
finally the motion will be given by the superimposition of the different effects. So, let us consider all the four cases:
1. Initial along-track offset y0 :

y0 ̸= 0 x0 = ẋ0 = ẏ0 = 0

For example, we are leaving the chaser as in the figure (we are not giving any initial velocity, so there is no
initial relative velocity), and we wait to see what happens. We have to check the terms in the solution: as
it is possible to see, in the radial coordinates, everything is zero. For the along-track, we can see that the
oscillations are zero and there is not 2 by 1 ellipse, there is no drift and we have only one term equal to y0 .
Space Flight Dynamics Michele Santarpia 121

Of course, z(t) does not change since it is independent. So, when we give this initial along-track position, the
consequence is that nothing will happen:

y(t) = y0 ∀t

Of course, this is what we obtain from the relative motion perspective. If we change our perspective considering
the absolute astrodynamics, we are saying something which is more than obvious, if we consider the inertial
view, we are saying that the chaser is behind the target because the chaser has only an along-track separation
which corresponds to an arc in absolute astrodynamics (of course, there is also a vertical component but if
they are close to each other, the arc becomes the horizontal, so the separation is not this angle, in the domain
of applicability of the Hill’s equations the ∆ν ∼ = y0 /rT , so it is a very small angle). So, if we have a chaser
on the same orbit of the target with a given anomaly difference nothing will happen in time and they will fly
together having always the ∆ν w.r.t. the target. The reason why the chaser is on the same orbit of the target
is because there is no initial relative velocity, so there is no difference in the absolute velocity! So, finally, if we
have an along-track offset, we will have that y(t) will be constant. There are some practical implications, like
for example in rendezvous operations: if we want to have some hold points (points where we can stay),
for example on the along-track axis we can set some hold points, so if we stay there and we have no relative
velocity w.r.t. the target, nominally we stay there without spending propellant (under the assumptions of
Hill’s equations), in the physical reality we will have some other effects and so it is not true that the chaser will
stay fixed in the hold points for a very long time, but nominally in the Hill’s equations the chaser stays fixed.
Imagine that we need to wait for an authorization from ground to jump to the next phase of the rendezvous
operation, because we want to have typically a number of checks in this phase, and then the ground control
will authorize us to pass to the next step. So, at least nominally, we can leave the chaser there and wait for
the ground command. This is true only in the along-track direction because we cannot leave the chaser on
the vertical (it does not stay fixed).
2. Initial radial (or vertical) velocity ẋ0 :

ẋ0 ̸= 0 x0 = y0 = ẏ0 = 0

In the ball example, we are throwing vertically up or down and all the other are zero. Let us draw this
situation in the HRF, where in red we have the initial positive radial velocity while in light blue we have an
initial negative radial velocity. Looking at the solutions, we can check which are the terms equal to zero, and
we can see that we activate some oscillations, one in radial and one in along-track and so we are activating a
2 by 1 ellipse, there is no radial offset and so no along-track drift, but instead there is an along-track offset.
Since there is no drift, so we have only a motion in periodic terms, the nice consequence is that when we try
to play with our ball we can throw up or down and the ball will come back because the only away to avoid
that it comes back is to activate the drift. So, we will have a 2 by 1 ellipse (which is visible because there
is no drift) with an along-track offset. So, throwing up we will activate this motion of the chaser and after
exactly one orbit (because the the period of the oscillations is equal to the target orbital period), the chaser
will come back from the nadir direction. If we increase the initial velocities, we are increasing the amplitudes
of the oscillations, and so the dimensions of the 2 by 1 ellipses (that will remain a 2 by 1 ellipse and with
always no drift), but the period of the motion will be the same. So, from the perspective of throwing the
CubeSat from ISS, we have not to throw it in the vertical direction otherwise it will come back. Let us try
to explain this in absolute astrodynamics perspective: if we consider that we are giving to the chaser a
radial positive ∆v (in red in the figure), and we have seen in the previous lessons that when we give a radial
impulse, we will have instantaneously a new velocity and so the velocity of the chaser will be in green. So,
the norm of the velocity of the chaser changes (vc ̸= vT ), but target and chaser are at the same distance, so
this means that we are giving a ∆ε, which means that we have a different semi-major axis and also different
period, so this means that the target will come back after one period while the chaser will come back to the
same point after a different period. So, the question is now:

How is it possible that they meet again if they have a different orbital period?

When we apply this maneuver, instantaneously we generate an eccentric orbit because now the new ve-
locity is not normal to the position vectors, and the point in which we apply the maneuver is not the perigee
or the apogee since we have instantaneously generated a flight path angle that is not zero, so we have finally a
∆e. The perigee is not in the point in which we apply the maneuver and we will have an ωnew (new perigee)
Space Flight Dynamics Michele Santarpia 122

and we will stay in an instantaneous new true anomaly νnew (we can compute analytically these terms). So,
now the question is:

Are these two conclusions (the one with Hill’s and the one with Keplerian) consistent?

For Hill’s solutions, the chaser comes back in the same point with the same period of the target while instead
for Keplerian this is not true. But in reality the two solutions are consistent. Let us see why: the Hill’s
solutions come from some approximations (we made some linear approximations), under the Hill’s approxi-
mations the idea is that the initial velocity ẋ0 is so small, that at 1st order the norm of the chaser velocity
does not change. The idea that we have in mind is that: the velocity of the target is in the order of 7 km/s
and the ∆v we are giving is in the order of 5 m/s and so we are considering that since 5 m/s is so small, we
can consider that the velocity of the chaser is equal to the one of the target and that’s why at 1st order the
orbital period is the same. So, we need to keep in mind that the applicability of the Hill’s equations is
linked to the close proximity assumptions. The close proximity assumptions does not mean that they
initially are together, it means that also in the long time they will be close to each other. So, for example,
if we give a ∆v = 1km/s we cannot apply the Hill’s equations because it determines a relative motion that
is not in close proximity. So, final take away is that we need to consider that this initial velocity has to be
small (bounded), so if we are throwing the CubeSat the Hill’s equations are good. So, if we want to launch
the CubeSat, this case is not a good idea.
3. Initial along-track velocity ẏ0 :

ẏ0 ̸= 0 x0 = ẋ0 = y0 = 0

We are launching forward or backward, we are activating in the radial the oscillations in cosine and the radial
offset, we will also have the along-track oscillations and the along-track drift (radial and along-track may only
oscillate together giving a 2 by 1 ellipse). Now, the motion is not purely periodic, we have a divergence, and
we will fly away from our target coming not back to the target. Imagine to have a positive along-track velocity,
and we generate a positive radial offset that implies a negative along-track drift, so if we throw forward finally
the motion will diverge in the other direction. This is the case where the 2 by 1 ellipse will become invisible
because it is covered by the drift, so we generate a 2 by 1 ellipse but it instantaneously will move in the
other direction. After one orbit since we have a negative drift, this is the sum now of a 2 by 1 ellipse plus a
negative drift, so from the point in which we have again the same relative velocity instantaneously, we will
follow again the same motion (periodic component) and so we will continue to drift backward. If we give an
initial along-track velocity in negative direction, initially we go backward but we generate a 2 by 1 ellipse with
negative radial offset and so we will drift forward. This shape is the sum of a 2 by 1 ellipse with the center
in the negative radial and with the drift which moves in the positive direction. If we want to pass the ball
behind, initially the ball will come towards the person behind me, but then it will go in the other direction.
So, this case, giving a velocity in the direction of the target is not a good idea. If we increase the velocity, we
are generating a larger ellipse (vertically we oscillate larger) and a larger radial offset which means a faster
drift. So, if we are throwing the CubeSat, we should use the along-track direction; if we do not want that
the CubeSat comes back, we need to use this component. We may also add some x0 , so we do not neces-
sarily we have to use a pure along-track velocity but we need some along-track component to activate the drift.

Nanoracks’20th CubeSat Deployment Mission on the ISS : finally, the typical direction in which we can launch
will be a combination of along-track and radial and so not really in along-track but we need some along-track
component necessarily to activate the drift in order to not come back to the station.

Drag effect: the discussion we have made are in Keplerian hypothesis. If we are in LEO, we will have
Drag but what is important is not the absolute Drag but the differential Drag between chaser and the
target, and so it is important the differential of the ballistic coefficients, because if they decay in the
same way for the same Drag effect, in relative terms we do not see a significant effect; but in general, because
of Drag, we do not follow the light blue trajectories. If we imagine that the chaser has a ballistic coefficient
larger than the target, it decays faster, and so the chaser w.r.t. the target will move down and forward. So,
instead of coming back to the same point with the same initial along-track velocity, we go down and forward
and so we are generating an altitude difference. So, we will also drift more in one orbit w.r.t. the Hill’s equa-
tions prediction. When we release CubeSat, we follow the magenta curve where there is the differential Drag
depending on the ballistic coefficient (Drag perturbation based on A/m). The magenta curve is a practical
Space Flight Dynamics Michele Santarpia 123

explanation.
4. Initial radial offset x0 :

x0 ̸= 0 ẋ0 = y0 = ẏ0 = 0

The oscillations are both activated for along-track and cross-track, and there will be also a radial offset, and
an along-track drift. Because of x0 ̸= 0, we have a divergence of the solution and it has a similar role of ẏ0 ,
the solution is not periodic. For example, if we have a positive radial offset of 10m above the target with
no relative velocity, there will be a drifting 2 by 1 ellipse. The drifting is happening backwards because the
radial offset is positive, so the 2 by 1 ellipse will become invisible in the solution. So, instantaneously at time
0, we are not moving due to zero relative velocity, then we will drift backward having that after one orbital
period we will be at exactly the same radial offset and we will be not moving again, and so for a moment we
will stop and then we will start for the 2nd jump. If we have a smaller component (green), the effect will be
smaller, so in one orbit we will drift less (every green point emphasized corresponds to a duration of one period).

How do we explain physically this drift? Why if we leave something with an offset there is
a drift?

The idea is that the initial radial offset is like a difference in the orbital radius: let us consider
in the absolute perspective to have the target in its orbit, and we imagine that instantaneously we are above
the target with the chaser and with a given offset, but it has instantaneously the same velocity of the target
in absolute term because in relative terms there is no relative velocity. If they are at a different distance from
the Earth, and with the same velocity, they will have different energy, the chaser has more energy, and it has
more semi-major axis, so the chaser is not on the same orbit of the target. So, now, the chaser since it is on
a higher orbit (not circular, but it is an ellipse), it will have a longer period, so the chaser is slower and
it will take more time to come back to this position. After one orbit, the target will be in an anomaly which
is different w.r.t. the chaser, so this means that the chaser w.r.t. the target is drifting backwards. We
need to come back on the ∆e since the orbit of the chaser is not circular.
• Radial offset x0 plus initial radial velocity ẋ0 : If we have both x0 and ẋ0 , in order to understand what
happens we can just combine the two effects considering them as independent. ẋ0 will be a 2 by 1 ellipse,
while the other will be a number of jumps, depending on which is the ratio between the terms we can obtain
different solutions. Since we have x0 all the solutions imply a drift.
• Initial radial offset x0 and initial along-track velocity ẏ0 : the two effects independently generate a
drift, if we combine them in a proper way we can nullify the drift.

7.2 Two-Impulse Rendezvous Problem

Two-Impulse Rendezvous Problem (it is a discussion of the rendezvous done in relative terms): let us consider
the HRF, with the target in the origin while the chaser is somewhere and it has initially a given position r0 and
a velocity v0 (-). The problem of the two-impulse rendezvous is to compute which is the maneuver that the chaser
has to perform in order to rendezvous with the target after a given amount of time τ . So, given a generic chaser
initial conditions r0 and v0 (-) w.r.t. the HRF, we set a time τ (it can be minutes, days, etc.) and we want to find
a maneuver with 2∆vs that will bring the chaser to the target after the time τ . The solution of this
problem is an application of the Hill’s equations.
The idea is that at time zero, we will not follow the natural evolution but we will apply a maneuver so that after
a time τ we arrive where is the target. The idea is that the chaser needs another velocity called v 0 (+) which is
obtained by applying a ∆v 1 : in this way, we will jump on a trajectory so that after a time τ we will arrive where
is the target. We will have a velocity v(τ ) and at this point we need to stop at the target because of course we do
not want to have a fragmentation event. So, as we arrive at the target, we need to reduce the velocity up to 0. So,
we will apply a ∆v 2 which is equal and opposite to v(τ ) so that instantaneously we will stop at the target location.

∆v 2 = −v(τ )

We can compute these two impulses analytically through the Hill’s equations. In the Hill’s analytical solutions, we
Space Flight Dynamics Michele Santarpia 124

have the following dependencies:


x(t) = fx (x0 , y0 , ẋ0 , ẏ0 , n, t)

y(t) = fy (x0 , y0 , ẋ0 , ẏ0 , n, t)

z(t) = fz (z0 , ż0 , n, t)


where x(t) and y(t), since they are decoupled from z(t), are not related to z0 and ż0 . n is the mean motion of the
target.
We can take the derivatives and write also that:

ẋ(t) = fẋ (x0 , y0 , ẋ0 , ẏ0 , n, t)

ẏ(t) = fẏ (x0 , y0 , ẋ0 , ẏ0 , n, t)

ż(t) = fż (z0 , ż0 , n, t)


So, if we know the initial conditions we know everything.

How do we solve the problem of the two impulses?
First of all, let us consider the 1st impulse. The only constraint is at time τ : what we want is that when we
compute the natural solution after the application of ∆v 1 , we want that after a time τ we are exactly at
the target (of course, we are assuming that the initial time t0 is zero):

x(τ ) = 0

y(τ ) = 0

z(τ ) = 0


If we consider these functions, we have that they depend on the initial positions and the desired initial velocities
ẋ0 (+), ẏ0 (+) and ż0 (+) which are the velocities that we need to bring the chaser to the target. The desired initial
velocities are the unknowns.

x(τ ) = 0 = fx (x0 , y0 , ẋ0 (+), ẏ0 (+), n, τ ) → ẋ0 (+)

y(τ ) = 0 = fy (x0 , y0 , ẋ0 (+), ẏ0 (+), n, τ ) → ẏ0 (+)

z(τ ) = 0 = fz (z0 , ż0 (+), n, τ ) → ż0 (+)


In the 3rd equation, the only unknown is ż0 (+) and so we obtain it, while from the first two equations we have two
unknowns. So, it is an algebraic problem from which we can get the the initial velocities ẋ0 (+) and ẏ0 (+).
Once we have the three components, we have v0 (+) and so we can have ∆v 1 :

∆v 1 = v 0 (+) − v 0 (−)

where v 0 (−) is the initial velocity (known initial condition). We have used so the position solutions of the
Hill’s equations to get ∆v 1 .
To get ∆v 2 , we can use again the solutions of Hill’s equations, but this time for velocities and we impose that:

∆v 2 = −v(τ )

and we write the velocity solutions which are function of the initial conditions that now are all known:

ẋ(τ ) = fẋ (x0 , y0 , ẋ0 (+), ẏ0 (+), n, τ )

ẏ(τ ) = fẏ (x0 , y0 , ẋ0 (+), ẏ0 (+), n, τ )

ż(τ ) = fż (z0 , ż0 (+), n, τ )


So, we can get the final velocity v(τ ) and evaluate the ∆v 2 .
Example 7.5 Curtis: we have a chaser and a target in the same 300km circular orbit. The chaser is 2km behind the
target. We want to compute a 2∆vs rendezvous maneuver so that we will rendezvous with the target in 1.49h=5364s.
We want to compute ∆v 1 and ∆v 2 and we just need to use the solutions of the two impulse problem.
Space Flight Dynamics Michele Santarpia 125

The initial velocities are zero since they are on the same orbits and so there is no relative velocity. So, v 0 (−) = 0.
Also the out-of-plane components are zero. We need to compute v 0 (+) = ∆v 1 . So, we use the equations in order
to find the initial velocity:

x(τ ) = 0 = fx (x0 , y0 , ẋ0 (+), ẏ0 (+), n, τ )

y(τ ) = 0 = fy (x0 , y0 , ẋ0 (+), ẏ0 (+), n, τ )

z(τ ) = 0 = fz (z0 , ż0 (+), n, τ )


3
The orbital period of the target 2π aµ ∼= 5431s. So, we are asking to have a rendezvous in a time that is one minute
less than the orbital period. The solution will be:
v 0 (+) = (−9.48 · 10−3 î − 1.22 · 10−1 )m/sĵ
We can find also the velocity by which we will arrive on the target which is equal to the 2nd impulse:
v(τ ) ∼
= (9.48 · 10−3 î − 1.22 · 10−1 ĵ)m/s = −∆v 2
The total ∆v tot is equal to the sum of the two norms:
∆v tot = |∆v 1 | + |∆v 2 | ∼
= 0.24m/s
In this problem, we are giving a ∆v which is like the one in the figure where the vertical one is much smaller
than the horizontal.
What happens applying this initial velocity?
We studied the evolution of the trajectory when we start from the origin but it is the same of course if we start from
another point in along-track. Because of the vertical component we will jump on a 2 by 1 ellipse and when
we combine in a proper way with the horizontal component, we will get a situation which is similar to just
along-track with some inclination, and if we do not do anything we arrive in the red point after one period. Again
we have a solution where there is a drift because we gave ẏ0 but this is the combination of an oscillation and a
drift. This geometry qualitatively explains the 2nd impulse because the final velocity v(τ ) has a positive component
vertical and so our 2nd impulse has to be opposite to this velocity. The reason why we have this type of geometry
(for this type of combination) is because we asked for a little bit less than the orbital period. After one orbital
period we will be in the point B, and then everything will repeat but since we have imposed to rendezvous in 1.49h
so less the orbital period, the maneuver will be before one entire orbital period. We can see that the impulses are
very small where vertical components are mm/s and fractions of m/s for along-track, remember that this is the
case of 2km. So, if we play with a smaller separation, we get smaller impulses.
All these ∆vs are computed in the HRF, which is a rotating frame but the ones we need for propellant
calculations are needed in the inertial reference frame. We can check that there is no difference. v c is the inertial
velocity of the chaser. We can write the equations before and after the maneuver (we can write without the subscript
0). The target does not maneuver, so the absolute velocity of the target is the same before and after the maneuver
and also for ΩH . Furthermore, rrel minus and plus are equal since the maneuver is impulsive so it is instantaneous.
When we take the difference between these two equations, we get that the ∆v in the inertial reference frame
is the same in the relative.

7.3 Limits of Hill’s Equations

What happens when we introduce perturbations in Hill’s equations?
Errors of Hill’s equations: they have a limited accuracy, since we have done some assumptions that lead to
approximations.
1. Close Proximity: due to this assumption, we apply a linearization. Even if the time scale is long, if we
continue to stay in close proximity the solutions are good (if the distances are small like few km in LEO will
be very good).
2. Target on Circular Orbit: of course, the target is never on a circular orbit, but it is characterized by a
certain eccentricity. If the target is on a high eccentricity orbit, all these discussions are totally not useful.
So, if we want to study the rendezvous with a Molniya satellite, Hill’s equations are not good. If the satellite
is almost circular instead there are some errors but we can use Hill’s equations.
Space Flight Dynamics Michele Santarpia 126

3. Keplerian assumption can be pretty good if the time scale is not so long, but in long time as in absolute
astrodynamics we will have problems and so divergence.

So, how can we modify Hill’s equations to go beyond Keplerian motion?

The idea to include perturbations is again to introduce something on orbital parameters, but now it is
important not the absolute effects of perturbations on the orbital parameters but the differential effects. The
differential effects will depend on the differences in the orbital parameters between the chaser and the target.
This problem can be described in terms of differences in orbital parameters. A very popular approach is based
on ’relative’ orbit parameters which are not exactly the difference between orbital parameters.
As we know, Hill’s equation are function of the initial conditions of the relative motion. We can now have an
approach where the solutions of the Hill’s equations can be written not in terms of initial conditions but
in terms of orbital parameters of the target and the difference between the orbital parameters of
chaser and target. So, the radial, along-track and cross-track components can be written as function of the
orbital parameters of the target satellite and of the ∆ between the orbital parameters:

x(t) = fx (aT , eT , iT , ΩT , ωT , MT , ∆a, ∆e, ∆i, . . . , t)

y(t) = fy (aT , eT , iT , ΩT , ωT , MT , ∆a, ∆e, ∆i, . . . , t)

z(t) = fz (aT , eT , iT , ΩT , ωT , MT , ∆a, ∆e, ∆i, . . . , t)


So, there are two different ways to describe the same relative dynamics (i.e. to approach the same problem) and
they are both useful. The ’relative’ orbit parameters approach is useful since we know what happens to the
orbital parameters due to perturbations and so it is easier to understand what happens to relative motion due to
perturbation.
Now, using relative orbital parameters, close proximity means that the differences between orbital parameters
are small:
∆a
<< 1 ∆e << 1 ∆i << 1 ∆Ω << 1 ∆(M + ω) << 1
aT
Now, what is important is the argument of latitude and not true anomaly or the argument of perigee, we may
have two satellites that are in close proximity and they have that the perigees are 180deg of difference, what is
important in this case is that their mean anomaly is equal and opposite and so for this reason what has to be small
is the difference of the sum.
Let us give a look at the equations of the relative motion of a chaser w.r.t. a target expressed as function of the
relative orbit parameters. The target is under the same assumptions of Hill’s equations. We find constant terms,
oscillations terms, radial and along-track offsets and an along-track drift term. What we can do is to compare with
Hill’s and we can find a one to one match. From the POV of orbital parameters, the reason why there is a 2 by 1
ellipse is that the chaser has some eccentricity w.r.t. the target. The reason why there is a cross-track motion is
that the two orbital planes are not exactly the same and so there is a difference in Ω and inclination.
Hill-Clohessy-Wiltshire equations: if we consider the differential model, it is a constant coefficient linear system
that can be solved using the classic control theory (PID, etc.) when we have the forcing functions. These solutions
are general.
In terms of natural solution, we have that the solutions depend on the initial conditions of the relative motion, so
we give these six scalars and we get the solutions. In fact, we study the shape of the different solutions.
Now, we can rewrite the same solutions in terms of the difference between the orbital parameters. From the Hill’s
equations POV, we have a given relative motion because we establish an initial relative position and an initial
relative velocity. When we consider the POV of difference between the orbital parameters, the idea is
that we have a relative motion because the two orbits are not exactly the same and we can express
the relative motion as a function of the δ. For example, under the assumptions similar to the Hill’s ones i.e.
Keplerian and circular orbit for the target, we obtain the equations where the δ represents the difference between
chaser and target. This is a different way of looking for the same phenomenon.
Comparison between the relative orbit parameters and Hill’s solutions: we can consider a one to one
match with the Hill’s solutions. The colors are useful to see which are the one to one match.
Space Flight Dynamics Michele Santarpia 127

In the Hill’s solutions we have that:

1. The radial offset xc exists due to a combination of the initial radial position x0 and the initial along-
track velocity ẏ0 .
2. The 2 by 1 ellipse exists due to an initial radial and along-track velocity, and due to an initial radial
position.
3. The along-track offset exists because we have an initial along-track position and/or an initial radial
velocity.
4. The drift exists due to an initial radial position and initial along-track velocity.
5. The cross-track motion exists due to the initial position and velocity in cross-track.
Instead in terms of the differential orbital parameters:
1. Radial offset exists due to the difference/delta of semi-major axis:
δa
aT

2. The 2 by 1 ellipse exists because the chaser is not on a circular orbit and, in fact, the cosine and sine depend
on δe (equal to the eccentricity of the chaser minus the eccentricity of the target which is zero):
−δe cos(MC0 + nC t) 2δe sin(MC0 + nC t)

3. Along-track offset: it is given by the terms in parenthesis (which are constant) and by ∆Ω i.e. by the
difference in RAAN between the chaser and the target:
(ωC + MC0 − uT 0 ) + δΩ cos iT

4. The drift in along-track exists due to the different mean motion of the two satellites that means
that they have a different orbital period and so they will fly away from each other in along-track, of course
everything we conclude on these equations is true as long as we stay in close proximity:
(δn)t

5. The following term represents the cross-track motion that exists due to a difference in RAAN and/or
in inclination:
−δΩ sin iT cos(ωC + MC0 + nC t) + δi sin(ωc + MC0 + nD t)

MC0 is the initial mean anomaly of the chaser, uT 0 is the initial argument of latitude of the target, ωC
is the argument of perigee of the chaser (ωC + MC0 is not the argument of latitude but it is called mean
argument of latitude).
The idea now is: since these two sets of equations are modeling the same phenomena, we have the one to one
match, if we image that we can equate the terms one to one, we can analytically match the difference in
orbital parameters and the target orbital parameters with the initial relative position and velocity.
This becomes a set of algebraic relations, so we can jump from the initial conditions to the difference in orbital
parameters or in the other direction.
In Hill’s equations, we have six scalars and in the relative orbit parameters equations we also have six differences in
orbital parameters. When we consider the link 1 to 1, the link also involves the absolute parameters of the
target. So, this means that the same set of initial conditions of the relative motion may correspond to
different sets of difference in orbital parameters because the link depends on the absolute parameters
of the target satellite. Let us imagine, for example, the cross-track motion where initially we gave a cross-track
velocity ż0 that may correspond to a difference in inclination or to a difference in RAAN or to a combination of
them depending on where we are giving this initial velocity because if we are at the equator this will correspond to
a difference in inclination, otherwise if we are close to the poles this will correspond to a difference in RAAN.
Let us analyze geometrically the different terms:
Space Flight Dynamics Michele Santarpia 128

1. Radial offset: xof f set in Hill’s equations (HCW) is given by:

ẏ0
xof f set = 4x0 + 2
n
while in difference between orbital parameters we have:

δa/aT

The geometric interpretation is trivial because we are saying that if we have our target circular orbit in black,
and then we have our chaser orbit in blue, if we draw the instantaneous position of our target on the target
circular orbit, we can plot the HRF, so the radial offset exists due to the δa. As we know:
3
ẏdrif t = − nxof f set
2
3
ydrif t = − nxof f set t
2
ẏdrif t is the rate of drift.

In the orbital parameters space, we can start from the mean motion n. We apply the differential, we can see
that a da implies a dn.
3
adn = − nda
2
dn is a difference in angular velocity, so if we multiply by a (the radius), we obtain a linear velocity in along-
track. So, adn=ẏdrif t .

So, the two ways to look the phenomenon are pretty consistent because we have written exactly the same. In
fact, we have written that:
3
adn = ẏdrif t = − nda
2
So, as before we have -3/2n and the offset da (before was xof f set ).

So, when we consider the equations in difference of orbital parameters, we have to imagine that we have
the function in da. So, we have that:
• The radial offset corresponds at 1st order to da.
• The drift in along-track in Keplerian mechanics at 1st order is proportional to da through the mean
motion n.
2. Along-track offset, in Hill’s equations we have that:
ẋ0
yof f set = y0 − 2
n
In terms of orbital parameters:

yof f set = a(ωc + MC0 − u0 + δΩ cos i)

(no subscript is for the target, with C is the chaser instead).

Again, the interpretation is trivial, we can plot the orbit, the target and the along-track offset conceptu-
ally is only an anomaly offset, consider the chaser, the angle is yof f set /a.

In orbital parameters, we have that the offset is in angle given by ωc + MC0 − u0 (difference in the mean
argument of latitude) but we have to understand why also δΩ cos i contributes i.e. why a difference in planes
contributes to this offset.
Space Flight Dynamics Michele Santarpia 129

3. How do we explain the 2 by 1 ellipse in orbital parameters, and then the difference in the
orbital planes (out-of-plane motion)?

For the 2 by 1 ellipse, if we consider the oscillations, in the orbital parameters we are saying that:
(
x∼= a(−δe cos(Mc0 + nc t))
y∼= a(2δe sin(Mc0 + nc t))

So, one is twice of the other and we have cosine and sine.

How do we explain the 2 by 1 ellipse by the chaser eccentricity?

We can interpret it geometrically: let us plot a target circular orbit, and let us consider a chaser that
has a little eccentricity w.r.t. to the target (no other difference in the other orbital parameters). Initially, we
consider a given position for the target and for the chaser which is at its perigee. This geometry is similar to
one we have seen before to see the effect of eccentricity on the longitude of the geostationary satellites, the
difference at perigee is aδe, initially they move with different angular rates since the chaser is faster, so there
will be an angle ∆ν (remember that we are in the same orbital plane), then we have a maximum separation
in B, in A we have again the angle equal to 0. In order to understand the relation with the relative motion,
we can plot the HRF which is a rotating frame. So, now we want to understand which is the motion of the
chaser in the HRF: what the chaser is doing is to fly on a 2 by 1 ellipse, in particular let us plot four ellipses.
The separation in B in along-track is equal to 2aδe. So, when we have some eccentricity, the result
is that the chaser is flying on a 2 by 1 ellipse and the interesting point is that the chaser is flying in an
opposite sense of rotation to the orbital motion of the target. In fact, let us imagine the situation:
in the starting point the chaser is below the target, after one quarter of the orbit it is forward, then after
another quarter of the orbit it is above and then it is backwards. So, in the HRF, the chaser is flying with
the sense of rotation reported in the figure. This should explain that the reason of the 2 by 1 ellipse in orbital
parameters is a slight eccentricity.
4. How do we explain the cross-track motion? Why do we have aδi and δΩ sin i, and why do we
find ωC of the chaser?

Imagine we have circular orbits and let us consider the spherical geometry. In z we have an oscillation
with amplitude aδi. If we have two orbits with different inclinations, they will touch different parallels and
let us imagine that they only have a difference in inclination and so they have the same AN.

So, where aδi comes from?

This δi will generate a maximum separation at the orbital poles in yellow (meridian), and this maximum
separation in linear terms (since they are circular orbits) is equal to aδi.

The first oscillating term is:

z∼
= aδi sin(ωc + Mc0 + nc t)

Conceptually, the term ωc + Mc0 + nc t is the instantaneous argument of latitude of the chaser (≈ uc )
(remember that we have Keplerian assumptions, that’s why we have the mean motion), this sine will be zero
at the nodes and maximum at the orbital poles. What we have obtained is a linearized model of the 1st order,
so that’s why we have the different effects 1 by 1, in this model we never have coupled effects because they
are higher orders (δiδΩ).

What about δΩ?

Let us consider the sphere, the orbits which differ only in RAAN, so they will touch the same parallel.
Target is red, chaser blue. With this choice δΩ is negative. δΩ will generate this lateral motion and this
lateral motion will not be equal to aδΩ and the reason is that we have to consider the projection from the
target to the orbit of the chaser considering the normal in violet, we are considering that δΩ is much lower
than 1 in radians (so 0.05deg), if we see this, we can confuse arcs with segments because everything is small,
Space Flight Dynamics Michele Santarpia 130

in this way we can identify a triangle, we write the different lengths. So, we understand why if we have a
difference in RAAN, we have a cross-track left / right oscillation whose amplitude is δΩ sin i.
z∼
= −aδΩ sin i cos(ωc + Mc0 + nc t)
As we can see from the equation, the separation is maximum at the equator and nullifies at the poles (remem-
ber that we are confusing arcs with segments, so there is a spherical triangle but we are using it as a normal
one).

With this explanation we are also understand why in the along-track offset there is δΩ cos i. In fact, if
we have two orbital planes with a difference in RAAN, and we have an inclination that is not 90deg, since
the HRF is aligned with x and y they are in the orbital plane, there will be the cosine of i projection, so the
difference in the orbital planes will be reflected in an offset of the 2 by 1 ellipse, or in general in an along-track
offset, so the motion will be not centered around 0. [non chiarissimo, min 50.30 (1)]

So, δΩ cos i derives from the simple projection of one orbital plane on the other.

7.4 Formation Flying

Formation Flying example: in this case, we will consider always the relative motion equations in terms of relative
orbit parameters. We still consider Keplerian mechanics as before, but now we have to use some slightly different
equations because we are considering the target not exactly on a circular orbit but on a slightly eccentric
one. This variation is introducing a new term both in the radial and along-track equations: if we check these two
terms, they both contribute to the 2 by 1 ellipse. These two terms do not exist due to the presence of a difference in
eccentricity but their amplitudes contain the difference in the initial mean anomaly between the chaser
and the target, so this means that these terms exist also if they have the same small eccentricity and these terms
are important if δM0 is not so small.
Be careful on a concept: we have said that when we have difference in orbital parameters, these differences have
to be small otherwise we do not have close proximity conditions. Instead, the mean anomaly and the argument of
perigee do not have to be small but the difference between the sum of the mean anomaly and argument of perigee
has to be small! So, this is the reason why δM0 can be any value and, of course, the sum of the two argument
of perigee has to be almost the opposite of the sum of the two initial mean anomaly in order to have a difference
which is close to zero.
Let us consider an example where we have two satellites: the chief (i.e. the target) is on a SSO frozen orbit
6PM 6AM (we have a radar as payload on-board the satellite). We want to fly a deputy (i.e. the chaser) close to
our chief, they have to fly together with a few hundreds of meters for the entire life and we would like to
minimize the need for formation control maneuver, we want to have the safest possible relative motion.
Selecting the relative motion means that the chief is where it is, we need to select the values of the δ in orbital
parameters! Let us give a look at the equations: if we want that they fly together, we do not want a
difference in the semi-major axis because even a small difference will make them to fly away (finally,
they will drift away because of our errors but this is maintenance).
δa = 0 → δn = 0
So, finally, δa = 0 (so, this means along-track drift equal to 0). If we want some lateral motion, we could ask for
a difference in inclination but we have to remember that J2 exists! If we give a different inclination to the 2nd
satellite, J2 will not be the same and so the nodes will drift from each other. For this reason, we want zero difference
in inclination in order to have stable orbits from the POV of nodal precession.
δi = 0 → δ Ω̇J2 = 0
We also assume that there is no difference in perigee (so, we select again also for the deputy perigee at 90deg) and
we set no initial difference in mean anomaly. So, δM0 = 0 and δω = 0 (ωc = 90◦ , where c stands for chaser). So,
finally, we have the chief on its orbit and we will set the 2nd satellite on an orbit that differs only in eccentricity
and RAAN. If this is true, only few terms are present: in particular, we have the 2 by 1 ellipse because we have δe
and then we have δΩ.
∂ Ω̇
δe → δ Ω̇J2 ∼
=0 =0
∂e e=0
Space Flight Dynamics Michele Santarpia 131

Which is the type of motion that we have?

Let us give a look at the equations in which we can understand the resulting motion: there are three oscillations,
zero mean, at the same frequency (at least in our Keplerian assumptions). The interesting point is that, inside the
cosine of the 3rd function (cross-track), we have the initial chaser argument perigee ωc0 and we select it to be equal
to 90 deg. Then, we choose the initial chaser mean anomaly to be equal to 0 at time t=0.

Mc0 = 0

This means that at time t=0, x=xmax =-aδe and z=0. Instead, at time t=T/4, x=0 and z=zmax .
Let us see the in-plane motion of the chaser in the xy plane, we have the 2 by 1 ellipse where there is a slight offset
i.e. the center has an offset equal to aδΩ cos i.
Let us see what happens instead in the xz plane (radial lateral plane) i.e. we are looking at the relative motion from
behind or forward. We have to consider the equations in x and z, which are two oscillations at the same frequency
but with a phase difference equal to 90deg. This means that when the radial is maximum negative, the z is zero
and vice versa. So, the geometry will be the one of an ellipse since we are combining two oscillations that are in
phase quadrature, and we fly the trajectory in a direction or in another one depending on the relative signs of δΩ
and δe. Since we are setting as our design parameters δΩ and δe, we can shape this ellipse as we like.
We like this geometry because it has an interesting feature, because whatever happens the deputy / chaser will
never fly in front of the target or behind the chief / target meaning that the chaser will never lie on
the velocity vector of the other satellite (it is like we have a tube around the target and the chaser does never
violate this tube, we are generating a sort of safety tube around the target). This type of geometry generates a
significant reduction of the collision risk. This is what we call a passively safe geometry and we have obtained
a (passively) safe ellipse. This is the idea of the so-called “safe ellipses”, which can be obtained by exploiting
the natural motion of the satellites. We call it passively safe geometry because if we lose completely control (so we
cannot apply any propulsion) and if there is an along-track drift, since the two satellites do not intersect (the two
orbits do not have intersection points) they cannot collide because we never find one satellite behind or in front of
the other one (we never cross the origin or fly close to it). This type of geometry is the one of the Terra SAR-X (the
initial orbit was selected for this satellite), TANDEM-X (the other orbit we have described is of this satellite). This
is what we call an helix formation. If we look at the motion of these two satellites in the inertial perspective, we
will see that they fly around each other similar to the helix of the DNA (double helix). Instead, when we look from
the POV of one of the two satellites, the other will fly around.
So, if we choose in a smart way the differences in orbital parameters, we can play with radial and cross-track
oscillations, with their relative phase, and generate the safe ellipse.
The complete relative motion is in xyz. To draw this motion, we have to consider that when the z is maximum,
the along-track is maximum. What we obtain as trajectory is a 2 by 1 ellipse that has undergone a yaw rotation
(imagine to have the 2 by 1 ellipse in the orbital plane xy, and then we maneuver to rotate by yaw). When we
project this trajectory in xy we obtain the 2 by 1 ellipse, and in xz we have the tube. If we see from the radial
direction from below or top, we see an inclined line.
So, we can build a safe geometry whatever is the orbit of the chief, and not only using δe and δΩ but there are some
other ’creative ways’ to be safe. Passively safe geometry is very popular whenever we need to stay in proximity and
we want to relax our control requirements: for example, thanks to this orbit, in German missions they were able to
realize ground-based control which sounds a little bit risky if we have two satellites at 500m but since the relative
motion is smooth and the risk of collision is reduced, we can wait for ground intervention.
Let us remember that we are able to set this safety tube around one satellite but this is true since ωc = 90deg,
because finally the shape depends on ωc . Of course, when time passes, ωc will change and so the geometry becomes
less and less safe as time passes. If we never control the perigee, what happens is that, even if there are no
differential effects (so, in terms of orbital parameters the two satellites do not fly away from each other), there are
some problems: the radial and the cross-track still oscillate with the same amplitude but now they are no more
in quadrature, so finally the relative motion will always be contained inside this rectangle but it becomes less and
less safe and, for example, we can get something that looks like the light blue (this is because x and z still oscillate
within the same intervals but the phase difference is no more 90deg). When the phase difference is 0 or 180deg,
this means that when one is maximum the same is for the other, the projection will be the bisector. In this case, it
Space Flight Dynamics Michele Santarpia 132

is not safe anymore because we pass exactly in front of the target. This is the reason why in time we need to pay
attention and typically we like to adjust the perigee in order to keep the same geometry.
So, this is not a problem of differential perturbations but just a problem of phase difference. So, if the absolute
orbit does not have the perigee at 90deg, this is not safe.
This is the same problem of the GEO satellites (remember the ’8’ and the inclined ’8’), the problem was depending
on the perigee because one motion depends where is the node and the other depends where is the perigee. In fact,
the radial oscillation is phased w.r.t. the perigee, while the z oscillation depends on the node. If we keep 90deg of
perigee, we have the same phase difference otherwise we will get a problem, so periodically we need to maintain
both the perigees.
Space Flight Dynamics Michele Santarpia 133

8 Interplanetary Transfers
Interplanetary Transfers: there are some missions in which, at a given moment, the spacecraft escapes the
region of space where the Earth is the dominant source of gravity (primary body) and so during this phase of the
mission the Sun becomes the dominant source of gravity. Based on this definition, Lunar missions are not exactly
interplanetary missions because we never escape the gravity of the Earth when we fly to the Moon.
Interplanetary Maneuvers from an Heliocentric Perspective: we need to pay attention to the time and
cost of interplanetary maneuvers. Let us consider that we want to fly from the Earth to another body of
the solar system, and let us consider it as a pure heliocentric problem, so we are not paying attention to
the gravity of the Earth. This is just to understand how large is the maneuver from an heliocentric perspective.
Let us consider that initially we are on the orbit around the Earth, and we give an Hohmann ∆v (tangential) to
jump to another orbit. The table contains the information about the Hohmann ∆v needed to fly to the orbit of
another planet in the solar system. For each destination, we have the required heliocentric velocity and so the
initial ∆v at the perigee of the transfer ellipse when we are in the Earth orbit (this corresponds to just the 1st ∆v).
If we want to fly to Mercury or Venus (inner orbits), the ∆v must be negative, instead for Mars, Ceres (Cerere,
pianeta nano, close to the asteroid belt), Jupiter, Saturn, Uranus, Neptune and Pluto (outer orbits), the ∆v must
be positive. As we can see, we have huge values of required ∆vs . When we consider outer planets, not only the
∆v is impressively large, but also the time is too high. For example, to fly to Neptune we will take 30 years. It
is clear that, when we want to go to outer planets, we cannot play directly with Hohmann and so with direct
transfers, like Voyager or New Horizon, but we need some intermediate steps. We need to save time and also
we require to steal some ∆v from other planets (we will see the gravity assist later).

8.1 Earth to Mars Transfer

Earth to Mars Transfer: as first approach to interplanetary missions, let us start by considering a direct
transfer, and in particular the classic case of Earth to Mars transfer, which requires several months (0.709 years).
If we consider the entire journey, there will be three sources of gravity (Earth, Sun and Mars). We cannot use
the solution of the two-body problem with Earth and Sun. As 1st level approach for mission analysis, we use the
two-body dynamics with the idea to switch on and off the bodies as needed. In fact, when we fly from Earth to
Mars we have three different phases:
1. Departure from Earth: we depart from Earth and, for some time, the Earth will be the dominant source
of gravity and so, in this phase, we can consider a two body problem w.r.t. the Earth (we are trying to escape
the Earth, so this phase has a limited duration).
2. Heliocentric Transfer: the gravity of the Earth will decrease as we fly away, the gravity of the Sun can be
considered almost constant, at a given moment the gravity of the Sun will become dominant w.r.t. the one
of the Earth, so we can imagine to cross an imaginary sphere in space called Sphere of Influence (SOI) of
the Earth. As we cross this volume, the Earth is not anymore the dominant source of gravity and the Sun
becomes the primary body. Then, we fly in the solar system, at a given moment we will start approaching
Mars, of course the gravity of Mars is present at any moment but, of course, for all the mission except for the
last days the gravity of Mars has a very limited influence.
3. Mars Arrival: as we approach Mars, its gravity will increase more and more. At a given moment, we will
cross the theoretical SOI of Mars and it becomes the primary body.
So, for mission analysis purposes we can split the mission in three phases:
1. Earth Departure.
2. Heliocentric Transfer.
3. Mars Arrival.
These three phases can be studied separately using three times the 2-body problem where, for each of them, we are
changing the primary body. So, we will have one conic around the Earth, one around the Sun and one around Mars.
These three conics have to be consistent with each other, so we have to patch the conics. This means that
we have to patch the Earth departure orbit with the heliocentric transfer, and then the heliocentric transfer with
the Mars arrival orbit. When we say patch, it means that the velocities have to be consistent which means
that the final velocity of the Earth departure phase has to be equal to the initial one of the heliocentric transfer
Space Flight Dynamics Michele Santarpia 134

conic, and the final velocity of the heliocentric transfer conic has to be equal to the initial one of the Mars arrival
conic. This method is the so-called patched conics method.
Tricky points of this discussion:
• The SOI is not an infinite volume and it has a finite radius. How do we compute the SOIs?
• What do we really mean for final velocity on the Earth orbit? There will be some apparent ambi-
guities, because on one hand the SOI is a finite volume, but on the other hand the velocity at the sphere of
influence is for us the velocity at the infinite w.r.t. the Earth.
As we can see from the figures, we will be in the sphere of influence of the Earth only for few days, and
most of the time is spent under the action of the Sun. As we can see, the Sun gravity is pretty constant even if it
decreases a little bit since the distance is increasing.
In the table, we can see the relative gravities in the different phases: when we are, for example, in Earth
departure phase, the relative gravity is w.r.t. the Earth, and for Earth is of course 1, for Sun is 6.26 · 10−4 , and for
Mars is 3.55 · 10−10 . In the table are reported also the perturbations, if we want to be very accurate we need to
consider also the other effects and we can consider which are their weights.
Strategy of the patched conics: it consists in starting from the design of the heliocentric transfer conic,
and then we build the departure conic and, finally, arrival conic and we patch the departure and arrival conics
with the heliocentric transfer conic.

8.1.1 Heliocentric Transfer Conic

For the heliocentric transfer conic, we use the porkchop plot ”letteralmente grafico a braciola” (or ”butterfly
plot”). We enter with the desired Earth Departure Date and with the desired Mars Arrival Date. For each
couple, we can compute which is the transfer and the cost in terms of energy. We obtain this diagram by solving
the heliocentric Lambert’s problem. The boundary conditions are:
1. At the Departure Date we are at the Earth position.
2. At the Arrival Date we are where at the Mars position.
Each point of this plot is obtained by solving a Lambert’s problem. The black dot represents the minimum
energy transfer (it is close to the Hohmann transfer with the right phasing).
Since we have to escape the gravitational field of the primary (i.e. the Earth) during departure, we will need a
departure hyperbola. Then, we will need a heliocentric conic for the transfer (an ellipse in the heliocentric
frame), and then the arrival hyperbola because from the POV of Mars we arrive from infinite because we arrive
from outside the SOI.
Sphere of Influence Definition: the design process starts from the heliocentric transfer conic, and the first
thing to do is to build the sphere of influence which can be defined for any body in the solar system (not for
the Sun, because we can talk about the sphere of influence for the Sun from an interstellar perspective). The SOI
is larger for bodies with larger mass and also it depends by the distance from the Sun. By definition, the
SOI is a volume in which the celestial body is the dominant source of gravity. For a quantitative definition, we can
use the Laplace definition where we use the 3rd body effects to define SOI. In particular, the SOI of a celestial
body is defined as the surface of a sphere over which the 3rd body effects of the Sun in a planet centric
perspective equals in relative terms the 3rd body effects of the planet in a heliocentric perspective
(if we are planet centric, the Sun is a source of 3rd body perturbation; if we are heliocentric, the planet is the source
of 3rd body perturbation).
To define the SOI, let us consider to have three bodies i.e. the satellite, the Sun and the Earth. We want to study
the motion of the satellite within the 2-body modeling and let us consider the two different perspectives:
1. The satellite moves w.r.t. the Earth and the Sun is the source of the 3rd body perturbation.
2. The satellite moves w.r.t. the Sun which is the primary and the Earth is the source of the 3rd body pertur-
bation.
Let us write the equations of motion in these two perspectives: the 1st term is primary body gravitational
attraction and the 2nd one is the 3rd body perturbation.
Space Flight Dynamics Michele Santarpia 135

After writing the equations, we can apply the definition of SOI: as we know B and D are the 3rd body perturbing
accelerations and we say that they have to be equal to each other in relative terms. In relative terms means in
terms of their ratio w.r.t. the primary.
Of course, the mass of the spacecraft is negligible w.r.t. the mass of the Earth and Sun. Let us make some definitions
and reasonable approximations:
1. Definitions: since this equation / equality represents the SOI, we can define the radius of the SOI as the
position vector from the Earth to the satellite:

rLsat → rSOI

while instead:

rsats → rs

2. Approximations: since the mass of the bodies that we consider are much smaller w.r.t. the
mass of the Sun, the sphere of influence is small in the scale of the solar system and this will be
true whatever is the body that we consider). The SOI of the Earth will be reasonable smaller than
the Earth orbit around the Sun. So, we can write our assumptions where:
(a) The norm of the position vector Sun to Earth is approximately equal to the semi-major
axis of the Earth orbit around the Sun (remember that the Earth is on almost a circular orbit).
(b) The norm of the position vector Sun to spacecraft (since the SOI is small) is equal to the
semi-major axis of the Earth (same order of magnitude).
(c) The radius of SOI rSOI is much smaller than aL .
These are proximity assumptions (proximity of the spacecraft w.r.t. Earth even when we are at the
boundary of the SOI).
Based on these assumptions, we can simplify the equations. We have to understand who is rs − rLs .
Let us give a look at the geometry:

rsL + rLsat = rssat → −rLs + rSOI = −rs

where we have used that:

rsL = −rLs rLsat = rSOI rssat = −rs

So, we can write that:

rs − rLs = −rSOI

So, finally, the radius of the SOI depends on the ratio between the mass of the planet and the mass of the
Sun, and then there is a linear dependence with the semi-major axis of the planet around the Sun:
 2
mL 5
rSOI = aL
ms

If theoretically we have the Earth on the orbit of Jupiter, the SOI will be larger due to the fact that the gravitational
field of the Sun is changing. In fact, the smaller is the gravitational attraction of the Sun, so the larger is the volume
in which the gravity of the planet is dominant, and so this is the reason why we have the semi-major axis which
is the scale factor for the gravity of the Sun. The SOI of the Earth is of about one million of km which is 145
times the radius of the Earth. The SOI of the Moon is well within the one of the Earth.
Important consideration: when we consider our departure trajectory, we will consider the final velocity of the
hyperbola which is the velocity at the infinite and we use this velocity in our computation because actually we will
consider at the same way the velocity at the SOI of the planet and the velocity at the infinite w.r.t. the
planet. The difference between the two velocities will be not considered.
Space Flight Dynamics Michele Santarpia 136

Porkchop plots: any point on the porkchop plot is found by solving the Lambert’s problem. As we know, first of
all, we work on the heliocentric phase. In the heliocentric problem, we ignore the spheres of influence. Our idea is
that we have our reference frame with Sun in the origin, and we want to explore the possible departure dates and
arrival dates and the possible trajectories. As input, we have time at departure and time at arrival and so we know
where the Earth is at td and where the final planet is at ta (this is known from planet ephemeris). So, this is a
classic problem where we say that we want to stay in a given position at departure date, and at the arrival date we
want to stay in another position. This is our heliocentric Lambert’s problem.
 µs
r̈ssat = − 3 r
rssat ssat








 rssat (td ) = rL (td )






rssat (ta ) = rh (ta )


We are solving 2-body problem for the spacecraft around the Sun. We can solve this problem for any combination
of departure and arrival dates. This problem admits a couple of solutions referred as type 1 and type 2, where
the type of solution depends on which is the ∆ν on the transfer conic that we obtain. If we have a short
transfer (type 1), ∆ν < 180deg otherwise for a long transfer (type 2) ∆ν > 180deg. So, we are saying that
we will find two possible trajectories that will satisfy our constraints, the shorter and longer one. As we have our
transfer conic by solving the Lambert’s problem, we compute the velocity of the spacecraft required on the transfer
conic i.e. the velocity of the spacecraft at departure and the one at the arrival. Lambert will give us heliocentric
velocities, and we have to convert in the planet centric velocities, because finally to understand how to patch
departure and arrival we like to work with planet centric velocities. In order to obtain this conversion, it is a
problem of relative motion (pure kinematics). We can write the velocity of the spacecraft w.r.t. Earth as
the difference between the velocity at the departure that we obtain from Lambert (velocity w.r.t. the
Earth we need at the initial point of the transfer) and the velocity of the Earth at time of departure. The
same for the velocity of the satellite w.r.t. Mars that we need at the final point of the transfer. These velocities
are the ones that have to be patched with the departure hyperbola and arrival hyperbola. Conceptually,
these are the velocities at the spheres of influence and they are the hyperbolic excess velocities. To patch the
conics, we will consider that these are the ones at infinite. These are the input information to build our departure
because this will be the objective of our departure hyperbola and the second will be instead the velocity at infinite
w.r.t. Mars i.e. the one for the arrival hyperbola. Of course, the velocity of the planets are known by ephemeris.

v ∞L = v sat (td ) − v L (td )

v ∞h = v sat (ta ) − v h (ta )

Departure Conic:

v2 µL v2
ε= − = ∞L
2 r 2
Arrival Conic:
v2 µh v2
ε= − = ∞h
2 r 2
These velocities are the ones that we obtain from the porkchop plots. In these plots, we generally have v∞,h and
2
c3 = v∞,L = 2ε where ε is the energy on the departure hyperbola. We plot the quantity c3 because we have a link
with propulsion subsystem: in fact, when we consider the launcher vehicles possibilities plot, we have the
possible values of c3 that can be provided by a launcher as a function of the maximum injected mass. So, if we
have a launcher and a given spacecraft mass, one parameter that characterizes the launcher will be the available
c3 which is the final energy that the launcher can provide to the spacecraft. So, in the porkchop plot
we can investigate what is required c3 , and then we jump in our catalogue of launchers and we see which is
the feasible one. We will need to play with the spacecraft mass because larger is the spacecraft mass for the same
launcher, lower will be the specific energy that can be given: so, in order to obtain a larger final energy, we can
reduce the mass of the spacecraft. So, the idea of porkchop is that we enter with the time of departure and arrival
and we obtain a grid of possibilities in terms of c3 and v∞ h and also the transfer duration. For each point on
Space Flight Dynamics Michele Santarpia 137

the porkchop plot, a Lambert problem has been solved (heliocentric problem) from which we obtain the
required heliocentric departure velocity and the heliocentric arrival velocity and we have to switch to the planet
centric velocities from which we obtain for each point v∞,h and c3 (L) = v∞,L and so we can select which is the
best strategy.
In the Earth Mars 2018 opportunity, in bold we have the arrival velocity on Mars, while in grey we have the square
of the departure energy c3 . If we want, for example, depart in May and we want to arrive very fast, it was physically
possible but paying in terms of departure energy and arrival velocity. Keep in mind that we like to minimize both
quantities, because we want to minimize the departure energy because we need to check our launcher w.r.t.
our mass, but we also like to minimize the arrival velocity if our scope is to arrive on Mars because finally
we need to slow down at Mars otherwise if we do not slow down in some ways we will fly away from Mars, so the
larger is the arrival velocity more expensive will be our maneuver to stop at Mars. Finally, we see that
there are two regions that correspond to the long and short time transfers, so the two types of solutions that
we are interested in (and that we obtain from Lambert). There are some regions with local minima and, for
example, the point in black is the one of minimum energy which corresponds to an arrival velocity less
than 3km/s and a departure energy of about less than 8km2 /s2 , and there is also another interesting point
on the other side but that corresponds to larger numbers.
When we work on this plot, we are not making simplifying assumptions meaning that here Mars and Earth are
not coplanar and are not circular (we are not applying these approximations) but we know that they are almost
coplanar and circular and so for this reason the minimum point is the one that emulates Hohmann. So, the best
solution is the one that emulates Hohmann, so that point corresponds to the moment in which Earth
and Mars are in the right positions i.e. the right relative alignment.
Here, the synodic period plays an important role, because if we lose this opportunity, then the geometry will
not be the same and so in theory we can build the porkchop plot for 2019 but it was more expensive in that year
and also in 2020 because we do not have the same relative geometry (and so it is worst), but of course we can still
launch but we need to accept that with the same launcher we can transport less mass of the spacecraft because we
can give less energy.
We can play this game for any couple of planets, and when we do not have direct transfers, this can be split in a
number of direct transfers and so we can use Porkchop plots to build each set.

8.1.2 Departure Conic

Departure Conic: so, we have decided our transfer, we need to build the departure conic which is an hyperbola.
We need to compute the ∆v to jump on the heliocentric (interplanetary) transfer. The hyperbola is characterized
by two parameters. Let us build the hyperbola, we take the dotted arc which identifies where we have to take the
planet, then we draw a dotted line to identify the perigee of the departure hyperbola, let us plot the hyperbola
in red, let us identify -a and rp , the b parameter, and the distance ∆ from the asymptote, the segment semi-focal
distance c. Conceptually, we are on a circular parking orbit, from this parking orbit one impulse is needed to jump
on this hyperbola, that is given in a point that becomes instantaneously the perigee of the hyperbola, so we give
a tangential ∆v dep jumping instantaneously on the departure hyperbola. It is clear that this hyperbola
will have the objective to acquire a given velocity at infinite that is v ∞,L , and so we give the ∆v dep and after some
time (conceptually when we cross the sphere of influence) we will have the velocity v ∞,L and so we consider that
we have escaped the gravity of the Earth. As we have said, two parameters identified the hyperbola:
1. One is v ∞,L . We know that ε is related to v ∞,L , so if we give the velocity, we are setting the energy and so
the semi-major axis.
2. The other parameter is typically the radius at perigee. So, if we know v ∞,L and rp we know everything.
In fact, for this reason, we know everything of the hyperbola and this can be explained by these following
equations:

rp = a(1 − e) → e

1 δ
 e = sin( 2 ) →
√
δ
b = ∆ = |a| e2 − 1 → b

Space Flight Dynamics Michele Santarpia 138

So, the key parameters are v ∞,L and rp . If we know everything, we can compute our departure ∆v.
∆v dep = v hyp,dep − v park,dep
In norm:
µL
r
∆vdep = vhyp,dep −
rp
rp is selected in advance because it is the radius of the parking orbit. In order to know vhyp,dep , we can use the
energy equation. Finally we have:
µL µL µL
r r r
∆vdep = vhyp,dep − = 2 2
+ v∞,L −
rp rp rp
2
Larger is v∞,L , larger is the ∆vdep , because we need an hyperbola with higher energy.
Parking Orbit and Oberth Effect: typically the parking orbit is characterized by a low altitude, generally in
the order 200km. This means that we need to wait just a little time otherwise we decay, so there will be a small
time interval between the launch and the interplanetary ∆v. We like a low parking orbit! In fact, when we give the
interplanetary ∆v we are increasing the energy ε. The scope of the game is to give as much as possible energy to
our spacecraft. We are playing with the following equation in an Earth centric perspective:
v2 µL
ε= −
2 r
This is non linear, and in order to understand what happens we differentiate the equation:
dε ∼
= vdv
If we give a dv, the change of energy is vdv. What the launcher gives is dv, and to maximize the increase of energy,
we need to pay attention that dε is equal to vdv i.e. dv is multiplied by the velocity v. So, if we stay as low as
possible, we maximize the increase of energy for the same ∆v since the velocity v at lower orbits is
higher: this is the reason why we like to have a very low parking orbit. The effect through which by
giving the same ∆v we maximize the increase of energy if we stay on a low parking orbit is called Oberth effect
(effect to maximize the benefit of our ∆v).
Departure Hyperbolas: let us remember that we need to obtain a vector v ∞,L . If we draw the ECI, we want
to obtain finally v ∞,L , and the direction of this velocity will be given by two angles i.e. the right ascension
at launch (RLA) and declination at launch (DLA), that will be always the impulse for our maneuver. We
want to understand how many trajectories will give us the required tridimensional velocity. The idea
is that: conceptually we have to jump at infinite, which means after a couple of days, we need to stay with this
velocity. The interesting point is that there are infinite hyperbolas that give us this velocity vector, and these
infinite hyperbolas exhibit an axial symmetry w.r.t. the direction of v ∞,L . So, we have studied before the
problem in the orbital plane of one hyperbola, and we can imagine that the geometry of this plane can be seen as
a problem that is axial symmetric. So, any hyperbola, where the orbital plane is different but that exhibits the
same direction with delta distance, will be good. This can be seen in tridimensional drawing. So, all the black
trajectories are all hyperbola which have the same parameters but with different orbital planes. What we have is a
bell (a surface) of all the possible hyperbolas around the direction required velocity. Each hyperbola
has a different injection point, and all the injection points lie on a circle which is centered w.r.t. the direction
of the required velocity. The questions now are:
1. From the mission POV does it make any difference if we select an hyperbola or another one?
No, they are all the same.
2. So, what does determine the hyperbola to use? It is a matter of launch because each hyperbola has a
different inclination and we know that when we launch from a site we have a set of possible inclinations. So,
we have our required velocity, and we need to intersect the circle of all the possible injection points with
the set of the possible inclinations that can be acquired with the launch. Since we have an interval of possible
inclinations, there is only a set of possibilities, the circle of the injection points will have a given intersection
with our possible parking orbits, and so selecting our LS, we have that the possible parking orbits are
in this interval of min and max, and this will determine the possible injection points and so our departure
hyperbolas. So, we have a possible set of orbital planes for the parking orbits and from the other side
we have the circle which is the locus of points from which we can jump on hyperbola.
Space Flight Dynamics Michele Santarpia 139

Launch Vehicle Possibilities: for New Horizon which requires high c3 we used the most powerful launcher Atlas
V 511, with a mass injected of 500kg. c3 can be also negative (sounds strange because we have said that it is the
square of velocity, but we have to see it in terms of mechanical energy), where c3 = 2ε is twice the mechanical
energy we can achieve, and this is a plot about launch possibilities and so not necessarily we will be able to fly over
an hyperbola. For example, if we take a point negative c3 , c3 can be negative because if we fly an ellipse around
the Earth the energy is negative (on an elliptical orbit we can launch larger mass).

8.1.3 Arrival Conic

Arrival Phase: as input, from the design of the transfer we know that we want a given v ∞,h , and so this is now
the velocity by which we start the arrival phase and the other parameter that characterize the arrival hyperbola is
∆ which characterizes the geometry of the arrival w.r.t. Mars. We can plot the hyperbola w.r.t. Mars: we start
drawing the asymptotes with an angle δ, we have Mars as body, we can build our hyperbola in red from which we
are arriving on Mars, we can define the arrival velocity v ∞,h .
The important quantity to be considered / checked is rp because we may have different situations: if we do not do
anything, which means that the radius at perigee is not so low, and we do not apply any maneuver, this is not an
arrival trajectory, meaning that this is an open trajectory, it is a fly-by of Mars, we will arrive at the perigee, and
from the POV of Mars, if nothing happens we will fly away exactly with the same speed with which we arrived
because energy is the same. So, from the POV of Mars, if we do not maneuver, this is a pure change of
direction of the velocity. We have an initial velocity along the arrival asymptote and finally we have the arrival
velocity along the other asymptote. Of course, if we are making an Earth to Mars mission, we want to leave this
arrival hyperbola and so at a given moment we have to jump on a closed trajectory. We can have different options:
1. We circularize the orbit applying a single tangential negative ∆v at the perigee of the hyperbola.
2. We perform the circularization in different steps: so, first of all we need a closed orbit, so we generate an
ellipse applying a negative ∆v, and then we will apply subsequent negative ∆vs to reduce the dimensions
of the ellipses, so it is a multi-step strategy, where finally we have our final orbit. In this process, we
can try to take advantage from the atmosphere of the planet, we need to slow down and we can try to use
Drag and this is the idea of aerobraking, which means that Drag is used as support in order to save
propellant. When we say aerocapture, we have that the Drag is the key force that allows us to reduce the
energy w.r.t. Mars that becomes negative jumping on a closed orbit and we do not escape the planet. If we
want to use aerobraking or aerocapture, the geometry of the trajectory becomes of key important, so we need
to properly adjust the trajectory but in general this fine adjustment of the trajectory will be always needed
as we are entering in our arrival phase because we will need to pay attention to the ∆ parameter because we
want to approach with the correct geometry and this may require a fine adjustment of the trajectory in the
final phase.
Imagine that we want to circularize the orbit in a single ∆v given by the difference between the velocity around
Mars on the arrival parking orbit and the velocity on the arrival hyperbola:

µh µh
r r
∆vcirc = vpark − vhyp,rp = − 2 2 ,h
+ v∞
rp rp

How do we compute rp ?
We have to use the usual relations, if we start from v∞,h then we have that:

v∞,h → ε →√a

b = ∆ = |a| e2 − 1 → e

rp = a(1 − e) → rp


So, from this relation we get the perigee and we can check which is our geometry and we can compute our ∆v.
Which is the error that we make if we model a trajectory with a patched conics method?
Patched Conics vs High-Fidelity: in the figure, we can see the error that we have if we consider a realistic
transfer to Mars vs what happens with the patched conics method: the nice point is that it is true that in some
phase of the transfer we have a huge error but finally, when we consider our global perspective, our approximations
Space Flight Dynamics Michele Santarpia 140

do not imply a substantial difference of the trajectory and this confirms that the patched conics trajectory makes
sense. As initial design, the patched conics can make sense, so with porkchop plots and with considerations about
departure and arrival for basic mission analysis.

8.2 Gravity Assists: Leading Side Fly-By and Trailing Side Fly-By
Gravity Assists (’effetto fionda gravitazionale’): gravity assists are needed because they are the only way to
perform several missions otherwise they will be impossible. Gravity assist is a way to solve the problem showed in
the initial table i.e. the problem of the huge ∆vs and time of the transfer. Since some maneuvers are too expensive
in terms of ∆vs and they require so much time, we try to get these ∆vs for free that means exploiting the gravity
of some other bodies. In this problem, we need some synchronization because the celestial bodies follow their
orbits and so we need to carefully design the trajectory to exploit the gravity assist. The most massive
objects are the most useful ones, because we exploit the gravity of the planet: huge planets, huge mass and
huge gravity so they give us larger ∆vs . This explains why we typically like to use Jupiter for any type of mission
to the outer planets and we can use also the Earth as body to perform the gravity assist.
When we perform the gravity assist, we are making an hyperbolic fly-by, we arrive from infinite and we fly to
infinite. So, let us draw our hyperbola. We are arriving with a velocity v ∞,in w.r.t. the planet and we are flying
away with a velocity v ∞,out w.r.t. planet. We got a ∆v which gives us a pure rotation of the velocity and this
rotation is equal to δ. The ∆v ∞ is along the axis of the hyperbola, and it is towards the planet. During the fly-by,
the energy is of course constant and the energy at beginning is equal to the one at the end:

|v ∞,in | = |v ∞,out |

As we know:
2
v∞ µ
ε= =−
2 2a
 
µ 1
rp = a(1 − e) = − 2 1 −
v∞ sin 2δ

We are writing the periapsis as a function of the velocity at infinite and δ:

 
µ 1
rp = 2 −1
v∞ sin 2δ

If we perform:

2
v ∞,in · v ∞,out
v ∞,in · v ∞,out = v∞ cos δ → cos δ = 2
v∞

In this problem, which are the two independent parameters that we use?
The energy (we can also talk about the angular momentum) does not change but this happens in a planet centric
perspective, and so w.r.t. the planet we have a pure rotation of the velocity vector. When we switch to
a heliocentric perspective, the situation is different and we have to consider the velocity of the planet around the
Sun. We can make an approximation: we consider that during the fly-by the velocity of the planet around
the Sun is constant, and this corresponds to the fact that the fly-by is reasonably short (faster) w.r.t. the
orbital period of the planet.
In green, we plot the velocity of the planet around the Sun (remember that in general it is not along the axis of the
hyperbola). Assuming that it is constant, let us see which is the velocity of the spacecraft w.r.t. the Sun before and
after the fly-by. After the fly-by the velocity of the spacecraft is changed since it has undergone a pure rotation.
When we are in a heliocentric perspective, the velocity of the satellite w.r.t. the Sun is given by the sum of
the satellite velocity w.r.t. the planet and the velocity of the planet w.r.t. the Sun (we are adding the velocity
of the planet w.r.t. the Sun), the effect of the fly-by is not only to rotate the velocity but we are also
changing the norm and we can change the norm in a substantial way.
Space Flight Dynamics Michele Santarpia 141

Leading Side Fly-By A possible geometry is called leading side fly-by: this type of geometry is defined by
the direction of the velocity of the planet w.r.t. the Sun and by the arrival hyperbola. In fact, before sketching
the velocity of the planet w.r.t. the Sun, we do not know if the geometry is defining a leading or a trailing fly-by.
With a leading side fly-by, we are slowing down our spacecraft w.r.t. the Sun and this happens because we
flew in front of the planet. If we look at the mathematical expression, we can see how the ∆v sat,Sun is equal to the
planet centric ∆v ∞ . So, with our fly-by, we are applying this ∆v ∞ to the heliocentric velocity of our spacecraft.

Trailing Side Fly-By The other type of geometry is the one in which we are passing behind the planet (behind,
because we are considering that we are flying behind w.r.t. the motion of the planet) and it is called trailing side
fly-by, and in this case we are increasing the velocity of the satellite w.r.t. the Sun (we are accelerating
it increasing its norm). We are rotating the velocity but also increasing the norm.
The gravitational pull of the planet is giving us a positive ∆v while when we pass in front the gravitational pull
is giving us a negative ∆v. We can understand the phenomenon by working with a kinematic perspective which
is the one for which we are just rotating the velocity vector with a ∆v towards the planet in the planet centric
discussion and then when we consider the velocity of the planet w.r.t. the Sun it will become also a change of norm.
Now, we can understand that we can design our fly-by based on the ∆v that we want, so we want a given heliocentric
∆v and we can try to explore which is the fly-by that we need. What is defining our fly-by are v ∞,in and
v ∞,out . We are selecting which are the velocities that we need in order to have a given ∆v and then we will compute
everything. So, our input quantities are these two velocities and then we can compute everything because once we
have the vectors we have the norm and δ. There is a basically check that we have to perform in order to verify that
this transfer is feasible: we have to verify that we do not hit the planet i.e. we have to check the radius
at periapsis rp and this is the reason why we have written rp as a function of v∞ and δ because of course in this
geometry the planet is a point but if rp is lower than the radius of the planet or if there is an atmosphere, we will
not have a fly-by but as said by Fasano ”a glorious end of the mission!”.
 
µ 1
rp = a(1 − e) = − 2 1 −
v∞ sin 2δ

The key concept of all this discussion is that: when we want to obtain a given v ∞,out and we have the v pl,Sun , if
we work on a carefully design of v ∞,out , the final velocity of the spacecraft in a heliocentric perspective will be the
velocity of the planet w.r.t. the Sun v pl,Sun plus the velocity of the satellite w.r.t. the planet v ∞,out , if we imagine
that we are able to design different directions for v ∞,out we will obtain a possible final velocity which is one of the
yellow vectors in a heliocentric perspective.

8.3 Earth to Saturn Mission

How do we really obtain the hyperbolic fly-by that gives the gravity assist we need?
Earth to Saturn mission: if we have multi-step mission, this will go through numerical optimizations (due
to the high number of combinations). We can consider the case Earth-Saturn (we never fly directly from Earth
to Saturn) and we typically exploits Jupiter to perform the gravity assist. The problem is faced using the following
steps:
1. Earth-Jupiter Heliocentric Trajectory Design (Porkchop Plots).
2. Jupiter-Saturn Heliocentric Trajectory Design (Porkchop Plots).
3. Matching of Porkchop Plots
(a) Match Dates: ta(j)1 = td(j)2
(b) Math Velocities: v∞j1 = v∞j2
4. Find Best Solution: Minimize C3L and v∞Saturn .
During the mission, there are two steps consisting of two direct transfers: Earth to Jupiter, and Jupiter to Saturn.
The 1st point is that we can study the two direct transfers through the porkchop plots (a porkchop plot for the 1st
transfer and another one for the 2nd transfer). As we have the porkchop plots, we need to match the two transfers
i.e. we have to perform a matching at Jupiter level. We need to have same dates which means that the
time of arrival at Jupiter that comes from the 1st porkchop plot taj has to be equal to the time of departure
Space Flight Dynamics Michele Santarpia 142

from Jupiter tdj that we obtain from the 2nd porkchop plot. Moreover, we need also to have the same velocities
at Jupiter: so, the arrival velocity on the conic Earth to Jupiter has to be equal to the departure velocity on the
conic Jupiter to Saturn. If we perform this matching, we will find what we want i.e. a number of combinations
where when we arrive at Jupiter only the direction of the velocity changes i.e. the norm of the velocity is the same.
In general, we will have a discrete number of possibilities (so not only one) and we need to select the best
one and, to do this, we will select the possible couples of transfers which will minimize the departure
energy from the Earth (C3 value Earth to Jupiter, minimum energy at launch) and the arrival velocity on
Saturn (v∞saturn Jupiter to Saturn, minimum energy at the arrival).
For multiple assists mission, we need to perform a numerical optimization.
Turning angle is the angle between the asymptotes. The nice point to focus on is the velocity before and after the
gravity assist with Jupiter: we have obtained a change of about 11km/s of the heliocentric velocity! Since
the problem is tridimensional, we can also change the inclination through a tridimensional ∆v and in fact we can
find in the image an heliocentric inclination change.
Cassini mission is an example of mission to Saturn where there have been different fly-by with the Earth and Venus.
The ∆vs obtained through Gravity Assists are not truly free: nothing is taken for free of course, the energy
has to be kept constant. The ∆v is not truly free: so, how is it possible to obtain this ∆v? We are getting
this free ∆v from the POV of the spacecraft. From the POV of the solar system, the energy is constant, because,
when we accelerate, we take energy from the planet, so the energy is not constant for the spacecraft which
is stealing energy from the planet and the planet is slowing down and so it loses energy. So, the balance is still
zero (transfer of energy from Jupiter to the spacecraft). Of course, the lose of energy and so reduction of velocity
for Jupiter is not visible since the masses are so different that the orbit of Jupiter does not change. Theoretically
speaking, this is what happens. When we have a leading side fly-by, the situation is opposite, the spacecraft slows
down and it gives energy to the planet which is accelerating.