OUTLINE : 24.

LAGRANGE MECHANICS

Classical Mechanics 23.5 Example 3 : an aircraft landing

LECTURE 24: 24.1 Lagrangian mechanics : Introduction

LAGRANGE MECHANICS 24.2 Introductory example : the energy method for the E of M

Prof. N. Harnew 24.3 Becoming familiar with the jargon

24.3.1 Generalised coordinates
University of Oxford 24.3.2 Degrees of Freedom
HT 2017 24.3.3 Constraints
24.3.4 Configuration Space

1 2

23.5 Example 3 : an aircraft landing An aircraft landing continued

The landing wheel of an aircraft may be approximated as a uniform
I Torque |τ | = |r × F| = aµM 0 g
M
circular disk of diameter 1 m and mass 200 kg. The total mass of the about O, where M 0 = 20 (ie. 10
aircraft including that of the 10 wheels is 100, 000 kg. When landing wheels, supporting 50% of mass)
the touch-down speed is 50 ms−1 . Assume that the wheels support I Angular momentum J = Iω
50% of the total weight of the aircraft.
where I = 21 ma2 (MoI of solid
Determine the time duration of wheel-slip if the coefficient of friction
disk where m is mass of a wheel)
between the wheels and the ground is 0.5, assuming that the speed
dJ
of the plane is not changed significantly. I τ= dt = I dω
dt where u = a ω
u is the speed of the wheel rim
I du
→τ = = aµM 0 g
a dt
R t 2 0g Rv
I Integrate : 0f a µM
I dt = 0 0 du
v0 is the speed of the aeroplane
v0 I v0 m
→ tf = a µM 0 g
2 = 2µM 0 g
50×200
I Putting in numbers: tf = 5 ∼ 0.2 s
2×0.5× 1×10
20
×9.8
3 4
 An aircraft landing continued An aircraft landing continued
Confirm the assumption that the speed of the plane is not Calculate the work done during wheel slip.
changed by calculating the speed at the end of wheel-slip in the
R
I Work done : W = τ dθ = τ θf
absence of other braking processes. (τ is constant, θf is the total turning
angle)
I Energy expended in getting the wheels 2µM g 0

up to speed: Ewheels = 21 Iω 2 × 10 (ie.

I From before : u = a dθ
dt = m t
R tf 2µM 0 g
10 wheels) = 25 mv02
R θf
I Integrate : 0 a dθ = 0 m t dt
[ Remember v0 = aω , I = 21 ma2 ] µM 0 g 2 v0 m
→ a θf = m tf where tf = 2µM 0 g
I Total energy of the aeroplane
E = 12 Mv 2
→ Energy loss: δE = Mv δv = 52 mv02
I Putting it all together:
5
mv02 5 M 0 gµ v0 m 2 1 v0 2
→ δv = 2
→ δv
= 2
m I W = τ θf = a1 (aµM 0 g)( )( ) = 2 I( a )
Mv0 v0 M | {z } m 2µM 0 g
5
×200 τ | {z }
δv
I Putting in numbers : v0 = 2
1×105
θf

→ a 0.5% effect I Hence W = Ewheel = 12 I ω 2 as expected !

5 6

Lagrange and Hamilton 24.1 Lagrangian mechanics : Introduction

I Lagrangian Mechanics: a very effective way to find the

equations of motion for complicated dynamical systems
using a scalar treatment
→ Newton’s laws are vector relations. The Lagrangian is a
single scalar function of the system variables
I Avoid the concept of force
→ For complicated situations, it may be hard to identify all
the forces, especially if there are constraints
I The Lagrangian treatment provides a framework for
relating conservation laws to symmetry
I The ideas may be extended to most areas of fundamental
I Joseph-Louis Lagrange (1736-1810) physics (special and general relativity, electromagnetism,
I Sir William Rowan Hamilton (1805-1865) quantum mechanics, quantum field theory .... )

7 8
24.2 Introductory example : the energy method for the 24.3 Becoming familiar with the jargon
E of M

I For conservative forces in 1D motion : 24.3.1 Generalised coordinates

Energy of system : E = 12 mẋ 2 + U(x) [Note dE
dt = 0] A set of parameters qk (t) that specifies the system
I Differentiate wrt time: mẋ ẍ + ∂U
ẋ = 0 configuration. qk may be a geometrical parameter, x, y , z,
∂x
a set of angles · · · etc
→ m ẍ = − ∂U
∂x = F
→ This is the E of M for a conservative force
I Take a simple 1d spring undergoing SHM : 24.3.2 Degrees of Freedom
1 2 1 2
E= 2 mẋ + 2 kx = constant
The number of degrees of freedom is the number of
dE
dt = 0 → mẋ ẍ + kx ẋ = 0 independent coordinates that is sufficient to describe the
→ mẍ + kx = 0 configuration of the system uniquely.
I Hence we derived the E of M without using NII directly.

9 10

Examples of degrees of freedom

24.3.3 Constraints

I A system has constraints if its components are not

permitted to move freely in 3-D. For example :
→ A particle on a table is restricted to move in 2-D
→ A mass on a simple pendulum is restricted to oscillate
I Pendulum whose pivot can at an angle θ at a fixed length ` from a pivot
I Ball rolling down an incline move freely in x direction
I The constraints are Holonomic if :
E = 12 mẋ 2 + 12 Iφ̇2 − mgx sin α I Pivot coordinates : (x, 0)
→ The constraints are time independent
I But ẋ = R φ̇ → x = Rφ I Pendulum coordinates :
→ The system can be described by relations between
I The problem is reduced to (x + ` sin θ, −` cos θ)
general coordinate variables and time
a 1-coordinate variable E= 1
m ẋ 2 + 12 m2

d
(x +
2
` sin θ) +
2 1 dt
2 → The number of general coordinates is reduced
q1 ≡ x and q̇1 ≡ ẋ

+ 12 m2 d
(−` cos θ) − mg` cos θ
dt to the number of degrees of freedom
I System has only 1 degree I This system has 2 degrees
of freedom : x of freedom : x and θ
11 12
 24.3.4 Configuration Space
Classical Mechanics
I The configuration space of a mechanical system is an
n-dimensional space whose points determine the spatial position
of the system in time. This space is parametrized by generalized LECTURE 25:
coordinates, q = (q1 · · · , qn )
I Example 1. A point in space determines where the system is;
the coordinates are simply standard Euclidean coordinates:
THE LAGRANGE EQUATION
(x, y, z) = (q1 , q2 , q3 )
I Example 2. A rod location x, angle θ - as it moves in 2D space is DERIVED VIA THE
passes through points (x, θ) in the configuration space
CALCULUS OF VARIATIONS
Prof. N. Harnew
University of Oxford
HT 2017
13 1

25.1 The Lagrangian : simplest illustration

OUTLINE : 25. THE LAGRANGE EQUATION DERIVED
The Lagrangian : L = T − U
VIA THE CALCULUS OF VARIATIONS
I In 1D : Kinetic energy T = 12 mẋ 2 No explicit dependence on x
Potential energy U = U(x) No explicit dependence on ẋ
25.1 The Lagrangian : simplest illustration
I Define the Lagrangian in 1D : L = 12 mẋ 2 − U(x)
∂L ∂L
25.2 A formal derivation of the Lagrange Equation
I
∂ ẋ
= mẋ and = − ∂U
∂x gives
∂x force F

The calculus of variations d ∂L
I Differentiate wrt time : dt ∂ ẋ
= mẍ = F
I Hence we get
 the Euler - Lagrange equation for x :
25.3 A sanity check d ∂L ∂L
dt ∂ ẋ = ∂x
I Now generalize : the Lagrangian becomes a function of 2n
25.4 Fermat’s Principle & Snell’s Law
variables (n is the dimension of the configuration space).
Variables are the positions and velocities
25.5 Hamilton’s principle (Principle of Stationary Action) L(q1 , · · · , qn , q̇1 , · · · , q̇n )
 
d ∂L ∂L
dt ∂ q̇k
= ∂qk
Next we expand on this concept.
2 3
 The calculus of variations, continued (1)
25.2 The calculus of variations
I We assume the unknown function f is a continuously
I Take 2 points A(x0 , y0 ) and B(x1 , y1 ) differentiable scalar function, and the functional to be minimized
I Curve joining them is represented by depends on y (x) and at most upon its first derivative y 0 (x).
equation y = y(x) such that y (x) I We then wish to find the stationaryR values of the path between
satisfies the boundary conditions : x
points: an integral of the form I = x01 f (y, y 0 , x)dx
→ y (x0 ) = y0 , y (x1 ) = y1
→ f (y, y 0 , x) is a function of x, y and y 0 (the first derivative of y )
I We want to find the function
y = y(x) subject to the above
I Consider a small change δy (x) in the
conditions which makes the closest function y (x) subject to the conditions that
path between the points a minimum. the endpoints are unchanged :

(note that this differs from what we → δy(x0 ) = 0 and δy(x1 ) = 0

are used to. We are not minimizing a I To first order, the variation in f (y, y 0 , x) is
∂f ∂f 0 2 02
set of variables here but a function). δf = ∂y δy + ∂y 0 δy + O(δy , δy )
d
where δy 0 = dx δy
I This is the calculus of variations. A branch of mathematics
I Thus the variation in the integral I is
that deals with functionals as opposed to functions. R x h ∂f i
∂f d
δI = x01 ∂y δy + ∂y 0 dx δy dx
4 5

The calculus of variations, continued (3)

The
h
calculus iof variations, continued (2) I So far we have used x as the independent
R x1 ∂f ∂f d
I δI = x0 ∂y δy + ∂y 0 dx δy dx variable with a functional f which is a
function of (y (x), y 0 , x)
I Integrate the second term by parts
h ix1 R
x1 d
  I Throughout we could have used other
∂f ∂f
R 2nd
term = ∂y 0 δy − x0 dx ∂y 0 δy dx variables, in particular time t and
x0
h i x1 generalized coordinates q1 , · · · , qn and
∂f
The ∂y 0
δy term = 0 due to the derivatives q̇1 , · · · , q̇n . The principles would
x0
conditions on the end points have been the same.
I Hence h I The integral
 i Rt
I = t01 f [q1 (t), · · · , qn (t), q̇1 (t), · · · , q̇n (t)] dt
R x ∂f d ∂f
δI = x01 ∂y − dx ∂y 0 δy dx
must be stationary wrt variations in any one
I For I to be stationary, δI = 0 for any
& all of the variables q1 (t), · · · , qn (t) subject
small arbitrary variation δy (x)
to the conditions δqi (t0 ) = δqi (t1 ) = 0
I This is only possible if the integrand vanishes identically I We get the n Euler-Lagrange equations for i = 1, · · · , n
I Hence we get out the Euler-Lagrange Equation ∂f d

∂f

  ∂qi
− dt ∂ q̇i
=0
∂f d ∂f
∂y
− dx ∂y 0
=0 The E-L equations give the conditions for the closest path between points
6 7
 25.3 A sanity check 25.4 Fermat’s Principle & Snell’s Law
The shortest distance between 2 points. Fermat : The actual path that a light ray propagating between one
point to another will take is the one that makes the time travelled
I Consider 2 neighboring points on the curve y (x) subject to between the two points stationary.
boundary conditions y(x0 ) = y0 , y (x1 ) = y1
p Question: at which point (x, 0) will the ray hit the interface between
I Distance between the points d` = dx 2 + dy 2 the two media to propagate from A to B?
p p
I d` = 1 + y 02 dx f ≡ 1 + y 02
  I Time taken from A to B :
∂f d ∂f
I The Euler-Lagrange Equation ∂y − dx ∂y 0 = 0 1
1 1
1
(x − x1 )2 + y12 2 + (x2 − x)2 + y22 2
 
t(x) = v1 v2
 
∂f ∂f √ y0 d ∂f
I
∂y = 0 , ∂y 0 = 02
, dx ∂y 0 = 0
I The Euler-Lagrange Equation
1+y  
∂t d ∂t
y0 − ∂x 0
=0 (where the second term = 0)
= constant, hence y 0 is constant
∂x dy
I Hence √
1+y 02 I ∂t
=0= 1 x−x1
− 1 x2 −x
∂x v1 1 v2 1
→ y = mx + c [(x−x1 )2 +y12 ] 2 [(x2 −x)2 +y22 ] 2
sin θ1 sin θ2
I We have proved that the shortest distance between 2 I
v1 − v2 =0
points is a straight line !
n1 sin θ1 = n2 sin θ2 Snell’s Law
8 9

25.5 Hamilton’s principle (Principle of Stationary Action)

I Consider for example a particle of mass m at point (xA , yA )

Classical Mechanics
moving under the influence of a force in the x − y plane.
We want to find the path that the particle will follow to LECTURE 26:
reach a point (xB , yB ).
I Hamilton’s principle: the path that the particle will take from
A to B is the one that makes the following functional
THE LAGRANGE EQUATION
stationary :
RB
I = A L (q1 (t), · · · , qn (t), q̇1 (t), · · · , q̇n (t)) dt
EXAMPLES
where L is the Lagrangian, I is called the action integral Prof. N. Harnew
I Hence the action integral I is stationary under arbitrary
variations q1 (t), q2 (t) · · · which vanish at the limits of
University of Oxford
integration ie. A and B. HT 2017

10 1
 OUTLINE : 26. THE LAGRANGE EQUATION : 26.1 Conjugate momentum and cyclic coordinates
EXAMPLES d

∂L

∂L
I The E-L equation is dt ∂ q̇k
= ∂qk with L = T − U
∂L
I Define conjugate (generalized) momentum : pk = ∂ q̇k
Note this is not necessarily linear momentum !
26.1 Conjugate momentum and cyclic coordinates
→ eg. simple pendulum L = 12 m`2 θ̇2 + mg` cos θ
∂L
→ ∂ θ̇
= m`2 θ̇ : which is angular momentum
26.2 Example : rotating bead
∂L
I Following on, E-L equation reads ṗk = ∂qk
I If the Lagrangian L does not explicitly depend on qk , the
26.3 Example : simple pendulum
coordinate qk is called cyclic or ignorable
26.3.1 Dealing with forces of constraint
26.3.2 The Lagrange multiplier method I With no qk dependence :
∂L ∂L
∂qk = 0 and pk = ∂ q̇k
= constant
The momentum conjugate to a cyclic coordinate is
a constant of motion
2 3

26.2 Example : rotating bead Example : rotating bead continued

A bead slides on a wire rotating at constant angular speed ω in
What happens if the angular speed is now a free coordinate ?
a horizontal plane
I L = 12 mṙ 2 + 12 mr 2 θ̇2
I Polar coordinates v = ṙ r̂ + r θ̇θ̂
I Two variables qk → r , θ
I L = T − U with U = 0
I r variable: as before
I L = 12 mṙ 2 + 12 mr 2 ω 2
→ mr̈ − mr θ̇2 = 0
I Single variable qk → r
d ∂L ∂L
d
 
∂L
I θ variable: dt = ∂θ
I E-L dt ∂ ṙ
= ∂L
∂ ṙ
∂ θ̇
∂L
  I
∂ θ̇
= mr 2 θ̇
∂L d ∂L
∂ ṙ
= mṙ → dt ∂ ṙ
= mr̈ ∂L
I
∂θ =0
∂L
∂r = mr ω 2 d
 
I E-L : mr 2 θ̈ = dt mr 2 θ̇ = 0
I E-L → mr̈ − mr ω 2 = 0
→ Conservation of angular
Central force Fcentral = mω 2 r
momentum
I r = Aeωt + Be−ωt
4 5
 26.3.1 Dealing with forces of constraint
26.3 Example : simple pendulum For the simple pendulum using Euler-Lagrange equation. The
Evaluate simple pendulum using Euler-Lagrange equation method did not get the tension in the string since ` was constrained.
If we need to find the string tension, we need to include the radial
I Single variable qk → θ term into the Lagrangian and to include a potential function to
I v = ` θ̇ represent the tension:
I T = 12 m`2 θ̇2
I ` → r , add 12 mṙ 2 , add V (r )
I U = −mg` cos θ
I L = 12 mṙ 2 + 12 mr 2 θ̇2 + mgr cos θ − V (r )
I L = T − U = 12 m`2 θ̇2 + mg` cos θ ∂L
= mṙ → dtd ∂L


  I
∂ ṙ ∂ ṙ
= mr̈
∂L
m` 2 θ̇ → d ∂L 2
= dt ∂ θ̇ = m` θ̈
I ∂L ∂V (r )
2
∂r = mr θ̇ + mg cos θ − ∂r
∂ θ̇ I

∂L
I
∂θ = −mg` sin θ I −
∂V (r )
∂r = (−T ) with T in the −r̂ dirn .
I E-L → m`2 θ̈ + mg` sin θ = 0 I E-L → mr̈ = mr θ̇2 + mg cos θ − T
g Reintroduce r̈ = 0 and r = `; v = r θ̇
→ θ̈ + ` sin θ = 0 I

mv 2
This is great, but note that the method T − mg cos θ
= |{z} as expected from NII
does not get the tension in the string r } | {z }
| {z Tension Weight
since ` is a constraint (see next slide). Centripetal force
6 7

26.3.2 The Lagrange multiplier method

An alternative method of dealing with constraints.
Back to the simple pendulum using Euler-Lagrange equation · · ·
Before : single variable qk → θ. This time take TWO variables x, y but
Classical Mechanics
introduce a constraint into the equation. L=T −U

I L0 = 12 m(ẋ 2 + ẏ 2 )+mgy + 12 λ(x 2 +y 2 −`2 )

LECTURE 27:
λ is the Lagrange multiplier
I d
 0
∂L
= ∂L
0
(including λ)
MORE LAGRANGE
dt ∂ q̇i ∂qi

x coord. → mẍ = λx (1) EXAMPLES

y coord. → mÿ = mg + λy (2)
λ coord. → x2 + y2 − `2 = 0 (3) Prof. N. Harnew
(which reproduces the constraint) University of Oxford
Comparing with Newton II : mẍ = − Tx ` ; mÿ = mg − Ty` . HT 2017
We see from the NII approach the Lagrange multiplier λ is
proportional to the string tension λ = − T` and introduces force
8 1
 27.1 The Lagrangian in various coordinate systems
OUTLINE : 27. MORE LAGRANGE EXAMPLES
I Cartesian coordinates

L = 12 m(ẋ 2 + ẏ 2 + ż 2 ) − U(x, y , z)
Already shown that E-L gives

27.1 The Lagrangian in various coordinate systems mẍ = − ∂U ∂U ∂U

∂x ; mÿ = − ∂y ; mz̈ = − ∂z
→ m r̈ = −OU

I Cylindrical coordinates
27.2 Example 1: the rotating bead
x = r cos φ ; ẋ = ṙ cos φ − r sin φ φ̇
y = r sin φ ; ẏ = ṙ sin φ + r cos φ φ̇
z = z ; ż = ż
27.3 Example 2: bead on rotating hoop
T = 21 m (ẋ 2 + ẏ 2 + ż 2 )
φ is cyclic if U = U(r ) only
= 12 m[(r 2 cos2 φ + r 2 sin2 φφ̇2 − 2r ṙ cos φ sin φφ̇)] ∂L
+(r 2 sin2 φ + r 2 cos2 φφ̇2 + 2r ṙ cos φ sin φφ̇) + ż 2 ]
→ pφ = ∂ φ̇
= mr 2 φ̇

L = 12 m(ṙ 2 + r 2 φ̇2 + ż 2 ) − U(r , φ, z) → constant angular

momentum
2 3

The Lagrangian in various coordinate systems continued

27.2 Example 1: the rotating bead
I Spherical coordinates
A bead of mass m slides on a frictionless wire which rotates about a
x = r sin θ cos φ ; ẋ = ṙ sin θ cos φ + r cos θ cos φ θ̇ − r sin θ sin φ φ̇ vertical axis at an angular velocity ω. The wire is tilted away from the
y = r sin θ sin φ ; ẏ = ṙ sin θ sin φ + r cos θ sin φ θ̇ + r sin θ cos φ φ̇ vertical by an angle α. Describe the motion of the bead.
z = r cos θ ; ż = ṙ cos θ − r sin θ θ̇
I Use spherical coordinates
T = 12 m (ẋ 2 + ẏ 2 + ż 2 ) I From before
 : 
T = 12 m Ṙ 2 + R 2 α̇2 + (R sin α)2 φ̇2
T = 21 m (ṙ 2 + r 2 θ̇2 + (r sin θ)2 φ̇2 )
I But α̇ = 0 , φ̇ = ω = constant
+ cross terms which all sum to zero  
1 2 2 2 2
I T = 2 m Ṙ + R ω sin α
L = 12 m(ṙ 2 + r 2 θ̇2 + (r sin θ)2 φ̇2 ) − U(r , θ, φ) U = mgR cos α (Take U = 0 at R = 0)
I L=T −U
 
L = 12 m Ṙ 2 + R 2 ω 2 sin2 α − mgR cos α

4 5
 The rotating bead, continued
27.3 Example 2: bead on rotating hoop
 
I L = 12 m Ṙ 2 + R 2 ω 2 sin2 α − mgR cos α
A vertical circular hoop of radius R rotates about a vertical axis at a
I Single generalized coordinate R constant angular velocity ω. A bead of mass m can slide on the hoop
I E-L equation: without friction. Describe the motion of the bead.
Use spherical coordinates again
  I
d ∂L ∂L
dt = ∂R
∂ Ṙ
I From before
 :
∂L
= mR ω 2 sin2 α − mg cos α 
∂R
    T = 12 m Ṙ 2 + R 2 θ̇2 + (R sin θ)2 φ̇2
d ∂L d
dt = dt mṘ = mR̈
∂ Ṙ I But Ṙ = 0 , φ̇ = ω = constant
I R̈ − R ω 2 sin2 α = −g cos α T = 12 m(R 2 θ̇2 + (R sin θ)2 ω 2 ) (NB. Ṙ = 0)
I Solution : R = Ae−λt + Be+λt + R0 where λ = ω sin α I U = −mgR cos θ (U = 0 at θ = 90◦ )
g cos α
[ P.I. → R̈ = 0 → R0 = ] I L=T −U
ω 2 sin2 α
If Ṙ = 0 at t = 0 , A = B ; then if R = R1 at t = 0, A = B = 21 (R1 − R0 ) L = 12 m(R 2 θ̇2 + (R sin θ)2 ω 2 ) + mgR cos θ

I If R̈ = 0 → R = R0 = g cos α
→ circular motion One single generalized coordinate : θ
ω 2 sin2 α
6 7

Bead on rotating hoop, continued

L = 12 m(R 2 θ̇2 + R 2 sin2 θ ω 2 ) + mgR cos θ
 
d ∂L
I E-L equation: dt ∂ θ̇
= ∂L
∂θ
   
d ∂L d 2 2
dt ∂ θ̇ = dt m R θ̇ = m R θ̈
I

∂L
∂θ = m R 2 sin θ cos θ ω 2 − mgR sin θ
→ θ̈ = sin θ cos θ ω 2 − Rg sin θ
→ θ̈ + ω02 − ω 2 cos θ sin θ = 0


g
where ω02 = R
I If ω = 0, θ̈ + ω02 sin θ = 0 → SHM, back to pendulum formula
I If ω 6= 0, for equilibrium, a necessary condition : θ̈ = 0
→ θ = 0 (stable equilibrium provided ω 2 R ≤ g ),
→ θ = π (unstable equilibrium)
ω2
→ cos θ = ω02 = ωg2 R
(stable equilibrium about a circle provided ω 2 R ≥ g )
8