Thiel 2/2/98

Physics 426

More on Geometrical Optics

Hecht: Chapter 6

I. Introduction—During the previous chapter, a simple model of the

imaging properties of optical systems was introduced. As in all of
physics, once a subject is grasped with a conceptual model, the
simplifying assumptions that limit the model’s usefulness are relaxed.
In this chapter the assumptions of “thin lenses” and paraxial rays will
be examined.

II. Thick Lenses—In order to deal with more complex optical systems, we
wish to develop a general method of analysis that will allow us to
reduce an optical system to a set of parameters that characterize its
effect on incident rays. This allows us to treat an optical system as a
“black box” that maps each point in object space onto a single point
in image space. As a starting point for developing this method of
characterization, we will examine a “thick lens” in which rays may be
significantly displaced from their entering position as they traverse
the lens.

Object Space Image Space

Optical System

(Thick lens,
series of thin
lenses, mirrors,
etc.)

The Six Cardinal Points—The traditional method of analyzing an optical

system is to locate six characteristic points known as the cardinal points.
These points are the front focal length (FFL), the back focal length (BFL),
the two principal points, and the two nodal points. The effective focal
length (EFL) and location of the vertices may be derived from these points.

1
 Thiel 2/2/98
Physics 426

1.) First, consider the points where parallel rays are focused. The distance
from the image side of the lens (the point where the lens surface crosses
the optical axis, sometimes called the vertex) to the focal point in image
space is the back focal length. The distance from the object side of the
lens to the focal point in object space is the front focal length.

2.) Now, consider points where the entering and

exiting rays from (1) would cross if extended.
These points will all lie on the two principal
planes of the system (actually, these points
will lie on the surface of a sphere centered on
the focal point, but we can approximate this
as a plane in the paraxial region). The
principal points are the points where the
principal planes cross the optical axis.
Notice that we can replace the complicated
behavior of the ray within the system with a
single bend at the principal plane without
altering the ray path outside of the lens.

Location of the
second principal
plane for a lens

P2

2
 Thiel 2/2/98
Physics 426

3.) Next, consider a ray that passes through the optical center of the lens.
We know from the inversion symmetry about the optical center that
the exiting ray must travel at the same angle as the incident ray, but
the rays will be displaced from each other. If we extend these two
rays, the points where they cross the optical axis are the two nodal
points. These nodal points will coincide with the principal points if
the index of refraction is the same in both object and image space.
Notice that since rays directed toward the nodal points exit at the
same angle, the lens may be rotated about either nodal point without
altering the image.

Ray Tracing—We will now utilize the cardinal points to trace rays through
the thick lens just as was done for the thin lens. In particular, we
would like to find a simple correspondence between the thick lens
and the thin lens equations. This may be done through the following
method.

1.) The first step is to trace rays as if all the bending occurred at the
principal planes. This implies that the nodal space (the space between
the principal planes) is ignored when tracing rays; all rays are translated
parallel to the optical axis between the principal planes. This is
equivalent to treating the system as a thin lens by acting as if the
principal planes were coincident.

3
 Thiel 2/2/98
Physics 426

2.) The effective focal length (f) is defined as the distance from either
focal point to the nearest principal plane. This distance is the same for
both sides of the lens (when the index is the same on both sides) and
corresponds to the focal length in thin lens theory.

3.) All distances are measured from the nearest principal plane; in
object(image) space distance is measured from the first(second)
principal plane.

These rules allow us to use the thin lens equation to calculate the image
1 1 1
position: + = (1)
so si f

We may also derive a thick lens version of the lens maker’s equation:
1 1 1 d ( n − 1) 
= ( n − 1) − + (2)
f  R1 R2 nR1 R2 

where d is the thickness of the lens and all the other quantities are as in the
thin lens version. Once the effective focal length is known, the position of
the principal planes may be calculated:
− fd ( n − 1) − fd ( n − 1)
h1 = h2 = (3)
nR2 nR1

where h1 and h2 are the distances to the right of the corresponding lens
vertex (if both planes were inside the lens, h1 would be positive and h2
would be negative). Notice that the three “easy to trace” rays may still be
used to locate images.

4
 Thiel 2/2/98
Physics 426

This method of analysis may be generalized to include all optical systems.

Thus, if we can find the location of the six cardinal points for any arbitrary
optical system, we may trace rays without the necessity of considering
individual components of the system—the system is reduced to a “black
box” (or, more appropriately, the system is reduced to a single thin lens)
that transforms object space into image space.

1.) The results of this analysis are particularly simple for a system composed
of two thin lenses separated by a distance d. Notice that these results
also hold for thick lenses if d is the distance between principal planes.

1 1 1 d fd fd
= + − h1 = h2 = − (4)
f f1 f 2 f1 f 2 f2 f1

III. Matrix Methods—We wish to develop a systematic approach to ray

tracing that is easily adaptable to computerized computation. The
linear form of the paraxial version of Snell’s law suggests the use of
matrix methods to represent the effect of an optical system upon an
incident ray (we may view an optical system as a linear
transformation that “maps” object space onto image space).

5
 Thiel 2/2/98
Physics 426

1.) In order to proceed, we must develop a general expression for refraction

of an arbitrary ray at a spherical surface and for translation of a ray
traveling through a medium of index n. For the sake of simplicity, we
will only consider meridional rays (rays which entirely lie in a plane
containing the optical axis) so that the problem is effectively reduced to
two dimensions. Let us examine refraction first, using the notation in the
diagram above.

ni sin θi = nt sin θt  → niθi = ntθt

paraxial
where θi = α i + α1 ; θt = α i + α1 (5)

y1
but, = sin α1 ≈ α1 from the central triangle with vertex C1 (6)
R1
y1 y  n − ni 
ni + niα i = nt 1 + ntα t 
→ ntθt = niα i −  t  y1 (7)
R1 R1  R1 
 nt − ni  1
and, from the lens maker’s equation,   = = D , where D is the (8)
 R1  f
power of the surface measured in diopters (m-1). So, in terms of the power,

⇒ ntαt = niαi − Dy1 (9)

This is the refraction equation for a ray incident at a height y1 and an angle
αi from the optical axis upon a surface with power D. This gives αt, the
transmitted angle.

6
 Thiel 2/2/98
Physics 426

Now, in order to represent an undeviated ray as a linear equation, we need

to find the slope of the ray and use the standard slope-intercept equation.
∆y
slope = = tan α t ≈ α t (10)
∆x

⇒ y2 = αt d + y1 (11)

This is the transfer equation which translates a ray a distance d along the
optical axis from y1. We are now ready to place this system of equations
into matrix form with use of the identities yi=yt and ni2αi2=nt1αt1. (12)

 ntα t   1 − D  niα i  ! !
  =  → rt = Rri where R is the Refraction Matrix (13)
 yt   0 1   y i 

 ni 2α i 2   d1 0  n α 
 t1 t1 → r! = Tr!
 = 1  y   where T is the Transfer Matrix (14)
 y i 2   n
i2 t1
t  t1

We may now trace a meridional ray through an optical system by repeatedly

applying R and T matrices to the original ray vector. The system matrix (A)
is the product of all the R and T matrices for an optical system (from the
first vertex to the last vertex). If we know the system matrix, we may trace
any meridional ray from object space to image space by applying the
appropriate transfer matrices to the system matrix: r!i = Ti ATo r!o (15)

2.) We may find the location of the cardinal points for an optical system
 a b
from knowledge of the system matrix. A =  
 c d
1 no (a − 1) ni (1 − d )
f =− ; h1 = ; h2 = (16)
b b b

3.) The same analysis that we made for a lens may be done for any optical
element. A table of such matrices from Yariv is given below. Notice
that the arbitrariness of the form of the ray vector causes a problem here
since Yariv uses a more common form than Hecht. The ray vector that

7
 Thiel 2/2/98
Physics 426

!  y
Yariv uses is r =   , so we need
α 
to apply a similarity
transformation to Yariv’s notation
to convert it to Hecht’s.
!  0 ni  Yariv  01 1
 ! Hecht (17)
ri Hecht =  A  0 ro
1 0 n 
o

5.) One of the most important uses

of this form of matrix optics is the
ABCD law. This law is used to
calculate beam waists and radii of
curvature for Gaussian laser
beams. If we define a complex
characteristic function (q) of the
Gaussian beam through the
1 1 λ
relation = −i ,
q( z ) R( z ) π nω 2 ( z )
we may calculate the new
characteristic function after
passing through an optical sytem
aq1 + b
from the relation q2 = , where a, b, c, and d are the elements of the
cq1 + d
system matrix in Yariv’s notation (R and ω are the radius of curvature and
beam waist for the Guassian beam at z). This law is extremely useful since
it allows us to calculate the diffraction limited beam waist and location for a
laser beam passing through any optical system for which the system matrix
is known.

8
 Thiel 2/2/98
Physics 426

6.) A simple example to illustrate how to use the matrix method:

Consider a lens with focal length ƒ. We want to find the new focal point f
when a dielectric plate of thickness d and index n is placed behind the lens
at a distance l.
f n

l d x
The first step is to find the system matrix:
 −1 
1 0 1 0  1 − 1  1 
A =  d    = f 
1  l 1  f 
 d d l
n  0 1  l + 1− − 
 n nf f
(notice that we don’t need R matrices for the plate since D=0 and thus R=I,
the identity matrix, for both surfaces of the plate—this is an advantage of
Hecht’s notation) now, we want to translate a ray from the plate a distance x
along the optical axis, so we must operate on A with a transfer matrix
 −1 
! !  1 0   1 !
f  ro
ri = TAro =  
 x 1  l + d 1 − d − l 
 
 n nf f
next, we want to find the value of x that causes the ray to cross the optical
axis for a ray entering parallel to the axis (this will be the focal point).
 −1   − yo   − yo 
α i   0   1 0  1   0   1 0    
f f f
  = TA  =   d   =     =
− xy o

 0  y o   x 1  l + 1−
d
−   yo   x 1  yo (1 −
l d
− ) 
l
+ yo (1 −
d
−
l
)

 n nf f  nf f   f nf f 
−y d
which gives us that αi = o and x = f − − l . These results are what we
f n
expect since αi is just the convergence angle, and n=1 gives us the original
focal distance. We see that due to the “extra” OPL from the plate, the
n − 1
second principal plane is shifted away from the lens by ∆ f = d  ,
 n 
which we observe as a shift in the position of the focal point relative to the
lens surface (the EFL is still f since the convergence angle has not changed).