George W. Hopkins
Optical Sciences Center, University of Arizona, Tucson, Arizona 85721
(Received 22 May 1976)
Algebraic raytrace equations for axially symmetric optical systems are expanded in terms of system
parameters and paraxial variables. The transfer parameter used is optical path, and equations are given for
second, fourth, sixth, and eighthorder differences in optical path of a ray from the axial value. Selected
rays are traced to a tilted reference sphere in the exit pupil, optical path differences of given order are equated
to the wave aberration polynomial of corresponding order and with proper coordinates, and the resulting
linear equations are solved for wave aberration coefficients.
INTRODUCTION
We have published a set of proximate raytrace equations which give the angular and transverse deviations
of a real ray from the corresponding Gaussian ideal. 1
These deviations were segregated by order and used to
determine transverse aberration coefficients.
An analogous set of equations is presented here, but
the angular variables used are direction cosines, and
optical path is used as the parameter for transfer operations. This offers a distinct advantage over previous methods based on similar principles. 1,2 Whereas
these provided transverse and angular errors for a ray,
this method also retains differences in optical path from
the axial optical path. These data can be used for calculating system wave aberration coefficients, including
piston error (the difference between optical path for a
chief ray and the axial value). In addition, direction
cosines can be preferred ideal coordinates. Coefficients based on such coordinates on pupil reference
spheres are the proper coefficients for merit functions
based on diffraction criteria, and satisfaction of the
ideal does not imply a violation of the sine condition, as
is the case with Gaussian optics. 3
The equations given here can be used for an ideal
mapping (i. e., a linear mapping represented by the
paraxial raytrace equations with finite angles) in either
a cosine space, a tangent space, or a mixed space,
such as the mapping of a plane object to a plane image
via pupil reference spheres. This is accomplished with
a technique used by Cox. 2 If the ideal coordinate is a
direction cosine, the initial value is the ideal value and
there are no errors. If the ideal coordinate is a direction tangent, the initial value is the ideal value, but
"errors " to appropriate order must be introduced to
represent the direction cosine of that ray as a series
expansion in the tangent.
(~3)+
(~5)+
(7) + *
(o
6c) + tjd~
c) +***v
(8
6dcc
cc.
where
[2] d(2c) = (c(I))2 + (c (1))2
[4] 6d(4)+= 2(C(1)6c(3)+ c(036c
cc
cc
c 5r:
~=1,
[6] 6d 4) =  6(d3c))2
Od06d
32
 4d2)dc
2
FdC:)
(+
+2(c
) +. . .
[0] w() =t
942
oz(2) 
[2]dwi)=
[3] 6Y (3)=
4]
[5]
[6]
[7]
 6C(4)W(O) 
c7)w(0)+ 6C(5)6W(2)
yv7) =6y7)+
[8] ow
C )w
*2y~l
ml=
[1]~~
[3] 6m3)=
c( )6w(6)
4)
) _ 5C(2)6w(6)
The equations for the x and y components of the surface normal, n, and n , and some needed rotational invariants are
CM )6w6)
3
[5] 5m(5 ) =  2[a*Sy(5) + 2b*(6y( )d2) + y
6z
0
8
 6c( )W( ) _ 6C(6)6w(2)
2
4
)
(d (2) 4
6C(2)6w(2)
6C (3)'64)+
(4d (4))2]
4
 6c( )6W(
[8] 5w( 8 ) = Sz~8 ) CZ6)6w(2)S
b*[2d s)Sd
2
+ 3c*(d 2)) ad 4) + d
(0)
o(2)
4) =z(4)
= a *6d(8) +
[8] Sz(8)
w)w
)(
y (Sw+
=S
[2]
)6ds4))
+ 3c*ysl)(d (2)) ]
[7] 6m(7)
[0] w() = o
(1))2
[3]
[2]
[2]
[3]
3
y~s3) = 5yV ) + c( 1)"w
[4] Sd
s4)
[4]
[5]
mm
+2y
l ) 5y0S3))
X )+
2(xl')xs
')
M (1)
[1]nl=
4) + b*(d(2a))
[6] 6d() 6m
3
+ )6vdn(4)+
 (m(
dd
d(2
+32]
2)
6W(2
o(4)
6W
z~S~~w
w 4 6Z Sz(4) _ 6C(2)
4
3
+6C(3)6W(2)+C(1)6W(4)
y5)=6y(5)
(5 = y
~~w
Y~
S
Sc Y~
y~S)
L7
,m(3)5d(2+
(2)=
n) =M(7X))2
O (7
38~'
x
n m4"Snl
6 [4
+"l6M (52
(48()
6M(3
(6y))+
n3
~+ Wx
mm,
X
+(nm'm)2
5
2
2
[6] Sd 6) = (6x3)) + (5yS3)) + 2(x(')6xs ) + y(l)6y(5))
[6]
6z)6=a*6d (6)+
6
c4*(d (2))3
2b*d 2S)6ds+
[4] ed
[6]
Y ) Y
3)
5
5
1
m
n~)c3
mtd+
cn2)+C2)6n3
c sn2eld
m
my3)(c(
, e)b(
1)t
6d(4)=2(,n3)6~n(3)+nce1)6n(3)+)lox)+x)C
c(1)6W( )
c+ nelnCl
+2
d2)=cyl,(ln5)
)6x 7) + Ysl)5ys )
xs3)Sx(5) + 5y(3)y(5))
6XS
S
REFRACTION EQUATIONS
The refraction equations are substantially simpler than those for the tangent proximate ray trace.
They are
[3] SC(3)'
Sc(3)[1/g]+
+6n(3)(j2)
+ (4At
943
2
5n(3)[(,ul)/ji]+nyl)(1/2gu )[(,u1) d2)+d (  1)d(2)+(2,u+2)d(2)]
s5)[
[(l)d
l+J(1)(l)
+ (A2,
+ 2)Sden]
2
2
3
2c)+ (A  l)d2) + ( 2u+ 2)d (2)]+n1)(1 4) [(_' 1) (d2)) + (A31) (d2))
(
George W. Hopkins
943
(2
1]]+l)
+ 6n(3)(.1) [(,i
1)6dYc+
A)6d
I+
+ ( 2,u + 2)d
]+n Y1)()
2,i 3
 2) (dC~d)
+ (
+ 4gl2
+ 6ng3)(3)
3)
+ 4u'? 
+ (12.
(4))+
) + ( 4
[(
n]+6n
(k)
(,u 346d~))
1) (2d+g6d
23) (d4)d ) 2
)2 +(d+2
(dn) +(l2
 4
12) d
cnl)dsc+ (j
2
(4.
44) (2d
2
+ 4) (d '6dcn + d6n )6d4)
+ ( 41)
d'6d
(?)6dnn)]
+4) (d
 l2)d
_ 1)6d(6n)+( 2i +2)6d(6)]
5 + 4P
(d (d
( 22c + 2)6d n
+ dn)~6d
) (d( )3 (
,l
2
(
(p' [31)
[(Pll1) (dC))
+ (4pA+
2
4)dccd
()I
) [(A  1)6d(6)+
+ (
2j,1
+ (P.
2) d
1) (dc)
+d
(4.
3
4, 2 + 6) (d4c) d4c
+ 46) (d n)dd)
4d62 d(2
4d
d2)
d](2)
[2]
[3]
62)
(2)
[4] d(2)
(y~'6)
+3
= (xol'))2
Yo =(4YO
)
,) 6
(8
Zo + 6z4 + 6z(
[4] 6d(4S) =
[4] zo
+ o
[0] w () =,
[1] Ys
=Yv
(x
6c?)6xw3)
+ Ys)6Y3))
3)+y(1)6y(3)
26d so4)
)P'2d'(1)]
[4] 6iW(4)
6W (2)
[6]
[5]
[6]
[6]
=6z(4)
6c
t(2)
Wp(2))2(
3
(,6d
6S)==(x3))2
+ (y3))2
( l
O6X ')+(6y
)++ 2(xy(1)6xy5)
5) =X 16xs+Yo
s)6Ys5 ))
(Ss)
)
(5)
6)=16
[6
[6]
6W (6) =
[7] 6y(7)=
[8]
(6)
(4)
6 ,
+ (6z.a
a+ p(2)
(2)2+1]
6) z
y(7) +
3
6c (c3
)6w(
)6w(44 )+ c(1)6w(6)
6Z (6) = C* (d (2))3
Fd(8)_=X(1)6X(7) + Y1)6Y(7)
[8] 6p
[8]
(Zs8)
= 2
+ [[
944
6z( 2
()
2
2P( )P(
)]
)+1+(
(_
6Z
M (6z
(.)2+2P(2)6P(4)(
)) )+p(2(
3z+(2)
26z(2) z(4)
(6z
))
George W. Hopkins
944
F[3(P(2))
2 6P(4)(I'\ + (p(2))3(
6z(8)
[8] 6w(8)
6C(65 6w(2)
6C(4)6W(4)
56z2o)] +(p(2))4(0
6
6c(2)6w( ) .
+2
[7]
RAY TABLE
(2)/t3
3(do(2))2
+8__
2
(2)
2 (2)do
^,4)
+ 36zo
0)
+X6(5Z12))
k()= 6z(
3
6k
6) =
 (6z 2
5t2
+ 83(d (20)N2 [
2tt
(2))
15U o
[5]
6c()
[7] 6C
k()
ye)  (1l))
cyl)(
[1]
[3] 6cy
The notation for an error consists of a letter, a number, and a letter. The first letter indicates either optical path difference (W) or transverse error at the
surface of interest (X or Y), the number indicates the
order of the error, and the final letter indicates the
particular ray from Table I.
= (yol) YS
= (y(l)  y(l))
6y(
 k
The coordinates used are normalized pupil coordinates, p, and py, and normalized image height H, taken
to lie in the meridional or yz plane. We also use p2
=px+p y. With a slight modification of the notation introduced by Hopkins, 5 the wave aberration polynomial
3 5 0
k(
6y(3)6k(2) 6y(5)k(0
5k6) 6y3)6k
y(
is
y 7
2 2
2
4
2
W=(Wo2op2+ WiiiHpy+ W2ooH )+(Wo 4 op + W131Hp py+ W 2 20 H p
+(Wo
60
6
p + W1 51 Hp py+ W 240 H
4
+ W2
+ W4 22 H p~
3
+ W 3 3 H p p5y
60 0
W440H p + W
W3 H p3 py
H py+ T
W 24 H p p+
4
3
W ,3 11H py+ W 4 00 H )
+ W 22 2 H p+
6
8
6
H ) + (W 080 p + WV71 Hp py+ T 260 H p + T 262 H p py2
4 4 2H
2
p p y+ W 444 H py+ WT53 1 H p p
W533H py
TVs 5 H p py
6
W 620 H p
+ T
622
H py
7
+ W800 H 8 )
+ W711 H p
The second, fourth, sixth, and eighthorder coefficients have been set off by parentheses for easy identification.
The solutions for these coefficients in terms of optical path errors at the exit pupil reference sphere are
[6] A=W6QW
06 0 W6G,
[6] B=2(W6P 
8W 0 6 0 
[2]
Wll = 2 (W2H
W6G),
[6] W 24 0 =2(AB),
W21),
W 420=A  W240
[6]
W4G,
 w
240
W4 2
 W6G],
George W. Hopkins
945
Ray
P0
[8] A=W8HW8If1(W8NW80),
A
B
C
D
E
F
G
H
I
J
K
L
M
0
1/v2
0
0
1/v2
0
1/v
I/v2
Ar
0
P
Py
1
[8]
1
5
i/V
1/v11
1
1
1
1
1
1
1
1
1
1
l/v'
2(AB),
[8] W 3 5 3 =A
1/v
1/f2
0
0
1
1
, 3/2

1
0
0
0
00
0
0
W 533 =
W53 3
W353 W533
[8] B=(W8JW8K)/f3
9 W35 3  w4 3 3
[8] C=(W8LW8M)/,
W3
5 312 W533
 / v21/vs
5 33
[8] E=4(AB)
*
1/vr1//
1 V
[8] F=2(AC),
[8] G=8DA,
[8]
H=45(GE),
[8] I= 2(GF),
[6] W24 =
3 2(A  B)
[8] W 17 =4(HI)
[6] W4 22 =A  W24 2
[8] W3 51= H YW 17 1
[6] B= (W6LW6M)/2 W
[6] C = (W6B  W6C)/V
333
W 333 ,
[6] W 1 ,=2A+
4B+ C,
4
5
[6] W
3 31 =5A4B4C,
[8]
W 08 0 = W8A ,
[8]
W800 =
W8 G + OPD8,
A transverse aberration polynomial in rectangular coordinates was chosen for its symmetry with the corresponding transverse pupil aberration polynomial. The
transverse image errors are
2W8G ,
[8]
W 440= 5A
 4B  4C
 W 4 4 0  W620
Ray:
 W080 W260
W8G
W080
W8G]
W26 0  W 4 4 0  W620 
[8] W 4 4 4 =2(AB) ,
946
x
x
xx
xx
X1
X3
X5
620  W8G]
X7
[8] W262 =
IV2
W4
W6
W8
26 0
AB
2 W26 0
~WG
2A+ 4B+ 8 C
CDE
FGH
IJK
LMN
OPQ
x
xxx
xx
xx
xx
xxx
x
x
x
xxx
xx
xxx
xxx
x
xxx
xxx
Error
xix
x
xx
Y1
Y3
xx
Y5
Y7
x
xx
xx
xxx
x
x
xxx
xxx
x
xx
xxx
George W. Hopkins
xx
xxx
946
E, =
(Glp, + G2H)
2
2
2
(PAp ,y + P2Hp
+ P 3 Hpy +
3
2
P 4 H2py + P 5 H p,
2
6H p
T7H p
T8 H p
4
) + (Slp py + S 2 Hp
2 2
2
+ S 3 Hp py + S 4 H p py
6
6
5
4
4
p2 + S 10 H p' + S 1 HI py + S 12 H ) + (Tlp py + T 2 Hp + T 3 Hp p2 + T 4 H p ty
4
2
py + T9H3p2 p + TJ 3 p + T,,H 4 p p + T12 H p + T 13 p 2 + T 1 4
+ S5 H P Py + S 6 H p3 + S 7 H p + S 8 H p2 + Sgf
T 5H
+ P6 H
7
6
T1 ,H p + T 16 H p2 + T1 7 H p2 + T1 8 H p6 + T 19 H p, + T 20 H )
5
5 2
and
E.=
4
3
2
2 2
2
2
+ P3 Hpp, + P5H px) + (Sjp4p. + S3Hp P.P' + S 5H p p. + S6 H p"p2 + S9H ppy + S 11H p
) + (TIp6 pP
6
5
4 2
3
3 2
2 2
2 4p
4
+ T3Hp pxpy + T5H p X+ T6H p pxpy + T 9 HIp pp + T OH p pY + T 13H1p pX + T 14H4pxpy + T 1 7H PXPY + T1 gI px)
[5]
S1 = A S4,
[7] T, = Y7A,
[7] T20 = Y7G
[7] A= Y7Q  T20 ,
[71 B = 2(Y7P T20)
[7] C= F2Y7F 8 T 20 ,
[7]
=2A+B+8C,
[1] G1 = Y1A ,
[7] T 7 =5A4B4C
[7]
[3] P1 =Y3A
[3] P6 = Y3G
[3] P 2 = Y3Q  P6 ,
[3] Ps=X3QP1
[7] D=5A4B4C
[7] E=2A+B+4C,
[5] S 1 =Y5A
[5] S1 2 =Y5G
[7]
[5] B= 2(Y5P S1 2 )
[7] T 1 6=E T 17
[5] S2 = 2(AB.)
[7] H= G T17 I
[5] S 7 =A  S 2 ,
[7] T8 =IH,
[7]
J=D
[5] S 3 =2(AAB),
[7]
Tl,=2(JH)
[5] Sg=X5NX50S 3
[7] Tg=J T1 o
[5] S 8 =AS 3 S g9
[5] A=X5QS 1 ,
[7]
=
B=
S,
[5] B= I'X5P4S
1
[7]
C=2(X7FT 1 ) ,
[5] S, = 2(AB.),
8C,
[7] T 5 =2A+4B+
[5]
[5] B= pL
[7]
T 1 5 =2A+83B+4C,
T
2T
20,
T15s
T20,
T 17 = 2(FG)
+T 7 + T15 + T20),
T8,
2X7P8T 1 ,
[7] T 13 =5A4B4C,
SI, =A S 5 ,
[5] S 6 =
947
[5] S 4 =2(AB)
SY5M)ocA.64S6
s 6 S
11
1
o29S5et2bS6er
T l g =2A+8B+4 C
 8T 1  I T 5  2T13  T 19
George W. Hopkins
947
11 t
T 13 "1
4T ,TrD
19
T 1 3  2 T19
[7] T 14 =2(F G) ,
[7] T 6 =FT1 4 ,
[7] T 4 =DT 6,
[7] H= E  T 14 ,
[7] I= (Y7N Y70)/
T 1  T 4  T 5  T6
T 13T3T
 2
14 T18T19,
T9
[7] T, 2 =2(HI)
[7] =H T12
The transverse pupil aberration polynomial to fifth
order is
6 y= (G1 H + 22py) + (P 1H 3 + P2H 2p
+
P3 H 22p + p4 Hp2
p py + S 1 0 Hp pY2 + S1 1 Hp
6. = (Gp.) + (P 2 H 2p.
+ P4 Hp.py +
+ S
p py),
: 6 p 2 px)
12
First, proximate rays were traced and ray coordinates and optical path differences were observed to
converge to real ray values. Comparison was made
with Cox's pseudoraytrace equations2 (as extended
to seventh order for spherical surfaces) and with the
tangent proximate raytrace equations (PROXI being instructed to "find" coordinates on a plane). Identical
surfacebysurface ray coordinates were obtained in
these two separate cases.
Next, the coefficients were checked for proper scaling as field and aperture specifications were changed.
This scaling was verified. An interesting test resulted
from an error by the author. A camera lens with a
100 km object distance and a 17 mm pupil radius was
being evaluated, and 17 km was accidentally entered as
the pupil radius. All real rays except the chief ray
missed the first optical surface; however, the coefficients, most of which were of the magnitude of intergalactic distances, computed and scaled down properly,
although with a slight loss of numerical accuracy!
Finally, proximate rays which were not included in
the ray set of Table I were traced, and the optical path
differences and transverse errors associated with these
rays were compared to values obtained from the aberration polynomials. Excellent agreement was obtained.
+ S 7H2pXp2 + S8 H PPX
+ ( 2 H Px + S4 H 3
+ S1 oHp pxpy +
p px).
SUGGESTED APPLICATIONS
The remarks made in Ref. 1 apply here, but the
method for obtaining surface contributions for wave
aberration coefficients will differ from the transverse
case.
The wave aberration induced on refraction is obtained from three operations. The first operation is
transfer from an objectspace reference sphere to a
surface. Optical path differences are stored. The
proximate ray is then refracted. The final operation is
transfer to an imagespace reference sphere. The resulting optical path differences are added to the objectspace values (with proper regard for sign conventions)
to give the wave aberration contribution.
The wave aberration induced on transfer can be obtained from optical path differences upon transfer between reference spheres. This is in contrast to present
theories which associate aberration contributions with
surfaces only.
Intrinsic and induced aberrations can be separated by
setting lowerorder ray "errors" to zero to obtain intrinsic values.
948
948
ACKNOWLEDGMENTS
I particularly appreciate the assistance of my dissertation director, Professor R. V. Shack, who suggested
the problem of determining higherorder wave aberration coefficients, suggested using a direction cosine
space, and sustained my enthusiasm throughout this
task. I would also like to thank 0. N. Stavroudis,
R. R. Shannon, R. A. Buchroeder, M. H. Kreitzer,
and W. C. Sweatt for encouragement, suggestions, and
stimulating discussions; and J. M. Rowe for preparing
the manuscript. This work was supported in part by
the Air Force Space and Missile Systems Organization,
Los Angeles, Calif. The State of Arizona provided computer funds.
949
949