H. Elagha Vol. 29, No. 12 / December 2012 / J. Opt. Soc. Am.

A 2679

Exact ray tracing formulas based on a nontrigonometric

alternative to Snell’s law

Hassan A. Elagha
High Institute of Optics Technology, El-Nozha, Heliopolis, Cairo 17361, Egypt
(Hassanelagha@gmail.com)

Received September 21, 2012; revised November 2, 2012; accepted November 4, 2012;
posted November 5, 2012 (Doc. ID 176667); published November 29, 2012
In this work, Fermat’s principle is applied to derive a simple exact formula for refraction (reflection) in terms of the
lengths of the incident and refracted rays. This formula is a nontrigonometric alternative to Snell’s law and is
general for all optical surfaces. It is used to derive the paraxial optics equations in a more simple and direct
way than that often used in the literature. It’s also applied to derive a new single, exact ray tracing formula
for the nonparaxial refraction (reflection) at a single optical surface. The obtained formulas are used to develop
a simple ray tracing procedure for meridional refraction through systems of spherical surfaces without the need to
use any form of Snell’s law. Numerical examples are provided and discussed. © 2012 Optical Society of America
OCIS codes: 080.1753, 080.2720, 120.5700, 120.5710.

1. INTRODUCTION directly derived from the refraction one by just setting

Fermat’s principle can be considered as the basis for geome- n
−1. This single formula is expected to have some theore-
trical optics. Its original formulation states that for a single tical and practical applications in optical design. Optics stu-
reflecting or refracting surface, the actual path that a light dents and optical designers can easily use it to compute
ray follows from one point to another via the surface is the position of the image formed by a paraxial or nonparaxial
ray in a short time using a pocket calculator. Therefore, it can
one for which the time required is a minimum. This particular
be considered an exact generalization for paraxial optics
formulation is called the principle of least time [1–3]. In the
equations. Although the present paper is devoted to the case
literature, this formulation is often used to derive Snell’s law,
of spherical surfaces, the obtained exact formulas can also be
which is considered the main step in the derivations of para-
applied for all other kinds of optical surfaces provided that
xial and nonparaxial ray tracing formulas. In the present
both the radius and the center of curvature corresponding
work, we apply Fermat’s principle in a different way to derive
to the point of incidence are known. Another goal of this work
an exact, nontrigonometric formula that can, successfully re-
is to derive a simple, exact ray tracing procedure for the case
place Snell’s law in many applications. This formula is written
of a centered system of spherical surfaces (Section 5.C). This
in terms of the lengths of the incident and refracted (reflected)
procedure consists of only five equations and doesn’t depend
rays and can be applied for all optical surfaces (spherical, as-
on any forms of Snell’s law, which makes it new and different
pheric, planar, etc.). Using this formula, we can easily derive
from other known procedures. All the results obtained in this
exact ray tracing formulas in both paraxial and nonparaxial
work are followed by numerical examples.
regions. Section 4 is devoted to the application of this formula
in deriving the well known paraxial (Gaussian) equations. The
technique of this derivation was found to be simpler and more
2. APPLYING FERMAT’S PRINCIPLE TO A
direct than the usual trigonometric technique based on Snell’s
GENERAL REFRACTING (REFLECTING)
law. This may indicate the usefulness of this formula in teach-
SURFACE
ing paraxial optics in the undergraduate level. Figure 1 shows a general surface (spherical, aspheric, etc.)
In the literature, the exact ray tracing of nonparaxial refrac- separating two media of refractive indices no and n1 . R is
tion (reflection) at a single optical surface is mainly divided the radius of curvature corresponding to the point of inci-
into two classes: meridional and skew. In both cases the tra- dence P. Let O and I be a point object and its image formed
cing of a single ray can be achieved by applying a procedure by the surface, respectively. The optical path lens of the ray
made up of a series of equations in which Snell’s law in its from O to I is
scalar or vector form is the cornerstone. There are several ex- OPL
no OP  n1 PI: (1)
cellent ray tracing procedures that can be used [4–7]. To the
best of the author’s knowledge, there are no published works Using the cosines law, we have
in which we can find a single, nontrigonometric formula for
q
















































meridional refraction and reflection at a single optical surface.
A
OP
R2  K 2 − 2RK cos ψ ; (2)
Such a single formula is missed in geometrical optics. An im-
portant goal of this work is to derive this single, exact formula
for cases of refraction [see Eq. (44)] and reflection [see q



















































Eqs. (64) and (65)]. Indeed, the reflection formula can be A0
PI
R2  K 02  2RK 0 cos ψ : (3)

1084-7529/12/122679-09$15.00/0 © 2012 Optical Society of America

2680 J. Opt. Soc. Am. A / Vol. 29, No. 12 / December 2012 H. Elagha

Fig. 2. Line OI can meet the radius at any point C 0 .

Fig. 1. Refraction by a single spherical surface.
surface, the point C0 is any arbitrary point located on the nor-
According to Fermat’s principle of least time, we have mal to the surface N, as in Fig. 3.
For a centered system of spherical surfaces, and taking sign
∂ n RK sin ψ n1 RK 0 sin ψ convention into account, we can apply Fermat’s principle for
OPL
o −
0: (4) the mth surface (Fig. 4) to get the general formula
∂ψ A A0
Am nm−1 K m
Hence, we obtain
: (10)
A0m nm K 0m
A n K

o : (5)
A0 n1 K 0 3. SIGN CONVENTION
1. Light rays travel from left to right.
In Fig. 1, we assumed that the refracting surface is converging
2. The radius of curvature, R, is positive for a convergent
(n1 > n0 ). A and A0 are the lengths of the incident and re-
(refracting/reflecting) surface and vice versa.
fracted rays, and K and K 0 are the distances between the cen-
3. The distance S o between the object and the vertex V of
ter of curvature C and both the object O and the image I,
the surface is positive when the object is located to the left of
respectively. A similar procedure can be followed for the case
V and vice versa.
of a diverging surface, where we finally get 4. The distance S 1 between the image and the vertex V of
the surface is positive if the image is located to the right (left)
A n K

− o 0: (6) of the refracting (reflecting) surface and vice versa.
A0 n1 K 5. The distance K between the object and the center of
curvature C for a (reflecting/refracting) surface is positive if
In fact Eqs. (5) and (6) exactly describe the refraction of any the object is to the left of C and vice versa.
meridional ray by an optical surface. The meaning of the ne- 6. The distance K 0 between the image and the center of
gative sign will be cleared in the sign convention (Section 3). curvature C for a (reflecting/refracting) surface is positive if
For the case of a diverging mirror, the application of Fermat’s the image is to the right of C and vice versa.
principle (following the same procedure) leads to the formula 7. The distance L between the object and the image is
positive if the former is located to the left of the latter and
A K

− 0; (7) vice versa.
A0 K 8. The length of a ray, A, is positive if the ray is real and
vice versa (i.e., negative for an extension of a ray).
while, for a converging mirror, we get
9. Heights above the optical axis are positive and
vice versa.
A K

: (8)
A0 K 0

It’s obvious that Eqs. (7) and (8) can be derived from Eqs. (5)
and (6), respectively, by setting n0
−n1 . Therefore, we have
an exact formula for reflection and refraction that can be writ-
ten in the general form

A n K

 o 0: (9)
A0 n1 K

It’s not necessary for K and K 0 to be on the optical axis. It’s so

easy to show that Eq. (9) remains correct if the line OI meets
the radius of curvature R (or its extension) at any arbitrary
point C0 . In this case, the image I will be formed at a point
I located on the extension of OC 0 . This can be shown in Fig. 2, Fig. 3. Refraction at a planar surface. The line OI can meet the nor-
where K
OC 0 and K 0
IC 0 . In case of a planar refracting mal N at any point C 0 .
H. Elagha
Fig. 4. Refraction of a meridional ray through the last two surfaces of
a centered system of m spherical surfaces.
10. The angle that the ray makes with the optical axis is
positive if the latter has to be rotated in the counterclockwise
direction (through the acute angle) to coincide with the ray
and vice versa.
N.B. To check the validity of the above sign convention in
various cases of reflection and refraction, we can use the fact
that the separation between the object and the image is
L
K  K 0 . i.e., we can substitute for K 0 in Eq. (10)
by K 0
L − K.
In the following sections, we apply Eq. (10) in the ray tra-
cing of both paraxial and meridional refraction and reflection
at spherical surfaces.
4. PARAXIAL RAY TRACING
In this section we shall see how the application of Eq. (10) in
deriving the well known paraxial formulas is easier and more
direct than the usual derivation procedure based on the
trigonometric form of Snell’s law.
A. Refraction at a Spherical Surface
In Fig. 1, the refraction of the meridional ray OP at a spherical
surface is exactly expressed by Eq. (5). In the paraxial region,
the point of incidence P is very close to the vertex V . So, we
can approximately write A
S o , A0
S 1 , K
S o  R and
K 0
S 1 − R. Substituting into Eq. (5), we get
no n1 n1 − no

: (11)
So S1 R
Equation (11) is the well known paraxial formula for a single
spherical surface. If the refracting surface is divergent, a simi-
lar procedure is applied, taking sign convention into account.
In this case we find that A
S o , A0
−S 1 , K
S o  R, and
K 0
S 1 − R. (K 0 is a negative quantity, since I is located to
the left of the center C.) Now, substituting into Eq. (6), we
again get Eq. (11).
B. Reflection at a Spherical Mirror
Figure 5 shows the meridional reflection at a convex spherical
mirror. Equation (5) can be used for the case of a convex
spherical mirror by setting n1
−n2 . It is reduced to the
Vol. 29, No. 12 / December 2012 / J. Opt. Soc. Am. A 2681
Fig. 5. Meridional reflection at a convex spherical mirror.
paraxial form as P comes very close to V , where A
S o ,
A0
−S 1 , K
S o − R, K 0
R − S 1 (according to the sign con-
vention, K 0 is negative, since I is to the left of C). Now, sub-
stituting into Eq. (5), we get
1 1 2

: (12)
So S1 R
Equation (12) is the expected paraxial form for a spherical
mirror. For a concave mirror, a similar procedure is applied
using Eq. (6) and we again get Eq. (12).
C. Refraction by a Centered System of Spherical
Surfaces
Equation (10) exactly describes the refraction of a meridional
ray by the mth surface of a centered system of spherical sur-
faces. This equation can be reduced to a paraxial form by mak-
ing use of Fig. 4. When P m comes very close to the vertex
V m , we can consider that Am
S m−1 − tm−1 , A0m
S m ,
K m
S m−1 − tm−1 − Rm , K 0m
S m − Rm . Then, by substituting
into Eq. (10), we get
nm nm−1 n − nm−1

m : (13)
S m tm−1 − S m−1 Rm
This is a general formula for a system of m spherical surfaces
in terms of S m and S m−1 . It can be easily used to trace a ray
refracted through a centered system of m spherical surfaces
in the paraxial region. In this case, the distances S m and S m−1
are measured to the vertices of the surfaces and not to the
principal planes. It should be noted that for the first surface
(m
1), Eq. (13) will have the form of Eq. (11). In this case to
can be seen as the separation between the first surface
(m
1) and a preceding (imaginary) surface (m
0) in con-
tact to the first surface, i.e., to
0. S o is the distance of the
image formed by the imaginary (m
0) surface. Because
it’s located to the left of the surface, it’s considered a virtual
image, and according to the sign convention S o must be a ne-
gative quantity, and Eq. (11) is obtained. Using Eq. (10) and
Fig. 4, the paraxial formula [Eq. (13)] can be rewritten in terms
of K m and K 0m as follows In the paraxial region we have
Am
K m  Rm ; (14)
A0m
K 0m  Rm : (15)
Now, substituting from Eqs. (14) and (15) into Eq. (10), we get
2682 J. Opt. Soc. Am. A / Vol. 29, No. 12 / December 2012 H. Elagha

nm nm−1 nm − nm−1 Thus Eq. (20) can be written as

−  0
: (16)
Km Km Rm
E n 1−n
−
E: (23)
Both Eqs. (16) and (13) are equivalent. They are general for f v S1 R2
divergent and convergent surfaces. For the first surface
(m
1), the distance K 1 in Eq. (16) should be negative, since Substituting from Eq. (22) into Eq. (23), we get
it represents the distance of a virtual image formed by an ima-
   
ginary surface (m
0) preceding the first surface. Since the 1 1 1 t1
image is located to the left of the curvature center, K 1 is ne-
n − 1 −  n − 12 ; (24)
fp R1 R2 nR1 R2
gative, and Eq. (16) can be written for the first surface as
where f p is the focal length measured from F to the second
n1 n n − n0 principal plane, while f v is the focal length measured from F
 0
1 : 17
K 1 K 01 R1 to the vertex V 2 , f p
f v ∕E.

Equation (17) is the paraxial form of a single spherical surface E. Thick Lens Equation in Terms of K
written in terms of K. The concept of “principal plane” plays a very important role in
paraxial optics. In Section 4.D, the appearance of the quantity
D. Paraxial Thick Lens Equation E
1 − n − 1t∕nR1  was natural and we assumed that
Using Eq. (13), we can easily derive the classical paraxial for- E∕f v
1∕f p . Now, using Eq. (16), we shall do the same steps
mula for a thick lens as follows. For the second surface applied in Section 4.D, and a new quantity C corresponding to
(m
2) we have E will appear. First, Eq. (16) is written for the second surface
of a thick lens in the form
n2 n1 n − n1

2 ; (18)
S 2 t1 − S 1 R2 n2 n n − n1
−  1
2 : (25)
K 2 K 02 R2
where S 2 is the distance of the final image formed by the lens
from the vertex V 2 , and t1 is the lens thickness (Fig. 6). Now, if In Fig. 7, we have K 2
K 01 − e, so we get
the object is located at infinity, i.e., S 0
∞, no
n2
1 (air)
and n1
n, in this case, the distance of the final image F from n2 n n − n1
−  1
2 ; (26)
the second surface will be the second focal length f v of the K 01 − e K 02 R2
lens measured from F to V 2 . Thus, Eq. (18) can be written as
where e is the separation between the two curvature centers
1 n 1−n of the lens,

; (19)
f v t1 − S 1 R2
e
t − R1 − R2 : (27)

1 n 1−n The equation of the first surface of the lens is

−
; (20) n∕K 1  1∕K 01
n1 − 1∕R1 , so if K 1
∞, then
f v S 1 1 − t1 ∕S 1  R2

where 1∕K 01
n1 − 1∕R1 : (28)

nR1 Assuming that no
n2
1 and n1
n, Eq. (26) becomes

S1
; (21)
n−1  
−1 n 1−n

: (29)
K 01 1 − e∕K 01  K 02 R2
 
n − 1t1
E
1 − t1 ∕S 1 
1 − : (22) Now let
nR1

C
1 − e∕K 01 : (30)

Fig. 6. Refraction of a paraxial ray through a thick lens. Fig. 7. Refraction of a paraxial ray through a thick lens.
H. Elagha Vol. 29, No. 12 / December 2012 / J. Opt. Soc. Am. A 2683

p
















Then Eq. (29) becomes  R2 − h2
cos ψ
: (39)
  R
−1 nC 1−n

C: (31)
K 01 K 02 R2 Substituting from Eqs. (36) and (39) into Eq. (38) and solving
for K 0 , we get
From Eqs. (28) and (30), we get
p




















p
















    β2 R2 − h2  R2 − h2
n−1 K0
 : (40)
C
1− e : (32) β2 − 1
R1
Multiplying the nominator and denominator by the conjugate
Substituting from Eqs. (28) and (32) into Eq. (31), we get of the nominator, we get
    
C n−1 1 1 2 e R2

−  n − 1 ; (33) K 0
p




















p















; (41)
K 02 n R1 R2 nR1 R2
β R − h2 − R2 − h2
2 2

which is the paraxial thick lens formula written in terms of K 02 . and the final form of S 1 is
Here, K 02 is the lens focal length measured from the focus F to
the center C 2 of the second surface. Now, as was done in
R2
Section 4.D, C∕K 02 can be replaced by 1∕f c where f c is the S 1
R  p




















p















: (42)
focal length measured from the focus F to a point that may β2 R2 − h2 − R2 − h2
be called the “principal center” in analogy to the concept
of the “principal plane.” So, Eq. (33) becomes In Eqs. (41) and (42) the positive sign is chosen if R is positive
(i.e., for a converging surface) and vice versa. Using Eq. (37)
     and Fig. 1, we can find β in the form
1 n−1 1 1 e

−  n − 12 : (34)
fc n R1 R2 nR1 R2 s



















































n p















2
β
h2  K  R2 − h2 : (43)
The location of that assumed “principal center” was found to K
be exactly that of the principal plane, i.e., f c
f p . This can be
p
















proved by substituting from Eqs. (27) and (29) into Eq. (34) The sign of the small root R2 − h2 is negative for a conver-
where we get 1∕f c
1∕f p which means that ging surface and vice versa. Now, substituting for β from
Eq. (43) into Eq. (42), we get
C E 1


: (35)
K 02 f v f p R2
S 1
R  r












































































h
:
p















i  2 p
















R2 2 2 2 2 h 2 2
In Section 4, we formulated the paraxial formulas in terms of n K 2 K  R  2K R − h − n − R − h
K, i.e., the distances measured to the center of curvature. This
new formulation of the equations may be useful in some ap- (44)
plications in which the vertex position is hard to determine. It
Equation (44) is a main goal of this work. It’s a single, exact
also facilitates the derivation of exact nonparaxial (meridio-
formula that can be used to calculate the position of the image
nal) formulas in simple forms, as will be seen in the next
formed by a nonparaxial (meridional) ray in both refraction
section.
and reflection. In the case of reflection, we just set n
−1,
where Eq. (44) is reduced to a very simple form [see
5. MERIDIONAL RAY TRACING Eq. (65)], as we shall see in Section 5.B.1.
A. Refraction at a Single Spherical Surface
1. Description 2. Paraxial Approximation
In Fig. 1, the meridional ray OP is refracted at the spherical In the paraxial region, the Gaussian formula [Eq. (11)] can
surface and the image I is formed at the optical axis. This be directly obtained from Eq. (44) by setting h
0 and
refraction was exactly described by Eq. (5), which can be K
S o  R. Also, we can get the paraxial form of a single
rewritten as spherical surface written in terms of K’s as follows. As h tends
to zero, Eq. (41) for a converging surface can be reduced to
A0
βK 0 ; (36) the form

1 β 1
nA
− ; (45)
β
; (37) K0 R R
K
and Eq. (43) reduces to
where n
n1 ∕no . Now, using the law of cosines we get
nR
A02
R2  K 02  2RK 0 cos ψ; β
n− : (46)
(38) K

where So, from Eqs. (45) and (46) we get

2684 J. Opt. Soc. Am. A / Vol. 29, No. 12 / December 2012 H. Elagha

n 1 n−1 Substituting for A into Eq. (53) and using Eq. (54), we get

: (47)
K K0 R
K oβ
h
sin θ: (55)
Equation (47) is the paraxial form for a single refracting sur- n
face. It’s general for both converging and diverging surfaces.
6. Summary
3. Parallel Incidence In terms of θ, the meridional refraction at a spherical surface
If K
∞ (parallel incidence), then Eq. (44) directly reduces to can be exactly traced by using the following three equations:
q

































R2
S 1
R  p




















p















: (48) β
n cos θ  n R∕K2 − sin2 θ; (56)
n R − h2 − R2 − h2
2 2

In this special case, the longitudinal spherical aberration exact Kβ

formula is h
sin θ; (57)
n

R R2 R2
L
∓ p




















p















: (49) S 1
R  p




















p















; (58)
n−1 n2 R2 − h2 − R2 − h2 β2 R2 − h2 − R2 − h2
4. Numerical Example where n, S 0 , R, θ, and K
S o  R are all given. The sign of the
In Fig. 1, let S o
10 cm, R
8 cm (i.e., K
root in (56) is opposite to the sign of the quantity (R∕K).
S o  R
18 cm), h
3 cm, and n
2 (n
n1 ∕no . To find
S 1 for the formed image I, we just substitute in Eq. (44) to
easily get S 1
41.8499969 cm. We’ve chosen the positive sign 7. Numerical Example
for the second term of Eq. (44) and the negative sign for the Using Fig. 1, let S 0
9.25 cm, R
19.5035816 cm (i.e.,
p
















root R2 − h2 , since R is positive. For the case of a concave K
S 0  R
28.7535816 cm), θ
26.565°, and n
1.5.
surface, we can consider the following numerical example: let The value of K 0 can be calculated by substituting in the three
S o
20 cm, R
−8 cm (i.e., K
S o  R
12 cm), h
3 cm, equations [Eqs. (56)–(58)]. The results are β
0.5766518,
and n
2. To find S 1 for the formed image, we just substitute h
4.9434327 cm, and S 1
−23.8971421 cm.
in Eq. (44) to get S 1
−11.43958214 cm. We’ve chosen the
negative sign for the second term of Eq. (44) and the positive B. Meridional Reflection at a Spherical Mirror
p
















sign for the root R2 − h2 , since R is negative. 1. Description
In Fig. 8, the meridional ray OP is reflected at a concave mir-
5. Refraction at a Single Spherical Surface in Terms of θ ror and a real image is formed at I. This reflection was exactly
In Fig. 1, we assume that the height h is unknown and the an- described by Eq. (8), which can be rewritten as
gle θ between OP and the optical axis is already given. Now
we have just two equations: Eqs. (42) and (43). To calculate S 1 A0
βK 0 ; (59)
from Eq. (42), we must first determine both h and β. To do so,
where
the following steps are suggested. Assuming that θ, S o , R, and
n (n
n2 ∕n1 ) are all given, we can write r

















































h p















i2
A 1
β

h2  K  R2 − h2 : (60)
K K
K
S o  R: (50)
Now, using the cosines law in Fig. 8, and following similar
steps to those of Section 5.A.1, we finally get
The normal r 1 is given by
R2
r 1
K sin θ; (51) K 0
 p




















p















: (61)
β R − h  R2 − h2
2 2 2

p
















A
K cos θ  R2 − r 2 ; (52)

h
A sin θ: (53)

The sign of the root in Eq. (52) was found to be negative if R is

positive (i.e., for a converging surface) and vice versa. Equa-
tions (50)–(53) can be summarized into just two equations as
follows. Substituting from Eq. (51) into Eq. (52), and then sub-
stituting for A into Eq. (37), we get
q

































β
n cos θ  n R∕K2 − sin2 θ: (54) Fig. 8. Reflection by a single spherical surface.
H. Elagha Vol. 29, No. 12 / December 2012 / J. Opt. Soc. Am. A 2685

p
















In Eq. (60), the sign of the root R2 − h2 is positive for the R2
convergent mirror and vice versa. In Eq. (61), the sign is po- K 0
p















: (68)
2 R2 − h 2
sitive for a convergent mirror and vice versa. Equations (60)
and (61) can be combined into a single equation for a sphe- The image distance S 1 from the vertex is
rical mirror in the form
R2
R2 S 1
R ∓ p















: (69)
0
K
 r







































































h i
: 2 R2 − h2
p















p
















R2 2  R2  2K R2 − h2 − h2  2 − h2
K2
K R
In this case, the exact formula for longitudinal spherical aber-
(62) ration is

Equation (62) can be directly derived from Eq. (44). This can R R2
be proved if we notice that for a spherical refracting surface L
−  p















: (70)
2 2 R2 − h2
K 0
S 1 − R. So, Eq. (44) can have the form
The sign of the second term in Eq. (69) is negative when R is
R2 positive and vice versa, while in Eqs. (68) and (70) it’s the
K 0
∓ r












































































h
:
p















i  2 p















same as that of R.
R2 2 2 2 2 h 2 2
n K 2 K  R  2K R − h − n − R − h

(63) 4. Numerical Example

Therefore, by setting n
−1 in Eq. (63), we get Eq. (62). The
larger root in the denominator of Eq. (62) can be easily rear-
ranged so that Eq. (62) can have the final simple form
s




















 2

−1 1 2 h
 0
1−
K K R R
Which can also be rewritten in terms of S 0 and S 1 as
; (64)
Let’s find the exact image distance S 1 for the convex mirror of
Fig. 5 if R
−7 cm (negative as the surface is divergent),
S o
15 cm, and h
4 cm. In this case we directly substitute
in Eq. (65) to get S 1
−1.709481753 cm (a virtual image).
5. Reflection at a Single Spherical Surface in Terms of θ
In Fig. 8, if θ is given and h is unknown, the same steps as in
Section 5.A.5 are applied for a spherical mirror, where we
finally obtain

s



















 q

































 2
h
cos θ  R∕K2 − sin2 θ K sin θ; (71)
1 1 2 h

1− : (65)
R − S0 R − S1 R R
s




















 2

To the best of the author’s knowledge, Eqs. (44), (64), and (65) −1 1 2 h
 0
1− : (72)
are all new and were never published before. They are also K K R R
general for all surfaces (planar, spherical, aspheric, etc.) pro-
vided that K and R (corresponding to the point of incidence)
are determined. The meaning of K in the case of a planar sur-
6. Numerical Example
face was shown in Fig. 4, where K
OC 0 and C 0 is any arbi-
Let’s find K 0 for the image formed by a spherical concave mir-
trary point located on the normal to the surface N. In this case,
ror (R
6 cm) for a point object placed at a distance S o

R is the distance PC 0 .
13.2 cm from the mirror vertex at the optical axis and
θ
22.5°. We simply substitute in Eqs. (71) and (72) to get h

2. Paraxial Approximation 4.585262251 cm and K 0
2.825833372 cm.
In the paraxial region, h tends to zero. So, Eqs. (64) and (65)
reduce to C. Exact Procedure for Meridional Refraction through a
Centered System of Spherical Surfaces
−1 1 2 1. Description

; (66) In the literature, there are several procedures for meridional
K1 K2 R
ray tracing [4–7]. In this section, another new procedure based
on the exact formulas [Eqs. (41) and (43)] is presented. The
1 1 2 procedure consists of five equations, and we derive it follow-

: (67) ing the same steps as in Section 5.A.5. Let’s consider the last
R − So R − S1 R
two surfaces of a centered system of m spherical surfaces as
Equations (66) and (67) are equivalent to the well known para- in Fig. 9. For the mth surface, we have
xial form [Eq. (12)]. They are general for convex and concave
mirrors.
r 1
K m sin θm−1 ; (73)

3. Parallel Incidence
q

















When the object is at infinity (K
∞), and for a given finite
Am
K m cos θm−1  R2m − r 21 : (74)
value of h, Eq. (64) reduces to
2686 J. Opt. Soc. Am. A / Vol. 29, No. 12 / December 2012 H. Elagha

Using Eq. (73), we can write Dividing Eq. (82) by Eq. (83) and equating the result to
Eq. (84), we get Eq. (81). In Fig. 9, we also find that
q








































Am
K m cos θm−1  R2m − K 2m sin2 θm−1 : (75)
K m
K 0m−1 − tm−1  Rm−1 − Rm : (85)

According to the sign convention, Am is a negative length,

2. Summary
since I m−1 P m is not a real ray. Thus, Eq. (75) should be
For the mth surface of the system, we have a procedure of five
written as
equations in the following sequence:
 q







































  
Am
− K m cos θm−1  R2m − K 2m sin2 θm−1 ; (76) nm−2 K m−1
sin θm−1
sin θm−2 ; (86)
nm−1 K 0m−1

hm
Am sin θm−1 : (77)

K m
K 0m−1 − tm−1  Rm−1 − Rm ; (87)
Using Eq. (37) we have
 q
















































nm A m nm
βm
: (78) βm
− cos θm−1  Rm ∕K m 2 − sin2 θm−1 ;
nm−1 K m nm−1
(88)
Substituting from Eqs. (76) into (78), we get
 q















































  
nm nm−1
βm
− cos θm−1  Rm ∕K m 2 − sin2 θm−1 ; (79) hm
K m βm sin θm−1 ; (89)
nm−1 nm

and using Eqs. (76), (77), and (79) we get

R2
 K 0m
 p
























m p


















: (90)
nm−1 βm Rm − h2m − R2m − h2m
2 2
hm
K m βm sin θm−1 ; (80)
nm
This procedure is different from other ray tracing procedures
where θm−1 is the angle between P m I m−1 and the optical axis. [4–7], since it doesn’t include any form of Snell’s law. The dis-
We can determine θm−1 in terms of θm−2 using the formula tances are measured to the centers of curvature not to the
vertices.
 
nm−2 K m−1
sin θm−1
sin θm−2 : (81)
nm−1 K 0m−1 3. Remarks
1. For the final image, the value of the distance S m is
To derive Eq. (81), we use Fig. 9 as follows: given by

r 2
K 0m−1 sin θm−1 ; (82) S m
K 0m  Rm : (91)

2. The sign of the root in Eq. (88) was found to be the

r 1
K m−1 sin θm−2 ; (83) same as that of the quantity (Rm ∕K m ).
3. The sign of Eq. (90) is positive for the convergent sur-
face and vice versa.
r 2 sin β2 nm−2 4. For the first surface (m
1), and according to the sign


: (84)
r 1 sin β1 nm−1 convention, the distance K 1 is negative, since we assume the
existence of an imaginary surface (m
0) preceding to
the first surface. In this case, the object O is assumed to be
the image formed by that (m
0) surface, and since this
image is located to the left of the surface center of curvature,
the distance K 1 is negative.

4. Numerical Example
In Fig. 10, we have a system of five spherical surfaces that
form a final image I 5 for a real object O located on the optical
axis at a distance S o from the vertex V 1 of the first surface.
Let’s calculate the exact distance S 5 of the final image from
the vertex V 5 of the fifth surface if we know that the incident
ray inclination angle is θo
17.30972443, S o
12 cm, R1

10 cm, R2
−8 cm, R3
12 cm, R4
−10 cm, R5
−8 cm,
Fig. 9. Refraction through the last two surfaces of a centered system t1
5 cm, t2
5 cm, t3
8 cm, t4
1 cm. no
1, n1
1.2,
of spherical surfaces. n2
1, n3
1.5, n4
1, n5
1.2. For each of the system
H. Elagha
Fig. 10. Centered system of five spherical surfaces.
surfaces (starting from the first one), we substitute into the
five Eqs. [(86)–(90)] as follows.
First surface:
θo
17.30972443° (given), K 1
−22 cm (given),
β1
−0.733293273,
h1
4 cm, and K 01
−33.12062834.
Second surface:
θ1
9.479589747°, K 2
−20.12062834 cm,
β2
−1.123526355, h2
4.467759332 cm,
K 02
−55.03140479 cm.
Third surface:
θ2
4.143784933°, K 3
−80.03140479 cm,
β3
−1.299007738, h3
5.008144154 cm,
K 03
37.33759883 cm.
Fourth surface:
θ3
−5.926743117°, K 4
51.33759883 cm,
β4
−0.552990504, h4
4.397067985 cm,
K 04
17.76842329 cm.
Fifth surface:
θ5
−26.58363535°, K 5
14.76842329 cm,
β5
−0.70684292, h5
3.892892507 cm,
K 05
22.16416375 cm.
The distance S 5 of the final image I 5 is given
by S 5
K 05  R5
14.16416375 cm.
Vol. 29, No. 12 / December 2012 / J. Opt. Soc. Am. A 2687
6. CONCLUSION
Fermat’s principle of least time was applied to derive an
exact equation in terms of the lengths of the incident and
refracted rays. This equation successfully replaced Snell’s
law and was found to be useful in deriving the well known
paraxial formulas in more simple steps. It was also applied
to derive exact single formulas for refraction and reflection
of nonparaxial (meridional) rays at spherical surfaces. The
author thinks that these exact formulas are missed in geome-
trical optics and that they may be useful in theoretical and
practical applications. Based on the obtained equations, a sim-
ple exact ray tracing procedure was derived that doesn’t de-
pend on Snell’s law. In this work, the positions of the object
and the images were determined relative to the curvature cen-
ters, which facilitated the analysis and simplified the resulting
formulas.
ACKNOWLEDGMENTS
The author would like to express his deep thanks to the High
Institute of Optics Technology, to Professor M. K. Fayek
(EAEA), Professor M. Medhat Abdel-rahman, and Dr. R. H.
Abdel-Maksoud, Faculty of Science, Ain Shams University,
for their constant encouragement.
REFERENCES
1. R. P. Feynman, R. B. Leighton, and M. L. Sands, “Optics: the
principle of least time,” in The Feynman Lectures on Physics
(Addison Wesley Longman, 1970), Vol. 1, Chap. 26.
2. F. L. Pedrotti and L. S. Pedrotti, “Geometrical optics,” in
Introduction to Optics (Prentice-Hall, 1993), pp. 34–36.
3. P. Mouroulis and J. Macdonald, “Rays and foundations of
geometrical optics,” in Geometrical Optics and Optical Design
(Oxford University, 1997), pp. 11–13.
4. R. Kingslake and R. B. Johnson, “Meridional ray tracing,” in Lens
Design Fundamentals (Academic, 2010), pp. 25–45.
5. D. Malacara and Z. Malacara, Handbook of Optical Design
(Marcel Dekker, 2004).
6. W. J. Smith, “Optical computation,” in Modern Optical Engineer-
ing (McGraw-Hill, 1990), pp. 308–404.
7. J. E. Harvey, “Exact ray trace procedure,” http://ebookbrowse
.com/3‑5‑exact‑ray‑trace‑procedure‑pdf‑d354859582.