Zernike polynomial fitting of lateral shearing

interferometry
Abstract. For a wavefront tested by shearing interferometer, the
Zernike polynomial coefficients of the wavefront are found in an analytic
form expressed by the Zernike polynomial coefficients of the shearing
interferograms. The validity of the new algorithm is tested by numerical
simulations and real interferograms. 1997 Society of Photo-Optical Instrumen-

Wei Shen
Ming-Wen Chang, FELLOW SPIE
Der-Shen Wan,* MEMBER SPIE
National Central University
Optical Sciences Center
Chung-Li, Taiwan 32054
E-mail: [email protected]

tation Engineers.

Subject terms: Zernike polynomials fitting; modal estimate; wavefront estimate;

shearing interferometry.
Paper 22076 received July 23, 1996; accepted for publication Oct. 6, 1996.

1 Introduction
Lateral-shearing interferometry is the interference between
two displaced images of the wavefront under test, and there
are a variety of approaches to accomplish the reconstruction of the phase of the wavefront under test from the
shearing interferogram. Basically, these phase reconstruction methods can be categorized as being either zonal or
modal, depending on whether the estimate is a phase value
in a local zone or a coefficient of an aperture function. A
great deal has been written on the zonal and modal estimations, for example, Refs. 1 to 6.
Whether the estimation is zonal or modal, its objective is
to find the relationship between the measured values and
the unknown wavefront. An example as the zonal estimate5
is shown in Fig. 1 to illustrate this point, where w 1 , w 2 , w 3 ,
and w 4 are the phases of the original wavefront and a 1 , a 2 ,
b 1 , and b 2 are the phases measured. Apparently the relationship between the wavefront under test and the measured
phases in this example is
a 1 5w 2 2w 1 ,

a 2 5w 4 2w 3 ,

b 1 5w 3 2w 1 ,

b 2 5w 4 2w 2 .

~1!

There are several possible solutions of w i and two of them

are
w 1 50,

w 2 5a 1 ,

w 3 5b 1 ,

w 4 5a 1 1b 2 ,

~2!

or
w 4 5b 1 1a 2 .
In principle, all the solutions are equivalent, but as specified by Rimmer,5 they are not because of the noise in the
measured data. Nevertheless, the problem of noise can be
reduced by the least-squares fitting of unknown phases. In
the case of a modal estimate, the aberration coefficients are
*Current affiliation: Massachusetts General Hospital, Boston, Massachusetts 02114.
Opt. Eng. 36(3) 905913 (March 1997)

0091-3286/97/$10.00

expressed by vectors, adopting the notations used in Ref. 6,

where a and b are the M 31 coefficient vectors of the x and
y shear data, c is the N31 coefficient vector of the unknown wavefront, and a and b are the relationship matrices
between c and a,b, respectively. The objective of a modal
estimate is to find c from the measurement data a and b
using the relationships a and b. However, the total number
of coefficients in a and b ~i.e., 2M ! is usually larger than
Nthe number of coefficients in cwhich means that if
we use the matrices a and b to list all the simultaneous
equations between the coefficients in a and b and the coefficients in c, there will be either redundant or contrary
equations, depending on the noise in data. Therefore, it is
also necessary to use the least-squares fitting algorithm to
solve for the unknown coefficient vector c.
On the other hand, we suggested an algorithm that resorts to averaging the contrary values to overcome the
problem due to the contrary equations.7 In this paper, we
give an explicit expression of the relationship between all
the coefficients in c and the coefficients in a and b, and
single out all the pairs of equations that might be contrary.
Then with the average algorithm, the number of equations
is reduced to the number matching the coefficients in vector
c. In this way, we have a unique solution of the unknown
vector c by solving the linear simultaneous equations. The
new algorithm is compatible with the least-squares fitting
~LSF! algorithm demonstrated by Rimmer and Wyant.6 The
comparison between the LSF and the new average algorithm is shown in Secs. 3 and 4, where computer-simulated
testing data and real shearing interferogram data are used.
The merit of this new algorithm is in its explicit expression
of the coefficient vector c under test and the measurement
coefficients a and b. The effects of the measurement errors
in a and b on the final result in the wavefront vector c are
discussed in Sec. 5.
2 Algorithm and Calculation Procedures
The coefficients in a and b in this paper are the fittings to
Zernike polynomials of the x and y shear interferograms,
respectively. We presume that all the coefficients in a and b
generated by this Zernike polynomial fitting algorithm8 are
independent. The average algorithm itself is independent of
1997 Society of Photo-Optical Instrumentation Engineers

Downloaded From: http://opticalengineering.spiedigitallibrary.org/ on 08/06/2015 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

905

Shen, Chang, and Wan: Zernike polynomial fitting . . .

where Z i (x,y) is the ith term of Zernike polynomials; and

c i , a i , and b i are the component coefficients in the vectors
c, a, and b, respectively. Here the wave aberration is represented by the Zernike polynomials expansion of 21 terms,
and the corresponding expansion of the small shear phases
has 15 terms, so m and n in Eqs. ~5! and ~6! and ~7! are 20
and 14, respectively. As Eqs. ~5!, ~6!, and ~7! are rearranged in terms of the monomials and plugged into Eqs. ~3!
and ~4!, two equations are obtained as
~ a 0 2a 4 1a 12! 1 ~ a 1 22a 7 ! x11 ~ 6a 1214a 131a 14! y 4

5s @~ c 1 22c 7 13c 17! 1 ~ 4c 4 22c 5 212c 1216c 13! x

11 ~ 5c 15115c 16110c 17! y 4 # ,

~8!

~ b 0 2b 4 1b 12! 1 ~ b 1 22b 7 ! x11 ~ 6b 1214b 131b 14! y 4

5t @~ c 2 22c 8 13c 18! 1 ~ 2c 3 26c 11! x1

1 ~ 50c 18125c 1915c 20! y 4 # .
Fig. 1 Illustration of 2-D shearing interferometry5: (a) wavefront, (b)
x shear interferogram, and (c) y shear interferogram.

the amount of the shear. In this section, however, we divide

the calculation into two categories: one for small shear and
the other for the general case of the finite shear, because for
small shear, the measured phase can be approximated by
the derivative of the unknown wavefront, which means that
the expression of the relationship does not involve the highorder terms of the shear. Thus it is easier and clearer to
demonstrate the algebra for solving the unknown coefficients c first in the case of small shear.

Wavefront Estimate from Small Shear

For small shear, the measured wavefronts can be approximated as the derivative of the original wavefront, which are
represented by the equations
2.1

] W ~ x,y !
,
]x

~3!

] W ~ x,y !
,
]y

~4!

DW s ~ x,y ! 5s3
DW t ~ x,y ! 5t3

where DW s (x,y), DW t (x,y), s, and t are the phases and

the shear in the x and y shear interferograms, respectively,
and W(x,y) is the wave aberration under test. These phases
are expanded by the Zernike polynomials9 as
m

W ~ x,y ! 5

( c i Z i~ x,y ! ,

i50

~5!

DW s ~ x,y ! 5

( a i Z i~ x,y ! ,

i50

~6!

DW t ~ x,y ! 5
906

( b i Z i~ x,y ! ,

i50

~7!

~9!

In Eq. ~8!, the coefficient of the same monomial on two

sides of the equal sign is equal, and therefore 15 equations
are derived. These equations are related to the x shear measurement. By the same process, another 15 equations are
derived from Eq. ~9!, which are related to the y shear measurement. Using the coefficient of the constant term as an
example, from Eq. ~8!, the equation
c 1 22c 7 13c 175 ~ a 0 2a 4 1a 12! /s

~10!

is found, and from Eq. ~9! another equation,

c 2 22c 8 13c 185 ~ b 0 2b 4 1b 12! /t,

~11!

is obtained. All 30 equations are listed in Table 1 in the

order of the monomials, where s i and t i include the linear
combinations of a i and b i like the terms shown on the
right-hand side of Eqs. ~10! and ~11!. Furthermore, we call
all these particular combinations of a i and b i shown in
these simultaneous equations, like the numerators in Eqs.
~10! and ~11!, as the variables g i and h i , and in Tables 2
and 3 the definitions of g i and h i and s i and t i are listed,
respectively. These variables will play the same role in the
derivation for the finite shear case, and only the relationships between s i and t i and g i and h i are different ~as
shown in Table 3!.
In the assumption of m520, 20 unknowns
(c 1 ,c 2 ,...,c 20) are to be solved from the 30 equations. As
mentioned in the previous section, these equations are in
addition to the necessity to solve the 20 unknowns. The
average algorithm will be applied to match the number of
equations with the number of unknowns. In Table 1 there
are 10 pairs of equations, as specified by a number from 1
to 10, which have the same function form of coefficients c i
on the left-hand side of equal sign but with different s j and
t l on the right-hand side. For a certain pair k, s j may not be
equal to t l due to the noise in data. The average algorithm
is to select the constant on the right-hand side of equations
as the average value of the corresponding s j and t l values
for these 10 pairs of equations. Therefore only 20 independent equations are left for solving the 20 coefficients in the
vector c. The solutions are shown in Table 4 and listed in

Optical Engineering, Vol. 36 No. 3, March 1997

Downloaded From: http://opticalengineering.spiedigitallibrary.org/ on 08/06/2015 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

Shen, Chang, and Wan: Zernike polynomial fitting . . .

Table 1 Simultaneous equations to solve all coefficients of unknown wavefront with all s i and t i listed in Table 3.

Table 2 Definitions of g i and h i .

Measurements in the x Direction

Wavefront Estimate in the x Direction

Terms of
Monomial
Polynomials
1

x
y
x2
xy
y2
x3
x 2y
xy 2
y3
x4
x 3y
x 2y 2
xy 3
y4

Equations

Contrary Pairs

c 1 22 c 7 13 c 175 s 0
2 c 4 2 c 5 26 c 1213 c 135 s 1
c 3 23 c 115 s 2
2 c 6 13 c 7 14 c 16212c 175 s 3
c 8 2 c 9 24 c 1814 c 195 s 4
c 6 1 c 7 24 c 1624 c 175 s 5
6 c 1224 c 131 c 145 s 6
2 c 1012 c 115 s 7
2 c 122 c 145 s 8
c 1012 c 115 s 9
c 1525 c 16110c 175 s 10
2 c 1823 c 191 c 205 s 11
2 c 151 c 1612 c 175 s 12
2 c 182 c 192 c 205 s 13
c 1513 c 1612 c 175 s 14

(T1.1.1)
(T1.1.2)
#1

(T1.1.3)

#2

(T1.1.4)
(T1.1.5)

#3

(T1.1.6)

#4

(T1.1.8)

#5
#6

(T1.1.9)
(T1.1.10)

#7

(T1.1.11)
(T1.1.12)

#8
#9

(T1.1.13)
(T1.1.14)

#10

(T1.1.15)

(T1.1.7)

Wavefront Estimate in the y Direction

Terms of
Monomial
Polynomials
1

x
y
x2
xy
y2
x3
x 2y
xy 2
y3
x4
x 3y
x 2y 2
xy 3
y4

Equations

Contrary Pairs

c 2 22 c 8 13 c 185 t 0
c 3 23 c 115 t 1
2 c 4 1 c 5 26 c 1223 c 135 t 2
c 8 2 c 9 24 c 1814 c 195 t 3
c 6 1 c 7 24 c 1624 c 175 t 4
3 c 8 1 c 9 212c 1824 c 195 t 5
2 c 1012 c 115 t 6
2 c 122 c 145 t 7
c 1012 c 115 t 8
6 c 1214 c 131 c 145 t 9
2 c 1823 c 191 c 205 t 10
2 c 151 c 1612 c 175 t 11
2 c 182 c 192 c 205 t 12
c 1513 c 1612 c 175 t 13
10c 1815 c 191 c 205 t 14

(T1.2.1)
#1

(T1.2.2)
(T1.2.3)

#2
#3

(T1.2.4)
(T1.2.5)

#4
#5

(T1.2.7)
(T1.2.8)

#6

(T1.2.9)
(T1.2.10)

#7
#8

(T1.2.11)
(T1.2.12)

#9

(T1.2.13)

#10

(T1.2.14)

(T1.2.6)

(T1.2.15)

the order from c 20 to c 1 ~c 0 is a constant that cannot be

found from the shearing interferometry!. Actually the averaging process can be regarded as the solution of the leastsquares fitting to determine a constant which meets the two
measured values s j and t l well.

Wavefront Estimate from Finite Shear

The wavefront estimated from the finite shearing interferogram is more complicated than the small shear case discussed, because it involves the high-order terms of the
shear. However, a set of simultaneous equations still can be
derived by the same calculation process described in the
previous section, and actually they are the same as derived
in Table 1. The proof is given later.

2.2

a 0 2 a 4 1 a 125 g 0
a 1 22 a 7 5 g 1
a 2 22 a 8 5 g 2
2 a 4 2 a 5 26 a 1213 a 135 g 3
a 3 23 a 115 g 4
2 a 4 1 a 5 26 a 1223 a 135 g 5
2 a 6 13 a 7 5 g 6
a 82 a 95 g 7
a 61 a 75 g 8
3 a 81 a 95 g 9
6 a 1224 a 131 a 145 g 10
2 a 1012 a 115 g 11
2 a 122 a 145 g 12
a 1012 a 115 g 13
6 a 1214 a 131 a 145 g 14
Measurements in the y Direction
b 0 2 b 4 1 b 125 h 0
b 1 22 b 7 5 h 1
b 2 22 b 8 5 h 2
2 b 4 2 b 5 26 b 1213 b 135 h 3
b 3 23 b 115 h 4
2 b 4 1 b 5 26 b 1223 b 135 h 5
2 b 6 13 b 7 5 h 6
b 82 b 95 h 7
b 61 b 75 h 8
3 b 81 b 95 h 9
6 b 1224 b 131 b 145 h 10
2 b 1012 b 115 h 11
2 b 122 b 145 h 12
b 1012 b 115 h 13
6 b 1214 b 131 b 145 h 14

(T2.1.1)
(T2.1.2)
(T2.1.3)
(T2.1.4)
(T2.1.5)
(T2.1.6)
(T2.1.7)
(T2.1.8)
(T2.1.9)
(T2.1.10)
(T2.1.11)
(T2.1.12)
(T2.1.13)
(T2.1.14)
(T2.1.15)

(T2.2.1)
(T2.2.2)
(T2.2.3)
(T2.2.4)
(T2.2.5)
(T2.2.6)
(T2.2.7)
(T2.2.8)
(T2.2.9)
(T2.2.10)
(T2.2.11)
(T2.2.12)
(T2.2.13)
(T2.2.14)
(T2.2.15)

For the finite shear case, the relationships between the

original and the sheared wavefronts are expressed by the
equations
DW s ~ x,y ! 5W ~ x1s, y ! 2W ~ x,y ! ,

~12!

DW t ~ x,y ! 5W ~ x, y1t ! 2W ~ x,y ! .

~13!

From Eqs. ~12! and ~13!, the relationships between c i , a i ,

and b i are established in the process as in the small shear
case, but these equations involve more terms. For example,
by equating the coefficient of the constant term in Eq. ~8!,
the relationship
s ~ c 1 22c 7 13c 17! 5a 0 2a 4 1a 12

~14!

is found, but the corresponding equation generated in Eq.

~12! is
Optical Engineering, Vol. 36 No. 3, March 1997

Downloaded From: http://opticalengineering.spiedigitallibrary.org/ on 08/06/2015 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

907

Shen, Chang, and Wan: Zernike polynomial fitting . . .

Table 3 Definitions of s i and t i for the cases of the small shear and
the finite shear with all g i and h i listed in Table 2.
Shearing Measurements in the x Direction (Shear5s )
Small Shear Case
Finite shear case

s 05 g 0/ s
s 1 5 g 1 /2s
s 2 5 g 2 /2s
s 3 5 g 3 /3s
s 4 5 g 4 /3s
s 5 5 g 5 /3s
s 6 5 g 6 /4s
s 7 5 g 7 /4s
s 8 5 g 8 /4s
s 9 5 g 9 /4s
s 105 g 10/5s
s 115 g 11/5s
s 125 g 12/5s
s 135 g 13/5s
s 145 g 14/5s

s 0 5 g 0 / s 2 g 1 /21 g 3 s /62 g 10s 3 /30

s 1 5 g 1 /2s 2 g 3 /21 g 6 s /4
s 2 5 g 2 /2s 2 g 4 /41 g 7 s /12
s 3 5 g 3 /3s 2 g 6 /21 g 10s /3
s 4 5 g 4 /6s 2 g 7 /61 g 11s /12
s 5 5 g 5 /3s 2 g 8 /61 g 12s /18
s 6 5 g 6 /4s 2 g 10/2
s 7 5 g 7 /12s 2 g 11/8
s 8 5 g 8 /12s 2 g 12/12
s 9 5 g 9 /4s 2 g 13/8
s 105 g 10/5s
s 115 g 11/20s
s 125 g 12/30s
s 135 g 13/20s
s 145 g 14/5s

(T3.1.1)
(T3.1.2)
(T3.1.3)
(T3.1.4)
(T3.1.5)
(T3.1.6)
(T3.1.7)
(T3.1.8)

t 05 h 0/ t
t 1 5 h 1 /2t
t 2 5 h 2 /2t
t 3 5 h 3 /3t
t 4 5 h 4 /3t
t 5 5 h 5 /3t
t 6 5 h 6 /4t
t 7 5 h 7 /4t
t 8 5 h 8 /4t
t 9 5 h 9 /4t
t 105 h 10/5t
t 115 h 11/5t
t 125 h 12/5t
t 135 h 13/5t
t145h14/5t

t 0 5 h 0 / t 2 h 2 /21 h 5 t /62 h 14t /30

t 1 5 h 1 /2t 2 h 4 /41 h 8 t /12
t 2 5 h 2 /2t 2 h 5 /21 h 9 t /4
t 3 5 h 3 /3t 2 h 7 /61 h 12t /18
t 4 5 h 4 /6t 2 h 8 /61 h 13t /12
t 5 5 h 5 /3t 2 h 9 /21 h 14t /3
t 6 5 h 6 /4t 2 h 11/8
t 7 5 h 7 /12t 2 h 12/12
t 8 5 h 8 /12t 2 h 13/8
t 9 5 h 9 /4t 2 h 14/2
t 105 h 10/5t
t 115 h 11/20t
t 125 h 12/30t
t 135 h 13/20t
t 145 h 14/5t

13c 7 14c 16212c 17! 1s 4 ~ 6c 1224c 131c 14! 1s 5 ~ c 15

25c 16110c 17! 5a 0 2a 4 1a 125g 0 ,

which indicates that more unknowns c i are now related to

the same measurement quantity of a 0 2a 4 1a 12 or g 0 , as
defined in Table 2. Similarly, by equating the coefficients
of the various order of x in Eq. ~12!, more equations are
found, in which for the x 4 term is
5s ~ c 1525c 16110c 17! 5g 10 ,

~16!

for the x 3 term is

(T3.1.10)
(T3.1.11)

4s ~ 6c 1224c 131c 14! 110s 2 ~ c 1525c 16110c 17! 5g 6 , ~17!

(T3.1.12)

for the x 2 term is

(T3.1.13)
(T3.1.14)
(T3.1.15)

3s ~ 2c 6 13c 7 14c 16212c 17! 16s 2 ~ 6c 1224c 131c 14!

110s 3 ~ c 1525c 16110c 17! 5g 3 ,

~18!

and for the x term is

(T3.2.1)
(T3.2.2)

2s ~ 2c 4 2c 5 26c 1213c 13! 13s 2 ~ 2c 6 13c 7 14c 16212c 17!

(T3.2.3)
(T3.2.4)

14s 3 ~ 6c 1224c 131c 14! 15s 4 ~ c 1525c 16110c 17!

(T3.2.5)
(T3.2.6)
(T3.2.7)
(T3.2.8)
(T3.2.9)
(T3.2.10)
(T3.2.11)
(T3.2.12)
(T3.2.13)
(T3.2.14)
(T3.2.15)

5g 1 .

~19!

Here all g i have the same definitions as in the small shear

case, and a similar case is found for h i in the y shear measurement. Note that in spite of the extra terms shown on the
left-hand side of Eqs. ~15! to ~19!, there is no new combination in c i for these extra terms, i.e., all the quantities in
the parenthesis on the left-hand sides of these equations
@Eqs. ~15! to ~19!# still can be represented by s i in Table 1.
It is equivalent to say that the equations in Table 1 are still
valid for the finite shear case.
In other words, as in the case of small shear, we solve all
s i and t i first instead of solving c i directly. For instance,

Table 4 Solutions of the simultaneous equations shown in Table 1. (The definitions of s i and t i are
listed in Table 3 for the small shear case and for the finite shear case, respectively.)

c 205(2 t 14210t 12210s 1315 s 1115 t 10)/32

c 195(2 t 1422 t 1222 s 1323 s 1123 t 10)/32
c 185(2 t 1412 t 1212 s 131 s 111 t 10)/32
c 175(2 s 1012 s 1212 t 111 s 141 t 13)/32
c 165(22 s 1012 s 1212 t 1113 s 1413 t 13)/32
c 155(2 s 10210s 12210t 1115 s 1415 t 13)/32
c 145( s 6 1 t 9 23 s 8 23 t 7 )/8
c 135( t 9 2 s 6 )/8
c 125( s 6 1 s 8 1 t 7 1 t 9 )/16
c 115( s 7 1 s 9 1 t 6 1 t 8 )/8

908

~15!

(T3.1.9)

Shearing Measurements in the y Direction (Shear5t )

Small Shear Case Finite Shear Case
3

s ~ c 1 22c 7 13c 17! 1s 2 ~ 2c 4 2c 5 26c 1213c 13! 1s 3 ~ 2c 6

c 105( s 9 1 t 8 2 s 7 2 t 6 )/4
c 9 5(32c 1912 t 5 23 s 4 23 t 3 )/8
c 8 5(32c 1812 t 5 1 s 4 1 t 3 )/8
c 7 5(32c 1712 s 3 1 s 5 1 t 4 )/8
c 6 5(32c 1622 s 3 13 s 5 13 t 4 )/8
c 5 5(6 c 131 t 2 2 s 1 )/2
c 4 5(12c 121 s 1 1 t 2 )/4
c 3 5(6 c 111 s 2 1 t 1 )/2
c 2 52 c 8 23 c 181 t 0
c 1 52 c 7 23 c 171 s 0

Optical Engineering, Vol. 36 No. 3, March 1997

Downloaded From: http://opticalengineering.spiedigitallibrary.org/ on 08/06/2015 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

Shen, Chang, and Wan: Zernike polynomial fitting . . .

from Eqs. ~15! to ~19! s 10 , s 6 , s 3 , s 1 , and s 0 can be solved

as follows:
c 1525c 16110c 175s 105g 10/5s,

~20!

6c 1224c 13 1c 145s 6 5g 6 /4s2g 10/2,

~21!

2c 6 13c 7 14c 16212c 175s 3 5g 3 /3s2g 6 /21g 10s/3, ~22!

2c 4 2c 5 26c 1213c 135s 1 5g 1 /2s2g 3 /21g 6 s/4,

~23!

c 1 22c 7 13c 175s 0 5g 0 /s2g 1 /21g 3 s/62g 10s 3 /30,

~24!

and the new relationships between s i , and t i and g i and h i

for the finite shear case are shown in Table 3.
In this way, we can see that for the finite shear case the
solutions of all coefficients in the vector c still can be
solved from the simultaneous equations shown in Table 1,
and only the definitions of s i and t i are different from the
ones for the small shear case. Therefore, the solutions
shown in Table 4 are also valid for the finite shear case,
provided that the definitions of s i and t i for the finite shear
are adopted ~Table 3!.
This section shows a systematic procedure of solving the
Zernike polynomial coefficients for an unknown wavefront
from the shearing interferometry. The analytic solutions
shown in Table 4 are derived in the case that the wavefront
under test is expanded to the fifth order of Zernike polynomials. The explicit expression is not only convenient in the
computation process but it is also easy for tracing the relationship between the Zernike coefficients ~vector c! of the
unknown wavefront and the Zernike coefficients ~vectors a
and b! of the shearing interferograms.
3

Computer Simulation

The performance of the new wavefront estimate algorithm

was tested by computer simulation. The comparison between the reduced results and the preset ones is served as a
check of the new algorithm. The simulation is also repeated
for the LSF for the purpose of comparison.
The flow chart of computer simulation is shown in Fig.
2. The wave aberration is described by the Zernike polynomials with the preset coefficients. Then using the definitions shown in Eqs. ~3!, ~4!, ~12!, and ~13!, the simulations
for the small shear and the finite shear in the x and y
directions are calculated, respectively. For the small shear
case, the sampling points of the shearing interferogram almost filled the whole exit pupil of the optics under test. In
the example shown here, there are 1257 samples. For the
finite shear case, the data are sampled only at the grids in
the overlap region of the sheared wavefronts ~1175 samples
here!. The sampling data are fitted to the Zernike polynomials. For the examples shown here, the wave aberration
represented by the Zernike coefficients are estimated by the
LSF and the new algorithms, respectively. The differences
between the reduced coefficients and the preset coefficients
will indicate if any error of the algorithm exists. In addition, the random noise is added to the calculated measurements as a simulation of the measurement noise.
In the simulations, the Zernike coefficients representing
the wave aberration are set ~in the unit of wavelength! as c 3
~the astigmatism with axis at 645 deg!50.1; c 4 ~the focus

Fig. 2 Flow chart of computer simulation.

shift!50.5; c 7 ~the third order coma along the x axis!

520.2; c 12 ~the third-order spherical aberration!51.0; and
the others are zeros. In the finite shear case, the shear is 0.1
both in the x and y directions. The conditions with noise
and without noise are set for both the small shear and the
finite shear. The added noise is generated randomly and
controlled by two parameters: the central level of the noise
and the deviation. They are preset as 0 and 20% of the root
mean square ~rms! value of the shearing wavefront. The
results are listed in Table 5, including the rms of the preset
wave aberration and the estimate results of two algorithms
under four conditions.
As shown in Table 5, before the noise was added, the
wavefronts estimated by the two algorithms are shown with
high accuracy in the cases of both the small shear and the
finite shear. The estimate error between the preset and the
reduced wave aberration is very small. The consistent results identify the correctness of the new algorithm. After
the noise was introduced, the reduced coefficients have
slight differences from the preset ones for the new algorithm and for the LSF, but the error of the rms value of the
estimated wavefront is still smaller than 0.2%. This small
estimate error in the noise condition results from the fact
that the Zernike polynomials fit only the data of low spatial
frequency. In summary, these simulations indicate that the
performance of the new algorithm is as good as that of the
LSF, but less computation is used.
4 Application On Measurements
Further comparison of the new algorithm with the LSF was
done by the phase reduction of the shearing interferograms
shown in Fig. 3, which are adopted from Ref. 6. The shears
in both the x and y directions are 0.36, as specified by Ref.
6, and the interferograms shown in Fig. 3 are rotated to
make the shear either in the horizontal or in the vertical
direction, in compliance with the calculation procedure.
Then by sampling the centers of the black fringes, two
interferograms sheared in the x and y directions were fitted
to Zernike polynomials. From the fitting coefficients of
Optical Engineering, Vol. 36 No. 3, March 1997

Downloaded From: http://opticalengineering.spiedigitallibrary.org/ on 08/06/2015 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

909

Shen, Chang, and Wan: Zernike polynomial fitting . . .

Table 5 Results of computer simulation.
Small Shear Case
Preset
Zernike Coefficients

Free from Noise

LSF
New

Finite Shear Case

Noise Added
LSF
New

Free from Noise

LSF
New

Noise Added
LSF
New

0.0

0.000

0.000

20.009

0.005

0.000

0.000

20.001

0.007

2
3
4

0.0
0.1
0.5

0.000
0.100
0.500

0.000
0.100
0.500

0.002
0.112
0.498

20.004
0.117
0.498

0.000
0.100
0.500

0.000
0.100
0.500

0.001
0.107
0.497

20.004
0.112
0.497

0.0

0.000

0.000

20.002

0.016

0.000

0.000

20.005

0.013

0.0

0.000

0.000

0.010

0.007

0.000

0.000

0.005

0.008

7
8
9

20.2
0.0
0.0

20.200
0.000
0.000

20.200
0.000
0.000

20.191
0.011
0.011

20.195
0.013
0.007

20.200
0.000
0.000

20.200
0.000
0.000

20.192
0.008
0.008

20.199
0.007
0.001

10

0.0

0.000

0.000

0.003

0.003

0.000

0.000

0.000

20.001

11

0.0

0.000

0.000

20.002

20.007

0.000

0.000

20.005

20.009

12
13
14
15

1.0
0.0
0.0
0.0

1.000
0.000
0.000
0.000

1.000
0.000
0.000
0.000

1.002
0.008
0.003
0.002

1.002
0.004
0.003
0.000

1.000
0.000
0.000
0.000

1.000
0.000
0.000
0.000

1.006
0.008
0.002
0.002

1.005
0.004
0.002
0.005

16
17
18
19
20

0.0
0.0
0.0
0.0
0.0

0.000
0.000
0.000
0.000
0.000

0.000
0.000
0.000
0.000
0.000

0.003
0.003
20.005
20.004
0.003

0.002
0.001
20.007
0.000
0.001

0.000
0.000
0.000
0.000
0.000

0.000
0.000
0.000
0.000
0.000

0.004
0.003
20.005
20.002
0.005

0.003
0.004
20.005
0.001
0.003

0.538

0.538

0.538

0.538

0.539

0.538

0.538

0.539

0.539

rms of wavefront

these measurements, the Zernike polynomial coefficients of

the wavefront under test were found by the new algorithm
and are listed in Table 6. The contour profile of the wave
aberration is shown in Fig. 4~a!, and the contour profile
generated by the LSF in Ref. 6 is shown in Fig. 4~b! for
comparison. The rms and the peak-to-peak ~P-P! values are
reported as 0.17 and 0.85 in Ref. 6, and in Fig. 4~a! the
corresponding rms and P-P values found by the new algorithm are 0.176 and 0.826, which shows no significant difference between the wavefronts estimated by the two algorithms.
5 Error Discussion
Table 4 gives the relationship between the Zernike polynomial coefficients of the wavefront under test and the measurement quantities s i and t i . This straightforward relationship can help us to estimate the effects of various
Table 6 Estimated Zernike polynomial coefficients (in the unit of
wavelength) of the wavefront from the shearing interferograms (Fig.
9(a) in Ref. 6). (The Zernike polynomials are according to Ref. 9.)

Fig. 3 Shearing interferograms [Fig. 9(a) in Ref. 6].

910

c1
c2
c3
c4
c5

0.000
0.000
0.029
0.000
20.005

c6
c7
c8
c9
c 10

20.017
0.207
0.041
0.006
20.002

c 11
c 12
c 13
c 14
c 15

20.010
0.355
20.020
20.007
0.003

c 16
c 17
c 18
c 19
c 20

Optical Engineering, Vol. 36 No. 3, March 1997

Downloaded From: http://opticalengineering.spiedigitallibrary.org/ on 08/06/2015 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

0.018
20.021
0.002
0.001
20.001

Shen, Chang, and Wan: Zernike polynomial fitting . . .

Fig. 5 Shearing interferogram (a) before rotation and (b) after 90deg counterclockwise rotation.

As shown in Fig. 5, the first interferogram is sheared in the

x direction, and after the lens or mirror is rotated counterclockwise by 90 deg, the second interferogram is sheared in
the y direction. The shearing interferogram before the rotation is represented by
DW s ~ x,y ! 5W ~ x1s, y ! 2W ~ x,y ! 1t px x1t py y,

~25!

and the interferogram after the rotation is represented by

DW T ~ x,y ! 5W ~ x,y1t ! 2W ~ x, y ! 1t py x2t px y,

Fig. 4 Contour of the wavefront estimated (a) by the new algorithm

and (b) by the least-squares fitting algorithm [Fig. 12(c) in Ref. 6].

measurement errors on the wave aberration under test. On

the other hand, the LSF does not give such an explicit
expression. In this section, the effects of the shear plate
wedge, the test wavefront variation in the x and y shear
interferograms, the errors from the measurement of the
shearing interferogram, and the errors in determining the
shear are discussed using the explicit expression of the
wave aberration coefficients.
5.1 Shear Plate Wedge
The shearing interferogram is obtained from the interference of the waves reflected from the two surfaces of the
shear plate. Any difference of these two surfaces is treated
as the source of error. Therefore a surface of good flatness
and parallelism of the shear plate are required. Although
the wedge problem can be solved by setting the wedge
direction perpendicular to the shear direction, we would
like to include this kind of error as part of a systematic
discussion.
The wedge of the shear plate introduces tilt aberration
into the shearing interferogram, and the additional aberration can be composed of two components of tilt t px and t py .

~26!

where the minus sign in front of t px in Eq. ~26! results from

the rotation of the coordinates. The effects of the shear
plate wedge on a 1 , a 2 , b 1 , and b 2 are found as Da 1 5t px ,
Da 2 5t py , Db 1 5t py , and Db 2 52t px . According to Eqs.
~T2.1.2!, ~T2.1.3!, ~T2.2.2!, and ~T2.2.3! in Table 2, the
errors in g i and h i are Dg 1 5t px , Dg 2 5t py , Dh 1 5t py , and
Dh 2 52t px . Then from Eqs. ~T3.1.2!, ~T3.1.3!, ~T3.2.2!,
and ~T3.2.3! in Table 3, the variables s i and t i in the finite
shear case have errors Ds 1 5t px /2s, Ds 2 5t py /2s,
Dt 1 5t py /2t, and Dt 2 52t px /2t, and the final errors in wave
aberration are estimated as
Dc 4 5

Dc 5 52
Dc 3 5

S D
S D
S D

t px 1 1
2 ,
8 s t

t px 1 1
1 ,
4 s t

t py 1 1
1 .
4 s t

~27!
~28!
~29!

As indicated in Fig. 5, the shear s and t of the two shearing

interferograms should be equal but with opposite sign.
Therefore the errors in c 3 and c 5 are eliminated and only
the error in c 4 is left as
Dc 4 5

t px
.
4s

~30!

The same result will occur in the case of rotating the test
wave 90 deg clockwise. This derivation indicates that, for
the same shear in the x and y directions, the tilt in the shear
direction results the error only in the defocus term of the
test wavefront, although this fact is well known beforehand.
Optical Engineering, Vol. 36 No. 3, March 1997

Downloaded From: http://opticalengineering.spiedigitallibrary.org/ on 08/06/2015 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

911

Shen, Chang, and Wan: Zernike polynomial fitting . . .

Test Wavefront Variation at 90-deg Rotation of

the Lens Under Test
When the wavefront under test is rotated to produce the
shearing interferogram sheared in the perpendicular direction, the wavefront may be different due to tilt or defocus
of the optical system. The effect of the additional aberrations on the wavefront estimate can be checked by the new
algorithm.
The first lateral shear is assumed in the x direction, and
the second measurement is in the y direction. If the defocus
aberration d w def is introduced after the rotation, Eq.
~T1.2.3! of Table 1 should be rectified as

Table 7 Magnification of the estimate error from the measurement

error with respect to various shearing quantities.

2c 48 1c 5 26c 1223c 135t 28 ,

Dc 5 5

5.2

~31!

where the defocus coefficient c 4 and the corresponding

measurement t 2 in the y shear interferogram are replaced
by c 48 ( 5 c 4 1 d w def) and t 28 ( 5 t 2 1 Dt 2 ). However, in the
wavefront estimate algorithm, the defocus term is assumed
to be the same in these two measurements, which means
c 48 in Eq. ~31! is still set as c 4 and only the variation in
measurement is considered. The final solution of c 4~sol!, as
indicated in Table 4, is
c 4 ~ sol! 53c 121 ~ s 1 1t 28 ! /4,

~33!

but according to Eq. ~31!, Dt 82 is equal to 2 d w def , which

results in the error of d w def/2 in c 4 . The half defocus error
is also apparent because of the average procedure adopted
here. In addition, the final solution of c 5 is affected by the
variation of t 2 and has the error
Dc 5 5 d w def .

0.01
0.05

~34!

The preceding analysis shows that the defocus introduced by the rotation only affects the defocus c 4 and the
astigmatism c 5 of the wavefront under test, and the other
aberrations of the wavefront under test are not affected.

D c 4 /D a 1

50
10

25
5

D c 5 /D a 1
50
10

0.1
0.2

5
2.5

2.5
1.25

5
2.5

0.5

0.5

1 Db 2 Da 1
2
.
4
t
s

~37!

Assuming the same shear is applied in the x and y shear

interferograms (s5t) and that the measurements of the tilt
of
the
OPD
have
the
same
errors
(Da 1 5Da 2 5Db 1 5Db 2 ), the variations of the estimated
errors with respect to Da 1 are
Dc 3 ~ max! 1
5 ,
Da 1
2s

~32!

and the estimate error of c 4 is

Dc 4 5Dt 28 /4,

D c 3 /D a 1

Dc 4 ~ max! 1
5 ,
Da 1
4s

and

Dc 5 ~ max! 1
5 .
Da 1
2s
~38!

Because s is a fraction of the aperture, the error introduced

is magnified by a factor of 1/s. Some examples showing the
magnification effects are listed in Table 7.

Error in Shear
The accuracy in the determination of shear is very critical
to the phase reduction of shearing interferometry. In this
new algorithm, the effects of shear on the aberration coefficients of the estimated wavefront can be found in the expression of s i and t i , as shown in Table 4. In the small
shear case, the relationship is straightforward, because s i
and t i are proportional to the reciprocal of the shear. If the
shear and its measurement error are the same in the x and y
shear measurements ~s5t and Ds5Dt!, the relative deviation for all the wavefront coefficients c i will be

5.4

5.3

Dc i
Ds
.
52
ci
s

Dc 3 5

From this result, it can be seen that the smaller the shear,
the larger the relative error in the aberration coefficients,
which indicates that the measurement error of the shear is
the major error source for the wavefront reconstruction in
the small shear case.
For the finite shear case, the effect of the error in determining the shear on the estimated wavefront is very complicated, since the estimate error cannot be expressed by a
simple 1/s relationship, as in the small shear case. However, as shown in Table 3, in the finite shear case the 1/s
and 1/t terms are still the dominant terms in s i and t i , so
the relative error of the estimated coefficient is still roughly
proportional to 2Ds/s.

Errors from Measuring the Optical Path

Difference
The effect of the error from measuring the optical path
difference ~OPD! of the interferogram can be analyzed by
the new algorithm. The measurement error of the tilt in the
OPD for example, is discussed in the following.
The measurement error of tilt of the OPD is represented
by the deviation of the corresponding fitting coefficients,
which are Da 1 and Da 2 ~tilt in the x direction, tilt in the y
direction! in the x shear measurement and Db 1 and Db 2 in
the y shear measurement, respectively. As the derivation in
Sec. 5.1, the final errors are

S
S

D
D

1 Da 2 Db 1
1
,
4
s
t

1 Da 1 Db 2
1
Dc 4 5
,
8
s
t
912

~35!
~36!

~39!

6 Conclusion
We derived an explicit solution for the aberration coefficients of the wavefront under test by using the coefficients

Optical Engineering, Vol. 36 No. 3, March 1997

Downloaded From: http://opticalengineering.spiedigitallibrary.org/ on 08/06/2015 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

Shen, Chang, and Wan: Zernike polynomial fitting . . .

of Zernike polynomials of the shearing interferograms. The

problem of having more equations than the number of coefficients to be fitted is treated by an averaging algorithm.
The test results reduced from the new algorithm and from
the least-squares fitting algorithm are almost the same, as
shown in the examples of numerical simulation and real
interferograms. The solution shown in Table 4 is a general
form independent of the magnitude of the shear. Except
that there is less calculation in the new algorithm, the main
advantage of the new algorithm is its explicit expression,
which can be used to trace the contribution of the measured
aberration coefficients to the aberration coefficients under
test. This property should be very important for analyzing
the sources of error in the shearing interferometry.

Acknowledgment
The authors are grateful to the National Science Council of
the Republic of China for support of this work under Contract No. NSC 84-2215-E008-005.

References
1. W. D. Humbel, D. E. Vandenberg, T. W. Dey, and J. G. Pitek, Foucault knife-edge test for quantitatively determining a wavefront aberration over the entire surface of an optical imaging device, Appl.
Opt. 30~22!, 3081 ~1991!.
2. N. H. Davis and T. A. Fritz, Application of Zernike polynomials to
reduction of wavefront-slope data, Internal Report, pp. 3942, Optical Sciences Center, University of Arizona ~1977!.
3. R. Cubalchini, Modal wave-front estimation from phase derivative
measurements, J. Opt. Soc. Am. 69~7!, 972977 ~1979!.
4. W. H. Southwell, Wave-front estimation from wave-front slope
measurements, J. Opt. Soc. Am. 70~8!, 9981006 ~1980!.
5. M. P. Rimmer, Method for evaluating lateral shearing interferograms, Appl. Opt. 13~3!, 623629 ~1974!.
6. M. P. Rimmer and J. C. Wyant, Evaluation of large aberrations
using a lateral-shear interferometer having variable shear, Appl. Opt.
14~1!, 142150 ~1975!.
7. W. Shen, M. W. Chang, and D. S. Wan, Wavefront estimate from
wavefront slope measurement by comparing their Zernike polynomials fitting coefficients, Proc. SPIE 2263, 186197 ~1994!.
8. E. R. Freniere, O. E. Toler, and R. Race, Interferogram evaluation
program for the HP-9825A calculator, Opt. Eng. 20~2!, 253255
~1981!.

9. D. Malacara and S. L. DeVore, Interferogram evaluation and wavefront fitting, Chap. 13 in Optical Shop Testing, D. Malacara, Ed.,
Wiley, New York ~1992!.
Wei Shen received his BS in electrophysics in 1988 from the National Chiao Tung
University and an MS in optical sciences in
1990 from the National Central University.
Currently he is working on his doctoral thesis at the Institute of Optical Sciences of
the National Central University. His current
research includes optical testing and holography.

Ming-Wen Chang is a professor at the Institute of Optical Sciences and directs the
Optical Sciences Center at the National
Central University, Taiwan. He received
his BS in physics from the National Cheng
Kung University in 1966, an MS in geophysics from the National Central University in 1968, and a PhD in optics from the
University of Arizona in 1975. His current
interests include holography, image processing, optical computing, and optical design and testing. He is a fellow of SPIE and OSA.
Der-Shen Wan received his PhD in optics
from the University of Arizonas Optical
Sciences Center in 1987. He has been
with the Chung Shan Institute of Sciences
and Technology in Taiwan since 1978,
working in the areas of optical testing and
optical design. His current research includes interferometry and image formation
theory.

Optical Engineering, Vol. 36 No. 3, March 1997

Downloaded From: http://opticalengineering.spiedigitallibrary.org/ on 08/06/2015 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

913