Reprinted from SPIE Vol.

192-1ntefferometrv
C 1979 by the Society 01 Photo-Opticollnstrumentatlon Engineers, Box 10. Bellinghem. WA 98227'()()10 USA

Differential technique for accurately measuring the radius of

curvature of long radius concave optical surfaces

Mark C. Gerchman. George C. Hunter

Zygo Corporation. laurel Brook Road
Middlefield. Connecticut 06455

Abstract
Described is a technique for accurately measuring the radius of curvature of very
long radius concave optical surfaces in a relatively short working length. Presented is
the basic theory of the technique and the detailed information necessary to provide for
its practical application. Using the technique, radii of up to 175 meters can be mea-
sured to an accuracy of better than 0.01 percent with a one-meter long scale and slide,
and in an overall working length of less than 5 meters. Generalizations of the technique
allow for the measurement of convex surfaces and even longer·radius concave surfaces. .
Pfte-t-ogpElphs; interferograms,-·c-alculations, and· an error analysis-illustrate the technique.
Introduction
The methods for measuring the radius of curvature of concave spherical optical sur-
faces are well known' and can be divided into two. major categories: direct measurement
methods in which the center of curvature is located and the distance between the surface
and center of curvature is measured, and indirect measurement methods where the radius
of curvature is calculated from a measurement of the sagitta of the surface. However,
significant problems arise with either method when measuring very long radii. For
example, direct measurement methods require large working distances and correspondingly
long measuring transducers. With indirect methods, small errors in the sagitta measure-
ment produce large errors in the calculated value of the radius.
The technique presented here provides a method for accurately calculating the
radius of curvature of long radius concave optical surfaces by measuring differences in
cavity lengths between successive-order confocal (retroreflecting) cavities formed by
the surface under test and a plano reference surface.
Description
In this differential measuring technique a cavity is formed between a plano surface
and the concave surface whose.rad.ius .we wish to measure. By varying the separation be-
tween surfaces we note that certain separations correspond to confocal cavity configura-
tions (i.e., those configurations where collimated light entering the cavity through
the plano surface is brought to a focus on one or the other of the surfaces). By mea-
suring the distance traveled between any two successive confocal configurations we can
calculate the radius of the concave surface. Because the separation between surfaces
is greatly reduced in the higher order confocal configurations and the distance travelled
between successive order configurations is smaller, surfaces with very long radii can be
measured in short working spaces, using a correspondingly short measuring slide and
transducer.
The successive orders of the confocal retroreflecting cavities are specified by a
configuration number n. The n = 1 configuration is the standard plano-concave cavity
where all rays incident normal from the plano surface reflect off the concave surface
and come to focus on the plano surface (Fig. 1). The n = 2 configuration reduces the
distance between the plano and concave surfaces so that all rays incident normal from
the plano surface reflect off the concave surface, fold off the plano surface, and come
to focus on the concave surface (Fig. 2). The n = 3 configuration further reduces the
distance between the plano and concave surfaces so that all rays incident normal from
the plano surface reflect off the concave surface, fold off the plano surface, again
reflect off the concave surface, and come to focus on the plano surfa~e (Fig. 3).
Still higher order configurations add additional folds and reflections so that all
odd order configurations focus the system on the plano surface and all even order con-
figurations focus the system on the concave surface.
Though many standard optical techniques can be used to examine the confocal cavity
for best focus, a particularly sensitive arrangement is to evaluate the cavity focus
interferometrically. This can be accomplished by having the plano surface of the cavity
be the beam splitting surface of the transmission element in a Fizeau interferometer.

SPIE Vol. 192 Interferometry (1979) / 75

 GERCHMAN. HUNTER

TRANS. FLAT
REFLECTIVITY = rl
ZI-----------------------------~.,~
!-
CENTER SPOT - - _ _ - .4
REFLECTIVITY=rl' _ _ _- - - TEST MIRROR-J
_ _ _ _ -- n=1 REFLECTIVITY=rl

--------------- .
FIGURE 1
---========-=~

' - - - - - - - - Z2 - - - - - - - - - - . 1 I - - - - - Z3 - - - - . . - j

------
f.....

------------~--.--=~=-~---

I-=:::====-
4---------- -------- n= 2

--------- -- ~~==~~--+_

---------
---

FIGURE 2
====--1

FIGURE 3

The technique can be generalized to provide for the measurement of convex surfaces
by using a non-plano Fizeau transmission element in place of the plano element.
Analysis
The equations that relate the cavity lengths, zn' and the radius of curvature of
the concave surface, R, as,a function of the configuration number, n, are derived from
a paraxial ray analysis. This is accomplished by repeated application of the Gaussian
form of the image equation
2/R = 1/s m + 1/s~ , (1)

and the conjugate recursion formula,

(2)

where m k, k-1, k-2, ... , 0, and:

for n odd, k (n-l) /2;

for n even, k (n-2)/2.

An appropriate initial condition is determined for each configuration depending upon
where the system comes to focus.
For n odd,
s k• (3)

76 / SPiE Vol 1921nterferometry(1979J

 DIFFERENTIAL TECHNIQUE FOR ACCURATELY MEASURING THE RADIUS OF
CURVATURE OF LONG RADIUS CONCAVE OPTICAL SURFACES

For n even,
(4)

The analysis involves calculating the cavity length, z , in terms of the radius of
curvature, R, and respective image and object conjugatRs, sand s " where each
conjugate is calculated in terms of successive lower order ~onjuga~es. Beginning
with the initial condition [Eq. (3) or Eq. (4)], the analysis continues relating
cavity length to lower order conjugates until the final object conjugate, so' is
reached. With the substitution that for a properly focused cavity the final object
conjugate is infinity (i.e., So = w), the cavity length and radius are related by
equations of the form,
Zn = CnR. (5)

The n 3 configuration is evaluated to demonstrate the analysis (Fig. 4).

---- ---- -- ---

-- -- -------- -- .

.,.-- --...

I--Z3=S( -J_
1 ........... - ---I

I~~~--------S~-----------~~~~--Sl~ I
FIGURE 4

From Eq. (3) we have the appropriate initial condition, that is:

Z3 = s1'· (6)

By applying Eq. (1) and rearranging, we have the result,

(7)
2z 3 - R
Upon application of Eq. (2), the result is,
RZ3
s~ = 2z3 - R (8)
2z 3 -
Applying Eq. (1), again with the condition that So =~, yields upon substitution
and rearrangement,
o. (9)

The solution of interest that gives the desired relationship is the root:
Z3 = 0.1464466R. (10)

From Eq. (5), Cn = 0.1464466.

This procedure for deriving equations is valid for all configurations, although
tedious in the higher order cases. Fortunately another means exists for obtaining
these equations. Due to the Specific form of the Gaussian image equation and the
conjugate recursion formula, the odd configuration equations are identical to the even

SPlEVolI92Interferometry(1979J/ 77
 GERCHMAN. HUNTER

order Chebyshev polynomials of the first kin~ Tn +1 (x), (n=1,3,5 ••• ) where x = Z /R.
Similarly. the even configuration equations are ioentical to the even Chebyshevnpoly-
nomials of the second kin~ Un (x),(n=2,4,6 •.• ). This allows any order equation to be
easily obtained from standard mathemati.cal handbooks.' The solutions of interest are
always the first positive root. The solutions for the first nine configurations are
given in Table 1.
Table 2 expresses the relationships between the differences in cavity lengths be-
tween successive configurations, zn - zn+l, and the radius of curvature, R, of the
surface under examination.
Table 1. Constants Table 2. Equations for relating radius
for relating cavity to differential cavity length
length to radius
zn = CnR R 4 (zC z 2)
n Cn R 9.65685 (z2- z 3)
1 0.5 R 19.62512 (z3~z4)
2 0.25 R 35 .• 08255 (z4- z 5)
3 0.1464466
R 57.23525 (z5- z 6)
4 0.0954915
5 0.0669873 R 87.29584 (z6- z 7)
6 0.0495156 R 126.47741 (z7- z 8)
7 0.0380603
R 175.99437 (z8- z 9)
8 0.0301537
9 0.0244717

Error Analysis
The error sources that must be considered when using this technique are the same as
those encountered with any direct measurement method. Mech~nical errors are primarily
associated with the measuring slide and transducer, while optical 'errors limit our ability
to sense the position of best system focus.
The mechanical errors associated with the measuring slide and transducer are well
understood and are covered in detail in other references.' However, because a much
shorter slide and transducer are required when using this technique, the selection of
these components with suitable accuracy is much less restrictive than with conventional
methods.
In the following analysis we will evaluate the error in determining the position of
best focus of the confocal cavity when using a Fizeau interferometer as the focus sensor.
Determination of the position of best focus of the interferometer cavity requires
that we be able to detect small amounts of power (i.e., defocus) in the wavefront re-
turning from the cavity. If the interferometer and interference cavity were perfect,
the interferenc.e pattern would consist of straight, equally spaced fringes. Power in
the return wavefront would appear as curvature of the fringes in the interference pat-
tern and would be quite easy to evaluate. The evaluation of the amount of curvature
of the fringes is complicated, however, by the presence o·f other aberrations. The in-
terference pattern will, in general, be comatic in appearance due to the retroreflecting
natur~ of the confocal cavity. In the retroreflecting cavity, the wavefront returning
from the cavity is sheared rotationally by 1800 with respect to the reference wavefront.
When this condition exists, any asymmetrical errors in the phase of the interferometer
wavefront "print thru" and produce an interference pattern that is asymmetrical, or
comatic, in nature. We can minimize the problem due to this aberration, however, by
adjusting the plano transmission element in the interference cavity to optimize the
orientation and spacing of the fringes. An orientation can be found for which the
aberrations in the interference pattern are symmetrical about a straight central fringe
and, once adjusted, the pattern will not change as we change the cavity configuration.
A second source of wavefront deformation is spherical aberration. Since we are
primarily concerned with measuring very long radii (and consequently are working at
very large f numbers), the amount of spherical aberration present is usually negligible.
However, if the amount of spherical aberration present is significant, the optical path

78 / SPIE Vol 192 Interferometry (1979)

 DIFFERENTIAL TECHNIQUE FOR ACCURATELY MEASURING THE RADIUS OF
CURVATURE OF LONG RADIUS CONCAVE OPTICAL SURFACES

difference due to the aberration can be measured in the interference pattern and the
corresponding system focus error can be calculated and a correction can be made.~
The accuracy in evaluating the radius of curvature is related to the error in
judging best cavity focus. The error in radius, dR, can be expressed as a function of
the error in focus that produces an amount of power, w, in the wavefront returning from
th~ system. The analysis from which this function is derived is based upon the recursion
relationships already discussed. The analysis consists of relating the power in the re-
turn wavefront, w, to an error in the final image conjugate, dsd. This is then related
to the error in cavity length, dz n , and finally to the error in radius, dR.
Starting from the sagitta equation,we have,
w = R _ (R 2 _ a 2 )1/2, (11)
w w
where Rw is the radius of the return wavefront and a is the radius of the limiting
aperture of the system. For a system close to focus, Rw » a, and the expression for
the sagitt·a simplifies to,
w = a 2 12Rw' (12)
From the Newtonian form of the image equation, f2 = xx', where the focal length, f,
is R/2, the object distance, x, is Rw, and the image distance, x', is dsd. Substi-
tuting, we can write the error in the final image conjugate as
ds'
o = R2 14Rw. (13)
Substituting for Rw from Eq. (12), Eq. (13) becomes,
ds o' = wRZ'2a 2 . (14)
Defining the effective f number, F, as,
F = R/4a, (15)

we obtain upon substitution the express.ion,

ds' = 8wF2. (16)
o
To calculate the error in cavity length, dz , as a function of the final image
conjugate, dsd ' we apply the Gaussian form of tRe image equation and the conjugate
recursion formula as described previously. Because the analysis concludes at ds'
we can combine the two expressions into a single recursion relationship, 0

s m' R
(17)
2s m'-R

Taking the differential, we have,

R2
ds'm-l = 2dz n + --------n- ds ' . (18)
(2s '_R)2 m
m
Applying this recursion relationship, with the same indices as used before, an
expression for dsd is obtained. When substituted into Eq. (16) we have for the
error in cavity length, dz n ,

for n odd, 8wF 2 ,(19)

dZ n
(N o') n

4wF2
for n even, dZ n (20)
(No') n
The constants (No')n are evaluated by applying the following recursion equations.

SPIE Vol 192Interferometry(1979J/ 79

 GERCHMAN. HUNTER

For n odd: (21)

Nm-1
' (22)

For n even: (23)

(24)

where Nk = Nk = 1 and the Cn values are obtained from Table 1. Grouping all the
constanEs together, we can write the error in cavity length, dz n , as a function of
return wavefront power, w, and f number, F, as,
dz n Cn' wF 2 '
= (25)
whe.:re ~Q.e values of C~ are given in Table 3.
Writing Eq. (5) in differential form, we have,
dz n = CdR. (26)
n
From Eqs. (25) and (26),
dR = (C~/Cn)wF2, (27)

where the ~alues of C~/Cn are given in Table 3.

Table 3. Constants for error analysis The accuracy of the technique
can be improved by t~o additional means.
dZn=C~ wF 2 dR=(C~ IC n )WF2 To more accurately determine the position
of best focus; the cavity can be slightly
defocused to both sides of best focus and
n C' a graph constructed of residual power in
n
the wavefront versus slide position.
1 8. 16. From this plot the position of best focus
can be more precisely located. Also, if
2 4. 16. more than one set of successive order
configurations can be measured (e.g.
3 2. ,)~13. 6569 n = 5 to 6 artd n = 6-to 7), the mea-
surements can be correlated and sta-
4 1.105573 11.5777 tistically averaged.
5 0.666667 9.9521
6 0.432968 8.7441
7 0.292892 7.6955
8 0.207960 6.8967
9 0.152787 6.2434

Visibility Analysis
Visibility in the interference fringe pattern imposes a limit on the highest order
of the test configuration that can be used. Because the first return reflection from
the confocal cavity is of significantly higher intensity than any subsequent return
reflections, we can consider the interference phenomena as being two beam in nature.
D~fining fringe visibility', V, for two beam interference in terms of the maximum and
minimum intensity in the fringe pattern, we have,
(28)

with, (29)

80 / SPIE Vol. 192Interferometry(1979}

~~--"----~~'----~-------~

DIFFERENTIAL TECHNIQUE FOR ACCURATELY MEASURING THE RADIUS OF 192-34

CURVATURE OF LONG RADIUS CONCAVE OPTICAL SURFACES

and, (30)

where 11 is the intensity of the reflection from the beamsplitter surface of the plano
transmission element and 12 is the intensity of the first return reflection from the
confocal cavity (Fig. 5).

We can express the intensities, 11 and

12' in terms of the initial intensity, 10
incident upon the beamsplitter surface
through the transmission element and the re-
flectivities of the cavity surfaces (Fig. 1).
Assuming no absorption, we have,
11 = r 1 I o ' (31)
and:
for n odd,

12
(32)

for n = even,
FIGURE 5
(33)

where rl is the reflectivity of the plano

beamspl1tter surface, r i i s the reflec-
FRINGE VISIBILITY (V) tivity of the small cenfral focus area of
the plano beamsplitter surface, and ri is
vs. the reflectivity of the concave test sur-
TRANS. FLAT REFLECTIVITY (r,) face. .

0.9
'.. .
1.0 r-r.t---,----.---::;;--oo:::::T---r--..-----r--0::::-:::
9---,
,.'.0.9
Upon substitution, Eq. (28) becomes:

. '.'·0.04 for n = odd,

0.8 2[(1-rl)2r~ r 1' r~+1]1/2

V (34)
I 2 n-l • n+1
]. --1 1.':1" + ( 1-:-rl ) r 1 r1 r2
0.1
I I
0.6 ! for n = even,
V 0.5 2[(1-rl)2r~+1 r~+l]ll2
V (35)
2 n n+1
0.4 r 1 + ( l-r 1 ) _rl r 2 .
13
0.3 ••
Figure 6 is a graph of visibility, V, as
0.2 a function of the reflectivity of the plano
beamsplitter surface, ri' where the reflec-
tivities of the central focus area and the
0.1 concave test surface, r1 and r2, are 90
percent. Examples of interferograms with
various visibilities are shown in Fig. 7.
°0 0.1 0.4 0.5 0.6 0.7 0.8 0.9 To optimize the visibility for higher
order cavity configurations, it is evident
from the graph that a relatively high re-
flectiv.ity coating\ is required on the
FIGURE 6 plano beamsplitter"surface and a high re-
flectivity coating is. required on the sur-
face being measured. In practice, a beam-
splitter with coating of 60 percent reflec-
tivity and a central spot of 90 percent reflectivity provides workable fringe visibility
up to the n = 9 order cavity configuration.

SPIEVol t92Interferometry(1979j/ 81
 GERCHMAN. HUNTER

v 0.42 v 0.28

v 0.17 v 0.08
FIGURE 6

Application
To demonstrate the application of the technique,consider the following example.
Assume we want to measure accurately the radius of curvature of a 200mm aperture concave
surface whose nominal radius is 15 meters. To perform the measurement we have available
a Fizeau interferometer with a 150mm aperture which is mounted on a table that provides
a l-meter long workspace as measured from the plano transmission element of the inter-
ferometer, and a sUitably coated plano transmission element.
To determine which cavity configuration will fit in the workspace available, refer
to Table 1. The lowest order configuration that we can use is the first one for which
zn is less than 1 meter for R = 15 meters. From Eq. (5), Zn = CnR, and we want zn/R
to be less than 1/15 or 0.067. We therefore require a configuration for which Cn is less
than 0.067. The n =6 configuration where Cn = 0.0495156 satisfies this requirement, so
we can make our differential measurement between the n = 6 and n = 7 cavity configurations.

82 / SPIE Vol t92lnterferometry(t979}

 DIFFERENTIAL TECHNIQUE FOR ACCURATELY MEASURING THE RADIUS OF
CURVATURE OF LONG RADIUS CONCAVE OPTICAL SURFACES

From Table 2, for the n = 6 to n = 7 case, R = 87.2958370 (z6-z7). For R = 15 meters

z6-z7 = 171.829mm. We would, therefore, require a measuring slide and transducer with
a 172mm range.
The accuracy of the measurement can be evaluated by calculating the error, dR, that
would result from an error in locating the pOSition of best focus of the confocal cavity.
From Eq. (27), dR = (C~/Cn)wF2. If we assume an error of O.OSA (A= 0.6328~m) in evalu-
ating the power in the interference pattern, then w = 0.3164 x 10-4mm. From Eq. (15),
F=R/4a, the f number can be calculated. For R=lS meters and a=7Smm, F=50. Note
that the limiting aperture for this case is that of the interferometer. Table 3 pro-
vides the constants (C~/Cn). For n=6, dR=0.692mm and for n=7, dR=0.609mm. If we
assume the worst case, where the errors add, the total error in the calculated radius
is 1.301mm, or 0.0087 percent.
An additional error in the calculated value of the radius will enter due to errors
in the measuring slide and transducer. Assume we can measure the position of the test
piece to an.accuracy of 10~m. The error in radius due to this error in position can be
calculated using Eq. (26) and Table 1. From Table 1, C6 = 0.0495156 and C7 = 0.0380603.
Substituting these values for Cn and 10~m for dZ n in Eq. (26), we _!lave: for n= 6,
dR = o. 202mm or 0.0013 percent, and for n = 7, dR = o. 263mrn or 0.0018 percent. In the
worst case, where all errors add, the--' total error in the calculated value of the
radius for our example is 1.766rnrn or 0.0118 percent. If we make the more reasonable
assumption that the errors add in a root sum square (RSS) fashion, the total error in
the calculated value of the radius would be 0.98Ornm or 0.0065 percent.
If \"e repeat the above example for the n = 7 to n = 8 case, we see that the travel
required for this measurement (z -z8) is 118.S99rnrn. The maximum error in the radius
calculation for this case is 1.749rnrn, or 0.0117 percent, and the RSS error would be
0.921mm, or 0.0061 percent.
Performing both the z6 to z7 and z7 to z8 measurements would provide two values for
the radius of curvature that can be averaged to improve the accuracy. Performing both
measurements with a single setup would require a measuring slide and transducer with a
290 Smm range.
Figure 8 shows a test setup for performing radius measurements using the dif-
ferential technique.

FIGURE 8

Conclusion
The differential measuring technique provides a convenient method for measuring
the radius of curvature of long radii concave optical surfaces. A short overall test
setup, th~ requirement for a correspondingly short measuring slide and transducer,
and high accuracy are the major advantages of the differential measuring technique.
Accuracies of better than 0.01 percent can be achieved.

SPIE Vol. 192 Interferometry (1979) / 83

 GERCHMAN. HUNTER

Acknowledgments
The authors wish to express their appreciation to Ms Connie Carpentiere and
Mr. Carl Bixby for assistance in the preparation of this manuscript.

References
1. Malacara, Daniel, Optical Shop Testing, John Wiley and Sons, Inc. 1978.
2. Abramowitz,.Milton and Stegon, Irene, Handbook of Mathematical Functions, National
Bureau of Standards, Applied Mathematics Series 55, 1964.
3. Moore, Wayne R., Foundations of Mechanical Accuracy, Moore Specia~ Tool Co. 1970.
4. Conrady, A. E., Applied Optics and Optical Design, Dover Publications, Inc., 1957.
S. Born, Max and Wolf, Emil, Principles of Optics, Pergamon Press, 1975.

84 / SPIE Vol. 192 Interferometry (1979J