Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

Mirror tilt immunity interferometry

with a cat’s eye retroreflector

Fabián E. Peña-Arellano* and Clive C. Speake


School of Physics and Astronomy, University of Birmingham, Edgbaston, Birmingham B15 2TT, United Kingdom
*Corresponding author: fepa@star.sr.bham.ac.uk

Received 5 October 2010; revised 28 December 2010; accepted 28 December 2010;


posted 4 January 2011 (Doc. ID 136121); published 18 February 2011

This paper describes the behavior of a cat’s eye retroreflector, which is incorporated in a novel way in a
double-pass homodyne polarization interferometer. The amount of mirror tilt immunity a cat’s eye pro-
vides is calculated within the paraxial approximation using 4 × 4 ABCD matrices. It is found that there is
a position of the target mirror in which the tilt immunity is at a maximum. A real cat’s eye, which is
affected by aberrations, is optimized and examined using Zemax software for optical design. The max-
imum amount of mirror tilt immunity is numerically calculated and written in terms of defocus and
spherical aberration. Finally, for the purposes of comparison, the amplitude of the Lissajous pattern
as the target mirror tilts is calculated for both an interferometer with an integrated cat’s eye and an
interferometer with a cube corner. © 2011 Optical Society of America
OCIS codes: 120.3180, 120.4570.

1. Introduction incidence. The most popular types of retroreflectors


Interferometers are useful devices, but they are dif- are cube corners and cat’s eyes. A cube corner is con-
ficult to align. In a Michelson interferometer, for in- structed with three mutually perpendicular reflect-
stance, the tilt of the mirrors produces straight ing surfaces. The most common realizations of the
fringes rather than circular ones. As the tilt in- cat’s eye consist of a lens with a mirror at the focal
creases, the fringes become narrower and their den- plane or a secondary mirror placed at the focal plane
sity increases, compromising the visibility of the of a primary mirror [8,9]. The most significant differ-
interference pattern. The maximum tilt for which ence between a cube corner and a cat’s eye is that
the visibility is not severely compromised corre- within a cat’s eye, the light is brought to focus at a
sponds to the angle for which the width of the reflecting surface, while the light going through a
straight fringes equals the width of the overlapping cube corner is not. In high-power applications, a cube
area of the interfering beams upon the surface of the corner may be more convenient to avoid focusing an
detector. For a single-pass Michelson interferometer, intense beam. In polarization interferometry, the use
where the mirrors are approximately 54 mm away of cube corners should be avoided since the reflec-
from the photodiode sensing the intensity of the in- tions upon its internal faces change the polarization
terference pattern and the beam is 1:3 mm in dia- state of the light (see [10], p. 87 in [11]), thus, in-
meter, such an angle is of the order of hundreds of creasing the error due to polarization mixing [12].
microradians. A cube corner appears to be the most frequently
Difficulty in aligning these devices can be eased by used type of retroreflector [3–7] since the individual
using retroreflectors instead of mirrors [1–7]. By de- elements that the cat’s eye comprises are affected by
finition, a retroreflector reflects the light back in the aberrations.
direction from which it came regardless of its angle of In this paper, the cat’s eye retroreflector is intro-
duced as a tool for mirror tilt immunity interferome-
0003-6935/11/070981-11$15.00/0 try. The analysis presented is part of a broader
© 2011 Optical Society of America effort to design an interferometer for measuring

1 March 2011 / Vol. 50, No. 7 / APPLIED OPTICS 981


the rotations of a spherical superconducting torsion
balance [13,14]. Although the amount of tilt of a
torsion balance may be of the order of microradians
(Fig. 9 in [13]), depending on the experiment,
the cryogenic environment imposes additional con-
straints that demand a more robust device than
the traditional Michelson interferometer. It is vital
to maximize the amount of tilt immunity since the
contraction of the components and their holders dur-
ing the process of cooling down could otherwise com-
promise the alignment. Furthermore, an optimum
amount of tilt immunity also makes possible the
use of holders that do not offer any type of adjust-
ment of the position of the optical components, mak-
ing the interferometer even more suitable for a
cryogenic environment. In a future paper, it will be
reported how to rearrange the interferometer
shown in Fig. 8 with the aim of measuring angular
motion [15].
In Section 2, four configurations of the polarization
Michelson interferometer are described. In Section 3, Fig. 2. Misalignment cancellation of the retroreflected beam in
the ability of the cat’s eye retroreflector for providing the configurations shown in Figs. 1(c) and 1(d).
mirror tilt immunity is examined by means of ABCD
matrices, and, in Section 4, a realization of the cat’s
eye is analyzed using Zemax software for optical de- larizing beam splitter, where the beam is divided into
sign. Finally, in Section 5, the performances of the the two arms of the device. The polarization direction
cat’s eye retroreflector in a Michelson interferometer of the light is at 45° with respect to the plane of in-
is analyzed by means of a numerical calculation cidence at the polarizing beam splitter, where the p
using Zemax. polarization component is transmitted and the s po-
larization component is reflected. After the beams
2. Different Configurations of the Michelson have sensed the position of the mirrors, they recom-
Interferometer bine at the beam splitter and propagate away toward
Figure 1 shows four different configurations of the a set of optical components (not shown in the dia-
polarization Michelson interferometer. Figure 1(a) gram) whose function is to create the interferograms
shows the traditional single-pass interferometer in from which the information about the relative posi-
which the target and reference objects hold flat mir- tion of the mirrors is extracted. In homodyne
rors. A linearly polarized beam propagates into a po- interferometry, these components are usually a non-
polarizing beam splitter, a quarter-wave plate, a half-
wave plate, and either polarizing filters or polarizing
beam splitters [4,12,16–21]. This interferometer is
immune to lateral displacements of the mirrors,
but not to mirror tilts.
Figure 1(b) shows a device in which the mirrors
have been replaced by cube corner retroreflectors
[1,4]. Such a replacement does provide some amount
of wavefront parallelism that becomes complete im-
munity when the cube corner rotates around its no-
dal point. The nodal point is the apparent location of
the apex as seen from outside the glass cube [22]. In
the case in which the rotation happens around
any other point, the beam remains parallel to its
original direction, but it is also displaced sideways,
decreasing the overlapping area between the inter-
fering beams and, therefore, reducing the visibility
of the interferogram. A similar effect is produced if
the cube corner moves sideways since the beam will
move twice as much as the cube corner in the same
direction. Obviously, this approach requires that the
cube corner retroreflector be mounted on the target
Fig. 1. Four different configurations of the polarization Michel- object, a requirement that is not always easy to fulfill
son interferometer. and is impossible in some cases.

982 APPLIED OPTICS / Vol. 50, No. 7 / 1 March 2011


The arrangement shown in Figure 1(c) provides, to
some extent, the advantages of both configurations
shown in Figs. 1(a) and 1(b). It is a double-pass inter-
ferometer with only one cube corner in the common
path of both arms. In this configuration, as in the pre-
vious ones, the polarizing beam splitter A divides the
light into the two arms of the interferometer. The p
polarization component is transmitted and the s po-
larization component is reflected. The ray, which pro-
pagates along the original optic axis, goes through
quarter-wave plate B, while the ray propagating Fig. 3. (Color online) Unfolded propagation diagram of a cat’s eye
along the perpendicular direction goes through retroreflector with a flat mirror. It is a symmetric system.
quarter-wave plate D. Both quarter-wave plates have
their fast axes at 45° with respect to the horizontal
direction, and, thus, the two light beams become cir- As in the configuration shown in Fig. 1(a), since
cularly polarized. One of the beams is reflected back the reflector mounted on the target is a flat mirror,
by mirror C, and the other beam is reflected back by this interferometer is also immune to lateral displa-
mirror E. As seen from the reference systems that cements of the object. It is important to notice, how-
describe the polarization vectors of the beams as they ever, despite the fact that the outgoing ray is indeed
propagate back after the reflections, the angles that parallel to the original incoming ray after the second
describe the orientation of the fast axes of wave reflection, they do not necessarily overlap. There is a
plates B and D are now −45°. The wave plates change lateral relative displacement between the two rays
the polarization of both beams back to linear. This that results from the light propagating with an angle
time, B makes the polarization direction of the light in between the two reflections on the tilted mirror.
perpendicular to the plane of incidence at the polar- The interferometer shown in Fig. 1(d) employs a
izing beam splitter A, while D makes the polarization cat’s eye rather than a cube corner in order to accom-
direction of the light parallel to the plane of inci- plish mirror tilt immunity. Section 3 is devoted to the
dence. The ray propagating along the original optic calculation of the lateral displacement between the
axis is now reflected by A, and the other one simply incoming and outgoing rays in this interferometer.
passes through. After the beams recombine, the light
encounters the retroreflector, which, in this config- 3. ABCD Matrix Description of a Retroreflector
uration, is a cube corner, and then is reflected back
to polarizing beam splitter A. Beam splitter A trans- A. ABCD Matrix of a Retroreflector
mits the horizontally polarized beam and reflects the
vertically polarized one. Both beams travel again The general expression of the ABCD matrix of a re-
along the interferometer arms, and, as before, by flector capable of providing mirror tilt immunity in a
passing twice through their respective quarter-wave double-pass interferometer can be calculated by con-
plates B and D, their polarization directions accumu- sidering a ray propagating into the retroreflector
late a new overall change of 90°. When they encoun- from the mirror placed at distance d from the retro-
ter polarizing beam splitter A for the fourth time, the reflector, and then back to the mirror. If the ray pro-
component traveling along the arm of mirror C is pagates away from the mirror with a slope yi from a
transmitted, while the other one is reflected. After point at a distance xi from the optic axis, the effect of
they recombine, the whole beam propagates toward the propagation through the whole system can be cal-
the detection unit as in the configurations shown in culated as the product
Figs. 1(a) and 1(b). In this configuration, the effect of
both of the double-pass and the retroreflector is to rout ¼ ðDd ÞðRRÞðDd Þðrin Þ; ð1Þ
cancel the misalignment produced by the tilt of any
of the mirrors. Figure 2 depicts a ray (shown as a so- where the vectors and matrices
lid line) propagating onto a tilted mirror with an an-      
gle of incidence ψ i . The ray is reflected with an angle xi xo 1 d
rin ¼ ; rout ¼ ; Dd ¼ ;
ψ r into a direction where a retroreflector (not shown) yi yo 0 1
is placed. When the retroreflected beam (shown as a  
A B
dotted line) encounters the tilted mirror again, the RR ¼ ; ð2Þ
angle of incidence ϕi will be exactly the same as C D
the angle of reflection ψ r . This, in turn, implies that
when the retroreflected beam is reflected by the mir- describe the initial and final states of the ray and the
ror again, it propagates away from the mirror with effect of the propagation through free space and
an angle ϕr equal to the initial angle of incidence through the system under examination, respectively
ψ i . In terms of equations, the relations ϕr ¼ −ψ i, [23]. For simplicity, let us consider C ¼ 0. In such a
ϕi ¼ −ψ r , ψ r ¼ ψ i − 2θ, and ϕr ¼ ϕi − 2θ hold, where case, the identity AD − BC ¼ 1 implies that AD ¼ 1
the sign convention is given by Appendix A in [21]. and the product [Eq. (1)] becomes

1 March 2011 / Vol. 50, No. 7 / APPLIED OPTICS 983


2   3  
  1 0
A þ A1 d þ B 5 xi Lf ¼ − 1
rout ¼ ðDd ÞðRRÞðDd ÞðrÞ ¼ 4A : f 1 ; ð8Þ
0 1 yi
A

ð3Þ the propagation through the distance f between the


lens and the mirror is represented by
From Eq. (3), the value of d that makes the depen-  
1 f
dence of the final displacement xo on the initial slope Df ¼ ; ð9Þ
yi vanish, is readily calculated as 0 1

and the action of the mirror whose radius of curva-


AB
d¼− : ð4Þ ture is S is described by the matrix
A2 þ 1
 
1 0
This means that all the rays that diverge from a MS ¼ 2 : ð10Þ
S 1
point on a plane at a distance d from the ABCD sys-
tem will converge at one point on the same plane
Notice that S > 0 for a divergent mirror (convex sur-
after the propagation. Such a plane defined by face) and S < 0 for a convergent one (concave sur-
Eq. (4) is called the sweet plane. The immunity de- face). The expression for a flat mirror is obtained
grades as the entrance plane moves away from the by making the radius of curvature approach infinity
sweet plane. If the entrance plane moves away a dis- S → ∞. Calculating the overall corresponding propa-
tance ϵ, the effect of the total propagation [Eq. (3)] gation matrix for this cat’s eye retroreflector takes
becomes only a simple matrix multiplication:
 
   RR ¼ ðLf ÞðDf ÞðMS ÞðDf ÞðLf Þ; ð11Þ
A A þ ϵ xi 1
rout ¼ A : ð5Þ
1 yi which yields the value of the parameter B in the
0 A
matrix (7):
8
From this expression, the condition that yields the >
>
> 2f  flat mirror;
maximum amount of immunity to the initial slope >
>
< 2 f þ f2 convex mirror;
yi is readily calculated by minimizing the quantity jSj
B¼   ð12Þ
A þ A1 . Such a minimum occurs when A ¼ −1, and >
>
>
> f2
then the total propagation [Eq. (3)] reads >
: 2 f − jSj concave mirror:
    
x −1 −2ϵ xi For instance, in the case in which a cat’s eye com-
rout ¼ o ¼ ; ð6Þ
yo 0 −1 yi prises a lens and a flat mirror, the condition B ¼ 2d
provides the optimum position of the target mirror
and the condition of mirror tilt immunity [Eq. (4)] for achieving the best mirror tilt immunity possible,
becomes B ¼ 2d. which is d ¼ f . How well such a condition is fulfilled
Since yo ¼ −yi , the whole system behaves as a ret- by the system with a cube corner can be analyzed
roreflector because the light is reflected back into the using the ABCD matrix of a cube corner given by
same direction it came from regardless of its angle of Wang and Ronchi [24] as
incidence. Furthermore, the unknown ABCD system  
is a retroreflector itself: −1 −2 hn
CC ¼ ; ð13Þ
0 −1
 
−1 B
RR ¼ ; ð7Þ where h is the length of the cube corner from the en-
0 −1
trance face to the apex and n is the index of refraction
of the glass. In this case, B ¼ −2 dn is always negative,
where the parameter B depends exclusively on the
and, therefore, the condition B ¼ 2d can never be ful-
particular properties of the individual components
filled. The mirror tilt immunity that the cat’s eye can
of this system. Notice that B must be always greater provide will always be better than the one provided
than zero to fulfill the condition B ¼ 2d. by the cube corner in the configurations shown in
In the case in which the retroreflector comprises a Figs. 1(d) and 1(c), respectively.
thin lens and a mirror, the mirror must be placed at
the focal plane of the lens in order to obtain a retro- B. Mirror Tilt Immunity Provided by the Cat’s Eye
reflecting effect [8]. The particular value of the para- The effect of the tilt of the mirror can be quantita-
meter B is readily calculated since the refraction by a tively taken into account by means of the aug-
convex lens of focal length f is quantified by the mented 4 × 4 ABCD matrices for misaligned optical
matrix components given by Wang and Ronchi [24]. As it

984 APPLIED OPTICS / Vol. 50, No. 7 / 1 March 2011


is shown in [21], if the parameters describing the po- The complete independence of the displacement xo
sition and the slope of the incident beam onto the from both angles θ and yo can only be achieved be-
tilted mirror are xi and yi , respectively, the corre- cause a cat’s eye produces a real image of any object
sponding parameters xo and yo describing the out- placed in front of it. The sweet plane defined by the
going beam are condition Brr ¼ 2d is the place where the plane of the
xo ¼ −xi þ ðyi − 2θÞðBrr − 2dÞ; ð14Þ real image coincides with the plane of the source ob-
ject itself. This amount of mirror tilt immunity is im-
yo ¼ −yi ; ð15Þ possible to achieve with a cube corner because such a
retroreflector does not produce real images, but only
where θ is the tilt angle of the mirror and Brr is the virtual ones. As can be seen in Fig. 4, after the rays
generic parameter B in Eq. (7). are retroreflected by the cube corner, they continue
The system comprising the tilted mirror and the diverging and never reach the same point again. It
retroreflector works as a whole as a retroreflector. is important to notice that since the points P1 and
The slope of the outgoing beam yo does not depend P2 are conjugate in a cat’s eye, the optical path length
at all on the tilt angle θ and its value is yo ¼ −yi. How- of any ray propagating between the two points is the
ever, the distance xo from the optic axis to the beam
same regardless of the trajectory of the individual
does depend on the tilt angle θ. From Eq. (14), it is
ray. Therefore, the mirror tilt immunity does not re-
also possible to notice that the displacement xo pro-
duced by having the target mirror misaligned is null fer only to the fact that the beam does not displace at
if Brr ¼ 2d. When this condition is fulfilled, the dis- all as the mirror tilts, but also to the fact that the
placement xo does not depend on the tilt angle θ or optical path length does not change either.
the angle of incidence yi . To explain this statement In the case in which the mirror is not placed at the
on intuitive grounds, let us consider the particular sweet plane, the optical path length can be calculated
case in which the cat’s eye comprises a thin lens using the matrix form of Eq. (14) given in [21] and the
and a flat mirror. As noted earlier, the condition Brr ¼ Eikonal equation given by [24]. If the distance from
2d and the corresponding value of Brr [Eq. (12)] for a the mirror to the sweet plane of the cat’s eye is s, the
flat mirror imply that d ¼ f . Figure 3 shows the un- Eikonal equation yields
folded ray-tracing diagram for the retroreflector with
the target mirror placed at the focal plane of the lens. L ¼ L0 − 4sθ2 þ sy2i ; ð16Þ
Let us imagine that only one incident ray (not pic-
tured in the figure) reaches the mirror from the left where L0 is the optical path length along the optic
at point P1 . Each of the rays diverging from P1 into axis. From Eq. (16), it follows that, when s ¼ 0, the
the retroreflector could be regarded as the reflection optical path length does not depend on the tilt of
of such a ray for different tilt angles of the mirror. the mirror.
Because the rays diverge from a point placed at the
focal plane, they all propagate parallel to each other
within the retroreflector. Therefore, after they are re-
fracted by the lens again they all converge onto the
same point P2 at the focal plane where the mirror is
placed. So far, this explains the independence of the
outgoing beam displacement xo from the tilt angle θ
when the mirror is placed at the focal plane of the
lens. The same argument may be used to prove that,
under the same conditions, xo does not depend on the
angle of incidence yo either. Indeed, it is only neces-
sary to consider that many incident rays reach point
P1 from the left from many different directions, and
that the rays that diverge from P1 represent the
reflection of those beams for only one tilt angle of
the mirror.

Fig. 4. Cube corners produce virtual images of objects. For


simplicity, refraction at the entrance face of the cube corner was Fig. 5. (Color online) (a) Afocal and (b) focal configurations of the
not considered in the diagram. cat’s eye retroreflector with a flat mirror.

1 March 2011 / Vol. 50, No. 7 / APPLIED OPTICS 985


Equation (16) is also valid for the system with the The system must be optimized using a beam of
cube corner. The value of the parameter s is the dis- rays rather than individual rays. Every single ray be-
tance from the mirror to the apparent position of the longs simultaneously to two configurations, one focal
apex of the cube corner as seen from the mirror. The and one afocal. Each focal configuration describes the
refracting effect of the quarter-wave plate and the po- misalignment of a single ray as the mirror rotates
larizing beam splitter must be taken into account around the point of reflection of that ray. The optical
when calculating this position. path length difference between one real ray and its
paraxial counterpart contributes equally to the aber-
4. Real Cat’s Eye rations in both configurations. The only distinction
Any realization of a cat’s eye retroreflector will be af- between the two quantities is the reference wave-
fected by aberrations. The properties of all the com- front from which this optical path length difference
ponents of a cat’s eye must be such that the total is considered. It should be possible, in principle, to
amount of aberrations in the retroreflected beam calculate for how much one ray is aberrated in one
must be always below the diffraction limit. The opti- configuration, if its aberrations in the other config-
cal system presented in this section was optimized uration and the optical path length difference be-
and analyzed using Zemax software for optical tween the two reference wavefronts are known.
design. An ideal reference wavefront is always tangent to
Retroreflectors are afocal systems. A collimated the point of intersection of the exit pupil and the chief
beam propagates into the system and a collimated ray, with its center of curvature coincident with
beam propagates away in the same direction from the ideal image point. Figure 5 shows the afocal
which the incident beam came. An appropriate figure (Fig. 5(a)) and focal (Fig. 5(b)) configurations to which
of merit for assessing the performance is the peak-to- one particular ray belongs. In both diagrams, the ray
valley wavefront distortion of the outgoing plane is shown as a black solid curve and propagates
wave, which must be below the diffraction limit. through the system from point P1 to point P2 . The
However, the cat’s eye retroreflector described in dashed curves represent other rays that belong to
Section 3 is also intended for providing immunity either configuration but not to both. In the afocal con-
to mirror tilts in an interferometer. A figure of merit figuration, the aperture stop is located at the en-
of how well a cat’s eye performs this task must be also trance plane, which is the plane of the mirror. The
included in the optimization procedure. In terms of entrance pupil is, therefore, located at this same
the paraxial approximation, how this immunity is plane, and the exit pupil coincides with the image
achieved is illustrated in Section 3 by means of plane. The chief ray is shown as a dotted–dashed
the unfolded propagation diagram (Fig. 3), and the curve that goes from point Q1 to point Q2 . The refer-
optical path length accumulated in going from point ence wavefront W is shown tangent to point Q2 , with
P1 to point P2 is written as Eq. (16). In Fig. 3, every its center of curvature at infinity, where the ideal im-
depicted ray is the reflection of one single ray for dif- age is located. In the focal configuration, the chief
ferent tilt angles of the mirror. Since all those rays ray, which is shown as a dotted–dashed curve propa-
converge into the same point P2 , the quality of the gating from point P1 to point P2 , is parallel to the op-
image point can be used as a figure of merit for tic axis and corresponds to the case in which the
the ability of the retroreflector to provide mirror tilt mirror is not misaligned. The symmetry of the cat’s
immunity. Because this system is intended to be used eye retroreflector implies that the chief ray is retro-
in an interferometer, the peak-to-valley wavefront reflected also parallel to the optic axis. The system is
distortion of the spherical wavefront is a better then double telecentric where the entrance and exit
choice than the rms radius of the image spot com- pupils are at infinity in object and image spaces re-
monly used in imaging systems. spectively, and the aperture stop is therefore located

Fig. 6. (Color online) Optimized cat’s eye retroreflector. The optics are 5 mm in size.

986 APPLIED OPTICS / Vol. 50, No. 7 / 1 March 2011


in Fig. 6 as different types of curves. The radius of
the beam is 0:65 mm, and the apodization is uniform.
In the focal configurations, three point sources were
located at the optic axis and at 0:65 mm away from
it, at points b, a, and c, respectively. The entrance
plane E is at the sweet plane where the mirror must
be in the real physical system. Although the cat’s eye
comprises only the lens L and the mirror M, in order
to be able to use the cat’s eye in a double-pass inter-
ferometer [25], a polarizing beam splitter cube P and
a quarter-wave plate Q must also be taken into ac-
count (see Section 5). The aspheric lens L is a com-
mercial product distributed by Thorlabs (Ely
Cambridgeshire, UK, part number AL1210). The
mirror M is the coated convex face of the lens (Thor-
labs, part number LA4249).
Figure 7 shows the rms wavefront error as a func-
Fig. 7. Root-mean-square wavefront error in waves with respect tion of the field coordinate of both configurations in
to the principal ray for both configurations. units of wavelength and with respect to the principal
ray for each field. The performance of the afocal con-
at the focal plane within the retroreflector. At point figuration is diffraction limited. The aberrations af-
P2 , the ideal wavefront is the ideal image point itself. fecting the focal configuration actually define the
Since the exit pupil in the focal configuration is real limits to the mirror tilt immunity. In terms of
located at infinity, it is more convenient to write a Fig. 3, in the paraxial approximation, P1 and P2
formal expression of the aberrations in the focal con- are conjugate points, and the optical path length be-
figuration W f ðh; θÞ in terms of the aberrations in the tween the two points is always the same regardless of
afocal configuration W a ðh; θÞ at its exit pupil plane, the amount of the tilt of the mirror θ. However, in the
where h is the coordinate of the ray at this pupil real cat’s eye, the aberrations introduce optical path
and θ is the tilt angle of the mirror. Such an equation length differences between the different rays travel-
is ing between P1 and P2 . An interferometer would be
able to measure those differences as the mirror ro-
W f ðh; θÞ ¼ W a ðh; θÞ þ ΔCðh; θÞ; ð17Þ tates if the signal produced is greater than the noise
in the system.
where W f ðh; θÞ ¼ OPLf − OPL, W a ðh; θÞ ¼ OPLa −
OPL, ΔC ¼ OPLf − OPLa , and OPL, OPLa , and
OPLf are the optical path lengths of the aberrated
ray and the chief rays in the afocal and focal config-
urations, respectively. The optical path length differ-
ence ΔC is subject to aberrations as well, and it must
be calculated using real rays. The paraxial contribu-
tion to this quantity, however, can be easily calcu-
lated using Eq. (16) as ΔCparaxial ¼ 4sθ2 − sy2i ,
where OPLa ¼ L and OPLf ¼ L0 .
A consequence of Eq. (17) is that whatever techni-
que is used to optimize the system in one configura-
tion will also optimize the performance of the system
in the other configuration. Thus, a cat’s eye retrore-
flector can be optimized, for instance, using the afo-
cal configuration only, and it should be expected that
if the position of the mirror is an optimization vari-
able, the mirror will be automatically located at the
sweet plane.
Figure 6 shows a cat’s eye optimized for light at a
wavelength of 1550:9 nm. Figure 6(a) shows the afo-
cal configuration, and Fig. 6(b) shows the focal one.
Although, in principle, for every ray traced in an afo-
cal configuration there must be a focal configuration, Fig. 8. Cat’s eye retroreflector within a homodyne polarization
in practice, only three fields were taken into account interferometer. A, laser; B, E, and K, polarizing beam splitters;
in each configuration. In the afocal ones, three colli- C, beam splitter; D, half-wave-plate; F, F 0 , quarter-wave-plate;
mated beams propagate into the retroreflector, one G, G0 , mirrors; H, aspheric lens; I, mirror; J, quarter-wave plate;
parallel to the optic axis and two at 2:5°, shown P1 , P2 , and P3 , photodiodes.

1 March 2011 / Vol. 50, No. 7 / APPLIED OPTICS 987


As can be seen in Fig. 7, the rms wavefront error terference pattern reaches either a maximum or
for the focal configuration decreases as the field minimum intensity [4,12,16–21]. The first compo-
coordinate increases. This is a consequence of the nent that the light encounters is polarizing beam
optimization procedure compensating the spherical splitter B. The light whose electric field is perpendi-
aberration and astigmatism with defocus with the cular to the plane of incidence is reflected and lost
aim of minimizing the merit function [see Fig. 6(b)]. (not shown in Fig. 8), and the light whose electric
As is seen in Section 5, the optimization procedure field is parallel to the same plane is transmitted into
introduces an extra optical path length in order to the rest of the interferometer. The following element
minimize, in the average, the spherical aberration in the path is beam splitter C. The transmitted beam
and astigmatism at the sweet plane by moving the is lost (not shown), and the reflected one propagates
source and image planes (the mirror) away from further onto half-wave plate D. The fast axis of the
the paraxial focal plane of the retroreflector. Such wave plate is oriented at 67:5° with respect to the
a cancellation is more effective for larger field coor- horizontal direction, and, thus, the polarization di-
dinates, where the rays are more affected by the rection of the light changes to be 45° with respect to
spherical aberration, leaving only the unbalanced de- the same direction. The next component, polarizing
focus as the dominant wavefront error for rays closer beam splitter E, divides the light into two beams of
to the optic axis. equal intensity that propagate along the arms of the
interferometer. The beams sense the position of
5. Interferometer with Mirror Tilt Immunity
the target and reference mirrors twice by means of
The goal of this section is to describe what the actual the cat’s eye retroreflector, as described in Section 1.
interferometer measures as the target mirror tilts. After the two beams recombine at polarizing beam
Although it would be useful to relate the aberrations splitter E, the light subsequently propagates again
present in the afocal configurations with what the onto half-wave plate D. The polarization directions
actual interferometer measures, the afocal config- of the perpendicular components of the electric field
urations are still an approximation to the real sys- change again by 45°. The beam splitter C divides the
tem. In the interferometer, the rotation takes place light into two beams. The transmitted beam reaches
around a point located on the optic axis. However, in wave plate J and, later, polarizing beam splitter K,
every afocal configuration, the mirror is assumed to where two interferograms are created. The intensi-
rotate around its corresponding point source, which, ties of the interferograms I 2 and I 3 are measured
with one exception, is not located along the optic by photodiodes P2 and P3 , respectively. The reflected
axis. The optimization procedure also takes into ac- beam at C propagates to polarizing beam splitter B,
count simultaneously afocal configurations whose where another interferogram is created upon reflec-
point sources are located at different distances from tion. The intensity of the interferogram I 1 is mea-
the axis. sured by photodiode P1 . Ideally, I 1 should be in
Figure 8 shows a diagram of the interferometer. quadrature with respect to I 2 and I 3 , which should
This device achieves displacement measurements be in phase opposition with respect to each other.
using reversible fringe counting with three interfer- However, this condition depends on the phase intro-
ence patterns, whose intensities are measured by duced by beam splitter C and the retardance of wave
photodiodes P1, P2 , and P3 . Using reversible fringe plate J [12,21].
counting, it is possible to measure displacements lar- It is important to point out that relating what the
ger than a quarter of a wavelength in a double-pass interferometer measures directly with the aberra-
interferometer without any ambiguity when the in- tions in the afocal configurations is not a straightfor-
ward task. The interferometer employs a Gaussian
beam and not just one single ray. Each of the three
output intensities of the interferometer is the result
of the interference of two Gaussian beams whose
wavefronts are not uniform across the beams due to
the aberrations. Furthermore, when the mirror is not
in the vicinity of the sweet plane, one of the beams
displaces with respect to the other one as the mirror
rotates. In practice, the intensities are calculated by
tracing 200,000 rays with a Gaussian intensity pro-
file. The photodiodes are 3:2 mm × 3:2 mm in size
with a resolution of 200 × 200 pixels. The intensity
at each pixel is calculated by considering the interfer-
ence of all the rays that arrive to that pixel, and the
total intensity at each photodiode is calculated by
Fig. 9. (Color online) Optical path difference measured by the in- adding together the intensities at all of its pixels.
terferometer when the mirror rotates around the sweet spot. The The quantities I 2 − I 1 and I 3 − I 1 represent the ab-
abscissa is the normalized pupil coordinate of the chief ray of the scissa and ordinate of a point in an ideally circular
Gaussian beam. Lissajous pattern centered at the origin. As the

988 APPLIED OPTICS / Vol. 50, No. 7 / 1 March 2011


optical path length changes, the point moves from chief ray of the Gaussian beam. The plots are sym-
one point to another along the Lissajous pattern, and metrical and can be characterized by the function
the displacement can be calculated according to the
formula WðρÞ ¼ Cρ2 þ Aρ4 ; ð20Þ
    where C represents defocus and A represents
1 λ x1 y2 − x2 y1
d¼ arctan ; ð18Þ spherical aberration. Unlike the system shown in
4 2π x1 x2 þ y1 y2
Fig. 6(b), in which there is astigmatism due to the
where ðx1 ; y1 Þ and ðx2 ; y2 Þ are the coordinates of the point sources a and c not being on axis, the system
two points. with the Gaussian beam on axis is not affected by as-
Figure 3 provides a qualitative idea of what the in- tigmatism due to the axial symmetry. The coeffi-
terferometer measures as the mirror rotates. The dif- cients C and A can be found by a linear least
ferent rays propagating away from point P1 are squares fit to calculated data.
shown in different types of curves in order to point The paraxial limit provides a model of how the de-
out the symmetry of the system. The five depicted focus varies as the mirror moves away from the sweet
rays accumulate different optical path lengths as plane. In such a limit, Eq. (20) can be written as
they propagate from point P1 to the focal plane R
in between the lenses. As they propagate from the 4C
plane R to the point P2 , each ray travels the same WðρÞ ¼ Cρ2 þ Aρ4 ≈ θ2 : ð21Þ
ðtan 2:5°Þ2
optical path length as the optical path length tra-
veled, from P1 to R, by the ray symmetrically placed Comparison of Eq. (16) with yi ¼ 0 yields the
on the other side of the principal ray, shown in black. expression
For instance, the ray d travels, from R to P2 , the same
optical path length as the ray e travels from P1 to R.  
s þ df
The symmetry also implies that the optical path C ¼ −ðtan 2:5°Þ2 ; ð22Þ
λ
length that the ray e travels from R to P2 equals
the optical path length that the ray d travels from where the term df has been included in order to
P1 to R. Similar relations exist for every pair of rays
quantify the distance between the sweet plane
symmetrically placed on the sides of the principal
yielded by the optimization procedure and the para-
ray. Therefore, regardless of the position of a ray xial sweet plane. The predicted value of the slope is
within the Gaussian beam, the optical path length
mp ¼ −ðtan 2:5°Þ2 =λ ¼ −1:229 mm−1 . The solid curve
traveled by the ray must be a symmetric function
of the tilt angle of the mirror. Therefore, it is expected in Fig. 10 shows the calculated value of C as a func-
tion of the distance s from the mirror to the sweet
that the measurement provided by the interferom-
plane. The parameter s is negative when the mirror
eter is an even function of the tilt angle. It is worth
is closer to the interferometer than the sweet plane
noticing that such a symmetry holds only when the
and positive when it is further away. A linear least
aberrations are small. If the image created at P2 is
squares fit yields the value of the slope as mLSF ¼ ð−
not small enough, the optical path lengths traveled
by the rays during the second pass cannot be ex- 1:228  3:275 × 10−4 Þ mm−1 . The maximum discre-
pancy between the predicted and the calculated
pected to be close in value to those in the first pass.
values of the slope can be estimated as mp − ðmLSF −
Figure 9 shows a plot of the optical path length dif-
ference measured by the interferometer when the ΔmLSF Þ, where ΔmLSF ¼ 3:275 × 10−4 mm−1 , yielding
mirror tilts around the sweet spot. With the excep- a value of 1:325 × 10−3 mm−1, which is negligible com-
tion of the black solid curve, which represents the pared with either the calculated or the predicted va-
data produced with a Gaussian beam, all the graphs lue of the slope. The least squares fit value of the
are produced by tracing individual rays. The optical actual defocus at the sweet plane is df ¼ ð304
path length difference is considered with respect to 1:727Þ × 10−3 mm. This is the distance from the real
the ray or beam that propagates when the mirror sweet plane to the paraxial sweet plane.
is not misaligned. The coordinate z that describes Figure 10 also shows the value of the spherical
each ray is the distance from the ray to the optic axis aberration coefficient A as a dashed curve. At the
as the ray propagates onto the mirror parallel to the scale shown, and compared with the value of the de-
axis. The abscissa is the normalized entrance pupil focus coefficient C along the whole range, the coeffi-
coordinate of the chief ray of the Gaussian beam cal- cient A is constant. At the sweet plane at s ¼ 0, C and
culated as A are approximately equal in absolute value but
opposite in sign. This is a consequence of the optimi-
zation procedure compensating the spherical aberra-
tan 2θ tion and astigmatism with defocus [see Fig. 6(b)] in
ρ¼ ; ð19Þ
tan 2:5° order to minimize the optical path length change
that the interferometer measures as the mirror
where θ is the tilt angle of the mirror. It is important rotates.
to notice that even the data for the individual rays is The spherical aberration sets the real limit to the
plotted as a function of the pupil coordinate of the ability of the cat’s eye retroreflector to provide

1 March 2011 / Vol. 50, No. 7 / APPLIED OPTICS 989


immunity to mirror tilts. For instance, according to
the graph in Fig. 9, the maximum optical path length
change that the interferometer measures as the mir-
ror tilts at the sweet plane is W max ¼ 0:11λ=4 ¼
43:77 × 10−9 m. This is still large compared with the
limit that the shot noise sets with an initial incoming
beam intensity of approximately pffiffiffiffiffiffiffi24:6 μW, which is
approximately 1:14 × 10−14 m= Hz [21]. The aberra-
tion limit can, nevertheless, be reduced by improving
the optical design.
Finally, Fig. 11 shows the mirror tilt immunity of
the device from the standpoint of the amplitude of
the Lissajous pattern for both retroreflectors, the
cat’s eye and the cube corner [see Figs. 1(c) and 1(d)].
The amplitude was calculated by displacing the
mirror by a quarter of a wavelength at different dis- Fig. 11. (Color online) From the standpoint of the amplitude of
the Lissajous pattern, a cat’s eye provides a better mirror tilt im-
tances s from the sweet plane for different amounts
munity than a cube corner. Parameter s is the distance from the
of tilt. As in Fig. 10, the parameter s is negative when mirror to the sweet plane. It is negative when the mirror is closer
the mirror is closer to the interferometer than to the interferometer than the sweet plane and positive when it is
the sweet plane and positive when it is further away. further away.
The displacement would produce one rotation
around the Lissajous pattern, and the amplitude of
the circle was calculated with a linear least squares
equations, yielding a phase change between −0:065°
fit to the calculated data. This result is a measure of
and 0:065°, depending on the values of the exit pupil
how much the Gaussian beam moves away from the
coordinates.
optic axis but still parallel to it as the mirror tilts. As
discussed in Section 3 with Figs. 3 and 4, a cat’s eye 6. Conclusions and Future Work
provides a larger tilt immunity than a cube corner The mirror tilt immunity that a cat’s eye retroreflec-
since the former produces real images and the latter tor provides in a double-pass homodyne polarization
only virtual ones. interferometer is examined. By means of 4 × 4 ABCD
A cat’s eye also has the advantage that its polari- matrices, an analytic expression of the misalignment
zation effect is small, if not negligible. Unlike a cube of the beam as the target mirror tilts is calculated
corner, a cat’s eye does not require large angles of in- [Eq. (14)]. It is found that there is a position of the
cidence on the surfaces of the optical components in target mirror, the sweet plane, in which the tilt im-
order to achieve its retroreflecting effect. For the munity is at its maximum. Furthermore, at the sweet
particular cat’s eye described earlier, the largest plane, the optical path length change introduced as
contribution to the phase change between the per- the mirror tilts is null [Eq. (16)].
pendicular components of the electric field of the ret- A real cat’s eye, which is affected by aberrations, is
roreflected light is introduced by the reflection on the designed with optics from stock and optimized with
mirror, where the angles of incidence are a maximum Zemax. Although as an afocal retroreflector the cat’s
of 3:77°. The overall polarization effect at the exit eye designed is diffraction limited (Fig. 7), as part of
pupil is easily calculated with Zemax using Fresnel the interferometer (Fig. 8), the mirror tilt immunity
it provides at the sweet plane is limited by spherical
aberration, with a maximum error of 43:77 nm with-
in the mirror tilt range considered between −1:25°
and 1:25°. The optical path length introduced by
the tilt for different positions of the mirror with re-
spect to the sweet plane is written in terms of defocus
and spherical aberration [Eqs. (20) and (22)], and the
amount of these aberrations is numerically calcu-
lated using real ray tracing (Fig. 10).
Also, as part of the interferometer, this retroreflec-
tor can maintain the alignment of the interfering
beams for a mirror tilt of at least 1:25°. The ampli-
tude of the Lissajous pattern decreases 5.3% for a tilt
of 1:25° at the sweet plane and 45.5% for the same
tilt when the mirror is þ6 mm away from the sweet
plane (see Fig. 11).
Fig. 10. Parameter C follows a linear dependence with s as pre- From the point of view of the design, the tolerance
dicted by the paraxial approximation [see Eq. (22)]. Compared analysis must still be made. Future work also in-
with the defocus, the spherical aberration remains constant. cludes reporting the experimental realization of a

990 APPLIED OPTICS / Vol. 50, No. 7 / 1 March 2011


version off axis of the interferometer shown in Fig. 7 13. G. D. Hammond, A. Pulido Patón, C. C. Speake, and
[26] and the application of the cat’s eye for interfero- C. Trenkel, “Novel torsion balance based on a spherical super-
metric angular motion measurement [15]. conducting suspension,” Rev. Sci. Instrum. 75, 955–961
(2004).
We appreciate the suggestions given by Armando 14. G. D. Hammond, C. C. Speake, A. J. Matthews, E. Rocco, and
Gómez Vieyra from the Centro de Investigaciones en F. E. Peña-Arellano, “Development of a second generation tor-
Óptica in León, Gto., Mexico. The work reported in sion balance based on a spherical superconducting suspen-
this article is subject to International patent pub. sion,” Rev. Sci. Instrum. 79, 025103 (2008).
no. WO 2009/010750 A1[27]. 15. F. E. Peña-Arellano, C. C. Speake, H. Panjwani, and
L. Carbone “An interferometer for measuring angular mo-
References tion,” in preparation.
16. E. R. Peck and S. W. Obetz, “Wavelength or length measure-
1. E. R. Peck, “Theory of the corner-cube interferometer,” J. Opt.
Soc. Am. 38, 1015–1024 (1948). ment by reversible fringe counting,” J. Opt. Soc. Am. 43,
505–509 (1953).
2. P. Hariharan and D. Sen, “The separation of symmetrical and
asymmetrical wave-front aberrations in the Twyman interfe- 17. P. L. M. Heydemann, “Determination and correction of quad-
rometer,” Proc. Phys. Soc. London 77, 328–334 (1961). rature fringe measurements errors in interferometers,” Appl.
Opt. 20, 3382–3384 (1981).
3. S. J. Bennett, “A double-passed Michelson interferometer,”
Opt. Commun. 4, 428–430 (1972). 18. K. W. Raine and M. J. Downs, “Beam-splitter coatings for
producing phase quadrature interferometer outputs,” J.
4. M. J. Downs and K W. Raine, “An unmodulated bi-directional
Mod. Opt. 25, 549–558 (1978).
fringe counting interferometer system for measuring displace-
ments,” Precis. Eng. 1, 85–88 (1979). 19. V. Greco, C. Iemmi, S. Ledesma, G. Molesini, and F. Quercioli,
“Multiphase homodyne interferometry: analysis of some error
5. M. J. Downs and J. W. Nunn, “Verification of the sub-
nanometric capability of an NPL differential plane mirror in- sources,” Appl. Opt. 34, 2207–2213 (1995).
20. M. A. Zumberge, J. Berger, M. A. Dzieciuch, and R. L. Parker,
terferometer with a capacitance probe,” Meas. Sci. Technol. 9,
“Resolving quadrature fringes in real time,” Appl. Opt. 43,
1437–1440 (1998).
6. P. de Groot, “Jones matrix analysis of high-precision displace- 771–775 (2004).
21. F. E. Peña-Arellano, “Characterization of polarization homo-
ment measuring interferometers,” in ODIMAP II (Pavia,
1999), pp. 9–14. dyne interferometers,” Ph.D. dissertation (University of
Birmingham, 2008).
7. H. J. Büchner and G. Jäger, “A novel plane mirror interferom-
22. N. Bobroff, “Recent advances in displacement measuring
eter without using corner cube reflectors,” Meas. Sci. Technol.
17, 746–752 (2006). interferometry,” Meas. Sci. Technol. 4, 907–926 (1993).
23. F. Pedrotti and L. Pedrotti, Introduction to Optics (Prentice-
8. J. J. Snyder, “Paraxial ray analysis of a cat’s-eye retroreflec-
tor,” Appl. Opt. 14, 1825–1828 (1975). Hall, 1996).
9. M. L. Biermann, W. S. Rabinovich, R. Mahon, and 24. S. Wang and L. Ronchi, “Principles and design of optical
arrays,” Prog. Opt. 25, 279–310 (1988).
G. C. Gilbreath, “Design and analysis of a diffraction-limited
cat’s-eye retroreflector,” Opt. Eng. 41, 1655–1660 (2002). 25. C. C. Speake and S. M. Aston, “An interferometric sensor
for satellite drag-free control,” Class. Quant. Grav. 22,
10. S. E. Segre and V. Zanza, “Mueller calculus of polarization
S269–S277 (2005).
change in the cube-corner retroreflector,” J. Opt. Soc. Am. A
20, 1804–1811 (2003). 26. F. E. Peña-Arellano, C. C. Speake, and S. M. Aston are prepar-
ing a manuscript to be called “Experimental realization of an
11. J. Dyson, Interferometry as a Measuring Tool (The Machinery
Publishing Company, 1970). interferometer with mirror tilt immunity.”
27. C. C. Speake, S. M. Aston, F. E. Peña-Arellano, and
12. F. Petrů and O. Čip, “Problems regarding linearity of data of a
T. P. E. Copland, “Improved interferometer,” International
laser interferometer with a single-frequency laser,” Precis.
Eng. 23, 39–50 (1999). patent, pub. no. WO 2009/010750 A1 (22 January 2009).

1 March 2011 / Vol. 50, No. 7 / APPLIED OPTICS 991

You might also like