Carousel interferometer

J. K. Kauppinen, I. K. Salomaa, and J. O. Partanen

A new type of a swinging interferometer, the carousel interferometer, is presented, and its properties are
studied and compared with other swinging interferometers. The new interferometer is built with five
plane mirrors. The optical path difference is accomplished by rotation of a system that consists of four
mirrors. The modulation is almost independent of the scanning of the interferometer. It is not
sensitive to external perturbations such as bending or other deformations of the mount. The
construction is very compact. Because of its stability and low cost it is very applicable to small
Fourier-transform spectrometers for any wave-number region from the far infrared down to the
ultraviolet.
Key words: Fourier-transform spectrometers, interferometers, ultraviolet spectroscopy.

1. Introduction A few years after the cat’s eye retroreflector was

More than 100 years ago, Michelson introduced his developed, cube-corner mirrors were introduced in
first interferometer. It was a plane-mirror interfer- infrared spectroscopy. The first really functional
ometer. In the 1950’s and 1960’s Fourier-transform very-high-resolution cube-corner interferometer was
spectroscopy was introduced, and for the first time the the Oulu interferometer,4,5 developed almost simulta-
Michelson interferometer was applied in spectroscopy. neously with the commercial Mattson low-resolution
Very soon the most difficult disadvantage of the spectrometer.6 Both the Oulu and the Mattson inter-
Michelson interferometer in Fourier-transform spec- ferometers are just Michelson interferometers in
troscopy was noticed. It was the tilting of the mov- which the moving mirror and the fixed mirror are
ing plane mirror during scanning. As is known, the cube corners. If the corners are perfect, the tilt
tilting angle b of the moving mirror should be less problem completely disappears. The only disadvan-
than approximately l@18D2, where l is the wavelength tage of an interferometer of this type is a shearing
of the light under study and D is the diameter of the problem, i.e., a lateral shift of the moving cube corner.
moving mirror. Thus driving the mirror is very This is the case in which the moving cube corner does
difficult, especially in the ultraviolet spectral region not follow the optical axis of the interferometer.4,5
and with high-resolution Fourier-transform infrared The lower the resolution, the more serious this disad-
spectrometers. vantage is. The next improvements were, for ex-
One of the first solutions for this problem was the ample, our interferometer with two cube corners
cat’s eye retroreflector,1 which consisted of a concave moving back to back,7 the Bomem swinging-type
paraboloidal mirror with a convex sphere used to interferometer,8 and the Perkin-Elmer System 2000.9
image the beam splitter back to itself. The use of two cube corners fixed back to back as a
Another solution was a dynamic alignment sys- single moving part eliminates tilting and lateral
tem,2,3 which measures b and tries to keep it at zero shifts completely, if the cube corners are perfect 190° 3
during the run. In practice, when the resolution 90° 3 90°2 and the corners coincide exactly. Further
increases, the probability of a malfunction in the improvements are coming up, namely our super-high-
dynamic alignment system is increased during a very resolution interferometer, with a resolution of 4 3
long run 1of the order of 1 m or more2. 1024 cm21, which has the moving cube-corner pair
back to back and the beam focus at the beam splitter.
This type of interferometer is almost ideal for very-
high-resolution infrared spectrometry.
The authors are with the Department of Applied Physics, Univer-
sity of Turku, FIN-20500 Turku, Finland.
However, there are still problems in the ultraviolet
Received 28 November 1994; revised manuscript received 3 May region with low resolution. The reason is that the
1995. pair of back-to-back cube corners has to be perfect.
0003-6935@95@276081-05$06.00@0. This means that, if we have to use six plane mirrors
r 1995 Optical Society of America. with, say, a flatness of better than l@20, the six

20 September 1995 @ Vol. 34, No. 27 @ APPLIED OPTICS 6081

angles formed by the mirrors cannot deviate more
than 0.001° from 90°, and the corners of the cubes
should coincide better than 0.1 mm. In principle, the
higher the resolution, the higher the performance of
the cube-corner pair has to be.
In this paper we present a new swinging interferom-
eter, the carousel interferometer, which has none of
the disadvantages described above. The new interfer-
ometer is compact in size and insensitive to mechani-
cal deformations.
2. Description of the Carousel Interferometer
The optical layout of the new carousel interferometer
is shown in Fig. 1. The interferometer consists of
beam splitter B and five plane mirrors, M1, M2, M3,
M4, and M5. The angle g, which determines the
directions of side mirrors M1 and M2, and the angle b,
which tells the directions of mirrors M3 and M4, are
shown in Fig. 1.
Beam splitter B and mirror M5 are fixed. Mirrors Fig. 2. Carrousel interferometer when the system of four mirrors,
M1–M4, has been rotated by the angle a around the axis O.
M1, M2, M3, and M4 are mounted on the carousel,
which rotates around the axis O. The axis O is
perpendicular to the plane of the figure.
The incoming beam hits the beam splitter at the tion when the carousel has been rotated by the
angle u. The transmitted beam travels in the left angle a.
section of the interferometer, first to mirror M1, then Let the distances of the mirrors be determined by R
to M3, and to the end mirror, M5. The beam hits M5 and D in Fig. 1. If the position of the left-side beam
perpendicularly, if the condition in mirror M5 is the point c at the distance of z,
measured from the center of the mirror, as shown in
2g 1 2b 5 u 112 Fig. 2, an inspection of the triangle abc 1Fig. 22 gives
the equation
is fulfilled, and returns to the beam splitter by the
same path. The reflected beam from the beam split- 1R@22 2 z
tan12b 2 2a2 5 .
ter follows a similar path in the right side of the D cos a 1 1z 2 D sin a2tan1b 2 a2
interferometer.
The optical path difference is created by the rota- 122
tion of the carousel on which mirrors M1, M2, M3, and
M4 are mounted. If the carousel is rotated counter- When z is solved, we find that z as a function of a is
clockwise, the optical path of the beam in the left side
of the interferometer is shortened and the path in the R
cot12b 2 2a2 2 D3cos a 2 sin a tan1b 2 a24
right side is lengthened. Figure 2 shows the situa- 2
z1a2 5 .
tan1b 2 a2 1 cot12b 2 2a2
132
The length of the optical path abcba is now

R R R

23 2
2 z1a2

sin12b 2 2a2
1
2
2 z1a2

tan12b 2 2a2
4 52
2
2 z1a2

tan1b 2 a2
.

142

The length of the optical path in the right side of the

interferometer is given by Eq. 142, where a is substi-
tuted with 12a2. Thus the optical path difference x of
the carousel interferometer is

x 5 R3cot1b 1 a2 2 cot1b 2 a24 2 2 cot1b 1 a2z12a2

1 2 cot1b 2 a2z11a2. 152

In Eq. 122 the small movement of point a, where the

Fig. 1. Optical layout of the new carousel interferometer: M1– beam hits the first mirror, was neglected. Thus Eq.
M5, plane mirrors; B, beam splitter. 152 is only an approximate formula of the path

6082 APPLIED OPTICS @ Vol. 34, No. 27 @ 20 September 1995

difference. The error between the exact and the
approximate values of the optical path difference is,
however, small. The exact path difference can also
be calculated easily with a computer. The optical
path difference as a function of the rotation angle is
practically linear. The path difference of a carousel
interferometer with typical dimensions is shown in
Fig. 3.
The maximum rotation angle is restricted by the
falling off of the beam from mirror M3 or M4 or end
mirror M5. The value of the maximum angle de-
pends on the angles g and b and the distances R and
D. In practice, it is of the order of 5–10°.
The rotation axis O does not necessarily have to be
situated in the center of end mirror M5, as in Fig. 1.
If the axis is in front of the mirror or behind it, the
optical path difference becomes less linear in large
rotation angles.
By modification of the optical layout of the carousel
interferometer, it is possible to build several varia-
tions of swinging interferometers. In all cases the Fig. 4. Examples of different types of swinging interferometers.
optical path difference is accomplished by the rotation
of a system of plane mirrors; only the path of the light
and the number of mirrors are changed. Four ex- created by the rotation of the parallel-mirror pair M1
amples of these variations are shown in Fig. 4. and M2 around the axis O 1see Fig. 62 perpendicular to
the plane of the picture.
3. Earlier Types of Swinging Interferometers If u is the angle at which the light enters the
A few other types of swinging interferometers have interferometer and R is the distance of the parallel
been introduced. One of them is the Dynascan in the mirrors, the optical path difference x as a function of
Perkin-Elmer System 2000 Fourier-transform infra- the rotation angle a has an expression given by
red spectrometer, in which the optical path difference
1 1 sin 2a 1 2 sin 2a

3 4
is also created by rotation. The optical layout of this .
interferometer is illustrated in Fig. 5. As one can see x 5 2R 2 162
1 2 1 2
u u
from Fig. 5, both the light beams use one pair of cos 90° 2 2a cos 90° 2 1a
parallel mirrors and one end mirror after the beam 2 2
splitter. The mirror systems for the two beams are
Despite the complicated expression, the optical path
completely separate.
difference x given in Eq. 162 is close to a linear function
The optical layout of another type of swinging
of a with small a values.
interferometer10 is illustrated in Fig. 6. Actually, it
is merely an improved modification of the Dynascan 4. Properties of the Carousel Interferometer Compared
interferometer, with the two pairs of parallel mirrors with other Swinging Interferometers
combined, thus making the construction mechani-
The new carousel interferometer is a typical plane-
cally more stable. The optical path difference is
mirror interferometer, in which a tilt of any mirror

Fig. 3. Optical path difference x of a carousel interferometer of

typical dimensions 1g 5 8.5°, b 5 14°, R 5 9 cm, D 5 7 cm2 as a Fig. 5. Schematic diagram of the optical layout of the Perkin-
function of the rotation angle a. Elmer Dynascan interferometer.

20 September 1995 @ Vol. 34, No. 27 @ APPLIED OPTICS 6083

 beam splitter. Distortions of type 1a2 can be effi-
ciently eliminated with a construction in which the
end mirror and the beam splitter are both fixed on the
same metal body in the immediate vicinity of one
another, thus minimizing their movements with re-
spect to each other.
If all the optical components are perfectly aligned,
the modulation of the carousel interferometer is inde-
pendent of the translational movements of the carou-
sel. When the carousel moves in the direction nor-
mal to the plane of the beam splitter, the only effect is
a change in the optical path difference. Movements
of the carousel in the direction of the beam splitter
have no effect. Distortions of mount type 1b2, tilt of
the rotation axis of the carousel, may be more signifi-
cant. Depending on whether the rotation axis is
tilted toward or perpendicular to the beam splitter we
call this distortion roll or pitch, respectively. The roll
of the carousel does not have a serious effect on the
Fig. 6. Optical layout of the rotating-parallel-mirror interferome- modulation, because the directions of the two interfer-
ter, in which the pair of parallel mirrors has been rotated by an ing beams do not change with respect to each other.
angle a around the axis O. The situation is analogous to a change of position of
the light source. Only the intersectional area of the
beams, where the fringes are observed, will be dimin-
generates a decrease in modulation m in the intensity ished. Unlike roll, the pitch of the carousel can
of the output beam, given by reduce the modulation, because the interfering beams
from the two arms of the interferometer will come out
x at different angles. However, in our carousel interfer-
3
Iout1x2 5 A 1 1 m cos 2p 1 l24 , 172 ometer even this kind of distortion should be rela-
tively insignificant because the axis is situated close
where l is the wavelength of the laser and x is the to the end mirror and to the beam splitter.
optical path difference. The coefficient A 5 I0RBTB, We have tested deformations of the mount with
where I0 is the intensity of the light entering the three types of interferometers: the Michelson inter-
interferometer and RB and TB are the reflectance and ferometer, the parallel-mirror interferometer, and the
the transmittance of the beam splitter, respectively. carousel interferometer. The mounts were distorted
We have built a small carousel interferometer 1Fig. 22, by the application of equal twisting and bending
an interferometer with rotating parallel mirrors 1Fig. forces to the mounts of each interferometer or of heat,
62, and a conventional Michelson interferometer with when hot air was blown on the mounts. Three types
plane mirrors. The performance of all has been of twisting or bending, illustrated in Fig. 7, were
tested by examination of the decrease in modulation applied to the mounts.
from the maximum value m 5 1 when the interferom- For the Michelson interferometer all these pertur-
eters were distorted both mechanically and by heat. bations caused a serious decrease in modulation.
A decrease in modulation is caused by any deforma- The parallel-mirror interferometer, instead, was im-
tion in the mirror system. In the carousel interferom- mune to distortion for the two first types of bending
eter there might be three possible sources of such a
deformation: deformation of the mount, deforma-
tion of the carousel, and external vibrations. The
most important reason for the modulation decrease,
in practice, is deformation of the mount.
An interferometer is fixed to its surroundings
through the mount, which is subject to external
pressure and temperature variations. Heat expan-
sion in particular often causes considerably strong
forces that bend or twist the mount. In the carousel
interferometer, however, the optical layout makes it
possible to position the beam splitter very close to the
Fig. 7. Diagrams of the three kinds of twisting and bending with
axis O and the end mirror, M5, which, furthermore, is which the mounts of the interferometers were tested: 112 bending
common for both of the interfering beams. There the mount in the plane of the beam splitter; 122 bending the mount
are two mechanical distortions that may result in a in the plane perpendicular to the plane of the beam splitter; and 132
decrease in modulation: 1a2 tilt of end mirror M5 twisting the mount, which rotates the ends of the mount in
with regard to the beam splitter and 1b2 tilt of the opposite directions. The intersection of the plane of the beam
rotation axis O of the carousel with regard to the splitter and the mount is indicated by the dashed lines.

6084 APPLIED OPTICS @ Vol. 34, No. 27 @ 20 September 1995

3Fig. 7, types 112 and 1224. However, type 132, twisting same feature in linear-motion interferometers, such
the mount, caused a decrease in modulation. The as the Michelson interferometer.
carousel interferometer was not affected at all by any
of these perturbations. Taken all together, the Mi- 5. Conclusions
chelson interferometer was approximately 10 times We have developed a new type interferometer, the
more sensitive than the parallel-mirror interferome- carousel interferometer, that uses five plane mirrors.
ter, and the parallel-mirror interferometer was also The optical path difference is created by rotation of
nearly 10 times more sensitive compared with the the carousel, which contains a system of four mirrors
carousel interferometer. fixed together. We have studied the performance of
The sensitivity of the mount to heat expansion the interferometer and its sensitivity to external
appeared to act in the same way. Blowing hot air on perturbations such as vibrations, bending, or other
the mount brought about the largest perturbations for deformation of the mechanical structure. The new
the Michelson interferometer, whereas the parallel- interferometer was compared with a patented rotat-
mirror interferometer was considerably less dis- ing-mirror-pair interferometer.
torted, having approximately 10 times less sensitivity We found several advantages to the carousel inter-
to heat expansion. The modulation of the carousel ferometer. Above all, the construction is very small,
interferometer changed the least, again approxi- compact, and cheap, because it is built of only plane
mately 10 times less than that of the parallel-mirror mirrors. The interferometer is very stable; the modu-
interferometer. Situating the end mirror, beam split- lation is independent of any deformation of the mount
ter, and rotation axis very close to one another makes of the interferometer. Thus it is immune to any
the carousel interferometer very stable and compact change in conditions, such as varying temperatures or
in size. The immunity of the carousel interferometer pressures. It is very suitable for low-resolution appli-
to mechanical deformations of the mount is an impor- cations in any wavelength range, including the ultra-
tant advantage. violet region.
In addition to deformations of the mount, misalign-
ment of the rotating mirrors might be another source References
of modulation decrease. However, the carousel is not 1. J. Connes and P. Connes, ‘‘Near-infrared planetary spectra by
vulnerable to external perturbations because it is Fourier spectroscopy. I: Instruments and results,’’ J. Opt.
fixed to its surroundings through only the two end Soc. Am. 56, 896–910 119662.
points of its rotation axis. The deformation of the 2. J. Kauppinen and V.-M. Horneman, ‘‘Dynamic alignment
carousel by inhomogeneous heat expansion would be system of a Michelson interferometer,’’ Rep. Series Turku-FTL-
R174 1Department of Physical Sciences, Turku, Finland, 19892.
possible only as a result of a strong heat flow through
3. Bomem, Inc., interferometers such as Models DA3-002, DA3-
the air. 01, and DA3-02.
Vibration has always caused malfunctions in inter- 4. J. Kauppinen and V.-M. Horneman, ‘‘Large aperture cube
ferometers, especially in linearly driven interferom- corner interferometer with a resolution of 0.001 cm21,’’ Appl.
eters, because the vibration usually has a component Opt. 30, 2575–2578 119912.
in the direction of the mirror movement. In interfer- 5. J. Kauppinen and P. Saarinen, ‘‘Line-shape distortions in
ometers that use rotational motion the problem is less misaligned cube corner interferometers,’’ Appl. Opt. 31, 69–74
serious, because the motion is not linear and the 119922.
effects of vibration, as well as of roll and pitch, can be 6. Mattson Instruments, Inc., for example, the Cygnus 100
interferometer.
eliminated with a properly fixed carousel.
7. E. Kyrö, J. Kauppinen, M.-L. Junttila, M. Linnavuo, P. Wallin,
The tests of the carousel interferometer and the and T. Valkeapää, ‘‘Fourier transform wavemeter,’’ Rev. Sci.
parallel-mirror interferometer confirm that the sensi- Instrum. 58, 1180–1184 119872.
tivity of the rotating-mirror system to external pertur- 8. Bomem, Inc., Michelson 100 FT-IR interferometer.
bations such as vibration and thermal expansion is 9. Perkin-Elmer Corp., System 2000 Dynascan interferometer.
orders of magnitude lower when compared with the 10. P. R. Brierley, U.S. patent 4,915,502 110 Apr. 19902.

20 September 1995 @ Vol. 34, No. 27 @ APPLIED OPTICS 6085