Multichannel Fabry-Perot spectrometer for infrared astronomy Donald E. Jennings and R. J. Boyle

Donald Jennings is with NASA Goddard Space Flight Center, Planetary Systems Branch, Greenbelt, Maryland 20771, and R. J. Boyle is with Dickinson College, Depart­ment of Physics & Astronomy, Carlisle, Pennsylvania 17013. Received 16 June 1986.

Fabry-Perot interferometer spectrometers are among the most sensitive high spectral resolution instruments used in infrared astronomy today.1-2 They are used to study atomic and molecular lines in planetary atmospheres, interstellar clouds, and stars. Fabry-Perot spectrometers have the ad­vantage that they achieve high resolution by producing a transmission fringe which is extremely narrow in spectral bandwidth. Radiation outside the transmission fringe is rejected and does not contribute noise to the spectrum. A cryogenically cooled Fabry-Perot spectrometer reduces the radiant background to the point where it is possible to oper­ate near the sensitivity limits of the detector and preamplifi­er. Although the Fabry-Perot spectrometer covers only lim­ited regions of the spectrum in a given observation, single molecular or atomic lines can be studied with maximum signal-to-noise. An additional advantage of the Fabry-Perot is that it is nondispersive, so that the source can be imaged in two dimensions in the focal plane. A detector array can therefore be used to produce spatial maps of all source points simultaneously.

Fabry-Perot spectrometers work by making use of the high finess (ratio between fringe separation and fringe width) and high throughput efficiency available with etalons made from modern infrared materials and dielectric reflection coat­ings.3 A low resolution filter or grating selects one fringe of the etalon and radiation transmitted by this fringe is focused at the detector. To produce a spectrum, the spectrometer is scanned by either tilting the etalon or by changing the etalon plate separation. The time required to scan a region is T = tN, where t is the integration time per resolution element and N is the number of resolution elements (fringe widths) being scanned.

4520 APPLIED OPTICS / Vol. 25, No. 24 / 15 December 1986

The primary disadvantages of Fabry-Perot spectrometers is that only one spectral element is accessed at a given time. Therefore a large fraction of the radiation from the telescope is rejected by the etalon, and flux containing useful informa­tion is wasted. A spectrometer scanning N resolution ele­ments uses only a fraction of 1/N of the flux which could contribute to the spectrum. If a technique were devised which would allow the rejected radiation to be used the instrument sensitivity would be improved.4

We have developed a design which makes use of the radia­tion normally rejected in a Fabry-Perot spectrometer. Fig­ure 1 is a schematic of a spectrometer which permits observa­tion of multiple resolution elements simultaneously. A bandpass filter limits the spectral bandwidth of incoming radiation to one free-spectral-range (fringe separation) of the etalon. The radiation enters the spectrometer, after being limited by a field aperture, and is collimated by M l and directed to the etalon. The etalon is oriented at an angle φ with respect to the incoming beam, and transmits a fringe pattern given by mλ = 2d cosθ, where m is the order (or fringe) number, λ is the wavelength of the radiation, d is the separation of the etalon surfaces, and the angle θ is equal to φ for the initial pass. The radiation is then focused at the detector element at one edge of a linear detector array. The radiation which is not transmitted by the etalon is reflected and focused at the surface of mirror M2 and directed to M3. The beam is then focused again by M3 at the edge of M2 near the entrance beam. The beam is returned to the etalon at an incident angle, i.e., φ – δ, which is slightly different from that of the initial beam. The angle is chosen to place the second-pass fringe at a spectral position adjacent to the first-pass fringe. The second-pass beam is focused at the second de-

Fig. 1. Optical schematic of the multiple-pass Fabry-Perot etalon spectrometer. Mirrors Ml and MA are off-axis parabolas, while M2 and M3 are spheres. The radiation path for the first cycle is shown with a dashed line. Radiation from the input focus and subsequent multiple-pass foci is collimated at mirror Ml and directed to the etalon, where one spectral element is transmitted. The etalon is tilted at a slight angle φ with respect to the first pass radiation beam. The radiation reflected from the etalon forms foci at M2 (open dots). Mirror M3 reimages the foci back at M2, forming a series of evenly spaced foci (solid dots). Each shift in position at M2 corresponds to a change in incident angle at the etalon, so that the transmitted spectral element is shifted in wavelength. The pattern at M2 is imaged at the focal plane, where a detector array can be used to

observe the multiple spectral elements simultaneously.

tector element in the array. The second-pass reflected beam follows the same optical sequence as did the first-pass beam and the cycle is repeated several times. The focal points at M2 (and at the array) are equally spaced, causing equal changes in angle (θ = φ,φ – δ,φ – 2δ, . . .) between etalon passes. The focal lengths of Ml and M2 are chosen to image the etalon at M3, and this causes the etalon to be reimaged on itself after each cycle.

Alternative optical designs for multipassing are possible, but this one has the advantage of minimizing the diameters of the etalon and optics, and permits a pupil stop to be located at the etalon. In particular, our design has advan­tages over the cases considered by Ascoli-Bartoli et al.4 Our use of spherical mirrors allows us to separate the images at M2, so that no radiation is lost through the entrance aperture and the beams can be completely collimated at the etalon.

Since each pass is shifted in angle by ∆θ = δ from the previous pass, the detector array samples a sequence of spec­tral elements which corresponds to a spectrum. The posi­tions of the spectral elements are not linear in λ, but instead follow the dependence cos(φ - Lδ), where L = 0,1,2, . . . . The entire spectrum is detected simultaneously. In the ideal case in which there are no losses at the etalon or mirrors the improvement in signal-to-noise in the spectrum is pro­portional to √n, where n is the number of detector elements in the array. The integration time necessary to record a spectrum decreases by 1/n. In practice the number of useful passes will be limited by the reflection efficiencies at the etalon and beam-cycling mirrors. The spectrometer is tuned by changing the etalon spacing or tilt φ. An advantage over conventional single-channel Fabry-Perots is that no mechanical tuning is necessary to obtain a spectrum, since all spectral points are detected simultaneously. Alternatively, however, the Fabry-Perot plate separation could be scanned to produce a complete spectrum in each channel.

We have constructed a laboratory version of the design in Fig. 1. Spherical gold-coated mirrors were used for M2 and MS, and these both had radii of 105 cm. The separation between M2 and MS was also 105 cm. The off-axis parabolic mirror M l was aluminum-coated and had a focal length of 46 cm. The etalon was placed 26 cm from Ml, which caused its image to be located at MS. The etalon was constructed of two parallel plates of ZnSe with dielectric 90% reflection coatings on the inner air-gap cavity surfaces and antireflec-tion coatings on the outer surfaces. The coatings were opti­mized for 12.5-µm wavelength. The plate separation was 1.1 cm, which produced a free-spectral range of 0.44 cm–1. An­other off-axis parabolic mirror MA with a 15-cm focal length formed the multiple field images in the focal plane. For our laboratory study we used a diode laser operating at 12.2-µm wavelength to provide a tunable monochromatic radiation source. This permitted a characterization of the throughput and fringe profile for each etalon pass as well as the beam-cycling efficiency. Instead of an array of detectors, a single-element 1-mm diam detector was moved among the multi­channel output positions in the focal plane and a separate fringe profile was recorded at each position.

Figure 2 demonstrates the fringe intensity and shape after each pass. The fringe spectral position shifts to longer wave­length and follows the c o s t ( φ – Lδ) (approximately quadrat­ic) dependence; as the angle decreases toward normal inci­dence on successive passes the change in a fringe position is progressively less for the constant angular change δ between passes. Adjacent passes begin to overlap in bandpass, caus­ing the fringe shape to become asymmetric and the intensity to decrease more and more rapidly. After normal incidence the fringe position would shift back toward short wave­lengths, but that radiation has already been passed by the etalon. The peak fringe intensities measured from the data in Fig. 2 were normalized to the laser intensities to produce

Fig. 2. Transmission fringe profiles for ten passes. A diode laser was used to characterize the fringe profiles and throughput efficien­cies. A single detector was moved in the detector focal plane to examine the multiple output foci (see Fig. 1) in sequence. These are raw data and have not been normalized to correct for a variation in diode laser intensity across the spectrum (the normalization was

applied before plotting the transmissions shown in Fig. 3).

Fig. 3. Absolute peak transmissions for ten passes. To obtain these values the peak fringe intensities in Fig. 2 were normalized with respect to the diode laser intensities measured with the etalon removed from the beam. The detector position is referenced to the

position of the first pass focus.

15 December 1986 / Vol. 25, No. 24 / APPLIED OPTICS 4521

absolute peak transmission values, and these are plotted in Fig. 3. These absolute transmissions demonstrate that as many as eight channels could be employed in the present spectrometer with adequate throughput. The total flux inte­grated during a multichannel measurement is proportional to Σi εi, the sum over the throughput efficiencies of all passes. In our laboratory demonstration depicted in Figs. 2 and 3 the measured efficiencies imply that using eight channels in­creases by a factor of 4 the flux integrated during a given time period compared with a single-channel instrument.

Because the spectrometer is nondispersive, the source can be imaged at each of the several output spectral positions. If a 2-D detector array were used, a spectral line could be mapped in the source simultaneously in spatial position and in spectral frequency. For example, a 10 × 50 element array could be used to create five maps of an astronomical source, each showing the intensity pattern at a different frequency (or velocity) resolved into a 10 × 10 grid. Alternatively, a 2-D array could be used to increase the spectral coverage; if a grating were used to isolate Fabry-Perot orders, these could be diffracted in the orthogonal array dimension, thereby using several free-spectral-range intervals of the etalon si­multaneously.

We expect to be able to increase the throughput in the multipass channels by improving the reflectivity of the mir­rors and etalon. Use in infrared astronomy beyond 5-µm wavelength will require cooling of the etalon, and perhaps the mirrors, to liquid nitrogen or helium temperature. De­tector arrays are now available in the mid-infrared which approach the sensitivity needed to fully implement a multi­channel Fabry-Perot.5

References 1. S. C. Beck et al., "High Spectral and Spatial Resolution Observa­

tions of the 12.28-µm Emission from H2 in the Orion Molecular Cloud," Astrophys. J. Lett. 253, L83 (1982).

2. M. Tanaka, Y. Yamashita, S. Sato, and H. Okuda, "A Fabry-Perot Spectrometer for Near-Infrared Astronomical Observa­tions," Pub. Astron. Soc. Pac. 97, 1020 (1985).

3. The Theory of Fabry-Perot spectrometers for astronomy has been discussed by, for example, F. L. Roesler, "Fabry-Perot In­struments for Astronomy," in Methods of Experimental Physics, Vol. 12, N. Carlton, Ed. (Academic, New York, 1974), p. 531; A. H. Vaughan, Jr., "Astronomical Fabry-Perot Interference Spectros­copy," Ann. Rev. Astron. Astrophys. 5, 139 (1967).

4. U. Ascoli-Bartoli, G. Benedetti-Michelangeli, and F. DeMarco, "An Improvement in Fabry-Perot Spectrometry," Appl. Opt 6 467 (1967).

5. See, for example, J. F. Arens, G. M. Lamb, and M. C. Peck, "Infrared Camera for 10 µm Astronomy," Proc. Soc. Photo-Opt! Instrum. Eng. 280, 50 (1981).

4522 APPLIED OPTICS / Vol. 25, No. 24 / 15 December 1986