June 15, 2000 / Vol. 25, No.

12 / OPTICS LETTERS

893

Prism interferometer for a compact Fourier-transform spectroscope

Takanori Kiyokura, Takahiro Ito, and Renshi Sawada
NTT Telecommunications Energy Laboratories, 3-1 Morinosato-Wakamiya, Atsugi, Kanagawa 243-0198, Japan Received January 27, 2000 As a step toward a f ield-use spectrometer based on Fourier-transform spectroscopy, we have developed a prismscanning interferometer. It consists of a sliding triangular prism with two mirror surfaces attached to a f ixed triangular-prism beam splitter. This design eliminates the effect of ambient air and simplif ies alignment, which can otherwise be time consuming. It also halves the stage-moving distance, because the optical path difference is doubled, since both mirrors move while the sliding prism is scanning. We have tested the basic operation of the prototype and found that the spectral resolution and the wavelength scale agree well with our simple calculation. 2000 Optical Society of America OCIS codes: 300.6300, 300.6190, 220.4830, 230.5480, 120.4570, 300.6340.

Technology for detecting environmental pollution is in wide demand, and spectroscopic analysis methods have the great advantage that they do not have to be in contact with what is being measured, so a compact spectrometer for field use will be a valuable tool. Although there are a variety of spectroscopes available, the Fourier-transform spectroscope is the most popular and is most often used as a laboratory tool because of its high throughput, resolution, efficiency, and accuracy.1 The main optics of the Fourier-transform spectroscope is a Michelson interferometer. To acquire spectra, the spectroscope obtains an interferogram by measurement of signals from the detector while the optical path difference (OPD) is varied by a moving mirror scan. The interferogram is then transformed into spectra by a Fourier-transform algorithm. The longer the OPD, the higher the wave-number spectral resolution, which, in the case of triangular apodization, can be roughly represented by the simple relation Ds 0.88OPD , (1)

where Ds is the wave-number resolution and 0.88 is a factor that varies depending on the apodization form.1 In most cases the OPD equals 2x, where x is the mirror-movement distance. Therefore, obtaining high spectral resolution is easier than with other spectroscopes, such as a grating spectrometer, which requires narrow slits. Spectroscopes that are used for field measurement should not require alignment and should be portable, robust, and unaffected by ambient air. Water vapor and carbon dioxide, both of which are absorptive in the infrared region, should not affect them either. However, the optical system of the Fourier-transform spectroscope is a large and fragile interferometer, which makes such a spectroscope unsuitable for field use. In this Letter we describe a compact Fouriertransform spectroscope that can be used in the field. To realize this portable spectrometer we developed a new prism-scanning interferometer that can be used in a variety of field conditions. Figure 1 shows a top view of our prism interferometer, and the inset is a picture of the interferometer.
0146-9592/00/120893-03$15.00/0

The interferometer consists of a sliding triangular prism with two mirror surfaces, a fixed triangularprism beam splitter, a def lection-mirror prism, and a parallel fiber collimator unit. The sliding triangular prism has two retroref lection mirrors on the isosceles plane, and it slides parallel to the hypotenuse plane of the prism. The beam-splitter prism and the def lection-mirror prism are stationary. The beam-splitter prism consists of two isosceles right-triangle prisms with beam-splitter film. The def lection-mirror prism is an isosceles right triangle prism with a def lection-mirror surface. To prevent misalignment caused by slips we parallelize the output beam to the input beam by use of the def lection mirror, which is parallel to the beam-splitter film plane. The parallel fiber collimator unit is rectangular and includes collimator and condenser lenses and fiber ends, which are located at the focus of each lens. The unit is connected to the light source by the input fiber and to the detector (via the sample) by the output fiber. In addition, the unit is bonded to the surface of the beam-splitter def lection-mirror prism, which enhances its robustness. A refractiveindex-matching liquid layer is located between the

Fig. 1. Schematic configuration of meter. As the sliding prism moves, to the two retroref lection mirror 22nx sin u , and the OPD varies as on the prism-translation distance x. 2000 Optical Society of America

the prism interferothe relative distances are 12nx sin u and 4nx sin u , depending

894

OPTICS LETTERS / Vol. 25, No. 12 / June 15, 2000

sliding prism and the beam-splitterdef lection-mirror prism, which prevents total ref lection at the interface. The dispersion of the refractive-index-matching liquid is customized to match the dispersion of the prismglass material. The input beam is divided into two beams, the transmitted beam and the ref lected beam at the beamsplitter film plane, and both beams are then ref lected by the retroref lection mirrors of the sliding triangular prism. After that, the two beams are returned to the beam-splitter film plane and integrated into the output beam. An interferogram is obtained by scanning of the sliding triangular prism by a linear translation stage, and the intensity of the output beam is measured by the photodetector, which is homologized with the stage position. This design has several advantages: First, the two mirrors for retardation, retroref lection mirrors 1 and 2, are integrated into one prism, and so time-consuming optical alignment and realignment after production are not necessary. Second, the parallel fiber collimator optics, the beam splitter, and the def lection mirror are integrated; therefore optical alignment of the collimator optics is not necessary. Third, all optical paths run inside the prisms, so ambient air has no effect on measurement. Fourth, the OPD, which is changed by means of driving the scanning prism, is double that of an ordinary interferometer. The OPD is given by OPD 4nx sin u , (2) where n is the refractive index of the prism, x is the prism-movement distance, and u is the angle of the sliding prism. The OPD is 4.2 times the prism-movement distance x when u 45 and n 1.5. This result shows that the movement distance is halved, or the wave-number resolution is doubled, compared with that of an ordinary interferometer. For example, 4-cm21 resolution requires only a 60.6-mm scan. Fifth, the prisms used are ordinary commercial prisms and thus inexpensive. Finally, the design is independent of the glass material and the device size. The optical paths run through the refractive medium, i.e., through the glass. Thus, changes in the refractive index of the glass, which depends on the wave number, have to be considered. The wave number in the glass is given by s 0 s ns , where s is the wave number in air and ns is the refractive index of the glass at wave number s . The wave-number dependence of the refractive index of the glass, s versus ns , and the wave-number scale-correction curve, s versus s 0 , are shown in Fig. 2. The refractive-index data and the dispersion formula for the Ohara S-BSL7 glass, which are equivalent to those for Schott BK7 and Hoya BSC7 glass, were provided by the manufacturer.2,3 The wave-number scale-correction curve was almost linear, with an inclination of 1.5. To correct the refractive index we measured the interferogram with prism-movement distance x, and a Fourier transform was performed on the interferogram with an OPD of 4x sin u in the glass. The spectrum that was obtained was then scaled by the wave number in the glass, s 0 , which equaled 0.884x sin u . After that, the wave-number scale in the glass was replaced with

the wave-number scale in air by use of the wavenumber scale-correction curve s versus s 0 . The deviation from the standard refractive-index data2,3 was 65 3 1024 , so the wave-number error was 65 cm21 at 15, 802 cm21 . However, if a standard glass block with an index inhomogeneity of 62 3 1025 is used and the refractive index of part of the block is measured with an error of 61026 for calibration,2,3 a prism interferometer Fourier-transform spectroscope can be mass produced from the glass block with a wave-number error of 60.2 cm21 at 15, 802 cm21 . To test the feasibility of the prism interferometer we measured spectra from various light sources. Figure 3 shows the measured spectrum of a HeNe laser. The spectral width was so narrow that we assumed that the measured spectrum represented the resolution itself, and the spectral peak position was very stable at 15, 802 cm21 . We used a Si photodetector that is sensitive to visible light. The stage was line scanned once over 675 mm, in sampling steps of 52 nm, which resulted in an OPD of 6321 mm and an OPD sampling step of 223 nm. The shortest

Fig. 2. Wave-number dependence of the refractive index of Ohara S-BSL7 glass (open squares) and the wave-number scale-correction curve (filled diamonds).

Fig. 3. Spectrum of a He Ne laser measured during a 675-mm sliding prism scan.

June 15, 2000 / Vol. 25, No. 12 / OPTICS LETTERS

895

Fig. 4. Spectrum of an InGaAsP semiconductor laser measured during a 6150-mm sliding prism scan.

Fig. 5. Spectrum and (inset) the interferogram of a tungsten-halogen lamp measured during a 6150-mm sliding prism scan.

observed wavelength was 446 nm 22, 421 cm21 . The wave-number resolution, estimated by use of Eqs. (1) and (2), with n 1.51509, was 27.6 cm21 . The measured spectral width was 25 cm21 , which is equivalent to a wavelength resolution of 1 nm. This result is consistent with the calculated results. The shift of the spectral peak position was 26 cm21 . We think that the cause of this shift was measurement error related to the prism position. From 10,000 to 22, 000 cm21 , the spectrum was acquired with a signal-to-noise ratio ranging from 1000 to 300. Next, to estimate the spectral resolution in the near-infrared region, we measured the spectrum of a Fabry Perot-type InGaAsP semiconductor laser used for telecommunications (Fig. 4). This laser emits near-infrared rays with a wave number of 6450 cm21 . We used a PbS photodetector that was sensitive to near-infrared rays. The stage was line scanned once over 6150 mm, in sampling steps of 104 nm, which resulted in an OPD of 6642 mm, in OPD sampling steps of 441 nm. The shortest observed wavelength was 883 nm 11, 327 cm21 . From 4500 to 10, 500 cm21 , the spectrum was acquired with a signal-to-noise ratio of 1000. The wave-number resolution, estimated with n 1.50065, was 13.9 cm21 . The Fabry Perot laser spectrum should have many subpeaks, but the spectral

resolution was too wide to separate them, and so the spectral width should be similar to the resolution. The measured spectral width was 13 cm21 , which is in good agreement with our simple calculation. Thus, in the ranges of both visible and near-infrared light, the spectral width of the lasers agreed well with the simple calculations of the wave-number spectral resolution. To test the feasibility of applying the prism interferometer as a spectroscopic analysis instrument we measured an interferogram and a spectrum of white light from a tungsten-halogen lamp. The inset of Fig. 5 shows the interferogram, which was measured during a one-time 6150-mm sliding prism scan. The corresponding wave-number resolution is 13.9 cm21 . We used a PbS photodetector for the near-infrared rays. Because of the symmetric nature of the optics, the envelope curve of the interferogram was almost symmetric. After Fourier transformation of the interferogram data, we obtained the spectrum of the white-light source (Fig. 5). Originally the white-light source had an almost f lat spectral distribution, which means that the spectrum represents the instrument transmission function. The spectrum shows a slight decrease near 5000 cm21 , but there is no characteristic absorption. We have thus conf irmed that this interferometer can work not only as an optical spectrum analyzer but also as a tool for absorption or ref lection-spectrum measurement with a white-light source. In conclusion, we have developed a prism-scanning interferometer and tested its basic operation by building a prototype. We confirmed that the prototype can be used as a Fourier-transform spectroscope. By integration of all the optical components, the optical paths are isolated form ambient air, and the need for timeconsuming alignment is reduced. The optical components are inexpensive prisms, so the spectrometer can be built inexpensively. The wavelength scale was corrected by use of the refractive-index dispersion curve. The prism interferometer was used to measure the spectra of lasers and a white-light source, and the resolution closely matched the calculated results. The current resolution is low because of the linear-stage performance. If microlinear stages with long moving distance can be used,4 the wave-number resolution will be improved. However, even with the present resolution, the spectra of plastics and organic liquids can be measured, and the prism interferometer can be used for chemometrics applications. T. Kiyokuras aecl.ntt.co.jp. References
1. P. R. Griffith and J. A. de Haseth, Fourier Transform Infrared Spectrometry (Wiley, New York, 1986), Chap. 1. 2. Ohara Glass Catalog (Ohara Inc., Sagamihara 229-1186 Japan, 1995). 3. http://www.ohara-inc.co.jp/. 4. S. D. Collins, R. L. Smith, C. Gonzalez, K. P. Stewart, J. G. Hagopian, and J. M. Sirota, Opt. Lett. 24, 844 (1999).

e-mail

address

is

kiyokura@