An infrared Lyot filter was fabricated by integrating a polarization beam splitter and two retarders into a single device. A liquid crystal layer was constructed between two silicon pentaprisms that were designed suitably so that light was incident on this layer at 28°. At this angle, the liquid crystal transmitted p-polarized light (Brewster’s angle) and reflected s-polarized light (total internal reflection). The p- or s-polarized light was directed to another liquid crystal layer (retarder) between the prism and a mirror, which induced a wavelength-dependent retardation in the reflection process. Consequently, the light that returned to the beam splitter was transmitted or reflected depending on wavelength.
© 2013 OSA
Applications of liquid crystals (LCs) are not limited to displays but extensive to the entire photonics fields . Various LC devices have been developed for uses in optical communications, photonic data processing, spectroscopy, switchable windows, etc. Most LC devices are dedicated to the visible or near-infrared region and rarely used in the mid- or far-infrared region beyond 2 μm wavelength. Development of infrared LC devices, however, has a long history [2–5], and optical properties of LCs have been studied in the wide infrared spectral range [6–8]. Difficulty in fabricating infrared devices arises from the following facts that are specific to infrared optical systems; i.e., 1) ordinary glasses, polymers, and even LCs themselves are absorptive, 2) optical elements, e.g., photodetectors or polarizers, are inefficient, nondurable, and expensive, 3) no strong lamps (incandescent light sources) exist and laser emission wavelengths are limited, and 4) optical alignment is difficult to achieve with invisible light.
Optical substrates for infrared elements are usually made of halide crystals (KCl, CsI, AgCl, TlBrI, BaF2, etc.), chalcogenide glasses or crystals (As2S3, ZnS, ZnSe, etc.), and semiconductors (Si, Ge, GaAs, etc.). Of these infrared transmitting materials Si and Ge provide the most reliable substrates, since they have sufficient chemical, thermal, and mechanical strengths as well as they are free from deliquescence, photodarkening, or toxicity. Electric conductivity of semiconductors is also advantageous when controlling LC orientation by voltage application [2–5, 9–12]. A disadvantage of these semiconductors is a high reflection loss that is caused by the high index of refraction. Since their refractive indices, i.e., 3.4 (Si) and 4.0 (Ge), are higher than any other transparent materials, the surface reflectance exceeds 30%, and accordingly, the substrate transmittance becomes approximately 50%; i.e., half optical power is lost. A throughput of an optical system, therefore, decreases remarkably every time a semiconductor-based element, e.g., a lens or a polarizer, is inserted in the optical path . Although an anti-reflection coating is effective to reduce the insertion loss, a low loss is not attainable in a wide spectral range. In addition to weak light intensity due to lack of strong light sources and the high reflection loss, invisibility of infrared light hinders optical alignment of multiple elements.
A solution to these problems is to integrate multiple elements into a single device. The element integration decreases the number of reflective surfaces and realizes an alignment-free system. It also improves the compactness and robustness of the optical system. A concept of photonic integrated circuits was proposed many years ago, and thereafter various opto-electronic devices have been developed for uses in the visible and near-infrared region. In the long-wavelength infrared region, however, the advantages of integration have not been fully utilized yet. LC is a suitable material for creating integrated devices, since its large birefringence and electric controllability induce various optical functions, e.g., polarization selection [5, 14], polarization rotation [9, 13, 15, 16], retardation [4, 14, 17, 18], phase shift , reflection , scattering [2, 3], diffraction [10–12], interference , and bistability . Si provides a suitable substrate for creating infrared LC devices owing to its infrared transparency, electric conductivity, and micromachining adaptability [5, 9–13, 19].
In this study, we fabricated an infrared Lyot filter by integrating a polarization beam-splitter and tunable retarders. The Lyot filter is a spectroscopic device that consists of a polarizer, a retarder, and an analyzer (another polarizer) [22–24]. When linearly polarized light passes through the retarder, its polarization state changes depending on wavelength (linear, elliptical, or circular polarization). After passing through the analyzer, therefore, the light exhibits a spectrum with a periodic oscillation. Theoretically, transmittance of the Lyot filter becomes 0% at specific wavelengths, and hence, a high-contrast spectrum is attainable. This is an advantage of the Lyot filter when compared with Fabry-Perot filters, which require high-reflectance cavity mirrors to attain a high-contrast interference spectrum. The transmission wavelength of the Lyot filter is electrically tunable if the retarder is constructed with an LC cell [14, 17, 18, 21, 25, 26]. A polarizer or a polarization beam splitter is also manufacturable by combination of LC and prisms [5, 14]. As contrasted with absorption-type polarizers, which cut off one polarization component, this polarization beam splitter provides both p and s components, and consequently, high throughput is attainable in measurements [14–16, 27]. The Si-LC combination is useful for creating an infrared polarization beam splitter, since most birefringent crystals or coating materials are opaque in the infrared region . If Si prisms are suitably designed, these polarization and retardation functions can be integrated in a single device. In the following sections, we describe an optical design of the integrated device and then report experimental results.
2. Device structure and principle
A nematic LC with a large birefringence (Merck, BL006) was selected to attain efficient polarization and retardation functions. The ordinary and extraordinary indices of refraction were no = 1.53 and ne = 1.82, respectively. As Fig. 1 shows, three LC layers were constructed on surfaces of two Si pentaprisms. The LC layer between the two prisms [Fig. 1(b)] acted as a polarization beam splitter that transmitted p-polarized light and reflected s-polarized light. The other two LC layers between a prism outer surface and a mirror (an Au-coated glass plate) [Fig. 1(c)] acted as a retarder for the branched p- or s-polarized light.
In the polarization layer shown in Fig. 1(b), LC molecules are oriented perpendicular to the Si surface. A light beam impinges on this layer with an incident angle θ. The s-polarized component of this light takes the ordinary index of refraction (no = 1.53). On the other hand, the p-polarized component takes an index between the ordinary and extraordinary indices (1.53<nL<1.82), which is determined by the following expression ; i.e.,Figure 2(a) shows the refractive index for s or p polarization as a function of the incident angle θ. As the incident angle increases, the index for p polarization increases gradually and reaches ne = 1.82 at 32°, which corresponds to the critical angle of the total internal reflection. Figure 2(b) shows an angular dependence of the reflectance that was evaluated by using the Fresnel reflection formula  with the refractive indices shown in Fig. 2(a). Since the refractive index is smaller for s polarization than p polarization, s-polarized light suffers the total internal reflection at a smaller angle (27°). The reflectance for p polarization becomes negligible in the vicinity of θ = 27–28°corresponding to the Brewster’s angle. If the incident angle θ is adjusted at ~28°, therefore, p-polarized light is transmitted and s-polarized light is reflected.
Figure 1(c) shows the structure of the retardation layer (the retarder 2) on the lower right of the prism, to which the p-polarized light is directed. LC molecules in this layer are aligned in the plane parallel to the Si surface (the mirror surface) and directed at 45° to the polarization direction. When light of wavelength λ makes a round trip in this LC layer of thickness d, it suffers a retardation Δφ; i.e.,Fig. 1(c), for example, light of wavelength λ1 or λ2 changes its state to s or elliptical polarization, while the polarization state is unchanged at λ3. Consequently, when the light beam returned to the polarization layer, the spectral components at λ1, λ2, or λ3 are totally reflected, partially reflected, or transmitted, respectively. This wavelength dependent reflection induces a high-contrast variation in the output beam spectrum. The same process takes place in the left prism of the device (the retarder 1); i.e., the s-polarized light that was directed to the lower left side suffers a phase change, and some spectral components emerge from the output surface as p polarized light.
If a voltage (V1 or V2) is applied to the retardation layers, the LC molecules change their orientation to the electric field direction, i.e., perpendicular to the Si (mirror) surface. This orientation change decreases the birefringence (Δn) or the retardation (Δφ), leading to a spectral change of the output beam. When the reorientation is complete by high voltage application, no retardation takes place, and consequently, no output light emerges from the device.
As mentioned earlier, the incident angle θ at the central LC layer has to be 28° to attain a low cross-talk (high contrast) between p and s polarizations. It is also desired to align the input and output beams on the same optical axis for facilitating optical system design and alignment. These requirements are satisfied, if the prism surface is inclined at a suitable angle α, as shown in Fig. 1(a). Since the angles of incidence and refraction at the input surface are θ0 = α and θ1 = α–θ, respectively, the Snell’s law gives the relation
3. Device fabrication
Two Si pentaprisms were prepared according to the optical design above. Although the element integration was effective to reduce the reflection losses at intermediate boundaries (surfaces of the in-between elements), reflection at the input and output surfaces were unavoidable. These surfaces were therefore coated with Al2O3 and ZnS films to reduce the reflection loss. The reflection loss was less than 10% in the 2–6 μm wavelength range.
As regards the retardation layers, the surfaces of the Si prism and the mirror were coated with a polyimide film (Nissan Chemical Industries, SE-410) to align the LC molecules parallel to the surface. The film was rubbed in the 45° direction with a rayon cloth to orient the LC director. The mirror was adhered to the prism so that the rubbing directions of the opposite surfaces were parallel to one another. Glass beads of 20 μm diameter were added to epoxy adhesive to create a gap between these two substrates. Then the LC was injected into this gap, and the periphery was sealed with the adhesive [5, 9].
The LC in the polarization layer had to be oriented perpendicular to the Si surface, as shown in Fig. 1(b). This orientation was attainable by coating the Si prisms with another type of polyimide film, i.e., a homeotropic alignment film (Nissan Chemical Industries, SE-1211). The refractive index of this film, however, was too low (1.54) to construct a polarization beam splitter, i.e., the total internal reflection occurred at 27° on the Si-film boundary regardless of the polarization direction. We therefore employed the electric orientation method (no alignment film). The two Si prisms were adhered to one another with the epoxy adhesive containing 5-μm beads to create a gap for LC injection. No thicker LC layer was needed for inducing the total internal reflection at the Si-LC boundary. This fact contrasted with the requirement for the retardation layer in which a large optical thickness was preferred to attain a sufficient phase change.
Electric wires were soldered to the prism and mirror surfaces by using an ultrasonic solderer. As shown in Fig. 1(a), an electric source generated a sinusoidal electric signal of 1 kHz, and amplifiers adjusted the signal voltage between 0 and 100 V (peak voltage). A signal voltage V2 was applied to the retardation layer on the right side (the retarder 2). Another signal voltage was divided into V1 and Vp to control the left retardation layer (the retarder 1) and the polarization layer, respectively. The voltage Vp was kept at 40 V in all experiments, since the LC reorientation was complete at this voltage.
4. Preliminary experiments
The performance of the fabricated device was examined by using a Fourier transformation infrared spectrometer (FTIR, Shimadzu, IR Affinity-1). The probe beam was collimated with apertures to reduce the focusing angle (deviation of the incident angle θ0) to Δθ0 = 3.6°. When θ0 was 39.0 ± 3.6°, the refraction angle was θ1 = 10.7 ± 0.9° and accordingly, the incident angle at the Si-LC boundary was θ = α–θ1 = 28.3 ± 0.9° (Fig. 1). In preliminary experiments, a BaF2 wire-grid polarizer (Edmund, WGP8203) was inserted in front of the sample. This polarizer was used only for evaluation of the polarization characteristic and removed in the final experiment, since the current device requires no additional polarizer. The transmittance was evaluated as the ratio of the probe light intensities that were measured before and after placing the device on the sample stage.
Figure 3 shows the transmission spectra that were measured for s polarization; i.e., the wire-grid polarizer was adjusted to create an input beam that was polarized in the direction perpendicular to the page in Fig. 1(a). The spectra in Fig. 3(a) were measured by applying a voltage V1 to the retarder 1 and no voltage (V2 = 0 V) to the retarder 2. No spectral change was visible below V1 = 1.4 V, since the voltage was too low to induce LC reorientation. As the lowest spectrum shows (1.0 V), peaks appeared at around 2, 3, 4, and 5–6 μm, and troughs were visible between them. The spectra were complicated because of strong absorption bands at 3.4 and 4.5 μm as well as weak absorption bands in the 4–7 μm range, which appeared in all spectra regardless of the applied voltage. The measurement range (2–8 μm) was limited by both the spectral intensity of the probe light and the transmission range of the wire-grid polarizer. As the voltage increased, the peaks shifted to shorter wavelengths, since the LC reorientation caused decrease in the birefringence Δn and the retardation Δφ. As the voltageincreased further exceeding 10 V, the spectrum gradually became flat, since the birefringence approached zero. By contrast to the voltage application to the retarder 1, no peak shift was visible when a voltage (V2) was applied only to the retarder 2 (V1 = 0 V); i.e., as shown in Fig. 3(b), all spectra were the same as the lowest one in Fig. 3(a). This fact indicated that the s-polarized probe light was reflected perfectly at the central LC layer (the polarization layer); i.e., no light beam was transmitted to the retarder 2.
Since the noisy absorption bands disturbed the spectra, the transmittance was replotted as a function of applied voltage V1. The black and gray lines in Fig. 4 show the dependences on the voltages V1 and V2, respectively. The transmittance oscillates as the voltage V1 increases, while no change is visible with the V2 variation. The highest transmittance is ~40% in the 2.5–5 μm range. The insertion loss of ~60% is attributed mainly to the reflection at the input and output ends (incomplete anti-reflection coating) as well as the Si-LC boundary of the retardation layer (no anti-reflection coating). Since the anti-reflection coating is less effective at 2 and 6 μm, the transmittance is lower at these wavelengths than the other wavelengths. As the voltage V1 (the retarder 1) increases exceeding 10 V, the transmittance decreases gradually to 0%, since the retardation Δφ (refractive index difference Δn) decreases by the LC reorientation. The rightmost peak in each curve corresponds to a retardation of π, and other peaks correspond to 3π, 5π, … from right to left. The troughs corresponding to 2π, 4π, … appear at voltages between these peaks. In Fig. 4(b), for example, the peaks and troughs at 5.5, 3.4, 2.5, 2.2, 2.0, 1.8, and 1.5 V correspond to π, 2π, 3π, 4π, 5π, 6π, and 7π, respectively. As wavelength becomes longer, the number of peaks decreases, since, as Eq. (2) indicates, the retardation is inversely proportional to the wavelength.
Next, transmission spectra were measured for p polarization by changing the direction of the wire-grid polarizer. Figure 5 shows the spectral change that was caused by voltage application to the retarder 1 or 2. In contrast to the former experiment, spectral change occurred only when a voltage was applied to the retarder 2. Although peak shift was visible in Fig. 5(b), the voltage dependence was milder than that for s polarization [Fig. 3(a)].
As Fig. 3(a) shows, no transmittance change takes place below 1.4 V. Such a low voltage cannot induce LC reorientation, and hence, the retarder 1 keeps the original birefringence of Δn = ne–no = 0.29. The spectra in this low-voltage range exhibit peaks at λ = 2.1, 2.6, and 3.8 μm. Taking into account the graphs in Fig. 4, these peaks are assumed to correspond to retardation (Δφ) of 9π, 7π, or 5π, respectively. By using Eq. (2), therefore, the thickness of the retarder 1 is evaluated to be d≈16 μm.
Similarly, LC in the retarder 2 is not reoriented with a voltage below 1.2 V, as shown in Fig. 5(b). In this voltage range, peaks are located at 2.6 and 3.8 μm. Since these peaks correspond to retardations of 5π and 3π, respectively, thickness of the retarder 2 is assumed to be 10–11 μm, which is smaller than that of the retarder 1. This is the reason that the peak shift is mild in the retarder 2. A thickness at the cell center can be smaller than the beads diameter (20 μm) if the prism and the mirror are pressed strongly in the adhering process. Or, if the LC director is not completely parallel to the substrate (anchoring angle: >0°), Δn becomes smaller than 0.29, and hence, yields a larger value for thickness; e.g., if Δn = 0.16 is substituted in Eq. (2), it gives d≈20 μm.
A voltage dependence of the retardation (Δφ) can be evaluated most accurately at 2 μm wavelength, since, as Fig. 4 shows, the data at this wavelength contain the largest number of peaks and troughs that correspond to Δφ = mπ (m = 1, 2, 3, …). The open circles in Fig. 6(a) show the data that were evaluated with the peaks and troughs in Fig. 4(a). As the voltage rises, Δn, and accordingly Δφ, decreases due to the LC reorientation. The data for the retarder 2 were taken in the same manner and plotted with closed circles in Fig. 6(a). The retarder 2 exhibits a smaller retardation, since its thickness is smaller than that of the retarder 1.
To attain the same spectrum with s- and p-polarized probe beams, the same retardation has to be created in the two retarders by applying different voltages to them. As Fig. 6(a) shows, for example, the retardation of π is attainable by applying V1 = 8 V to the retarder 1 and V2 = 7 V to the retarder 2. The correspondence of V1 and V2 was plotted in Fig. 6(b). This graph contains data that were taken for other wavelengths. The regression line gives the relation,
The complicated voltage application process above can be avoided if the fabrication process is improved to create the retardation layers with the same thickness. Otherwise, if a device containing a single retardation layer is fabricated, as shown in Fig. 7, the tuning process will become simple and reliable. In this structure, however, electric voltages on the retardation and polarization layers are difficult to control independently. We are currently considering the method of suitable voltage application.
6. Evaluation of the device performance
In the preliminary experiments above, the probe light was linearly polarized with the wire-grid. The wire-grid polarizer was removed in the following experiment to demonstrate the polarization-independent performance. The voltages V1 and V2 were adjusted according to Eq. (4). Figure 8(a) shows the transmission spectra for nonpolarized input light. The applied voltage V1 is shown beside the spectra. Spectral peaks shifted with the voltage increase in the range below 10 V. Thereafter the transmittance decreased gradually in the entire spectral range, since the retardation approached zero as the voltage rose. In comparison with the preliminary experiments, both the input and output light intensities were doubled by superimposition of s- and p-polarized beams. (The transmittance was therefore unchanged.)
Figures 8(b)–8(e) compare two types of Lyot filters; i.e., (b) a conventional discrete type and (c) the current integrated type. The conventional type consisted of three discrete optical elements; i.e., two BaF2 wire-grid polarizers and a retarder that was constructed by sandwiching the LC with two Si plates in the same manner as described in Sec. 3 (LC layer thickness: ~20 μm). Transmittance was evaluated as the ratio of input and output light intensities, I/I0. The gray lines in Figs. 8(d) and 8(e) show transmission spectra of this conventional filter before and during a voltage application process (6 V). By voltage application, a spectral peak at 2.5 μm shifted to a wavelength below 2 μm, and another broadpeak appeared at around 5 μm. The transmittance was lower than 10% because of the following reasons, i.e., extinction of one polarization component (3 dB), the transmission loss of the two polarizers (0.5 × 2 dB), the reflection losses at the two Si-air boundaries (1.5 × 2 dB) and the two Si-LC boundaries (0.7 × 2 dB), etc. The black lines in Figs. 8(d) and 8(e) show transmission spectra of the integrated device. To facilitate the comparison with the discrete system, similar spectra (peak wavelength: 2.5 or 5 μm) were created by applying suitable voltages (V1 and V2) to the retarders. The maximum transmittance of the current integrated device exceeded 40%, which was five times larger than that of the discrete system.
Integration of optical elements is effective to enhance a throughput of infrared devices that use materials with high index of refraction. Integrated devices also facilitate optical alignment with invisible probe light and realize robust measurements. Birefringence and electric controllability of LC yield various optical functions in the infrared region, when these features are combined with infrared transparency and electric conductivity of Si. A Lyot filter was fabricated by using LC and Si pentaprisms, and its spectral tunability as well as high-throughput characteristic was demonstrated in the 2–6 μm wavelength range.
The authors acknowledge the provision of alignment-coating materials by Nissan Chemical Industries, Japan.
References and links
1. I. C. Khoo, Liquid Crystals, 2nd ed. (Wiley, New York, 2007).
2. A. F. Fray, C. Hilsum, and D. Jones, “Some properties of liquid crystals as infrared modulators,” Infrared Phys. 18(1), 35–41 (1978). [CrossRef]
3. J. G. Pasco and W. Elser, “Liquid crystal infrared modulation,” Opt. Eng. 20, 970–975 (1981).
5. M. Saito and T. Yasuda, “An infrared polarization switch consisting of silicon and liquid crystal,” J. Opt. 12(1), 015504 (2010). [CrossRef]
6. S.-T. Wu, “Birefringence dispersion of liquid crystals,” Phys. Rev. 33(2), 1270–1274 (1986). [CrossRef]
7. D. B. Chenault, R. A. Chipman, K. M. Johnson, and D. Doroski, “Infrared linear diattenuation and birefringence spectra of ferroelectric liquid crystals,” Opt. Lett. 17(6), 447–449 (1992). [CrossRef] [PubMed]
9. M. Saito, R. Takeda, K. Yoshimura, R. Okamoto, and I. Yamada, “Self-controlled signal branch by the use of a nonlinear liquid crystal cell,” Appl. Phys. Lett. 91(14), 141110 (2007). [CrossRef]
10. J. Moore, N. Collings, W. A. Crossland, A. B. Davey, M. Evans, A. M. Jeziorska, M. Komarčević, R. J. Parker, T. D. Wilkinson, and H. Xu, “The silicon backplane design for an LCOS polarization-insensitive phase hologram SLM,” IEEE Photon. Technol. Lett. 20(1), 60–62 (2008). [CrossRef]
11. W. A. Crossland, I. G. Manolis, M. M. Redmond, K. L. Tan, T. D. Wilkinson, M. J. Holmes, T. R. Parker, H. H. Chu, J. Croucher, V. A. Handerek, S. T. Warr, B. Robertson, I. G. Bonas, R. Franklin, C. Stace, H. J. White, R. A. Woolley, and G. Henshall, “Holographic optical switching: the “ROSES” demonstrator,” J. Lightwave Technol. 18(12), 1845–1854 (2000). [CrossRef]
17. J. R. Andrews, “Low voltage wavelength tuning of an external cavity diode laser using a nematic liquid crystal-containing birefringent filter,” IEEE Photon. Technol. Lett. 2(5), 334–336 (1990). [CrossRef]
19. J. Pfeifle, L. Alloatti, W. Freude, J. Leuthold, and C. Koos, “Silicon-organic hybrid phase shifter based on a slot waveguide with a liquid-crystal cladding,” Opt. Express 20(14), 15359–15376 (2012). [CrossRef] [PubMed]
20. J. S. Patel and S.-D. Lee, “Electrically tunable and polarization insensitive Fabry-Perot étalon with a liquid-crystal film,” Appl. Phys. Lett. 58(22), 2491–2493 (1991). [CrossRef]
21. N. A. Clark and S. T. Lagerwall, “Submicrosecond bistable electro-optic switching in liquid crystals,” Appl. Phys. Lett. 36(11), 899–901 (1980). [CrossRef]
22. B. Lyot, “Le filtre monochromatique polarisant et ses applications en physique solaire,” Ann. Astrophys. (Paris) 7, 31–79 (1939).
23. J. W. Evans, “The birefringent filter; a correction,” J. Opt. Soc. Am. 39(5), 412 (1949). [PubMed]
24. J. W. Evans, “Solc birefringent filter,” J. Opt. Soc. Am. 48(3), 142–145 (1958). [CrossRef]
25. R. Chang, “Application of polarimetry and interferometry to liquid crystal-film research,” Mater. Res. Bull. 7(4), 267–278 (1972). [CrossRef]
26. Y.-H. Lin, H. Ren, Y.-H. Wu, Y. Zhao, J. Fang, Z. Ge, and S.-T. Wu, “Polarization-independent liquid crystal phase modulator using a thin polymer-separated double-layered structure,” Opt. Express 13(22), 8746–8752 (2005). [CrossRef] [PubMed]
27. J. S. Patel and M. W. Maeda, “Tunable polarization diversity liquid-crystal wavelength filter,” IEEE Photon. Technol. Lett. 3, 73–740 (1991).