In this study, we demonstrate a polarization sensitive pixel for a complementary metal-oxide-semiconductor (CMOS) image sensor based on 65-nm standard CMOS technology. Using such a deep-submicron CMOS technology, it is possible to design fine metal patterns smaller than the wavelengths of visible light by using a metal wire layer. We designed and fabricated a metal wire grid polarizer on a 20 × 20 μm2 pixel for image sensor. An extinction ratio of 19.7 dB was observed at a wavelength 750 nm.
© 2013 Optical Society of America
The polarization properties of light play an important role in optical measurements. An image sensor that can facilitate the observation and measurement of different polarization components at each pixel enables polarization imaging. This technique is expected to be a powerful tool to realize highly functional imaging, which cannot be achieved using optical intensity imaging methods. Image sensors with on-pixel polarizers or polarization-sensitive pixel structures have been proposed and demonstrated by using metal wire grids [1–9], liquid crystal, materials [10–12], polymers  or photonic crystal structures . In particular, metal wire grid polarizers can satisfy both the conditions of a high extinction ratio and a reduced device thickness. Further, such a polarizer is compatible with complementary metal-oxide-semiconductor (CMOS) technology, and it can be fabricated monolithically by using metal layers for wiring. Thus, such polarization image sensors have been reported by some groups [1–4]. However, the reported extinction ratios were less than 10 (= 10 dB); this value is not as high as that obtained with the polarizers fabricated separately.
This is essentially because the CMOS technologies used in these studies did not enable a sufficiently fine grid design, thereby limiting the extinction ratio. This problem can be solved by using recently developed deep-submicron CMOS technologies, which allow the design of metal patterns finer than 100 nm. The angle of the polarizer on each pixel can be designed arbitrarily in the range of the design rule. Moreover, in CMOS image sensors, it is feasible to integrate mixed signal circuits and perform signal processing of the polarization image in a real-time manner. This feature is very suitable for carrying out large-scale parallel optical measurements.
In our previous paper, we first reported the fabrication of CMOS image sensor pixels with an on-chip polarizer developed using 65-nm standard CMOS technology . The extinction ratio of this polarizer was 43.7 (= 16.4 dB) at a wavelength of 633 nm. The wavelength for measurement was fixed and not optimized to this sensor. In this study, we investigated the sensor’s spectral characteristics. At the optimum wavelength, an extinction ratio of approximately 20 dB was achieved.
2. Design of CMOS image sensor pixel on-chip polarizer
In this work, we designed metal wire grid polarizers on image sensor pixels by using 65-nm standard CMOS technology (Fujitsu). Figure 1 shows the schematic diagram of the pixel structure. By the use of a typical metal wire grid polarizer, the polarization component with its electric field perpendicular to the grid (TM polarization) is transmitted and that parallel to the grid (TE polarization) is reflected. This is because the electrons in the metal are suppressed by the line shape of the grid in the electric field oscillation direction of TM polarization. In contrast, the electrons oscillate along the direction of the TE polarization electric field and the transmitted component is cancelled out by the electric field as a result of the oscillation of the electrons.
The characteristics of a metal grid polarizer depend on the pitch of the grid and the type of material used. In order to obtain a high extinction ratio, the grid pitch must be considerably less than the wavelength of the incident light. A metal with high absorption is not suitable for use as the grid material because the optical electromagnetic fields are concentrated on the surface of the metal.
Figures 2 show the simulated characteristics of metal wire grid polarizers. The simulation was performed by using a rigorous continuous wave analysis method (Diffract MOD, RSoft). A metal wire grid was modeled using a periodic boundary condition. It is assumed that the metal wires are composed of Cu surrounded by a dielectric with a refractive index of 1.5. In a practical CMOS chip, there is a barrier metal positioned between the metal wire and the dielectrics. However, the barrier metal is not included in this simulation. And, the intermediate layers in deep-submicron technologies are usually composed of low dielectric materials, whose refractive indices are estimated lower than 1.5. The wavelength is set in the range of 450–900 nm which can be detected by a Si-based photodiode. The simulation results shown in Fig. 2(a) indicate that a fine pitch is required to obtain broad spectral band characteristics. Further, the extinction ratio is improved as the grating line and space widths are decreased or the wavelength of incident light is increased. For a grating pitch less than 200 nm (line/space = 100 nm/100 nm), an extinction ratio greater than 102 (= 20 dB) is expected in infra-red region.
The transmission spectra of TE and TM polarizations and the extinction ratio are plotted as a function of wavelength in Fig. 2(b). The line/space of the grating is assumed to be the finest that can be achieved using 65-nm technology. The transmission of TM polarization is drastically decreased at wavelengths less than about 600 nm. This is because we assumed that the metal wires are composed of Cu, which is a typical wire material used in deep-submicron CMOS technologies. Thus, because of the use of Cu, it is difficult to cover the entire range of visible light wavelengths. In order to obtain high TM transmission at visible wavelengths, a special process is required to fabricate a wire grid polarizer with less absorptive metal, such as aluminum.
The designed pixel layout and the specifications are shown in Fig. 3 and listed in Table 1, respectively. The pixel size is 20 × 20 μm2. The area of the photodiode is 13.2 × 13.2 μm2, which corresponds to a fill factor of 43.6%. The wire grids are composed of the bottom metal layer. The grating is set to the finest pitch allowed by the design rule, whose line/space is smaller than 100 nm/100 nm. If the interlayer dielectric films are assumed to be composed of SiO2 glass, the effective wavelength of visible light is approximately from 300 nm to 500 nm. Thus, it is observed that the grating pitch is less than visible-wavelength range. The pixel consists of a 3-transistor active pixel sensor and the photodiode is composed of n-well/p-sub layers. The supply voltage for the circuit (VDD) is set to 3.3 V using high-voltage transistors in the 65-nm technology. To reduce the crosstalk between the pixels, the photodiodes are enclosed by guard rings composed of n-well/deep n-well layers, and their voltages are pulled up to VDD.
3. Experimental results
The fabricated pixels with and without polarizers were observed by using a microscope with polarized white light. The micrographs are shown in Fig. 4. The color of the pixel with metal wire grid polarizer is different from that of the gridless pixel and depends on polarization direction.
The polarization characteristics of the fabricated image sensor pixels were measured as described below. The experimental setup is shown in Fig. 5. The light source is a halogen lamp and the output wavelength is chosen by using a monochromator (Micro HR; HORIBA). A two-dimensional slit (SLX-1; Sigma Koki) is used as a rectangle aperture. After a plano-convex lens as a collimator, a polarizer (SPF-30C-32; Sigma Koki) is inserted to generate linearly polarized light. The light beam is focused on a single pixel by using an objective lens (M Plan Apo 5×; Mitsutoyo). Because the pixel is based on a 3-transistor active pixel sensor architecture, the output signal is obtained by measuring the discharge slope of a pixel after pixel reset for each incident light intensity. The light from the monochromator is not perfectly unpolarized. The intensity spectrum of each polarization direction was measured and used for correction of sensor output signal.
Figure 6 shows the normalized output from the gridless pixel as a function of wavelength. The peak wavelength is 710 nm. With fine CMOS technologies, a peak sensitivity of an image sensor pixel is typically at a wavelength in the range of 500 to 600 nm, which is less than that of a discrete photodiode owing to its relatively thin depletion layer. The fabricated pixels exhibit a peak at around 700 nm in the near-infrared region. This result is in agreement with the result reported by Ikeda et al.. The result suggests that the interlayer dielectrics work as an interference filter because they are composed of several types of materials with different refractive indices.
Figure 7 shows the normalized outputs from a pixel with a 0- or 90-degree wire grid as a function of the polarization angle of the incident light at 750 nm, for which value the nearly maximum extinction ratio is achieved. For both polarizer angles, the observed results were nearly identical. The extinction ratios for the 0- and 90-degree polarizers were 19.7 dB and 19.3 dB, respectively. These values are considerably greater than those obtained using 0.35-μm [1, 2] and 0.18-μm technologies .
The normalized outputs from the pixel with the polarizer and the extinction ratio as functions of the incident wavelength are shown in Fig. 8. Here, TE and TM indicate the polarization directions of the incident light parallel and perpendicular to the wire grid, respectively. The result indicates that the transmittance for TM polarized light is higher than that for TE polarized light in the measured region; that is the grid works as a polarizer. The normalized output spectrum for TM light shows a correlation with the simulation result presented in Fig. 2(b). However, the output range for TM at wavelengths shorter than 700 nm is considerably less than that shown in Fig. 2(b). Therefore, the proposed metal wire polarizer exhibits extinction ratio as high as approximately 20 dB from 700 nm to 800 nm. For TE polarization, the output intensity increases as the wavelength becomes greater than approximately 800 nm. This is not observed in the simulation result of the simple grating model. In order to examine the consistency of results between the experiment and simulation in this case, further studies are needed. It is a matter of concern that some of the incident light is diffracted at the edge of the grating. If this diffraction can be suppressed, a higher extinction ratio would be achieved.
Polarization-sensitive CMOS image sensors have previously been reported by certain groups. Tokuda et al. have demonstrated the functioning of an image sensor based on 0.35-μm technology [1, 2]. The finest grating pitch that can be obtained with the 0.35-μm process is approximately 1 μm, which is greater than the wavelengths of visible light in the dielectric layers. Thus, the transmittance of TE polarization is higher than that of TM. This characteristic is in contrast to that of the conventional metal grid polarizer. Sarkar et al. have reported an image sensor based on 0.18-μm technology . The grating pitch in this case was 0.48 μm. This value is probably comparable to visible-light wavelengths because the metal wires of the grating are surrounded by an isolation dielectric. In our present device, the extinction ratio is as large as 19.7 dB at 750 nm. Although the wavelength used in the measurement is different from those used in the abovementioned studies, this value is more than 10 times greater than those for the sensors based on 0.35-μm and 0.18-μm technologies, and it is comparable to that used in the sensors with fine polarizers (0.14-μm pitch) fabricated via a post process performed on a charge coupled device (CCD) image sensor . Thus, our result suggests that on-chip nano-photonic devices can be designed by using recently developed deep-submicron CMOS technologies. One of the methods to improve the extinction ratio is to use finer CMOS technology. Another approach is to design a multiple-metal layer polarizer [8, 15], although there are certain design limitations.
From the result observed in Fig. 6, our sensor exhibits peak sensitivity at a wavelength of approximately 700 nm. This result indicates that the isolation dielectric layers act as a band rejection filter at visible wavelengths. The normalized output spectrum of the pixel with the polarizer indicates that the device sensitivity at wavelengths up to 700 nm is reduced further by the presence of the metal wire grid layer. This is probably because the metal wire grid is made of Cu, which has a relatively high absorption coefficient at the above-mentioned wavelengths. Such a low sensitivity leads to a low effective extinction ratio. This is because the offset of dark current in the photodiode cannot be neglected when the signal level of photoelectric conversion is low. In order to realize an image sensor with a polarizer for visible light, the optical filter characteristics of the interlayer dielectric layers should be designed taking care of the layer thicknesses, the number of layers, and their refractive indices. In addition, the polarizer is required to be made of metals with low absorption at visible wavelengths such as Al or Ag. On the other hand, if the target wavelength is limited to the infrared region, the optical characteristics obtained using the standard 65-nm CMOS technology are acceptable.
The proposed metal wire grid polarizer is fabricated with a CMOS process, so that it is possible to increase the number of pixels to the order of millions by using the current state-of-the-art CMOS technology. The large extinction ratio characteristics in the near-infrared region can potentially be used in polarization imaging with a laser source, such as optical coherence tomography [16, 17] and terahertz/millimeter-wave/microwave electric field imaging [18–21]. In these types of measurement, the detection of weak polarization variation is required. Several such systems employing image sensors have been reported to date [16–21]. However, in these cases, polarization-sensitive imaging is realized using the combination of a high-speed image sensor and a bulk polarizer. The achieved sensitivities in these cases are not as large those obtained with the single point measurement systems although a large imaging throughput is realized. This reduction in sensitivity is due to the fact that it is difficult to apply signal processing techniques for noise reduction by using a balanced photodetector and a two-phase lock-in amplifier, which are used in state-of-the-art single point measurement systems. Our sensor pixel provides a large extinction ratio that is suitable for such measurement systems. In addition, a wavelength range of 700 nm to 800 nm with a large extinction ratio is sufficient for this purpose. By integrating analog CMOS circuits with the pixels, an on-chip balanced detection pixel for highly sensitive measurement can be realized using the combination of a pair of polarization-sensitive pixels for TE and TM polarizations and on-chip lock-in detection.
We demonstrated the functioning of a CMOS image sensor pixel with a metal wire grid polarizer using a metal layer based on 65-nm standard CMOS technology. An extinction ratio of 19.7 dB was obtained at a wavelength of 750 nm. This polarization imaging pixel array can potentially be used in highly functional polarization imaging because of its large extinction ratio and the feasibility of analog signal circuit integration.
This work was supported in part by Japan Science and Technology Agency, Core Research for Evolutional Science and Technology (JST-CREST), a Grant-in-Aid for Scientific Research (B) # 24310101, and a Grant-in-Aid for Scientific Research on Innovative Areas # 24106729. This work was also supported by the VLSI Design and Education Center (VDEC), The University of Tokyo with the collaboration of Cadence Corporation and Mentor Graphics Corporation, and partially performed by the author for STARC as part of the Japanese Ministry of Economy, Trade and Industry sponsored “Silicon Implementation Support Program for Next Generation Semiconductor Circuit Architectures.”
References and links
1. T. Tokuda, S. Sato, H. Yamada, K. Sasagawa, and J. Ohta, “Polarization-analyzing CMOS photosensor with monolithically embedded wire grid polarizer,” Electron. Lett. 45(4), 228–230 (2009) [CrossRef] .
2. T. Tokuda, H. Yamada, K. Sasagawa, and J. Ohta, “Polarization-analyzing CMOS image sensor with monolithically embedded polarizer for microchemistry systems,” IEEE Trans. Biomed. Circuits Syst. 3(5), 259–266 (2009) [CrossRef] .
3. M. Ikeda and Y. Kim, “Measurement and analysis on characteristics of transmission and polarization for 12ML 65nm CMOS,” in Proceedings of the IEEE Sensors (Institute of Electrical and Electronics Engineers, New York, 2010), pp. 548–551.
4. M. Sarkar, S. Member, D. San, S. Bello, C. V. Hoof, and A. Theuwissen, “Integrated polarization analyzing CMOS image sensor for material classification,” IEEE Sensors J. 11(8), 1692–1703 (2011) [CrossRef] .
5. S. Shishido, T. Noda, K. Sasagawa, T. Tokuda, and J. Ohta, “Polarization analyzing image sensor with on-chip metal wire grid polarizer in 65-nm standard complementary metal oxide semiconductor process,” Jpn. J. Appl. Phys. 50(4), 04DL01 (2011) [CrossRef] .
6. P. B. Catrysse and B. A. Wandell, “Integrated color pixels in 0.18-μm complementary metal oxide semiconductor technology,” J. Opt. Soc. Am. A 20(12), 2293–2306 (2003) [CrossRef] .
9. M. Guillaumée, L. A. Dunbar, C. Santschi, E. Grenet, R. Eckert, O. J. F. Martin, and R. P. Stanley, “Polarization sensitive silicon photodiodes using nanostructured metallic grids,” Appl. Phys. Lett. 94(19), 193503 (2009) [CrossRef] .
10. F. Boussaid, A Bermak, and V. G. Chigrinov, “Thin photo-patterned micropolarizer array for CMOS image sensors,” IEEE Photon. Tech. Lett. 21(12), 805–807 (2009) [CrossRef] .
11. X. Zhao, A. Bermak, F. Boussaid, and V. G. Chigrinov, “Liquid-crystal micropolarimeter array for full Stokes polarization imaging in visible spectrum,” Opt. Express 18(17), 17776–17787 (2010) [CrossRef] [PubMed] .
12. X. Zhao, F. Boussaid, A. Bermak, and V. G. Chigrinov, “High-resolution thin “guest-host” Micropolarizer arrays for visible imaging polarimetry,” Opt. Express 19(6), 5565–5573 (2011) [CrossRef] [PubMed] .
14. T. Sato, T. Araki, Y. Sasaki, T. Tsuru, T. Tadokoro, and S. Kawakami, “Compact ellipsometer employing a static polarimeter module with arrayed polarizer and wave-plate elements,” Appl. Opt. 46(22), 4963–4967 (2007) [CrossRef] [PubMed] .
16. Y. Watanabe, Y. Hayasaka, M. Sato, and N. Tanno, “Full-field optical coherence tomography by achromatic phase shifting with a rotating polarizer,” Appl. Opt. 44(8), 1387–1392 (2005) [CrossRef] [PubMed] .
17. M. Akiba, K. P. Chan, and N. Tanno, “Full-field optical coherence tomography by two-dimensional heterodyne detection with a pair of CCD cameras,” Opt. Lett. 28(10), 816–818 (2003) [CrossRef] [PubMed] .
18. Z. Jiang and X.-C. Zhang, “Terahertz imaging via electrooptic effect,” IEEE Trans. Microwave Theory Tech. 47(12), 2644–2650 (1999) [CrossRef] .
19. M. Usami, M. Yamashita, K. Fukushima, C. Otani, and K. Kawase, “Terahertz wideband spectroscopic imaging based on two-dimensional electro-optic sampling technique,” Appl. Phys. Lett. 86(14), 141109 (2005) [CrossRef] .
20. K. Sasagawa, A. Kanno, T. Kawanishi, and M. Tsuchiya, “Live electrooptic imaging system based on ultra-parallel photonic heterodyne for microwave near-fields,” IEEE Trans. Microwave Theory Tech. 55(12), 2782–2791 (2007) [CrossRef] .
21. K. Sasagawa, A. Kanno, and M. Tsuchiya, “Instantaneous visualization of K-Band electric near-fields by a live electrooptic imaging system based on double sideband suppressed carrier modulation,” J. Lightwave Technol. 26(15), 2782–2788 (2008) [CrossRef] .