Abstract
Pixelated filter arrays of Fabry-Perot (FP) cavities are widely integrated with photodetectors to achieve a WYSIWYG (“what you see is what you get”) on-chip spectral measurements. However, FP-filter-based spectral sensors typically have a trade-off between their spectral resolution and working bandwidth due to design limitations of conventional metal or dielectric multilayer microcavities. Here, we propose a new idea of integrated color filter arrays (CFAs) consisting of multilayer metal-dielectric-mirror FP microcavities that, enable a hyperspectral resolution over an extended visible bandwidth (∼300 nm). By introducing another two dielectric layers on the metallic film, the broadband reflectance of the FP-cavity mirror was greatly enhanced, accompanied by as-flat-as-possible reflection-phase dispersion. This resulted in balanced spectral resolution (∼10 nm) and spectral bandwidth from 450 nm to 750 nm. In the experiment, we used a one-step rapid manufacturing process by using grayscale e-beam lithography. A 16-channel (4 × 4) CFA was fabricated and demonstrated on-chip spectral imaging with a CMOS sensor and an impressive identification capability. Our results provide an attractive method for developing high-performance spectral sensors and have potential commercial applications by extending the utility of low-cost manufacturing process.
© 2023 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
From classical RGB Bayer filters to multi/hyperspectral integrated color filter arrays (CFAs), on-chip spectral sensors have been a key characterization technology for spectral imaging in remote sensing [1–7]. They allow the miniaturization of bulky spectrometers into compact and portable equipment. Spatially and spectrally-variable narrowband filters are integrated into the image sensor and aligned with underlying pixels [8–14]. Various integrated filter schemes and fabrication approaches based on thin-film Fabry-Perot (FP) cavities [15–24], photonic crystals [25], waveguide resonators [26], and metasurfaces [27–34] have been exploited, which are differentiated by their principles, configurations, and materials. Among these micro/nano filters, FP microcavities have the most mature mechanism and the simplest tuning method, i.e., by varying their cavity length. Although they have been used commercial products, FP-cavity spectral sensors have suffered from a trade-off between a high spectral resolution and a wide working bandwidth, which limits designs using currenet configurations. All-dielectric FP cavities with alternating high- and low-index multilayer films can obtain an extremely narrow passband corresponding to a high spectral resolution due to their high reflectance by a Bragg reflection mirror. However, they fail to work over a wide bandwidth range due to their limited stop-bandwidth, especially in the visible region, as well as their fast dispersion of reflection phase in the Bragg-mirror stopband. By contrast, metal-based FP cavities have a very wide stopband, along with weaker reflection-phase dispersion, but they are also very lossy at optical frequencies, which decreases the spectral resolution and transmittance of the passband.
Recently, reconstruction algorithms [35–38] have been introduced in computational spectral imaging by using back-end computation to overcome the deficiencies in device performance. In this way, compressive sensing or machine-learning techniques provide a promising method to improve the spectral resolution, as the FP-cavities arrays usually have a low spectral correlation. However, additional computing time and power consumption are necessary during reconstruction, resulting in non-real-time spectral imaging. Reconstruction may also fail even with low-level noise. Extra efforts to handle noise will increase the computing time and may induce undesired artifacts if they are not properly employed.
Complex, high-cost fabrication methods are another limitation on the number of spectral channels. Conventionally, FP-based CFAs with different cavity lengths require separate lithography processes assisted with combinatorial deposition/etching method. Typically, running N times of processes can achieve 2N integrated filter arrays [39,40], which is cumbersome, expensive, and inflexible. A potential solution has been proposed to utilize one-step grayscale lithography to flexibly control and customize the cavity length, which is a cost-effective, versatile, and wafer-scale framework. In the visible range, the unit size of a CMOS pixel is typically on the order of (or less than) one micrometer, which requires high-precision electron-beam lithography (EBL) fabrication. Therefore, the issue of on-chip spectral imaging with excellent spectral resolution and multi-channel in a wide spectral bandwidth remains a great challenge from the perspective of CFAs.
In this work, we propose a novel configuration of FP-based CFAs consisting of multilayer metal-dielectric-mirror FP microcavities, which enables snapshot spectral imaging with a favorable spectral resolution of ∼10 nm over an extended visible band from 450 nm to 750 nm. An appropriate FP-microcavities array, composed of the multilayer mirrors (silver Ag/silicon dioxide SiO2/titanium dioxide TiO2) and a PMMA resist spacer layer, was designed. The exact operation that we can realize depends on – By adding another two dielectric layers on the metallic film, the broadband reflectance of the metal mirror was greatly enhanced along with an expected as-flat-as-possible reflection-phase dispersion. Moreover, we performed a fast and CMOS-compatible manufacturing process by using grayscale EBL lithography for multi-channel integration. The experimental results showed that the prepared 16-channel CFAs approached hyperspectral resolution (∼15 nm) over the whole bandwidth and easily supported snapshot spectral imaging, with high optical performance and identification capabilities.
2. Theoretical design and assessment
Snapshot spectral imagers acquire a complete dataset in one detector integration period over a sensor. Integrated CFAs are integrated into the photodetector and then a CMOS imaging sensor is converted into a compact spectral sensor, as shown in Fig. 1(a). Each filter should be aligned with one pixel so that wavelength-specific information is resolved. FP optical cavities with a spacer layer sandwiched between two mirrors are classic compact color filters. Incident light passes through two reflectance mirrors and then emits after interference. If the reflectance of the mirrors is sufficiently high, high spectral resolution can be obtained. The central wavelength of the transmission band in FP cavity can be described by the following equation:
where n and d are the refractive index and physical thickness of the spacer layer, respectively, m is the order of transmission, and $\varphi $ is the reflection phase (RP) of two mirrors which can be approximatively considered as equality. It can be seen that, for the specific cavity length $nd$, we can have the zero-order transmission ($m$=0) at ${\lambda _1}$ that is defined by ${\lambda _1}({{\varphi_{{\lambda_1}}}/\pi } )= 2nd$ and the first-order transmission ($m$=1) at ${\lambda _2}$ defined by ${\lambda _2}({1 + {\varphi_{{\lambda_2}}}/\pi } )= 2nd$. We can see that the final attainable spectral coverage area $\mathrm{\Delta }\lambda = {\lambda _1} - {\lambda _2}$ is deterministically dependent on the mirror phase dispersion $\mathrm{\Delta }\varphi = {\varphi _{{\lambda _1}}} - {\varphi _{{\lambda _2}}}$, that is, $\Delta \lambda $ and $\Delta \varphi $ are actually inversely proportional. Therefore, the designed mirror films need to reduce $\Delta \varphi $ as much as possible, thus increasing $\mathrm{\Delta }\lambda $ to cover a wider bandwidth. The spectral resolution, i.e., full width at half maxima (FWHM) can be described as follows:We can see that, when the interference level is determined, the FWHM of the FP filter depends on the reflectivity R of the mirrors. The higher the reflectivity, the smaller the FWHM, and the higher the monochromaticity and spectral resolution of the transmitted light.
The transfer matrix method was utilized to simulate and verify the calculated ${\lambda _0}$ and ${\varphi _\lambda }$ values. We compared the performance of all-dielectric films with SiO2 and TiO2 multilayers and metallic films with Ag reflectance mirrors over the extended range of 450 nm - 750 nm. The simulation curves are shown in Fig. 2(a-b). Although all-dielectric films have an extremely high spectral resolution (∼5 nm), they have a smoother gradient trend in the RP curve, meaning a wider reflectance broadband. This conclusion can also be verified by the reflectance broadband of the two type of mirrors. However, the slightly lower reflectance than all-dielectric films’ because of losses led to a wide FWHM. Therefore, due to the respective limitations of metallic and all-dielectric films, we proposed metal-dielectric compound material multilayer films shown in Fig. 2(c). The novel mirror improved the reflectivity, but it also needs to ensure phase dispersion. The new reflectance R of compound material films (CMFs) can be described by the equation:
3. Fabrication process and results of compound material CFAs
Considering the accuracy of the pixel-level filter size, height control during grayscale exposure, etc., the compound material CFAs were fabricated using one-step 3D profile grayscale EBL technology assisted with electron-beam evaporation (EBE) on fused-silica substrates. After cleaning the substrates in acetone followed by iso-propyl alcohol (IPA), the Ag, SiO2, and TiO2 layers were deposited by EBE. Then, a 250-nm-thick PMMA e-beam photoresist (MW: 950 K) was spin-coated and baked on a hotplate, followed by a conductive resist atop the PMMA photoresist. An array of 6.9 µm (x-y dimensions) square pixels was assigned increasing doses so that each channel had a different final thickness with a constant development time. As shown in Fig. 3(a), each dose-modulated of spectral channel was determined based on the contrast curve, and the resultant transmission mode for each channel spectrally red-shifted from 450 nm to 750 nm. The power of the electron beam is 100 kV, the aperture size is 300 µm and the optimized exposure doses is 50-148 µC/cm2. After e-beam exposure, PMMA photoresist was developed in the mixed solution of deionized water and IPA and we obtained spacer layers with different thickness. The thicknesses of PMMA for color filter ranges from 30 nm to 220 nm. Finally, symmetrical high-reflectance mirrors were deposited on the PMMA dielectric spacer layers with different thicknesses.
Figure 3(b) shows the halogen-lamp microscope images of the 16 channels with different thickness spacer layers, with a photograph of the compound material CFAs in the upper-left. The 16 channels (6.9 µm each pixel) were arranged as a 4 × 4 matrix, and the scale bar was 100 µm. Color changes between different spectral channels were observed. Then, the optical performance of 16 channels filter elements ranging from 450 nm to 750 nm was measured by a universal measurement spectrophotometer (UMS, Agilent, CARY-5000). Figure 3(c) shows the normalized transmission spectra and narrow bandwidths (∼10–15 nm), of which the absolute optical efficiency was almost 48%.
4. Imaging experiment of the 16-channel spectral system
Due to the difficulty of buying custom CMOS, commercially-available CMOS have protective layers on the surface that cannot be applied for EBL overlay exposure, making on-chip integration experiments difficult. Therefore, we characterized the compound material CFAs by the alternating optical system depicted in Fig. 4(a-b). First, we used incandescent lamps as the broad illumination source. Instead of placing an image sensor in the CFAs’ image plane, a camera lens (Sony with 1 mm focal length and 1.8 F value) and relay lens (Edmund) were utilized to relay images to the CMOS sensor, i.e., the filter arrays were set on the secondary image plane. All devices were mounted on XY or XYZ axis translation stage, and the accuracy of CFAs on the precision rotation mount was up to 5 arc minutes, which met the alignment requirements of CFAs and image elements. The whole test optical path was equivalent to that of the on-chip integration spectral system.
A leaf of a tree was chosen as the captured input object, as shown in Fig. 4(c). The green area represents a healthy part with chlorophyll, while the yellow area represents low chlorophyll parts. Figure 4(d) shows the spectral response curves of the two different parts. For the green parts with sufficient chlorophyll, there were distinct characteristic peaks near 475 nm and 650 nm, which are consistent with the spectral peaks of chlorophyll. For the yellow part, the characteristic peak near 475 nm disappeared, while the characteristic peak near 650 nm was blue-shifted, indicating that leaf aging and other factors were present in this region [41,42]. Finally, Fig. 4(e) shows a 16-channel image array and the center wavelength of each channel increased from top to bottom and left to right over the range of 450–750 nm. Channels in this image were brightness-normalized to correct for uneven illumination and spectral variance in the image sensor’s sensitivity. A false-color image was used to simultaneously visualize the data in the three different channels, which showed subtle differences.
5. Discussion
The measured spectral resolution and efficiency in Fig. 3(c) were slightly lower than the theoretical values because the fabricated TiO2 film had some absorption, which was limited by our deposition process conditions. However, this type of film defect could be improved by using more advanced equipment and more mature CMOS-capable fabrication processes. The non-uniform distribution of the 16 channels in different rows, may be a result of the exposure inhomogeneity and thickness control precision. However, the results show that the 16 spectral channels were in the extended visible bandwidth, and the channels of all integrated filters were ultra-narrow. In fact, the resolution we achieved was close to the high spectral resolution Δλ/λ ∼0.01. If we continue to increase the thickness of the metal layer, we may obtain a higher spectral resolution, but the transmittance will drop further, which will require increasing the integration time to ensure a high signal-to-noise ratio. But even so, our results already show a great improvement in the combined performance of resolution and bandwidth compared with the original metallic and all-dielectric FPs. Although we only showed multispectral results of 16 channels, further hyperspectral performance with more channels could be achieved. the high-precision EBL 3D grayscale process has great potential to provide more spectral channels. In industrial production, grayscale masks with multi channels filters for stepper lithography can also be fabricated by EBL, thus improving the efficiency and stability of filter production.
6. Conclusion
Here, we have proposed a design concept for obtaining FP color filters with a high spectral resolution for wide broadband applications via a simplified low-cost fabrication process. We determined the key parameters limiting the broadband and spectral resolution. We designed FP color filters with metal-dielectric mirrors with great optical performance over a wide broadband range. By introducing another two dielectric layers on the metallic film, the broadband reflectance of the FP-cavity mirror was greatly enhanced, along with an expected as-flat-as-possible reflection-phase dispersion, resulting in balanced spectral resolution (∼10 nm) and spectral bandwidth with a range of 450–750 nm. The 16-channel (4 × 4) CFAs fabricated by one-step grayscale EBL successfully measured the spectral wavelengths and took photos of green leaves with a narrow band (∼15 nm) over an extended visible band (450–750 nm). Although the design concept of metal-dielectric FP filters presented in this work is for the extended visible band, it can be further used for other wide broadband regions. One-step grayscale lithography is a low-cost fabrication process and can be adapted to high-volume ultraviolet mask-based photolithography. Therefore, our device provides a comprehensive concept for snapshot spectral imaging with great optical performance over a wide broadband range. It also provides a low-cost, compact, and portable solution for many applications, from diagnostic medical imaging to remote sensing.
Funding
Fundamental Research Funds for the Central Universities (2021SHZDZX0100); Shanghai Pujiang Program (20PJ1414200); China Postdoctoral Science Foundation (2020TQ0227, 2021M702471); Shanghai Municipal Education Commission (17SG22); Science and Technology Commission of Shanghai Municipality (17JC1400800, 20JC1414600, 21JC1406100); National Natural Science Foundation of China (61621001, 61925504, 6201101335, 62020106009, 62105243, 62111530053, 62192770, 62192772, 62205248).
Disclosures
The authors declare that there are no conflicts of interest related to this article.
Data availability
Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.
Supplemental document
See Supplement 1 for supporting content.
References
1. T. Kuroda, Essential Principles of Image Sensors (CRC Press, 2017).
2. J. W. Choi, E. Sahin, B.-U. Sohn, G. F. Chen, D. K. Ng, A. M. Agarwal, L. C. Kimerling, and D. T. Tan, “High spectro-temporal compression on a nonlinear CMOS-chip,” Light: Sci. Appl. 10(1), 130 (2021). [CrossRef]
3. G. Lu and B. Fei, “Medical hyperspectral imaging: a review,” J. Biomed. Opt 19(1), 010901 (2014). [CrossRef]
4. P.-J. Lapray, X. Wang, J.-B. Thomas, and P. Gouton, “Multispectral filter arrays: Recent advances and practical implementation,” Sensors 14(11), 21626–21659 (2014). [CrossRef]
5. N. A. Hagen and M. W. Kudenov, “Review of snapshot spectral imaging technologies,” Opt. Eng 52(9), 090901 (2013). [CrossRef]
6. J. A. Winkelmann, A. Eid, G. Spicer, L. M. Almassalha, T.-Q. Nguyen, and V. Backman, “Spectral contrast optical coherence tomography angiography enables single-scan vessel imaging,” Light: Sci. Appl. 8(1), 7–9 (2019). [CrossRef]
7. B. E. Bayer, “Color imaging array,” (Google Patents, 1976).
8. B. Geelen, N. Tack, and A. Lambrechts, “A compact snapshot multispectral imager with a monolithically integrated per-pixel filter mosaic,” in Advanced fabrication technologies for micro/nano optics and photonics VII(SPIE, 2014), pp. 80–87.
9. D. M. Kita, B. Miranda, D. Favela, D. Bono, J. Michon, H. Lin, T. Gu, and J. Hu, “High-performance and scalable on-chip digital Fourier transform spectroscopy,” Nat. Commun. 9(1), 4405 (2018). [CrossRef]
10. X. Gan, N. Pervez, I. Kymissis, F. Hatami, and D. Englund, “A high-resolution spectrometer based on a compact planar two dimensional photonic crystal cavity array,” Appl. Phys. Lett. 100(23), 231104 (2012). [CrossRef]
11. M. Faraji-Dana, E. Arbabi, H. Kwon, S. M. Kamali, A. Arbabi, J. G. Bartholomew, and A. Faraon, “Hyperspectral imager with folded metasurface optics,” ACS Photonics 6(8), 2161–2167 (2019). [CrossRef]
12. B. Boldrini, W. Kessler, K. Rebner, and R. W. Kessler, “Hyperspectral imaging: a review of best practice, performance and pitfalls for in-line and on-line applications,” J. Near Infrared Spectrosc. 20(5), 483–508 (2012). [CrossRef]
13. A. A. Gowen, C. P. O’Donnell, P. J. Cullen, G. Downey, and J. M. Frias, “Hyperspectral imaging–an emerging process analytical tool for food quality and safety control,” Trends Food Sci. Technol. 18(12), 590–598 (2007). [CrossRef]
14. T. Pügner, J. Knobbe, and H. Grüger, “Near-infrared grating spectrometer for mobile phone applications,” Appl. Spectrosc. 70(5), 734–745 (2016). [CrossRef]
15. S.-W. Wang, D. Liu, B. Lin, X. Chen, and W. Lu, “16× 1 integrated filter array in the MIR region prepared by using a combinatorial etching technique,” Appl. Phys. B 82(4), 637–641 (2006). [CrossRef]
16. J. Castracane, P. L. Clow, S. Wegener, and G. Seidler, “64 channel fiber optic spectrometer for a dual wavelength interferometric pressure sensor array,” Rev. Sci. Instrum. 66(6), 3668–3671 (1995). [CrossRef]
17. S. W. Wang, M. Li, S. Xia, H. Q. Wang, X. S. Chen, and W. Lu, “128 channels of integrated filter array rapidly fabricated by using the combinatorial deposition technique,” Appl. Phys. B 88(2), 281–284 (2007). [CrossRef]
18. S. Saxe, L. Sun, V. Smith, D. Meysing, C. Hsiung, A. Houck, M. Von Gunten, C. Hruska, D. Martin, and R. Bradley, “Advances in miniaturized spectral sensors,” in Next-Generation Spectroscopic Technologies XI (International Society for Optics and Photonics2018), p. 106570B.
19. P. Gonzalez, B. Geelen, C. Blanch, K. Tack, and A. Lambrechts, “A CMOS-compatible, monolithically integrated snapshot-mosaic multispectral imager,” NIR news 26(4), 6–11 (2015). [CrossRef]
20. N. Tack, A. Lambrechts, P. Soussan, and L. Haspeslagh, “A compact, high-speed, and low-cost hyperspectral imager,” in Silicon Photonics VII(International Society for Optics and Photonics2012), p. 82660Q.
21. S. Tisserand, “Custom Bayer Filter Multispectral Imaging: Emerging Integrated Technology,” in 2019 10th Workshop on Hyperspectral Imaging and Signal Processing: Evolution in Remote Sensing (WHISPERS) (IEEE2019), pp. 1–4.
22. M. Commandré, B. Vial, S. Tisserand, L. Roux, H. Dallaporta, F. Bedu, G. Demésy, A. Nicolet, and F. Zolla, “Design, fabrication and characterization of resonant metamaterial filters for infrared multispectral imaging,” Thin Solid Films 592, 296–304 (2015). [CrossRef]
23. P. L. Dillon, A. Brault, J. Horak, E. Garcia, T. Martin, and W. Light, “Fabrication and performance of color filter arrays for solid-state imagers,” IEEE Trans. Electron Devices 25(2), 97–101 (1978). [CrossRef]
24. P. Gonzalez, K. Tack, B. Geelen, B. Masschelein, W. Charle, B. Vereecke, and A. Lambrechts, “A novel CMOS-compatible, monolithically integrated line-scan hyperspectral imager covering the VIS-NIR range,” in Next-Generation Spectroscopic Technologies IX (SPIE2016), pp. 129–137.
25. Y. Zhu, X. Lei, K. X. Wang, and Z. Yu, “Compact CMOS spectral sensor for the visible spectrum,” Photonics Res. 7(9), 961–966 (2019). [CrossRef]
26. I. Zegaar, A. Hocini, and D. Khedrouche, “Plasmonic stop-band filter based on an MIM waveguide coupled with cavity resonators,” in Journal of Physics: Conference Series (IOP Publishing2022), p. 012025.
27. S. Jahani and Z. Jacob, “All-dielectric metamaterials,” Nat. Nanotechnol. 11(1), 23–36 (2016). [CrossRef]
28. Q. Chen, X. Hu, L. Wen, Y. Yu, and D. R. Cumming, “Nanophotonic image sensors,” Small 12(36), 4922–4935 (2016). [CrossRef]
29. A. Kristensen, J. K. Yang, S. I. Bozhevolnyi, S. Link, P. Nordlander, N. J. Halas, and N. A. Mortensen, “Plasmonic colour generation,” Nat. Rev. Mater. 2(1), 16088–14 (2016). [CrossRef]
30. T. Xu, Y.-K. Wu, X. Luo, and L. J. Guo, “Plasmonic nanoresonators for high-resolution colour filtering and spectral imaging,” Nat. Commun. 1(1), 59 (2010). [CrossRef]
31. P. Genevet, F. Capasso, F. Aieta, M. Khorasaninejad, and R. Devlin, “Recent advances in planar optics: from plasmonic to dielectric metasurfaces,” Optica 4(1), 139–152 (2017). [CrossRef]
32. A. McClung, S. Samudrala, M. Torfeh, M. Mansouree, and A. Arbabi, “Snapshot spectral imaging with parallel metasystems,” Sci. Adv. 6(38), eabc7646 (2020). [CrossRef]
33. Y. Intaravanne, M. A. Ansari, H. Ahmed, N. Bileckaja, H. Yin, and X. Chen, “Metasurface-Enabled 3-in-1 Microscopy,” ACS Photonics 10(2), 551 (2023). [CrossRef]
34. Y. Intaravanne, R. Wang, H. Ahmed, Y. Ming, Y. Zheng, Z.-K. Zhou, Z. Li, S. Chen, S. Zhang, and X. Chen, “Color-selective three-dimensional polarization structures,” Light: Sci. Appl. 11(1), 302 (2022). [CrossRef]
35. D. S. Jeon, S.-H. Baek, S. Yi, Q. Fu, X. Dun, W. Heidrich, and M. H. Kim, “Compact snapshot hyperspectral imaging with diffracted rotation,” (2019).
36. W. Zhang, H. Song, X. He, L. Huang, X. Zhang, J. Zheng, W. Shen, X. Hao, and X. Liu, “Deeply learned broadband encoding stochastic hyperspectral imaging,” Light: Sci. Appl. 10(1), 108 (2021). [CrossRef]
37. C. A. Garillos-Manliguez and J. Y. Chiang, “Multimodal Deep Learning and Visible-Light and Hyperspectral Imaging for Fruit Maturity Estimation,” Sensors 21(4), 1288 (2021). [CrossRef]
38. L. Huang, R. Luo, X. Liu, and X. Hao, “Spectral imaging with deep learning,” Light: Sci. Appl. 11(1), 61 (2022). [CrossRef]
39. C. Williams, G. S. Gordon, T. D. Wilkinson, and S. E. Bohndiek, “Grayscale-to-color: scalable fabrication of custom multispectral filter arrays,” ACS photonics 6(12), 3132–3141 (2019). [CrossRef]
40. O. Sandfuchs, M. Kraus, and R. Brunner, “Structured metal double-blazed dispersion grating for broadband spectral efficiency achromatization,” J. Opt. Soc. Am. A 37(8), 1369–1380 (2020). [CrossRef]
41. I. Filella and J. Penuelas, “The red edge position and shape as indicators of plant chlorophyll content, biomass and hydric status,” International journal of remote sensing 15(7), 1459–1470 (1994). [CrossRef]
42. F. Boochs, G. Kupfer, K. Dockter, and W. Kühbauch, “Shape of the red edge as vitality indicator for plants,” Remote Sens. 11(10), 1741–1753 (1990). [CrossRef]