Design and Fabrication of an Artificial Compound Eye for Multi-Spectral Imaging

The artificial compound eye (ACE) structure is a new type of miniaturized, lightweight and intelligent imaging system. This paper has proposed to design a multi-spectral ACE structure to enable the structure to achieve multi-spectral information on the basis of imaging. The sub-eyes in the compound eye structure have been designed as diffractive beam splitting lenses with the same focal length of 20 mm, but with the different designed center wavelengths of 650 nm, 532 nm, and 445 nm, respectively. The proximity exposure lithography and reactive ion etching process were used to prepare the designed multi-spectral ACE structure, and the spectral splitting and multi-spectral imaging experiments were carried out to verify the multi-spectral imaging function of the structure without axial movement. Furthermore, the structure can be designed according to actual requirements, which can be applied to covert reconnaissance, camouflage identification, gas leakage or other fields.


Introduction
Compound eyes of arthropods such as ants, flies and bugs have attracted extensive research interests due to their unique features such as wide fields of view (FOV), high sensitivities to motions and infinite depths of field [1][2][3][4][5][6][7][8]. These compound eyes are integrated, lightweight and smart optical imaging systems composed of multiple individual lenses arranged on a hemispherical surface. Optical devices inspired by natural compound eyes exhibit great potentials in various applications such as surveillance cameras on micro aerial vehicles, high-speed motion detections, endoscopic medical tools, and image guided surgeries [9][10][11].
For years, several attempts were made to develop artificial compound eyes (ACEs) to realize imaging [12][13][14][15], which typically analyzes the spatial characteristics of the target by studying the shape information of the target. In this paper, the spectroscopy technology has been proposed to apply into the ACE structure combined with imaging. Based on the original two-dimensional spatial imaging, one-dimensional spectral information has been added in the structure. Compared with the traditional imaging technology, it can improve the accuracy and sensitivity of target detection, and expand the detection capability of traditional imaging systems for target detection, camouflage identification, complex background suppression, life state observation, environmental monitoring and other fields.
Multi-spectral imaging technology needs to divide the incident full-band or wide-band spectrum into several narrow-band spectrum so as to obtain images of different spectral bands by the corresponding detectors. Thereinto, spectral splitting technology is very important. Spectral splitting technology mainly includes interference spectroscopy [16,17] and traditional filter spectroscopy [18][19][20].

Principle and Structural Design
The ACE is composed of several sub-eye imaging channels, as shown in Figure 1. Different subeye imaging channels are designed as diffractive beam splitting lenses (D ) with different designed parameters. The lenses have the same focal length (F) and different central wavelength ( ). In the imaging process, spectral information of different wavelengths can be captured by different sub-eyes at the same image plane. The arrangement of the sub-eye lenses in the multi-spectral ACE is shown in Figure 2. Along the Y direction, the sub-eye lenses are the same. In the X direction, the sub-eye lenses are designed for different wavelengths. The designed wavelengths ( , i = 1,2,3) are 650 nm (R), 532 nm (G) and 445 nm (B), respectively. The aperture (Di, i = 1,2,3) are the same as 2 mm, and the focal length (F) is also the same as 20 mm. Three sub-eye lenses with different central wavelengths are arranged periodically in the X direction with a period of 6 mm to form a complete ACE structure. Different sub-eyes are composed of different annular structures. The maximum number of annular bands of the corresponding sub-eye lenses is determined by the aperture (Di) of the sub-eye lens, the design wavelength ( ) and the focal length (F), which can be calculated by Formula (1). Then, according to Formula (2), the radius ( ) corresponding to different number of annular bands ( ) can be calculated to obtain the complete distribution of annular bands structure for different subeye lenses, and some radius values are shown in Table 1:   The arrangement of the sub-eye lenses in the multi-spectral ACE is shown in Figure 2. Along the Y direction, the sub-eye lenses are the same. In the X direction, the sub-eye lenses are designed for different wavelengths. The designed wavelengths (λ i 0 , i = 1,2,3) are 650 nm (R), 532 nm (G) and 445 nm (B), respectively. The aperture (D i , i = 1,2,3) are the same as 2 mm, and the focal length (F) is also the same as 20 mm. Three sub-eye lenses with different central wavelengths are arranged periodically in the X direction with a period of 6 mm to form a complete ACE structure.

Principle and Structural Design
The ACE is composed of several sub-eye imaging channels, as shown in Figure 1. Different subeye imaging channels are designed as diffractive beam splitting lenses (D ) with different designed parameters. The lenses have the same focal length (F) and different central wavelength ( ). In the imaging process, spectral information of different wavelengths can be captured by different sub-eyes at the same image plane. The arrangement of the sub-eye lenses in the multi-spectral ACE is shown in Figure 2. Along the Y direction, the sub-eye lenses are the same. In the X direction, the sub-eye lenses are designed for different wavelengths. The designed wavelengths ( , i = 1,2,3) are 650 nm (R), 532 nm (G) and 445 nm (B), respectively. The aperture (Di, i = 1,2,3) are the same as 2 mm, and the focal length (F) is also the same as 20 mm. Three sub-eye lenses with different central wavelengths are arranged periodically in the X direction with a period of 6 mm to form a complete ACE structure. Different sub-eyes are composed of different annular structures. The maximum number of annular bands of the corresponding sub-eye lenses is determined by the aperture (Di) of the sub-eye lens, the design wavelength ( ) and the focal length (F), which can be calculated by Formula (1). Then, according to Formula (2), the radius ( ) corresponding to different number of annular bands ( ) can be calculated to obtain the complete distribution of annular bands structure for different subeye lenses, and some radius values are shown in Table 1:   Different sub-eyes are composed of different annular structures. The maximum number of annular bands of the corresponding sub-eye lenses is determined by the aperture (D i ) of the sub-eye lens, the design wavelength (λ i 0 ) and the focal length (F), which can be calculated by Formula (1). Then, according to Formula (2), the radius (r i n ) corresponding to different number of annular bands (n i ) can be calculated to obtain the complete distribution of annular bands structure for different sub-eye lenses, and some radius values are shown in Table 1: r in = n i Fλ i0 , n i = 1, 2 · · · N i ;

Fabrication
The preparation technology of the proposed multi-spectral ACE was studied. The preparation process of diffractive beam splitting lens with a central wavelength of 532 nm is described in detail below. When the central wavelength is 532 nm, the etching depth (h) can be obtained by Formula (3). The refractive index (n) of silica material is 1.461 at 532 nm, so the etching depth is calculated as 577 nm: h = λ 2(n − 1) , n = 1.461@532m; The fabrication process flow mainly includes photoresist spin coating, exposure, development and etching, as shown in Figure 3. Firstly, the silica material (JGS1) was selected as the substrate material, and the photoresist AZ1500 (20cp) (AZ 1500 Photoresist (20cp), Merck Electronic Materials, Shanghai, China) was spun on the surface of the substrate at a speed of 6000 rpm for 30 s as the initial structure (Figure 3a,b). Key parameters such as prebake temperature and prebake period were 90 • C and 10 min, respectively. Secondly, the annular bands structure was encoded as the mask data according to the design parameters, and the mask was prepared using the high-precision laser direct writing technology, as shown in Figure 4a. The UV lithography machine (URE-2000S/A) with a central wavelength of 365 nm was used. The exposure mode was proximity exposure with total exposure time of 10 s at an exposure power density of 2.5 mW/cm 2 ( Figure 3c). Third, the development and post-exposure bake were carried out. The photoresist was developed in a solution of AZ300MIF (AZ300MIF Developer, Merck Electronic Materials, Shanghai, China) for 25 s, followed by 2 h post-exposure bakes at 120 • C, and the photoresist multi-spectral compound eye structure was obtained as shown in Figure 3d. Finally, etching of the silica was conducted, which transferred the patterns of the multi-spectral compound eye structure to the silica substrate (see Figure 4b). A reactive ion etching machine was used to transfer the structure into the substrate. The etching gases were SF 6 and CHF 3 in the proportion of 1:20 and the etching time was 15 min (see Figure 3e). patterns of the multi-spectral compound eye structure to the silica substrate (see Figure 4b). A reactive ion etching machine was used to transfer the structure into the substrate. The etching gases were SF6 and CHF3 in the proportion of 1:20 and the etching time was 15 min (see Figure 3e).
A step profilometer (Stylus Profiler System, Dektak XT, Bruker, Karlsruhe, Germany) was used for the measurement of the fabricated structure, and the cross-sectional profile was drawn by the scientific drawing and data analysis software of OriginLab (OriginPro 8.0, Northampton, MA, USA) with etching depth of 574 nm (see Figure 4c), which was consistent with the design result.  The structure of different diffractive beam splitting lenses corresponding to different wavelength was realized by multiple overlay exposure and etching technology. The lenses with different depths were obtained by changing etching times during the etching process.
The diffraction efficiency of the structure is related to the step number of the actual structure. The higher the step number is, the higher the diffraction efficiency is. When the number of steps is 2, the diffraction efficiency is about 40%; when the number of steps is 4, the diffraction efficiency is more than 80%; when the number of steps is more than 8, the diffraction efficiency can be more than 90%; were SF6 and CHF3 in the proportion of 1:20 and the etching time was 15 min (see Figure 3e).
A step profilometer (Stylus Profiler System, Dektak XT, Bruker, Karlsruhe, Germany) was used for the measurement of the fabricated structure, and the cross-sectional profile was drawn by the scientific drawing and data analysis software of OriginLab (OriginPro 8.0, Northampton, MA, USA) with etching depth of 574 nm (see Figure 4c), which was consistent with the design result.  The structure of different diffractive beam splitting lenses corresponding to different wavelength was realized by multiple overlay exposure and etching technology. The lenses with different depths were obtained by changing etching times during the etching process.
The diffraction efficiency of the structure is related to the step number of the actual structure. The higher the step number is, the higher the diffraction efficiency is. When the number of steps is 2, the diffraction efficiency is about 40%; when the number of steps is 4, the diffraction efficiency is more than 80%; when the number of steps is more than 8, the diffraction efficiency can be more than 90%; A step profilometer (Stylus Profiler System, Dektak XT, Bruker, Karlsruhe, Germany) was used for the measurement of the fabricated structure, and the cross-sectional profile was drawn by the scientific drawing and data analysis software of OriginLab (OriginPro 8.0, Northampton, MA, USA) with etching depth of 574 nm (see Figure 4c), which was consistent with the design result.
The structure of different diffractive beam splitting lenses corresponding to different wavelength was realized by multiple overlay exposure and etching technology. The lenses with different depths were obtained by changing etching times during the etching process.
The diffraction efficiency of the structure is related to the step number of the actual structure. The higher the step number is, the higher the diffraction efficiency is. When the number of steps is 2, the diffraction efficiency is about 40%; when the number of steps is 4, the diffraction efficiency is more than 80%; when the number of steps is more than 8, the diffraction efficiency can be more than 90%; if the number of steps is 32, the diffraction efficiency can be as high as 99%. Of course, with the increase of the number of processing steps, the difficulty of preparation process will also increase. In order to verify the principle of the proposed structure from the optical effect, the structure of the fabricated structure is only two steps, and the diffraction efficiency is slightly low at 40%. If the practical application of the structure requires higher diffraction efficiency, the structure with higher steps can be prepared.

Imaging Verification
The imaging light path was constructed to verify the spectral characteristics of the prepared multi-spectral ACE structure. The structure was irradiated by red, green and blue lasers of 650 nm, Micromachines 2019, 10, 208 6 of 10 532 nm and 445 nm, which were consistent with the design center wavelength, as shown in Figure 5. During the experiment, the expanding and collimation of the laser beam were firstly realized, so that the laser beam with a diameter of more than 6 mm can simultaneously illuminate the area with different designed center wavelengths on the multi-spectral compound eye structure. Secondly, the three-color beams were combined to make the three-channel beam coaxially illuminate the multi-spectral compound eye structure. Finally, by adjusting the distance Z between the photodetector and the structure, the spot distributions of different wavelengths at different distances were observed.
practical application of the structure requires higher diffraction efficiency, the structure with higher steps can be prepared.

Imaging Verification
The imaging light path was constructed to verify the spectral characteristics of the prepared multi-spectral ACE structure. The structure was irradiated by red, green and blue lasers of 650 nm, 532 nm and 445 nm, which were consistent with the design center wavelength, as shown in Figure 5. During the experiment, the expanding and collimation of the laser beam were firstly realized, so that the laser beam with a diameter of more than 6 mm can simultaneously illuminate the area with different designed center wavelengths on the multi-spectral compound eye structure. Secondly, the three-color beams were combined to make the three-channel beam coaxially illuminate the multispectral compound eye structure. Finally, by adjusting the distance Z between the photodetector and the structure, the spot distributions of different wavelengths at different distances were observed. During the experiment, three laser beams were independently used to illuminate the multispectral compound eye structure. At the design focal length of 20 mm, laser beams of different wavelengths were converged by different sub-eye lenses, as shown in Figure 6. The sub-eye lens structure marked with λ10 corresponds to the designed wavelength of 650 nm, so when the red laser beam irradiated the structure, the detector detected the focus at 20 mm converged by the corresponding sub-eye structure, as shown in Figure 6c.
Accordingly, when the green laser beam irradiated the structure with λ20 corresponding to the designed wavelength of 532 nm, the detector detected the focus at 20 mm converged by the corresponding sub-eye structure, as shown in Figure 6e. When the blue laser beam irradiated the structure with λ30 corresponding to the designed wavelength of 445 nm, the detector detected the focus at 20 mm converged by the corresponding sub-eye structure, as shown in Figure 6g.
When tricolor laser beams of red, green and blue illuminate the multispectral structure at the same time, it can be seen that different laser beams of different wavelengths can be converged by the corresponding sub-eye structures of the same designed wavelengths, while the other laser beams disperse, as shown in Figure 7a. It is verified that the structure can realize the spectral splitting function for different wavelengths by different sub-eye imaging channels without axial movement. The intensity distributions of the focal spot sizes were analyzed and shown in Figures 7b-d. The ranges of the energy of the focal spots reduced from the highest to 1/e were calculated as the sizes of the focal spots. The focal spot sizes of the wavelengths of 650 nm, 532 nm and 445 nm were quantified as 88.8 μm, 66.6 μm and 59.2 μm, respectively. During the experiment, three laser beams were independently used to illuminate the multi-spectral compound eye structure. At the design focal length of 20 mm, laser beams of different wavelengths were converged by different sub-eye lenses, as shown in Figure 6. The sub-eye lens structure marked with λ 10 corresponds to the designed wavelength of 650 nm, so when the red laser beam irradiated the structure, the detector detected the focus at 20 mm converged by the corresponding sub-eye structure, as shown in Figure 6c. Under the same focal length and aperture, the longer the wavelength is, the larger the focal spot size is. The experimental results are in agreement with the theory. Therefore, with the increase of focal spot size, the imaging quality will become worse. In addition, because the scattered spots of other wavelengths are superimposed on the focal spots, the sizes of the captured focal spots which affects the imaging quality are larger than that of the theoretical Airy spots. From the energy distribution curve, it can be seen that, for the focal spot of red light, the intensity of the diffuse spots produced by other wavelengths is larger than the other colors. This is also the reason for the poor effect of red image in the later imaging experiment.
At the same time, if the detector is moved axially, the spectral splitting function of each sub-eye lens can still be observed. For example, a sub-eye lens with a central wavelength of λ10 can converge red light of 650 nm at the distance of 20 mm, as shown in Figure 6c. With the increase of distance Z, the sub-eye lens converges green light at the distance of 24.4 mm, as shown in Figure 6f. As the distance Z increases more, the blue light converges at the distance of 29.2 mm, as shown in Figure 6i. This is determined by the spectral splitting characteristics of the sub-eye lens. Its focal length is inversely proportional to the wavelength. With the increase of distance, the wavelength of the converged light decreases. The wavelength decreases from 650 nm to 532 nm, and then to 445 nm in the experiment, which is consistent with the spectral splitting theory.
Similarly, the sub-eye lens with a central wavelength of λ20 can converge green light of 532 nm at the distance of 20 mm, as shown in Figure 6e, With the increase of distance Z, the sub-eye lens converges shorter wavelength of blue light at the distance of 24 mm, as shown in Figure 6h. With the decrease of distance Z, the sub-eye lens converges longer wavelength of red light at the distance of 16.4 mm, as shown in Figure 6b. A sub-eye lens with a central wavelength of λ30 can converge blue light of 445 nm at a distance of 20 mm, as shown in Figure 6g. With the decrease of distance Z, subeye lens converges longer wavelength green light at the distance of 16.7 mm, as shown in Figure 6d. With the more decrease of distance Z, sub-eye lens converges the longest wavelength red light at the distance of 13.7 mm, as shown in Figure 6a.  Accordingly, when the green laser beam irradiated the structure with λ 20 corresponding to the designed wavelength of 532 nm, the detector detected the focus at 20 mm converged by the corresponding sub-eye structure, as shown in Figure 6e. When the blue laser beam irradiated the structure with λ 30 corresponding to the designed wavelength of 445 nm, the detector detected the focus at 20 mm converged by the corresponding sub-eye structure, as shown in Figure 6g.
When tricolor laser beams of red, green and blue illuminate the multispectral structure at the same time, it can be seen that different laser beams of different wavelengths can be converged by the corresponding sub-eye structures of the same designed wavelengths, while the other laser beams disperse, as shown in Figure 7a. It is verified that the structure can realize the spectral splitting function for different wavelengths by different sub-eye imaging channels without axial movement. The intensity distributions of the focal spot sizes were analyzed and shown in Figure 7b Furthermore, in order to verify the imaging characteristics of the red, green and blue multispectral compound eye structure, the prepared structure was used to image the luminous object "E". The luminous object was white light in a wide band, and the image was captured by the photodetector. The overall experimental light path was shown in Figure 8. The object distance (l) between "E" and multi-spectral compound eye structure was set as 1850 mm. The distance between the detector and the multi-spectral compound eye structure was defined as the imaging distance (l'). According to the Gauss imaging formula, the imaging distance was calculated as 20.22 mm with the focal length of 20 mm for various wavelengths. The detector was moved along the axis to record the image, which was located on the corresponding imaging plane of the structure. The imaging effect of the prepared structure was effectively verified as shown in Figure 9. The detector obtained different wavelengths of red, green and blue spectral information at the same image plane, which verified the multi-spectral imaging performance of compound eye structure. It is worth mentioning that the Under the same focal length and aperture, the longer the wavelength is, the larger the focal spot size is. The experimental results are in agreement with the theory. Therefore, with the increase of focal spot size, the imaging quality will become worse. In addition, because the scattered spots of other wavelengths are superimposed on the focal spots, the sizes of the captured focal spots which affects the imaging quality are larger than that of the theoretical Airy spots. From the energy distribution curve, it can be seen that, for the focal spot of red light, the intensity of the diffuse spots produced by other wavelengths is larger than the other colors. This is also the reason for the poor effect of red image in the later imaging experiment.
At the same time, if the detector is moved axially, the spectral splitting function of each sub-eye lens can still be observed. For example, a sub-eye lens with a central wavelength of λ 10 can converge red light of 650 nm at the distance of 20 mm, as shown in Figure 6c. With the increase of distance Z, the sub-eye lens converges green light at the distance of 24.4 mm, as shown in Figure 6f. As the distance Z increases more, the blue light converges at the distance of 29.2 mm, as shown in Figure 6i. This is determined by the spectral splitting characteristics of the sub-eye lens. Its focal length is inversely proportional to the wavelength. With the increase of distance, the wavelength of the converged light decreases. The wavelength decreases from 650 nm to 532 nm, and then to 445 nm in the experiment, which is consistent with the spectral splitting theory.
Similarly, the sub-eye lens with a central wavelength of λ 20 can converge green light of 532 nm at the distance of 20 mm, as shown in Figure 6e, With the increase of distance Z, the sub-eye lens converges shorter wavelength of blue light at the distance of 24 mm, as shown in Figure 6h. With the decrease of distance Z, the sub-eye lens converges longer wavelength of red light at the distance of 16.4 mm, as shown in Figure 6b. A sub-eye lens with a central wavelength of λ 30 can converge blue light of 445 nm at a distance of 20 mm, as shown in Figure 6g. With the decrease of distance Z, sub-eye lens converges longer wavelength green light at the distance of 16.7 mm, as shown in Figure 6d. With the more decrease of distance Z, sub-eye lens converges the longest wavelength red light at the distance of 13.7 mm, as shown in Figure 6a.
Furthermore, in order to verify the imaging characteristics of the red, green and blue multi-spectral compound eye structure, the prepared structure was used to image the luminous object "E". The luminous object was white light in a wide band, and the image was captured by the photodetector. The overall experimental light path was shown in Figure 8. The object distance (l) between "E" and multi-spectral compound eye structure was set as 1850 mm. The distance between the detector and the multi-spectral compound eye structure was defined as the imaging distance (l'). According to the Gauss imaging formula, the imaging distance was calculated as 20.22 mm with the focal length of 20 mm for various wavelengths. The detector was moved along the axis to record the image, which was located on the corresponding imaging plane of the structure. The imaging effect of the prepared structure was effectively verified as shown in Figure 9. The detector obtained different wavelengths of red, green and blue spectral information at the same image plane, which verified the multi-spectral imaging performance of compound eye structure. It is worth mentioning that the spectral image received by the detector is the overlap image of the focus image formed by one wavelength on its focal plane and defocus images of other wavelengths at this position. Therefore, the light from other spectral channels will cause interference and blur to the image. A series of algorithms are needed to process the image in the following steps to obtain the target spectral image data cube. In order to eliminate spectral blurring and restore image quality, image space linearization based on a translation invariant model is needed, that is, the relative wavelength of spectral image should have an equal interval and equal frame. Then, the three-dimensional deconvolution and other reconstruction image processing techniques in three-dimensional optical slice microscopy technology need to be used to obtain clear spectral image information. Furthermore, in order to verify the imaging characteristics of the red, green and blue multispectral compound eye structure, the prepared structure was used to image the luminous object "E". The luminous object was white light in a wide band, and the image was captured by the photodetector. The overall experimental light path was shown in Figure 8. The object distance (l) between "E" and multi-spectral compound eye structure was set as 1850 mm. The distance between the detector and the multi-spectral compound eye structure was defined as the imaging distance (l'). According to the Gauss imaging formula, the imaging distance was calculated as 20.22 mm with the focal length of 20 mm for various wavelengths. The detector was moved along the axis to record the image, which was located on the corresponding imaging plane of the structure. The imaging effect of the prepared structure was effectively verified as shown in Figure 9. The detector obtained different wavelengths of red, green and blue spectral information at the same image plane, which verified the multi-spectral imaging performance of compound eye structure. It is worth mentioning that the spectral image received by the detector is the overlap image of the focus image formed by one wavelength on its focal plane and defocus images of other wavelengths at this position. Therefore, the light from other spectral channels will cause interference and blur to the image. A series of algorithms are needed to process the image in the following steps to obtain the target spectral image data cube. In order to eliminate spectral blurring and restore image quality, image space linearization based on a translation invariant model is needed, that is, the relative wavelength of spectral image should have an equal interval and equal frame. Then, the three-dimensional deconvolution and other reconstruction image processing techniques in three-dimensional optical slice microscopy technology need to be used to obtain clear spectral image information.