Infrared detection technology，due to its advantages such as long detection range，day and night imaging capabilities，and atmospheric penetration，finds extensive applications in both military and civilian domains. Due to adjustable bandgap of Mercury cadmium telluride（Hg_1-xCd_xTe）material by carefully selecting the composition，it offers the flexibility to fabricate infrared detectors with adjustable cutoff wavelengths ^［1-2］. The development of third-generation infrared focal plane arrays（IRFPAs），characterized by their large-scale，multicolor，and high integration features，has been ongoing for nearly 20 years^［3-5］. To achieve farther detection range，higher operational temperatures，improved spectral resolution，and lower costs，a new generation of detectors has been developed for diverse fields such as military reconnaissance and identification，space remote sensing，airborne remote sensing，meteorological monitoring，and environmental/resource monitoring.

The detection range of IRFPAs is directly influenced by the instantaneous field of view of the pixels. Consequently，the development of small-pitch and high-resolution FPAs within a fixed field of view becomes crucial for increasing the detection range. For instance，when it changes from a 30 μm pitch 320×256 FPA to a 15 μm pitch 640×512 FPA can enhance the MW IRFPA's detection range by approximately 50% at F=2 ^［6］. Consequently，high-resolution FPAs have become an integral component of third-generation infrared focal plane detectors^［7-8］.

Many institutions are doing research and developing the high-resolution FPAs. In order to meet the needs of IR detection systems with higher spatial resolution，Sofradir has developed a Jupiter model operating in the MWIR band with a 1280×1024 format and a pixel size of 15µm，and is cooled by Thales Cryogenics' linear flexure bearing split Stirling cooler^［9］. Additionally，Teledyne's Hawaii-2RG（H2RG），which is based on the focal plane array with an 18μm pixel pitch and a 2048×2048 array，finds applications in space and ground-based equipment，including the James Webb Space Telescope ^［10-11］. Moreover，significant process improvements have been made by researchers，to enhance the practical application potential of HgCdTe photodetectors. A micro-mesa array technique has been employed by Hu et al. and selective B+ implantation to fabricate HgCdTe LW/MW two-color infrared detectors ^［12］. Additionally，the surface quality of typical n+-on-p HgCdTe LWIR photodiodes has been improved by Hu et al. through hybrid surface passivation，effectively suppressing trap-assisted tunneling currents ^［13］. Furthermore，Hanxue Jiao et al. designed and fabricated a high-performance room temperature polarization-resolved MWIR photodetector using HgCdTe/bP van der Waals heterojunction. This design effectively suppresses dark current，enabling outstanding MWIR detection capability at room temperature^［14］.

After successively manufacturing 30μm pitch 320×256 and 15μm pitch 640×512 MW IRFPAs，Zhejiang Juexin Microelectronics Co.，Ltd. has also conducted research and development on the manufacturing technology of 10μm pitch 1280×1024 MW IRFPAs. The key points in the research and development process are to overcome the impact of thermal stress between HgCdTe chips and ROICs on the performance of IRFPAs，as well as to solve the problems such as large-area material uniformity，small pixel process technology，and high-density In bump bonding technique. By using CdZnTe as the substrate and removing it to release the thermal mismatch，and improving the structure of the In bump to enhance the interconnection strength，10μm pitch 1280×1024 HgCdTe MW IRFPAs with high performance has been developed successfully. This paper introduces the preparation and related properties of the medium-wave 1280×1024（10 µm）HgCdTe infrared detector made by Zhejiang Juexin Microelectronics Co.，Ltd.

With the continuous progress of HgCdTe IRFPAs technology，the preparation techniques for IRFPAs with small pixel sizes have reached a level of maturity，facilitating the development of high-resolution IRFPAs. Nevertheless，it is crucial to acknowledge that the advancement of new manufacturing technology is accompanied by a range of challenges attributed to the reduction in pixel size and the expansion of the FPA area.

The vertical Bridgman method was employed to grow CdZnTe crystals as substrates for the HgCdTe epitaxial layer. The CdZnTe substrates were polished，and Hg1-xCdxTe material（with x~0.3）was grown on the（111）B CdZnTe substrate using liquid phase epitaxy（LPE）. The resulting HgCdTe epilayers，with a etch pit density lower than 5×10⁴ cm^-2，were obtained through a stepwise cooling process. These epilayers，measuring 40 mm×30 mm，exhibited high surface flatness，composition uniformity，and low defect density. Subsequently，the epitaxial layer was annealed to form P-type HgCdTe materials for chip processing.

To ensure the smoothness of the HgCdTe epitaxial layer，the surface flatness of the CdZnTe substrates was controlled within 1 μm through processes such as chemical mechanical polishing（CMP）and chemical polishing（CP）. The surface profiles of the HgCdTe materials were measured using a Bruker ContourGT-X interferometer，as shown in Figure 1. The maximum height difference across the entire 1280×1024 FPA chip surface was found to be smaller than 0.5 μm. This optimization of surface morphology allows for a wider process window in subsequent chip processing steps，particularly for applications involving small pitch and large-scale arrays，such as lithography patterning uniformity and the preparation of uniform indium bumps.

The performance of HgCdTe infrared detectors is closely tied to the structure of the p/n junction^［15-17］. In this study，planar junction technology，based on B+ ion implantation and passivation，was utilized for the fabrication of HgCdTe infrared detectors. Furthermore，through a series of chip processes including coating（involving thermal evaporation，electron beam evaporation，and magnetron sputtering），wet etching，and flip-chip bonding，1280×1024 arrays with a pitch of 10 μm were achieved. The pixel structure of the 1280×1024 array is illustrated in Figure 2.

The pixel size of the fabricated 1280×1024 arrays in this study is 10 μm，which allows for a smaller，lighter，and more compact system. Additionally，it contributes to reduced power consumption and cost. Moreover，reducing the pixel pitch enables more FPAs to be obtained from the same material substrate ^［18］. The processing of small-sized pixels，particularly the fabrication of small In bumps，is a crucial technique. In bumps are soft metals with low melting points and excellent ductility，making them ideal for chip bonding ^［19］. Therefore，the HgCdTe focal plane chips and readout circuit chips are typically bonded using the In bump flip-chip interconnection technique for signal readout ^［20］. In this work，the chips were bonded using FC 150 flip-chip welding equipment. Through optimization of the In bump structure，lithography，and In deposition processes，In bump arrays with excellent consistency were achieved. The uniformity of In bump heights exceeded 95%. The morphology of the In bumps，as measured by ContourGT-X，is depicted in Figure 3. The use of uniform In bumps and advanced flip-chip bonding technology resulted in exceptional connectivity for the 1280×1028 FPAs，with a bonding success rate exceeding 99.999%.

HgCdTe IRFPAs consist of several components，including the HgCdTe chip，In bump interconnection area，Si readout circuit，and circuit boards. These components are fabricated at room temperature and operate at low temperatures（typically 77~120 K）. However，due to the mismatch in thermal expansion coefficients among these materials，thermal stress can arise during the cooling process of FPA devices. This can lead to issues such as chip fracture and fatigue damage of solder joints，resulting in degraded FPA performance ^［21］.

To address these challenges，the gap between the HgCdTe chip and the readout circuit is filled with low-temperature glue. In this study，an optimized glue filling process was adopted to ensure reliable interconnection and prevent incomplete filling at the edges. To achieve uniform glue distribution，the capillary effect was utilized. Additionally，a three-stage variable temperature baking process was employed to prevent excessive stress caused by rapid glue curing. The process involved an initial bake at 45°C for 2 hours，which is below the glass transition temperature of the adhesive. Subsequently，the glue was cured by baking at a temperature above the glass transition temperature for 1 hour，followed by a final bake at 45°C for 12 hours. Furthermore，a slotting process was implemented to mitigate device failures resulting from stress. After chip metallization，the cutting process using a diamond blade can generate microscale edge chippings，leading to stress concentration and device failure during thermal shocks. Various methods such as wet etching，laser etching，or dry etching can be employed to create slots around the chip，effectively reducing edge chippings during cutting. In this study，dry etching was chosen due to the expansion of corrosion associated with wet etching and the thermal effects induced by lasers. Figure 4 is the bad pixel mapping for 1280 × 1024 MWIR FPAs，indicating that in Figure 4（a），unslotted FPA develop cracks due to stress，whereas the FPA with edge slotted in Figure 4（b）do not exhibit cracks.

Through a series of process improvements，10 μm pitch 1280×1024 MW HgCdTe infrared focal plane arrays were successfully fabricated. The FPAs were then mounted and wire-bonded in a leadless chip carrier（LCC）within a dewar and coupled with a Stirling cooler，as depicted in Figure 5. Finally，the performance of the FPAs was systematically evaluated using an infrared FPA evaluation system at Zhejiang Juexin Microelectronics Co.，Ltd.（ZJM）.

The spectral response of the detector was tested using a monochromator at an operating temperature of 85 K，as shown in Figure 6. The figure illustrates that the measured device exhibits a spectral response ranging from 3.67 to 4.88 μm.

The responsivities and NETDs of the detectors were determined by measuring the output voltages of the detector using a black body as the background at temperatures of 20℃ and 35℃. The measurement employed an integration time of 20 ms. Figure 7 displays the grayscale image of the responsivity of the detector operating in IWR（LG）mode，demonstrating its uniformity with a responsivity non-uniformity of 5.07%. Figure 8 illustrates the NETDs of the detectors at a background temperature of 20℃. Figure 8（a）represents the grayscale image of the NETD，while Figure 8（b）shows the histogram of the NETD. At an operating temperature of 85 K，the histogram exhibits symmetrical characteristics without any tails，indicating a high level of operability for the detector. The average NETD is measured at 15.56 mK with a 50% well fill. Defective pixels are defined as those falling outside 30% of the mean responsivity，signal，or within the NETD range of 0 to 60mk. The effective pixel count for this detector is 99.95%.

In addition，the dark current of the FPA was tested at an operating temperature of 85 K. The results of the dark current measurements are shown in Figure 9. It can be observed that the average dark current of the device is 2.06×10^-14A，with a corresponding dark current density of 2.06×10^-8A/cm².

Table 2 presents a performance comparison of 10 μm pitch MW IRFPAs with major IR-detector manufacturers. The data in the table clearly demonstrates that the FPA developed by ZJM exhibits a lower NETD and higher array operability compared to other manufacturers.

Finally，to compare the differences between the 10 µm pitch MW 1280×1024 array and the 15 µm pitch 640×512 array，the same optical system design was employed for both detectors. The optical aperture of the system is F/4，and the optical field of view is 14.59°×11.69°. As depicted in Figure 10，the structure of the target in the image obtained by the 1280×1024 array is clearer than that of the 640×512 array. The words on the billboard and the details of the crane can be recognized by the 1280×1024 10 µm MWIR detector.

The 10 μm pitch 1280×1024 HgCdTe MWIR FPAs were successfully fabricated by Zhejiang Juexin Microelectronics Co.，Ltd.. The height difference of the HgCdTe surface less than 0.5 μm by the optimization of substrate CMP and CP processing. And successfully developed the processing technique of 10 μm pixels based on B⁺ injected n-on-p planar junction and small size In bump bonding technique. The performance of 1280×1024 HgCdTe MWIR FPA were measured at 85 K and evaluated. The results show that the FPA has average value of NETD of 15.56 mK and operability of 99.95%. The average value of dark current is 2.06×10^-14A. The imaging of 1280×1024 HgCdTe MWIR FPAs with high performance was also successfully demonstrated. The fabrication technology developed in this work has been transferred to the production line at ZJM to produce the assemblies of 10 μm pitch 1280×1024 MW HgCdTe FPAs.