Electron bombardment CMOS (EBCMOS) is an advanced micro-optical imaging technology. The back-side bombarded CMOS (BSB-CMOS) imaging chip, when integrated with electron bombardment imaging systems, facilitates digital imaging of targets under extremely low illumination conditions. EBCMOS is better than traditional micro-optical imaging devices considering it combines high gain and signal-to-noise ratio (SNR) of vacuum devices with rapid response, miniaturization, and digital output and transmission features of solid-state devices. This chip advancement offers considerable potential in military applications, single-photon detection, and medical imaging, among others. The EBCMOS structure primarily has BSB-CMOS, a photocathode, and a vacuum tube. The noise characteristics of EBCMOS devices are affected by various factors, such as dark current, readout, photon, and multiplication noises; among these, dark current plays a critical role in determining the imaging quality of EBCMOS, especially under low light conditions.

This study presents a comprehensive analysis of noise sources and their influencing factors in the EBCMOS imaging system based on the operational principles of an EBCMOS. A theoretical model for SNR calculation was developed by incorporating the existing EBCMOS gain model. The roles of the passivation and electron multiplication layers in the dark current generation in EBCMOS devices is emphasized. Key parameters, including the SNR within a pixel, the number of total noise electrons within a pixel (N_pixel), the number of multiplying electrons (N_M) within a pixel, and the number of dark current electrons per unit pixel (N_dark), describe the noise characteristics. The simulations examined the effects of various passivation layer materials, passivation layer thickness, incident electron energy, and substrate temperature on these noise characteristics. This study provides a theoretical foundation for the development of low-noise and high-performance EBCMOS.

Results and discussions Applying Al₂O₃ as a passivation layer material on the BSB-CMOS surface, with a density of 3.8 g/cm³, notably reduces the interfacial state density and increases N_M, thereby improving charge collection efficiency. Therefore, the SNR is substantially improved, with an SNR value of up to 116 in the 5 pixel×5 pixel area (Fig. 2). Increasing the thickness of the Al₂O₃ passivation layer effectively reduces the interfacial density of states because Al₂O₃ has a high negative fixed charge density, which can inhibit charge traps at the interface, thereby decreasing N_dark. However, increasing thickness of the passivation layer decreases the N_M of the device. Thus, when the thickness of the Al?O? passivation layer is 15 nm, a lower interfacial state density can be achieved while maintaining a higher N_M, resulting in the highest SNR of the device (Fig. 3). Increasing the incident electron energy can substantially improve the N_M and device SNR without considerably affecting the N_dark. Selecting an incident electron energy that matches the passivation layer thickness results in high SNR. When the incident electron energy is 6 keV and Al₂O₃ passivation layer thickness is 15 nm, the SNR can reach up to 158 in the 5×5 pixel area (Fig. 4). As the substrate temperature increases, N_dark exhibits a notable increase, in addition to decreasing N_M. Thus, reducing the substrate temperature of the device improves the SNR. By optimizing the passivation layer material, thickness, incident electron energy, and substrate temperature, Al₂O₃ can be selected as the passivation layer with a thickness of 15 nm, incident electron energy of 6 keV, and substrate temperature of 260 K. This configuration achieves the maximum SNR of 188 (Fig. 5).

This study develops a theoretical noise model for EBCMOS based on the principles of micro-optical imaging and semiconductor materials. This study investigates the effects of the passivation layer material, passivation layer thickness, incident electron energy, and substrate temperature on the noise characteristics of the device. The simulations and analyses indicate that the passivation layer in an EBCMOS device remarkably reduces surface recombination, thereby decreasing the dark current. Furthermore, the passivation layer plays a crucial role in determining the photoelectric conversion efficiency of the device. Optimizing the passivation layer enhances the collection efficiency of photogenerated electrons, improving the device SNR. Results indicate that using Al?O? as the passivation layer is particularly effective in reducing dark current and increasing the number of multiplied electrons, because of its low interfacial state density and small material density, thereby improving the device SNR. Selecting an optimal thickness of Al₂O₃ passivation layer effectively reduces the interfacial state density and decreases the surface recombination rate of electrons. This combined effect suppresses dark current and increases the number of multiplied electrons, resulting in high SNR. In addition, increasing the incident electron energy and lowering the substrate temperature enhances electron multiplication while minimizing recombination, thereby improving the SNR. After optimization, the SNR in the central pixel area of the device reaches 188, whereas the number of dark current electrons per pixel decreases to 100; these findings highlight the importance of noise characterization in EBCMOS devices and provide valuable theoretical insights for developing high-SNR imaging systems.