The quantum efficiency spectrum of a CMOS image sensor refers to the response efficiency of the sensor to light at different wavelengths. Physically, the energy of a photon is inversely proportional to its wavelength, so the response efficiency of photons with different wavelengths to a CMOS image sensor also varies. The quantum efficiency spectrum can reflect the response ability of the image sensor at different wavelengths, helping people understand characteristics of the image sensor such as sensitivity and color reproduction ability. Typically, the quantum efficiency spectrum of an image sensor will show different characteristics within the visible light band, such as peaks and valleys. These characteristics also directly affect the image quality of the image sensor.
● BSI processing design
● Optical Crosstalk inspection
● Color filter quality and performance
● Si wafer THK condition in BSI processing
(1) How BSI works
BSI stands for Back-Side Illumination. It refers to the manufacturing process of “back-illuminated” image sensors. Compared to traditional “front-illuminated” (FSI) image sensors, it can improve the optical performance of image sensors, especially the significant increase in photosensitivity at various wavelengths. In the BSI process, pixels are placed on the back of the silicon substrate, and light enters the photosensitive pixels through the silicon substrate, reducing interference from the transmission layer and metal wires in front and improving light utilization and diffraction effects, thereby enhancing resolution and sensitivity of the image sensor.
Fig.1 How BSI works
(2) How traditional “front-illuminated” image sensors work
FSI is a traditional image sensor process technology. After the light passes through the lens, it illuminates the photosensitive surface of the image sensor from the front of the image sensor. Therefore, some circuits and metal wires need to be placed above the photosensitive surface (yellow box, Silicon). These components will block part of the light and reduce the light utilization of the image sensor, affecting image quality. In contrast, BSI technology fabricates photosensitive elements on the back side of the photosensitive surface, which is the back side of the substrate. This allows light to directly enter the photosensitive surface, thus maximizing light utilization and improving image quality. Without the blockage of additional circuits and metal wires, higher pixel density and faster image reading speeds can also be achieved.
(3) Why is BSI process important?
The BSI process is one of the important manufacturing technologies that can significantly improve the photosensitivity and quantum efficiency of CMOS image sensors, and is very helpful for image acquisition in low light environments.
The BSI process can also improve the resolution, dynamic range, signal-to-noise ratio and other performances of image sensors, resulting in higher image quality.
As image applications become more and more widespread today, the requirements for image quality and performance are also increasing. Therefore, the BSI process plays an important role in the manufacturing of modern image sensors. Currently, BSI technology has become one of the mainstream process technologies for high-end image sensors, and is widely used in various high-end imaging products.
(4) How quantum efficiency spectrum evaluates the quality of BSI process
As aforementioned, photons of different wavelengths have different photosensitive capabilities for the image sensor chip during CMOS image sensor manufacturing. Therefore, the quantum efficiency spectrum is a method that can detect the photosensitivity of image sensor chips. The quality of BSI processes can be evaluated using quantum efficiency spectra.
As shown in Fig. 2, TSMC (Taiwan Semiconductor Manufacturing Company) used quantum efficiency spectra to analyze the differences in photosensitivity of RGB pixels between front-illuminated FSI and back-illuminated BSI processes. The results show that the BSI process can significantly improve the photosensitivity of pixels, increasing the approximately 40% quantum efficiency of the original FSI to nearly 60%.
Fig. 2 TSMC used Wafer Level Quantum Efficiency Spectrum to analyze the differences in photosensitivity of RGB pixels under different wavelengths between 1.75um FSI and BSI processes. Quantum efficiency spectrum analysis can help engineers determine the impact of different processes on photosensitivity and identify the advantages of BSI processes.
(5) Using quantum efficiency spectra to analyze the impact of different BSI processes on the photosensitivity of CMOS image sensor chips
As shown in Fig. 3, a CMOS image sensor manufacturer used wafer-level quantum efficiency spectra to analyze the impact of different processes on the photosensitivity of CMOS image sensor chips during mass production using TSMC 65nm technology. In the quantum efficiency spectrum comparison between BSI-1 and BSI-2 processes with 1.4um pixel size, it can be clearly determined that the quantum efficiency of BSI-2 has nearly 10% improvement over BSI-1. This represents that the BSI-2 process can increase the absolute photosensitivity of the CMOS image sensor chip by 10%.
Fig. 3 A CMOS image sensor manufacturer used wafer-level quantum efficiency spectra to analyze the impact of different processes on the photosensitivity of CMOS image sensor chips during TSMC 65nm technology mass production.
In addition, the quantum efficiency spectrum is an important tool for optimizing CMOS image sensor chip manufacturing. For example, when applying BSI-2 to the 1.1um pixel process, the comparison with the 1.4um pixel shows that in terms of blue pixel, BSI-2 can provide higher photosensitive efficiency, while the photosensitivity of green and red pixels is similar to that of 1.4um pixels.
This result shows that under the premise of maintaining pixel size, the BSI-2 process can improve the photosensitivity of the CMOS image sensor chip, thereby improving image quality. Therefore, comparing the effects of different processes on CMOS image sensor chips using quantum efficiency spectra can provide important references for CMOS image sensor manufacturing optimization.
(1) What is Optical Crosstalk?
Optical crosstalk of CMOS image sensors refers to the phenomenon where light traveling in the image chip causes interference between adjacent pixels due to refraction, reflection and other reasons.
Fig. 4 Optical Crosstalk
(2) Why is Optical Crosstalk inspection important?
In CMOS image sensor chips, optical crosstalk is an important issue because it affects image quality and accuracy. Optical crosstalk is generated due to the optical interaction between pixels, causing interference between the optical signals of adjacent pixels, which in turn affects the distinguishability and contrast between pixels. Therefore, reducing optical crosstalk is one of the important goals for improving CMOS image sensor chip quality.
(3) How to use quantum efficiency spectrum to detect optical crosstalk of CMOS image sensors?
Quantum efficiency spectra can be used to detect crosstalk issues in CMOS image sensors. When there is a crosstalk problem in the CMOS image sensor, abnormal quantum efficiency may be observed at certain wavelengths. In this case, appropriate measures can be taken to reduce crosstalk, such as optimizing the CMOS image sensor design or improving the process.
Scaling down pixel size is absolutely necessary for high-resolution imaging and quantum image sensors.
As shown in Fig. 4 above, TSMC used advanced 45nm CMOS technology to produce 0.9um pixel stacked CMOS image sensors. Optical crosstalk has significant effects on signal-to-noise ratio (SNR) and imaging quality.
Therefore, TSMC adopted a pixel technology to improve this optical crosstalk. The structure is shown in Fig. 5 below.
Fig. 5 Cross-sectional schematic diagrams of a pixel (a) control pixel; (b) crosstalk suppression pixel.
Structure (a) is the control pixel. The light path is ML (Microlens), CF (Color Filter), PD (Photodiode, photosensitive layer). The schematic diagram of the influence of optical crosstalk is shown by the green line trajectory. After photons from the adjacent pixel unit enter, due to the refraction of the multilayer structure, they are incident on the intermediate PD photosensitive region, causing crosstalk signals. TSMC designed the structure (b) “Deep Trench Isolation (DTI)” technology to suppress optical crosstalk without sacrificing dark performance in parallel. It can be seen from figure (b) that the trench formed by DTI can isolate the photons that would have caused optical crosstalk from being incident on the intermediate photosensitive photodiode region, suppressing crosstalk and improving SNR.
Fig. 6 This figure shows the quantum efficiency spectra of 0.9um pixels, where the dashed line represents the control 0.9um pixel (a), and the solid line represents the improved 0.9um pixel (b). Due to the slight change in the optical aperture area of the grid structure, optical crosstalk is greatly suppressed. Direct evidence of optical crosstalk suppression is reflected in the quantum efficiency spectrum. The three yellow arrows indicate evidence of crosstalk suppression in the R, G, and B channels. The blue and red channels show a slight decrease, but the quantum efficiency of the green channel is improved through the newly developed color filter materials. (Reference: tsmc CIS)
Wafer-level quantum efficiency spectrum technology can directly demonstrate the suppression of optical crosstalk. For different CMOS image sensor chips, quantum efficiency spectrum testing can be used to compare their quantum efficiencies at different wavelengths to distinguish whether optical crosstalk is suppressed.
(1) What is the Color filter in CIS?
The Color filter in CIS is an optical filter used for CMOS image sensors. It is used to adjust the spectral response of each pixel in the image sensor so that the CMOS image sensor can sense and separate different colors of light and convert them into digital signals. Color filters usually include three basic color filters – red, green and blue. As for the various color filter arrays (CFAs) formed by different filter arrangements, please refer to the information below. The most common CFA is the Bayer filter arrangement, where each unit has one B, one R, and two G filters arranged.
Color filters play a very important role in CMOS image sensors, and their quality directly affects the color reproduction effect of the image. To ensure the performance of the Color filter meets the design requirements, precise spectral analysis and quality inspection are needed. The transmittance spectrum can evaluate the optical performance of different Color filters; the quantum efficiency spectrum can detect the matching degree between the Color filter and the photodiode. Only through strict quality inspection can ensure that the CIS chip outputs high-quality images.
Fig. 7 How Color filters are combined in the “Pixel” sensor. A pixel unit consists of three main components: Micro Lens + CFA + Photodiode.
The main function of the Color filter is to decompose the incident white light into different color lights, and selectively block certain color lights, thereby achieving selective photosensitivity for light of different wavelengths.
(2) Why is Color filter inspection important?
In CMOS image sensors, each pixel has a color filter to selectively sense RGB three color lights, thereby realizing the capture and processing of color images. If the performance of the color filter is poor, it will affect the photosensitivity and spectral response of the pixels, thus affecting the image quality and accuracy. Therefore, optimizing the performance of the color filter is crucial to improving the quality of CMOS image sensors.
Inspecting the color filter is very important because the quality and stability of the color filter will directly affect the color accuracy and contrast of the CMOS image sensor, and thus affect the overall image quality and clarity. If there are defects or non-uniformities in the color filter, it will lead to problems such as offset, distortion, and uneven color in some colors in the image. Therefore, strict inspection of the color filter can help manufacturers ensure that its performance and quality meet the design requirements, thereby improving the production efficiency and reliability of the CIS image chip.
(3) How to use QE spectrum to inspect Color filter quality of CIS?
The Color filter of CIS is usually made of a substance called “organic dye or pigment”. These organic dyes can selectively absorb light of specific wavelengths to produce the desired color filtering effect. These organic dyes are usually deposited on glass or silicon substrates by coating techniques to form color filters.
The quantum efficiency (QE) spectrum can measure the photosensitivity of CIS at different wavelengths, thereby determining the quality and performance of the Color filter. Under normal circumstances, the Color filter should be able to properly separate lights of different wavelengths and produce less crosstalk during the optical process. Therefore, if the quantum efficiency at a specific wavelength is lower than the expected value, it may be caused by Color filter quality or performance issues. By analyzing the quantum efficiency (QE) spectrum, it can be determined whether the performance of the Color filter meets the design requirements, and corresponding adjustments and optimizations can be made in advance.
Fig. 8 TSMC uses Wafer Level Quantum Efficiency Spectrum technology to inspect different green filter materials to evaluate their effects on photosensitivity and optical crosstalk of CMOS image sensors.
As shown in the figure above, TSMC’s CIS process uses the spectral technology of Wafer Level Quantum Efficiency Spectrum to inspect different green filter materials to evaluate their effects on the photosensitivity and optical crosstalk of CMOS image sensors. The wafer-level quantum efficiency spectrum shows the characteristics of three different Color filter materials (Green_1, Green_2 and Green_3).
By comparing these three materials, it can be found:
(1) The main green peak is shifted to 550nm
(2) The optical crosstalk phenomenon between the green and blue channels is significantly reduced
(3) The optical crosstalk phenomenon between the green and red channels is significantly increased.
By analyzing the quantum efficiency (QE) spectrum, it can be determined whether the performance of the Color filter meets the design requirements, and corresponding adjustments and optimizations can be made in advance. This ensures that the characteristics of the filter materials meet the design requirements and guarantees image quality and accuracy, improving the reliability and stability of CMOS image sensors.
(1) What is Si wafer thickness control?
When manufacturing BSI CMOS image sensors, a process called “thinning down” is used to make the wafer thinner. The wafer thickness after thinning directly affects the photosensitivity of the CIS chip, so the wafer thickness has a great influence on the photosensitive performance and quality of the image chip.
In order to ensure that the image chip can work properly, we need to use the “Si wafer thickness control” process to accurately control the wafer thickness. This can ensure that the thinned wafer thickness meets the design requirements, and also improve the product yield of the image chip.
Fig. 9 Process flow of BSI. CMOS image sensors manufactured using the BSI process have an important process of “thinning down”, which is to reduce the wafer thickness to a certain extent.
(2) Quantum efficiency detection is very important in Si wafer thickness control process monitoring
In the manufacture of CMOS image sensors, the control of the Si wafer thickness control process has a direct impact on the photosensitivity of the sensor. This impact can be observed through the quantum efficiency spectrum to ensure that the thinned CMOS image sensor has the optimal photon-to-electron conversion quantum efficiency. The thinned wafer will have an optimal thickness value that can ensure the CMOS image sensor has the optimal photoelectric conversion quantum efficiency. The quantum efficiencies of the red, green, and blue channels can be tested using wavelengths of 450nm, 530nm, and 600nm. The experimental results showed the changes in the quantum efficiency values of the red, green, and blue channels of the CMOS image sensor at different thinning thicknesses. Deviations in the thinning thickness have a direct effect on the photosensitivity of the CMOS image sensor, which in turn affects the value of the quantum efficiency. Therefore, quantum efficiency detection is crucial for monitoring the Si wafer thickness control process to ensure stable and consistent quality of the manufactured CMOS image sensors.
Fig. 9 shows the changes in the quantum efficiency values of the blue, green, and red channels of the CMOS image sensor at different thinning thicknesses. The quantum efficiency value of the blue channel was measured using a wavelength of 450nm. When the thickness after thinning was 0.3um thicker than the standard thickness, its quantum efficiency value decreased from 52% to 49%; when the thickness after thinning was 0.3um less than the standard thickness, the quantum efficiency of the blue channel was only slightly lower than 52%. The quantum efficiency value of the red channel was measured using 600nm wavelength. It was found that the performance of the red channel was opposite to that of the blue channel at different thicknesses. When the thickness after thinning was 0.3um less than the standard thickness, the quantum efficiency of the red channel decreased significantly from 44% to 41%. Under thicker conditions (+0.3um), the quantum efficiency of the red channel did not change significantly. The quantum efficiency value of the green channel was measured at 530nm wavelength. Under the three thickness conditions (STD THK ± 0.3um), the quantum efficiency of the green channel did not change significantly.
Fig. 10 CMOS image sensor quantum efficiencies were tested using different Si wafer thicknesses (THK) of 600nm, 530nm and 450nm, and the quantum efficiencies of the red, green and blue channels were evaluated.
The results show that for the green channel, within the ±0.3um range of Si wafer thickness variation, the quantum efficiency at 530nm did not change significantly. However, for the red channel, as the Si wafer thickness decreased, the quantum efficiency decreased significantly. In the case of 450nm blue channel, the quantum efficiency decreased significantly as the Si wafer thickness decreased. These results indicate that the Si wafer thickness has an important influence on the quantum efficiency of CMOS image sensors, and the extent of the influence varies for different channels. Therefore, precise control of the Si wafer thickness is required in the manufacture of CMOS image sensors to ensure product quality and performance.