Heavy Ion Single Event Effects in CMOS Image Sensors: SET and SEU

Abstract

High-energy particles in space often induce single event effects in CMOS image sensors, resulting in performance degradation and functional failure. This paper focuses on the formation and morphology of transient bright spots in CMOS image sensors and analyzes the formation process of transient bright spots by conducting heavy ion irradiation experiments to obtain the variation law of transient bright spots with heavy ion linear energy transfer values and background gray values; in addition, we classify the single event upset that occurred in the experiments according to the state of transient bright spots and extract the characteristics of different single event upsets. The failure mechanisms of different single event upsets are analyzed according to their characteristics and are combined with the information given by transient bright spots. This provides an essential reference for rapidly evaluating single event effects and the reinforcement design of CMOS image sensors.

1. Introduction

Complementary metal oxide semiconductor (CMOS) image sensors (CISs) are widely used in spacecraft imaging systems such as remote sensing imaging and star-sensitive vehicles due to their excellent system performance in terms of power consumption, size, and quality [1,2]. However, the high-energy particle radiation environment in space, such as high-energy protons and heavy ions, can induce single event effects (SEEs) in CISs, which can affect the performance of CISs and even cause functional failure [3,4,5].

Due to the frequent occurrence of SEEs in CISs in the space environment, a series of studies have been carried out in China and abroad about single event transient (SET) and single event upset (SEU) [6,7,8,9,10,11,12,13,14]. Hopkinson et al. found SET in the pixel arrays during heavy ion evaluation of the radiation-resistant STAR-250, which appear as transient white bright spots that disappear in the following image [6]. Lalucaa et al. investigated the charge collection process of bright spots by conducting heavy ion irradiation experiments and discussed the effect of blooming, whereby the additional charges of the saturated diode diffuse into neighboring diodes [7]. Yang et al. discovered that the distribution of the total collected charge of each bright spot could be well-fitted by Landau distribution [8]. SEU is a single high-energy particle incident on a semiconductor device that causes a flip in the logic state of that sensitive unit [9]. Beaumel et al. conducted heavy ion irradiation experiments on the HAS2 CIS, which illustrated potential SEU-sensitive cells in the readout circuit of this CIS [10]. Virmontois et al. tested more than 30 registers in the readout circuit in a heavy ion experiment. They found that not all registers triggered image anomalies, and only a few registers (e.g., gain or integration time registers) corrupted the image [11].

The main focus of the current study is on the description and failure explanation of single transient phenomena (SET or SEU). However, SETs and SEUs often occur simultaneously in the experiment, but these two anomalies have yet to be linked and analyzed simultaneously in previous studies. In this paper, by analyzing the morphological size of the transient bright spots, the SEU phenomena that appeared in the experiment are classified into bright spots that disappeared, were unaffected, and were affected. The failure mechanism of the SEU is analyzed accordingly, and the anomalous phenomena that appeared in this experiment are elaborated on and classified, providing an experimental basis and a theoretical foundation for the systematic study of SEUs in CISs.

2. Samples and Irradiation Conditions

The sample in this experiment is a four-transistor active pixel sensor (4T-APS) with a resolution of 2048 × 2048 and a pixel size of 5.5 μm × 5.5 μm. Figure 1 shows the block diagram of the selected CIS, whose readout circuit includes the analog front-end, addressing circuit, Serial Peripheral Interface (SPI) registers, and LVDS block. The driver board has a 10-bit pixel depth and eight data output channels.

The experiment setup is shown in Figure 2a. The CIS is mounted on a test board and fixed to the irradiation board, which is moved to the particle beam irradiation position by laser positioning and guide rails during the experiment. The rest of the test board was protected with a shield layer, except for the exposed irradiated CIS. The CIS is remotely controlled by a signal line, such as a camera-link cable, and the online test operation is performed outside the irradiation room. During the test, the CIS was in global exposure mode with a frame rate of 180 frames/s and an exposure time set to 1000 lines. Figure 2b shows the schematic diagram of 4T-APS, whose pixel unit mainly consists of a pinned photodiode (PPD), a reset transistor (Trst), a source follower (SF) transistor, a selector transistor (Tsel), a transmission gate (TG) transistor, and a floating diffusion (FD) area.

The heavy ion irradiation experiments were conducted at the HI-13 Tandem Accelerator at the China Institute of Atomic Energy Science [15]. The encapsulated optical glass window was removed in advance. The heavy ions were incident vertically, and the heavy ion beam spot was 30 mm × 30 mm, which could simultaneously cover the entire surface of the CIS chip.

Table 1. Experimental heavy ion types and energies.

Accelerator	Ion Species	Initial Energy (MeV)	LET (MeV cm2 mg−1)	Range (μm (Si))
HI-13	16O	100	3.01	95.2
	28Si	135	9.3	50.7
	35Cl	150	13.4	42.8
	48Ti	160	22.2	32.9
	74Ge	205	37.37	29.95

shows the ion species, range, and linear energy transfer (LET) values used in the experiment.

3. Results and Discussion

3.1. SET Bright Spots

The SET bright spot is a transient phenomenon unique to optoelectronic imaging devices. It manifests as a bright white spot at the particle incidence location of the acquired image, and the bright spot at that location disappears in the following image.

In order to understand the formation process of the SET bright spot, it is necessary first to understand the signal readout process of the pixel, as shown in Figure 3. Before exposure, Trst and TG are on to reset the PPD so that the PPD is depleted and in reverse bias; during exposure, Trst and TG are off and the light signal is irradiated at the PPD to generate a photogenerated charge, which is collected by the depletion zone; after a period of integration, Trst is turned on to reset the FD to clear the residual charge before the photogenerated charge is transferred to the FD via TG; immediately afterwards, TG is turned on and the photogenerated charge accumulated on PD is transferred to FD, and TG is turned off after the charge transfer is completed; then, the photogenerated charge transferred to FD is converted into a voltage by the parasitic capacitor of FD, which is amplified by SF and output to the column output bus; the output reset voltage and the optical signal voltage are passed through the CDS circuit to eliminate the noise, and then through the amplifier for signal amplification; the amplified voltage signal is turned into a digital signal by the AD converter, which is processed by a specific image signal inside the sensor and finally output to the external [16].

The appearance of SET bright spots occurs during the exposure period. The electron-hole pairs generated by heavy ions in the sensitive layer of the sample are collected by the photodiode depletion region, causing a change in the potential in the PPD region, followed by a readout of the changed potential through the transistor, which is expressed in the image as a bright spot with a gray value greater than the background value. One particle incidence changes the current potential in the PPD. After the current potential change in the PPD is read out, the potential in the PPD is reset to a high level before the subsequent integration. Therefore, the bright spots in the current image will disappear in the following image.

As shown in Figure 4, the SET bright spot is not a dot, but a circular-like cluster formed at the center of the incident point, with the maximum gray value at the center and the decreasing gray value at the edges as the distance from the center increases. The reason is that the neighbor pixels in the pixel array are interconnected and isolated only by a shallow trench (STI). Therefore, the electron–hole pairs incident on the particle traces in the sensitive layer of a pixel, in addition to being collected by the PPD of that pixel unit, will also move to neighbor pixels by drifting and diffusion and thus appear on the image as a bright spot instead of a dot. A single-pixel unit charge collection sensitive body is a region that collects charge into an integrating capacitor. The sensitive body of the photodetection region consists of the depletion region of the p-n junction and the epitaxial layer part. The PPD collects the electron–hole pairs generated in the depletion layer by drifting under the action of the electric field. The electron–hole pairs generated in the epitaxial layer outside the depletion region are partially collected by random diffusion to the adjacent PPD through thermal diffusion, and the rest are compounded or captured. Moreover, at the edge of the bright spot, due to the blooming effect, the diffusion of charges is not collected in the saturated photodiode into the neighboring photodiodes.

3.2. Effect of Different Conditions on SET Bright Spots

Previous works have described the SET bright spot size concerning the conditions of integration time, process, and incident angle [8,9,10], and this paper focuses on the size of the SET bright spot concerning the LET value of the incident heavy ions and the background gray value.

The collected charge increment and size are two critical characteristic parameters of SET bright spots. The collection charge increment is obtained by subtracting the background gray value from the total gray value of all covered pixel cells of the SET bright spot. The size refers to the total number of covered pixels. During the test, the collected image signal value is the gray value, and its unit is DN (digital number), which indicates the digital signal value obtained directly by AD conversion.

$$\Delta N_e=\left({\mu }_{DN}-{\mu }_{DN.dark}\right)/CVG$$

(1)

The formula $\Delta N_e$ denotes the collected charge increment, ${\mu }_{DN.dark}$ denotes the total gray value of the background, and ${\mu }_{DN}$ denotes the total gray value of an SET bright spot. CVG (charge voltage gain) indicates the gain of the charge collected in the photodiode into a digital signal through the readout circuit.

3.2.1. LET Value

To reasonably characterize this energy transfer, the physical quantity LET (linear energy transfer) is introduced [17], which is the energy transferred per unit distance of the incident particle in the target material,

$$\mathrm{LET}=-\frac{1}{\rho }\mathrm{dE}/\mathrm{dx}$$

(2)

where $\rho$ is the density of the incident material, E is the energy, and x is the transmission distance. Different LET values were obtained by changing the heavy ion species in the experiment. In Si materials, the ionization energy of Si, the energy required for an electron outside the nucleus to leave the nucleus of a silicon atom to be a free electron, $E_{Si}$= 3.6 eV/pair, so that the number of electrons and holes produced by an incident particle per unit path is expressed as:

$$\frac{d{N}_e}{dx}=\frac{d{N}_h}{dx}=\frac{1}{E_{Si}}\frac{dE}{dx}=LET·\frac{\rho }{3.6\mathrm{e}\mathrm{V}}$$

(3)

N_e and N_h are the numbers of electrons and holes produced by ionization, respectively. From Equations (2) and (3), it can be seen that as the value of LET is larger, more electron–hole pairs are generated per unit distance.

As shown in Figure 4, the size of the SET bright spot increases as the LET value increases. Figure 5 shows that the size and charge collected by the SET bright spot increase rapidly and then slowly with the LET value of the heavy ions. The more extensive the LET value, the larger the energy loss per unit distance of the incident heavy ions. Therefore, the larger the number of charges generated, the increasing size and charge collected by the SET bright spot. However, it eventually saturates because it is limited by the carriers’ diffusion length and the pixel’s full well charge (FWC).

3.2.2. Background Gray Value

Figure 6 and Figure 7 show the variation of the bright spot size and the collected charge for different background gray values under ³⁵Cl irradiation. It can be seen that the size of the SET bright spot increases with the increase of the background gray value in a nearly proportional relationship. In contrast, the total collected charge increases first and then decreases.

By the Fick law of diffusion [18]:

$${\left.\frac{d\Delta n}{dt}\right|}_x=-D_n\frac{d\Delta n}{dx}$$

(4)

$$L_d=\sqrt{D\tau }$$

(5)

For large injections, the carrier lifetime is:

$${\mathsf{\tau }}_{\mathrm{n}}={\mathsf{\tau }}_{\mathrm{p}}=\frac{1}{\left(G_{th}/n_i^2\right)·\Delta n}$$

(6)

In Equations (4)–(6), ∆n is the injected electron concentration, $D_n$ is the electron diffusion coefficient,$L_d$ is the diffusion length, ${\mathsf{\tau }}_{\mathrm{n}}$ is the electron lifetime, $G_{th}$ is the heat generation rate, and $n_i$ is the intrinsic electron concentration.

Since the charge generated by the incident heavy ions are much larger than the number of photogenerated charges in the current environment, it meets the large injection condition (∆n = ∆p > n and p). As the background gray value increases, the number of charges required to saturate the pixel unit decreases, and the excess charge density becomes larger. Since the diffusion coefficient D is proportional to the concentration gradient, D becomes larger; since ∆n remains constant, the carrier lifetime remains almost unchanged. In summary, the diffusion length becomes larger. Therefore, the bright spot size keeps increasing with the background gray value.

The total collected charge is affected by the amount of charge generated by the incident heavy ions and the lifetime of a few carriers. When the background gray value is small, it is limited by the diffusion length, resulting in the diffused charge being compounded and challenging to be collected effectively; when the background gray value is large enough, the saturation level of the readout circuit is smaller than the saturation level of the PPD, resulting in the calculated charge on the image being smaller than the actual amount, so the total number of collected charges increases and then decreases with the background gray value.

3.3. Classification of SEU Phenomenon by SET Bright Spot

As the most common type of SEE, SET bright spots often coexist with other anomaly phenomena, such as SEUs. Since the features of SET bright spots are easy to extract and the morphology of transient bright spots is correlated with readout circuits, evaluating the SEU events that occur in experiments is meaningful.

According to the description above, the transient white bright spots appearing on the image are due to electron–hole pairs generated by the ionization of heavy ions across the pixel units. The pixel unit collects and spreads these charges to form circular bright spots of similar size. Its size is related to the exposure time, background brightness, and heavy ion LET value. Several SEUs, such as the row, column, and output anomalies, were seen during the experiment. We classify some of the SEU phenomena according to the morphological size of the transient bright spots on the images and discuss each type of SEU failure mechanism based on the bright spot information.

3.3.1. Disappeared Bright Spot

The disappeared bright spot means that the transient bright spots in the image disappear and can be restored only after power-off and restart. As shown in Figure 8, pixel outputs on the image were from 135 DN (gray value) to zero, and all bright spots disappeared in three consecutive pictures.

The global disappearance of the transient bright spot in the figure and the fact that the outputs were zero indicated an output anomaly. The reason may arise from the process of pixel cell signal readout. According to the process of transient change in Figure 8b, it can be seen that the row readout mainly causes the anomaly. Hence, this phenomenon is induced by heavy ion bombardment of the row address decoder within the CIS. When a row of pixels in the array is selected, the row address decoder can control the pixel operation of the row by controlling the selection of the transfer gate, reset transistor, and selector transistor. However, when a heavy ion hits the row address decoder and SEU occurs, it may cause the transfer gate, reset transistor, and selector transistor to be generally on or off. Therefore, the pixels of that row each output the same signal during the signal sampling and reset phases. After the associated double-sampling process, the output value after subtracting the two is zero. Therefore, no transient bright spots appear on these images, and the global gray value is zero.

3.3.2. Unaffected Bright Spot

The unaffected bright spot means that after the SEU phenomenon, the size and shape of the bright spot in the image remain the same as usual, and there is no change. As shown in Figure 9, the global gray value of the image dropped from the original 170 to about 30, and the bright spots did not disappear. Comparing the bright spots in the before and after images, as shown in Figure 9c,d, it can be found that the size of the bright spots does not change, having sixty saturated pixels. Since the size of the transient bright spot is in connection with the average gray value of the background, but the figure shows the parallel size of the bright spot, it means that the cause of the abnormality has nothing to do with the row or column readout circuit. Therefore, the gray value of the background dropped due to a flip in one of the registers in the analog front end, causing a change in the global gray value, not due to external light leakage or other reasons.

According to the failure characteristics, this global gray value drop was caused by the failure of the offset register. The function of the offset register in CISs is to calibrate the output to compensate for the output signal dark level value [19], as shown in Equation (7).

Dark-level output = 70 + setting − 16,383

(7)

When heavy ions bombard the register and SEU occurs, the offset register setting will no longer be the correct value. When the offset register setting value decreases from the original 16,323, it causes the gray values of all pixels to decrease, as shown in Figure 9. There are still bright spots on them because heavy ions generate a large amount of ionized charge on the incident traces to be collected, increasing the pixel output dark level. Because enough charge is collected, the pixel output reaches 1023 DN (maximum value) at the bright spot location. The ionizing charge formed by a single heavy ion can affect multiple pixels near the incident location by diffusion, drift, and the funnel effect. Hence, the bright spot still exists under this anomaly.

3.3.3. Affected Bright Spot

The affected bright spot means that the size of the bright spot in the image is affected, mainly as the bright spot becomes larger or smaller. The impact on the bright spot mainly occurs in the row and column anomalies, as shown in Figure 10; we can observe two white vertical stripes with a constant gray value of 934 DN, much higher than the average gray value of 159 DN in the background. The positions of the two anomalous vertical stripes are X127 and X255. We can get more information through the transient bright spot situation, such as the misalignment problem at the anomalous column. The X127 column is down one wrong, and the X255 column is up one wrong, precisely equivalent to the two columns in the middle of the data line signal value read out in the following line. Secondly, comparing the bright spot size, it is found that the cut bright spots on the left and right add up to the same size as the typical bright spot, which indicates that the X127 column and X255 column data are invalid data, and the value is locked. Zooming in at Y = 0, as shown in Figure 11, verifies that the signal value in the middle of the abnormal column is panned down one row.

With the abnormal characteristics described above, it can be inferred that the abnormality is due to an abnormality in the signal output. This CIS has 18 LVDS output channels, including 16-pixel output channels, one clock channel, and one control channel. The driver board used in the test had eight output channels. The anomalous column in Figure 4 is located in the last of the two parallel segments of 128 pixels on LVDS channel one in columns 127 and 255. This anomaly may be due to SEU occurring on LVDS channel one during pixel data transfer, causing an anomaly in the row addressing register and resulting in a misalignment of the output values between them.

Figure 12 illustrates one type of row anomaly; the gray value on the anomalous row (90 DN) is smaller than the regular row (125 DN), as shown in Figure 12b. The transient bright spot in the anomalous row is smaller than the bright spot on the normal background outside the row, caused by the lower gray value of the background.

This anomaly is similar to one SEU phenomenon reported by Beaumel in 2013 [10], which is thought to be a flip in the row reset register or row read register. The read or reset register pointer jumped to a non-corresponding line, resulting in visible disturbances in the integration time of some lines on the image. As a result, the gray values at the disturbed rows are abnormal, as shown in Figure 12.

In conclusion, a summary of the classification of bright spots in SEU phenomena is shown in

Table 2. The classification of SEU events with the bright spots.

Category	Characteristic	SEU Phenomenon	Failure Localization
Disappeared	Bright spot which disappeared	Global gray value drop to 0	Row address decoder
Unaffected	No change in bright spot size	Global gray value drop	Offset register
Affected	Change in bright spot size	Column anomalies, row anomalies	LVDS block,
			row reset register, or row read register

.

4. Conclusions

This paper investigated the relationship between the variation of SET bright spots with different experimental conditions by conducting CIS heavy ion irradiation experiments. The size of the SET bright spot increases with the heavy ion LET value and then tends to saturate; it gradually increases with the increase of the background gray value. Meanwhile, a variety of SEU events were observed in the experiment, and the SEU events were classified into three categories: disappeared bright spot, affected bright spot, and unaffected bright spot, by feature extraction of the SET bright spots in the SEU events. The above classification narrowed down the failure location of SEU. The disappearance of the bright spot indicates a failure at the readout circuit of the CIS; the unaffected bright spot indicates a failure at the analog front-end; and the change of the bright spot indicates an abnormality in the readout circuit of the CIS or the LVDS block.

In summary, this paper proposes to identify and classify SEUs using SET bright spot characteristics and establishes a fast identification method to analyze SEU patterns and sensitive areas based on transient bright spot size, background gray value, and other parameters. It provides an essential reference for the rapid evaluation and reinforcement design of the SEE of CISs. In the future, this analysis method can be combined with machine learning methods to collect relevant SEE experiment data, which can be used for online identification analysis and failure localization of SEEs in optical imaging devices.