1. Introduction
The Complementary Metal-Oxide-Semiconductor (CMOS) is a solid-state imaging device that converts optical signals into electrical signals through a photodiode array [
1]. Compared to Charge Coupled Devices (CCDs), CMOS devices reduce system complexity and power consumption while offering high quantum efficiency, wide dynamic response, and superior anti-crosstalk performance [
2]. These advantages have led to CMOS devices gradually replacing CCDs in applications such as intelligent surveillance [
3], space remote sensing [
4], and automotive applications [
5]. However, the multilayer structure of CMOS devices, composed of metal and silicon materials, is susceptible to damage from high-intensity laser irradiation. This can cause signal distortion, structural damage, and interlayer delamination [
6]. Research on the mechanisms of laser-induced damage in CMOS devices is essential for designing robust imaging systems and developing laser protection strategies for photodetectors.
Researchers have conducted experimental studies to investigate the impact of different laser parameters on the damage characteristics of the detector. Wang et al. compared continuous-wave (CW) lasers with single-pulse nanosecond lasers and found that pulsed lasers had a significantly lower energy density threshold for CMOS damage [
7]. Schwarz et al. used the 1-on-1 test method to measure damage thresholds of picosecond and nanosecond pulsed lasers on CMOS cameras, with picosecond lasers showing a lower damage threshold [
8,
9]. Westgate et al. demonstrated that back-side illuminated (BSI) CMOS requires two orders of magnitude higher laser energy density for complete damage than front-side illuminated (FSI) CMOS, due to its circuitry placement beneath the photosensitive layer [
10]. Cao et al. demonstrated that combined nanosecond/picosecond laser irradiation induced damage to CCD detectors, achieving a complete damage threshold of 103 mJ/cm
2—significantly lower than thresholds observed under single nanosecond or picosecond laser pulses [
11]. Yoon et al. found that CMOS damage from CW near-infrared laser irradiation had reversible interference at low energy densities and irreversible damage at higher energy densities [
12]. Théberge et al. used different CW lasers to damage silicon-based cameras, categorizing damage into three stages based on pixel response degradation, circuit layer melting, and sensor failure [
13]. Wang et al. studied CMOS devices with mid-infrared lasers, observing saturation, black screen, and other damage phenomena [
14]. Xu et al. conducted CW laser experiments on BSI CMOS devices, observing various types of damage and determining power density thresholds [
15]. Zhang et al. investigated CMOS device damage from CW and pulsed laser irradiation, finding different damage mechanisms for lens damage, point damage, line damage, and stress damage [
16]. Han et al. demonstrated that irradiation of a CCD detector with a 1.06 μm continuous laser induces a multi-stage progression of damage across its multilayer structure, characterized by sequential phases of microlens melting, aluminum film delamination, silicon electrode fusion, and destruction of the N-Si substrate layer [
17]. Niu et al. achieved millisecond-level online evaluation of laser-induced damage in CCD detectors by fusing features from three data sources (cat’s eye echo signals, plasma flash data, and surface image characteristics) and using a probabilistic neural network algorithm [
18].
In numerical simulations, there are currently few referenceable theoretical models for laser-irradiated CMOS sensors [
19,
20]. Dai et al. created a thermomechanical model for thermal damage in CMOS detectors, showing that mechanical damage mainly occurs at the photosensitive surface [
19]. Qian et al. conducted thermodynamic simulations on CMOS damage under pulsed lasers with different pulse widths and wavelengths. Combined with experimental data, they found that the damage threshold of CMOS under 532 nm pulsed lasers was significantly lower than that under 1064 nm lasers [
20]. Numerical methods developed for CCDs [
21,
22,
23,
24,
25], another class of imaging sensors, have also provided critical insights for CMOS laser damage studies. Li et al. simulated the process of millisecond pulse laser irradiation on Si-CCD by establishing a thermal-stress coupling model. They found that damage originates in the color filter layer. As energy density increases, this progressively leads to microlens loss, melting of the photosensitive area, and ultimately functional failure due to channel damage in the N-Si layer [
21]. Ren et al. simulated thermal damage in CCDs under repetitive pulsed laser irradiation, showing temperature accumulation and staged damage progression [
22]. Han et al. simulated CCD damage under CW laser irradiation, calculating time thresholds for different damage stages [
23]. Kou et al. confirmed that electrode short circuits from insulation layer ablation were the main failure mechanism in CCDs [
24].
Although significant progress has been made in understanding the damage mechanisms of the CMOS under laser irradiation through experimental studies, numerical simulations still require further exploration to characterize device damage characteristics comprehensively. Current two-dimensional thermomechanical coupling models struggle to accurately represent the interaction between lasers and the CMOS, often neglecting the influence of detector fill factor on thermal damage processes and lacking a detailed analysis of multilayer structural damage progression. Given this background, this study aims to investigate the thermal damage characteristics of the FSI CMOS detector under 532 nm nanosecond pulsed laser irradiation. Experimental analysis was conducted by output images, optical microscopy, and scanning electron microscopy (SEM) observations to understand the damage mechanisms at different stages under varying laser energy densities. Numerical simulations were also performed using a three-dimensional thermal–mechanical coupling model of the CMOS multilayer structure, incorporating internal heat conduction, material phase transitions, and thermal expansion to simulate energy deposition, heat diffusion, and mechanical responses. The study seeks to improve the understanding of nanosecond pulsed laser interactions with CMOS detectors.
2. Theoretical Model
The CMOS detector used in the experiment is the NVC-50-M-CL model, which is equipped with an FSI CMOS device (GMAX0505), as shown in
Figure 1. The device has a resolution of 1280 × 1024 and a pixel size of 5 µm × 5 µm. As depicted in
Figure 2a, the multilayer structure of the FSI CMOS along the incident light path includes the following layers from top to bottom: a 2 µm thick microlens made of polyimide for focusing light, a 1 µm thick silicon dioxide buffer layer to reduce thermal stress, a 0.5 µm thick silicon nitride passivation layer for corrosion protection, a 1 µm thick metal routing layer primarily made of aluminum for electrical signal transmission, a 0.5 µm thick silicon dioxide insulation layer for electrical isolation, and a 25 µm thick epitaxial layer containing silicon photodiodes for converting optical signals to electrical signals.
As shown in
Figure 3, a three-dimensional finite element model was developed for the CMOS structure. As shown in
Figure 2b, for the FSI CMOS structure, incident light must pass through the openings in the metal routing layer to reach the photosensitive surface of the epitaxial layer, with partial light being blocked by the metal routing. Therefore, the model considered the fill factor, which is the ratio of the effective photosensitive area to the total pixel area. The experimental CMOS detector has a fill factor of approximately 70%. The model consists of 25-pixel units, each sized 5 µm × 5 µm.
The laser irradiation process on CMOS devices involves the interaction of optical, thermal, mechanical, and electrical fields. This study focuses on analyzing thermal damage effects induced by lasers, making the following assumptions for finite element calculations: (1) Simplified to pure thermal conduction processes, ignoring non-thermal mechanisms. (2) Laser energy distribution follows a Gaussian spatial pattern. (3) Material properties remain constant with temperature variations. (4) All layers are considered continuous, isotropic media with uniform properties.
Upon reaching the CMOS surface, the laser beam’s energy is converted into a heat source with specific temporal and spatial distributions. This leads to the generation of a transient heat flow field within the CMOS multilayer structure, a process that is modeled using Fourier’s heat conduction equation [
23].
where
T is the instant temperature,
Q is the laser heat source,
,
c, and
k are the material density, specific heat, and heat conductivity, respectively.
When the CMOS is exposed to the nanosecond pulsed laser, the polyimide, silicon dioxide, and silicon nitride do not absorb laser heat directly. Therefore, most of the laser beam will directly illuminate the photosensitive surface of the epitaxial layer. Since silicon can absorb laser energy volumetrically, the laser heat source in this layer is considered a volumetric heat source [
24]. The corresponding laser energy deposition function is expressed as follows:
Given the fill factor of the CMOS, a portion of the laser irradiates the metal routing layer. As the thickness of the metal routing layer in our model greatly surpasses its laser absorption depth (10
−7–10
−8 m
−1), the laser heat source in this layer is classified as a surface heat source [
22]. The corresponding laser energy deposition function is presented as
where
represents the peak power of the laser, and the expression is
In Equations (2) and (3), is the radius of the laser spot, t represents time, corresponds to the absorption coefficient of silicon for the laser (α ≈ 7500 cm−1), R is the reflectivity of the photosensitive region in the epitaxial layer (R ≈ 0.3), is the absorptivity of aluminum to the laser (β ≈ 8%), is the single pulse energy and is the pulse width.
The temporal distribution of the pulsed laser can be expressed as follows:
The corresponding boundary conditions for convective heat exchange between the CMOS surface and ambient environment are as follows:
where
h represents the convective heat transfer coefficient,
is the CMOS surface temperature, and
is the ambient temperature set at 298 K.
When laser irradiation begins, the initial temperature of the model is set to 298 K.
During laser irradiation, a transient and non-uniform temperature field distribution develops across CMOS layers, with spatial temperature variations inducing distinct deformation and thermal stress in each layer. This physical process can be described by the thermoelastic equilibrium equation [
21,
22].
Here, , , and represent the displacement components in the three orthogonal directions, while , , and are the material’s Poisson’s ratio, volumetric strain, and thermal expansion coefficient, respectively.
The stress components in three orthogonal directions are:
The stiffness matrix components
C11,
C12, and
C13 can be determined using Young’s modulus [
22].
The thermodynamic properties of the materials in each layer of the CMOS are listed in
Table 1.
3. Experiments
The experimental system shown in
Figure 4 was used to investigate the damage characteristics. The laser source is a custom developed setup operating at a wavelength of 532 nm with a pulse width of 150 nanoseconds. The fundamental 1064 nm beam originates from a diode-pumped Nd: YAG laser module. This beam exits the laser cavity via mirrors M1 and M2. It then passes through a half-wave plate (HWP1) and a polarizing beam splitter (PBS) to produce P-polarized light. The polarized light undergoes frequency doubling to 532 nm using a half-wave plate (HWP2) and a KTP crystal. The resulting 532 nm laser beam is directed sequentially through dichroic mirror M3, energy attenuators, and a beam splitter, ultimately irradiating the CMOS target surface. Energy attenuators, including four absorptive neutral density filters and one continuous neutral density filter, were used to regulate the laser energy intensity. The beam splitter (with an approximately equal~1:1 splitting ratio) divides the laser beam into two paths. The transmitted beam irradiated the CMOS target surface, while the reflected beam was monitored by the energy meter (Ophir PE50-DIF-C) for real-time energy measurement. Output images from the CMOS detector were acquired by the computer. The experiments employed a single pulse operating in manual trigger mode. The displacement platform moved the CMOS target to unexposed areas after each pulse to avoid the thermal accumulation effect of laser energy. Damage stages were determined based on abnormal features in the CMOS output images, with corresponding laser energy density values recorded for each stage. All experiments were conducted under darkroom conditions to eliminate ambient light interference.
The 532 nm laser beam spot distribution was measured using a beam profiler (DataRay Beam’R2), as shown in
Figure 5. The spot exhibits a typical Gaussian intensity distribution and its beam diameter is approximately 600 μm, with the peak intensity located at the center of the beam.
Based on the pulse energy measured by the energy meter and the spot size obtained from the beam profilers, the laser energy density on target was determined.
According to the law of error propagation, the relative uncertainty of a variable raised to a power is scaled by the corresponding exponent:
The relative error of the energy density is calculated as 4.24%, where δE denotes the error of the energy meter and δR represents the error of the beam spot radius.
The determination of damage stages relied on observed abnormalities in CMOS output images. As shown in
Figure 6, three distinct damage stages were observed in the detector output images as the laser energy increased progressively: point damage, line damage, and complete failure. At a laser energy density of 78.9 mJ/cm
2 (the single-pulse energy was measured at 0.223 mJ), an irreversible white bright spot appeared in the output image (
Figure 6a), indicating localized pixel damage, and approximately 1% imaging function was lost across the device. When the energy density increased to 241.9 mJ/cm
2 (the single-pulse energy was measured at 0.684 mJ), mutually perpendicular irreversible black lines emerged along both horizontal and vertical orientations in the irradiated area (
Figure 6b), demonstrating full pixel failure along these linear regions while other areas maintained normal imaging function, and approximately 8% area of the device lost its imaging function at this time. Ultimately, when the energy density reached 2005.4 mJ/cm
2 (the single-pulse energy was measured at 5.670 mJ), the entire sensor area failed to output any valid image (
Figure 6c), marking total device failure.