Method and equipment for improving sensitivity of diode junction of infrared PD chip

Abstract

The present invention relates to the technical field of infrared PD chip design, and discloses a method and device for improving the sensitivity of the diode junction of an infrared PD chip, wherein the method comprises: performing multi-structure modeling on the PN junction of the infrared PD chip to construct a multi-structure PN junction model; calculating the mapping relationship between the carrier separation efficiency and the PN junction structure parameters; creating a carrier enhancement structure model; inputting the geometric parameters of the carrier enhancement structure model as state variables into a deep reinforcement learning model, optimizing the parameters with a photoelectric response model as a reward function, and obtaining the optimal PN junction structure parameters; performing responsiveness calculation based on the optimal PN junction structure parameters, and evaluating the sensitivity improvement index of the infrared PD chip, wherein the method is used to construct a comprehensive performance scoring system including normalized detection rate, detection probability and false alarm rate, and realizes a comprehensive evaluation of the performance of the infrared PD chip in practical applications.

Description

Technical Field

[0001] The present invention relates to the technical field of infrared PD chip design, and in particular to a method and device for improving the sensitivity of an infrared PD chip diode junction. Background Art

[0002] The performance requirements of infrared detectors are constantly increasing, especially in terms of detection capability and signal-to-noise ratio under low light conditions. Traditional PN junction structure infrared photodiode (PD) chips have been unable to meet the growing demand in terms of sensitivity and response speed. Although the use of a single structural optimization or material improvement can improve performance in some aspects, it often leads to a decline in other performance indicators, making it difficult to achieve comprehensive performance improvement.

[0003] The traditional infrared PD chip diode junction structure often shows insufficient sensitivity when facing weak signal detection. The existing infrared PD chip design methods mostly rely on experience and repeated experiments, and lack systematic optimization methods. This design method is not only time-consuming and labor-intensive, but also difficult to fully tap the potential of materials and structures. Especially in nanoscale structural design, traditional optimization methods have difficulty dealing with complex multi-physical field coupling problems, resulting in design results that are often not ideal. Existing infrared PD chip performance evaluation methods usually only focus on a single indicator and ignore the comprehensive performance in practical applications. This evaluation method cannot fully reflect the actual performance of the chip under different working conditions, especially the detection capability under low signal-to-noise ratio conditions. Summary of the invention

[0004] The present invention provides a method and device for improving the sensitivity of the diode junction of an infrared PD chip, which are used to construct a comprehensive performance scoring system including normalized detection rate, detection probability and false alarm rate, thereby realizing a comprehensive evaluation of the performance of the infrared PD chip in practical applications.

[0005] In a first aspect, the present invention provides a method for improving the sensitivity of an infrared PD chip diode junction, the method comprising:

[0006] Carry out multi-structure modeling on the PN junction of the infrared PD chip and construct a multi-structure PN junction model;

[0007] Based on the multi-structure PN junction model, the electric field intensity distribution and the carrier concentration distribution under different reverse bias voltages are calculated to obtain a mapping relationship between the carrier separation efficiency and the PN junction structure parameters;

[0008] According to the mapping relationship between the carrier separation efficiency and the PN junction structure parameters, a carrier enhancement structure model consisting of a nano-metal particle array, a quantum well structure and a gradient doped transition layer is created;

[0009] The geometric parameters of the carrier enhancement structure model are input into the deep reinforcement learning model as state variables, and the photoelectric response model is used as the reward function to perform parameter optimization to obtain the optimal PN junction structure parameters;

[0010] Constructing a quantum efficiency evaluation model based on the optimal PN junction structure parameters and calculating quantum efficiency distribution data;

[0011] The responsivity of the quantum efficiency distribution data is calculated, and the detection probability and false alarm rate under different light intensities and wavelengths are evaluated in combination with an equivalent circuit model to obtain the sensitivity improvement index of the infrared PD chip.

[0012] In a second aspect, the present invention provides a device for improving the sensitivity of an infrared PD chip diode junction, the device for improving the sensitivity of an infrared PD chip diode junction comprising:

[0013] A modeling module is used to perform multi-structure modeling on the PN junction of the infrared PD chip and construct a multi-structure PN junction model;

[0014] A calculation module, for calculating the electric field intensity distribution and carrier concentration distribution under different reverse bias voltages based on the multi-structure PN junction model, and obtaining a mapping relationship between the carrier separation efficiency and the PN junction structure parameters;

[0015] A creation module is used to create a carrier enhancement structure model composed of a nano-metal particle array, a quantum well structure and a gradient doped transition layer according to the mapping relationship between the carrier separation efficiency and the PN junction structure parameters;

[0016] An optimization module, used to input the geometric parameters of the carrier enhancement structure model as state variables into a deep reinforcement learning model, perform parameter optimization using a photoelectric response model as a reward function, and obtain optimal PN junction structure parameters;

[0017] A construction module is used to construct a quantum efficiency evaluation model based on the optimal PN junction structure parameters and calculate quantum efficiency distribution data;

[0018] The evaluation module is used to calculate the responsiveness of the quantum efficiency distribution data, and evaluate the detection probability and false alarm rate under different light intensities and wavelengths in combination with the equivalent circuit model to obtain the sensitivity improvement index of the infrared PD chip.

[0019] The third aspect of the present invention provides an infrared PD chip diode junction sensitivity enhancement device, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory so that the infrared PD chip diode junction sensitivity enhancement device executes the above-mentioned infrared PD chip diode junction sensitivity enhancement method.

[0020] A fourth aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the above-mentioned method for improving the sensitivity of the diode junction of an infrared PD chip.

[0021] In the technical solution provided by the present invention, by integrating nano-metal particle arrays, quantum well structures and gradient doped transition layers into PN junctions, the comprehensive enhancement of carrier generation, separation and transport is achieved, and the quantum efficiency and responsiveness of infrared PD chips are significantly improved. The deep reinforcement learning model is used for parameter optimization, which breaks through the limitations of traditional empirical design, can quickly find the optimal solution in high-dimensional parameter space, and greatly improves the design efficiency and performance. By constructing an electromagnetic field-semiconductor coupling equation group, the accurate simulation of the complex physical process inside the infrared PD chip is achieved, providing a reliable theoretical basis for performance evaluation. The quantum efficiency calculation in the wavelength range of 300nm to 14μm is achieved by using Fourier transform and discrete sampling methods, providing reliable performance data for wide spectrum applications. By constructing a comprehensive performance scoring system including normalized detection rate, detection probability and false alarm rate, a comprehensive evaluation of the performance of infrared PD chips in practical applications is achieved. By introducing temperature-related quantum efficiency distribution data, the method can adapt to different environmental conditions, improving the applicability and reliability of infrared PD chips. By calculating thermal noise, shot noise and 1 / f noise in detail, an accurate evaluation of the noise characteristics of infrared PD chips is achieved.

[0022] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description, claims and drawings.

[0023] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 A schematic diagram of an embodiment of a method for improving the sensitivity of an infrared PD chip diode junction in an embodiment of the present invention;

[0025] Figure 2 A schematic diagram of an embodiment of a device for improving the sensitivity of an infrared PD chip diode junction according to an embodiment of the present invention;

[0026] Figure 3 It is a schematic diagram of an embodiment of a device for improving the sensitivity of an infrared PD chip diode junction in an embodiment of the present invention. DETAILED DESCRIPTION

[0027] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0028] The terms "including" and "having" and any variations thereof mentioned in the embodiments of the present invention are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device end including a series of steps or units is not limited to the listed steps or units, but may optionally include other steps or units that are not listed, or may optionally include other steps or units that are inherent to these processes, methods, products or device ends.

[0029] To facilitate understanding of this embodiment, a method for improving the sensitivity of the diode junction of an infrared PD chip disclosed in an embodiment of the present invention is first introduced in detail. Figure 1 As shown, the method comprises the following steps:

[0030] 101. Carry out multi-structure modeling on the PN junction of the infrared PD chip and construct a multi-structure PN junction model;

[0031] It is understandable that the execution subject of the present invention may be an infrared PD chip diode junction sensitivity enhancement device, or a terminal or a server, which is not limited here. The embodiment of the present invention is described by taking a server as an execution subject as an example.

[0032] Specifically, the PN junction of the infrared PD chip is modeled in layers. The PN junction is divided into a P-type semiconductor layer, an N-type semiconductor layer and a depletion layer to construct an initial three-layer structure model, namely the first PN junction hierarchical model. The various regions of the PN junction are separated so that the electrical and optical properties of different layers can be analyzed separately. On the basis of the first PN junction hierarchical model, in order to optimize the carrier separation efficiency and response speed, a gradient doping transition layer is introduced between the P-type semiconductor layer and the N-type semiconductor layer. The transition layer smoothly connects the electrical characteristics between the P-type and N-type materials, which can reduce the interface resistance and help to form a more uniform electric field distribution near the center of the PN junction. In this way, a second PN junction hierarchical model with gradient doping characteristics is constructed, thereby forming a more stable charge carrier transmission environment. The thickness, doping concentration and band structure of each layer structure in the second PN junction hierarchical model are parameterized. These variables are parameterized to construct a PN junction parameter set covering material properties. Through the PN junction parameter set, the electron and hole concentrations of each layer in the PN junction and their corresponding electric field characteristics are fully described. Based on the PN junction parameter set, the Poisson equation is constructed to describe the relationship between the charge distribution and the electric field in the PN junction. Due to the high computational complexity of the Poisson equation between different layers, the finite difference method is used to numerically solve the equation to obtain the potential distribution function inside the PN junction, which contains the potential change information between the layers of the PN junction. By calculating the first-order derivative of the potential distribution function, the electric field intensity distribution function of the PN junction is obtained to describe the spatial distribution of the electric field intensity in the PN junction. In order to improve the response efficiency of the PN junction to the optical signal, a metal-dielectric-semiconductor structure is introduced on the surface of the second PN junction hierarchical model, and a surface plasmon effect is stimulated by forming a nanoscale metal structure on the surface. The surface plasmon effect can enhance the local electric field, thereby amplifying the response of the PN junction to photon excitation and forming a surface plasmon enhanced PN junction model. By calculating the local field enhancement factor of the surface plasmon in the surface plasmon enhanced PN junction model, its local field distribution function is obtained to describe the degree of enhancement of the electric field near the metal nanostructure. The electric field intensity distribution function is convolved with the local field distribution function, and the enhancement effects of the two are integrated to obtain a comprehensive multi-structure PN junction model.

[0033] 102. Based on the multi-structure PN junction model, the electric field intensity distribution and carrier concentration distribution under different reverse bias voltages are calculated to obtain the mapping relationship between carrier separation efficiency and PN junction structure parameters;

[0034] Specifically, different reverse bias voltages are applied to the multi-structure PN junction model to explore the electric field and carrier distribution characteristics of the PN junction under various voltage conditions. A set of discrete reverse bias voltages is applied to the multi-structure PN junction model to obtain a set of reverse bias voltages. For each voltage value in the set, the electric field strength analysis is performed in the model. The electric field strength distribution under each bias voltage is calculated by the finite difference method or other numerical methods to obtain the electric field spatial distribution characteristics of the PN junction under different voltage conditions. Based on the electric field strength distribution, the drift-diffusion equation is used to analyze the carrier concentration distribution under each reverse bias voltage. The drift-diffusion equation describes the migration process of carriers under the action of the electric field and the influence of the diffusion phenomenon. By solving the drift-diffusion equation, the carrier concentration distribution under each reverse bias voltage condition is obtained. According to the concentration distribution data, the mobility and lifetime of the carriers in the PN junction are calculated to obtain the kinetic parameters reflecting the dynamic characteristics of the carriers. These parameters include the moving speed and direction of the carriers under the electric field and the time they can be maintained in the PN junction. In order to evaluate the response of PN junction to incident light, the absorption process of incident photons in the multi-structure PN junction model is simulated. The generation rate distribution of photogenerated electron-hole pairs is obtained by modeling the absorption and reflection behavior of photons in each layer of the PN junction structure. It represents the photon absorption of incident light at different positions and wavelengths, as well as the spatial distribution characteristics of photogenerated carriers. The carrier kinetic parameters and the generation rate distribution of photogenerated carriers are combined and substituted into the continuity equation. The continuity equation is an equation that describes the generation, migration, and recombination process of carriers under steady-state conditions. It is solved by numerical methods to obtain the separation efficiency of carriers in the PN junction under steady-state conditions, that is, the ability of photogenerated electron-hole pairs to be separated and migrate to different regions under specific bias voltage conditions. The key structural parameters of the multi-structure PN junction model are parameterized. The structural parameters include the thickness of each layer, the doping concentration, and the gradient characteristics of the transition layer. These parameters are integrated into the PN junction structure parameter vector, and combined with the calculation results of the carrier separation efficiency, a multivariate regression analysis model is used to establish the mapping relationship between the two. The multivariate regression analysis captures the trend and influence of carrier separation efficiency with different structural parameters, thus obtaining the mapping relationship between carrier separation efficiency and PN junction structural parameters.

[0035] 103. Based on the mapping relationship between carrier separation efficiency and PN junction structural parameters, a carrier enhancement structure model consisting of nano-metal particle array, quantum well structure and gradient doped transition layer is created;

[0036] Specifically, the size and spacing of the nano-metal particle array are optimized and calculated to improve its surface plasmon resonance characteristics. By adjusting the size of the metal particles and the spacing of the array, the surface plasmon resonance enhancement of infrared light of a specific wavelength is achieved. The resonance characteristics are simulated and optimized by the finite time-domain difference method, and the interaction process between the metal nanostructure and the incident light is calculated. The results obtained by the simulation of the finite time-domain difference method are used to obtain the local field enhancement factor distribution of the optimal nano-metal particle array on the PN junction surface. The local field enhancement factor distribution indicates the degree of increase in the local electric field intensity near the metal particles. According to the local field enhancement factor distribution, the effective light absorption cross section of the PN junction surface is calculated, and the light field distribution in the multilayer structure is solved by the transfer matrix method to obtain the target light absorption model. The transfer matrix method is an effective tool for processing light transmission and absorption of multilayer thin film structures. The method is used to simulate the influence of the nano-metal particle array on the propagation path of the incident light in the multilayer structure to obtain the target light absorption model. This model describes the influence of the enhanced electric field generated by the metal particle array on the PN junction surface on the photon absorption efficiency. The band layout of the quantum well structure is designed based on the target light absorption model. The band layout design of the quantum well structure needs to meet specific energy levels and wave function overlaps to optimize the transition probability and lifetime of carriers in the quantum well. By solving the Schrödinger equation, the energy levels of electrons and holes in the quantum well and the corresponding wave functions are obtained. Based on the energy level and wave function distribution, the carrier transition probability and the carrier lifetime in the quantum well structure are calculated, and a quantum well carrier dynamics model is constructed. This dynamics model can quantify the behavior characteristics of carriers in the quantum well, and then evaluate the influence of the quantum well structure on the separation efficiency of photogenerated carriers. The quantum well carrier dynamics model is combined with the mapping relationship between the carrier separation efficiency and the PN junction structure parameters to optimize the doping concentration distribution function of the gradient doped transition layer. By optimizing the gradual distribution of the doping concentration, the carrier concentration in the transition layer is gradually increased or decreased, and the separation efficiency of carriers at the PN junction interface is improved. By gradually adjusting the doping concentration distribution function, the optimal gradient doped transition layer structure is obtained, thereby enhancing the separation ability and mobility of carriers. The optimal gradient doped transition layer structure is used to simulate the electric field distribution, and the drift-diffusion model is used to calculate the carrier transport characteristics. The drift-diffusion model takes into account factors such as the electric field, carrier concentration distribution, and doping concentration gradient during the calculation process, and can provide the carrier movement speed, direction, and distribution state at different positions. The optimized carrier separation efficiency is obtained. The optimization results are integrated with the optimal nanometal particle array parameters, the band layout of the quantum well structure, and the optimal gradient doped transition layer structure to construct a complete carrier enhancement structure model. This model can significantly improve the carrier separation efficiency of the PN junction through the synergistic effect of the quantum structure and the metal-semiconductor interface, thereby effectively improving the sensitivity and response speed of the infrared PD chip.

[0037] 104. The geometric parameters of the carrier enhancement structure model are input into the deep reinforcement learning model as state variables, and the photoelectric response model is used as the reward function to optimize the parameters and obtain the optimal PN junction structure parameters;

[0038] Specifically, the geometric parameters in the carrier enhancement structure model are encoded, and the parameters are input into the deep Q network in the deep reinforcement learning model in the form of a state vector, and an initial policy function corresponding to the state and action is established. Through the policy function, the deep Q network gradually explores and learns the photoelectric response under different geometric parameter combinations. A set of PN junction structure parameters is generated based on the initial policy function, and the PN junction structure parameters are input into the photoelectric response model. The photoelectric responsivity value is calculated through the model, and a reward function is constructed based on this. The photoelectric responsivity value reflects the response degree of the carrier enhancement structure to the light signal under specific geometric parameters. The higher the value, the higher the sensitivity of this structure to the light signal. By setting a reward function related to the photoelectric responsivity value, the DQN is guided to continuously adjust the geometric parameters during the reinforcement learning process to improve the photoelectric response effect. The initial policy function is sampled to generate multiple sets of PN junction structure parameters, and their corresponding reward values ​​are calculated. The reward value describes the photoelectric response intensity under different geometric parameter combinations, thereby forming a state-action pair sequence. These sequences are stored in the experience replay buffer. By randomly extracting samples from the buffer, a training set covering a wide range of parameters is formed for training the deep Q network. Through the experience replay mechanism, the data correlation in the parameter update is effectively reduced, so that the model has better generalization ability on different structural combinations. During the training process, the Q value function of the deep Q network is updated by using the data samples in the experience replay buffer. Then, based on the updated Q value function, the ε-greedy strategy is used to select the optimal action, that is, to select the PN junction structure parameter combination that can maximize the reward function. By setting a certain exploration and utilization balance parameter ε, the model can fully explore new geometric parameter combinations in the early stage, and can also focus on fine-tuning the structure with better performance in the later stage, thereby improving the final optimization effect. Each time a new PN junction structure parameter is selected, it is used as a new state vector and input into the target network to calculate the target Q value. The calculated target Q value is compared with the output Q value of the current deep Q network to obtain the temporal difference error. The temporal difference error quantifies the difference between the selected action and the target action under the current strategy. By minimizing the error, the parameters of the deep Q network are continuously updated to approach a better policy function. The temporal difference error is used to perform gradient descent optimization on the deep Q network, and the network parameters are updated by back propagation to obtain the optimized policy function. This policy function gradually approaches the parameter combination that can achieve the highest photoelectric responsivity. This process is executed cyclically in the model, that is, by continuously repeating the above steps, the strategy is gradually optimized during the training process of reinforcement learning. After multiple rounds of iterations, as the number of iterations increases, the policy function gradually converges, and the photoelectric responsivity also tends to the optimal state. When the preset number of iterations or convergence conditions are reached, the final target policy function is output.This target strategy function reflects the optimal PN junction geometric structure parameters under the existing model structure and photoelectric response function conditions. According to the target strategy function, the optimal PN junction structure parameters are determined, and then an infrared PD chip with the best sensitivity is constructed.

[0039] 105. Construct a quantum efficiency evaluation model based on the optimal PN junction structure parameters and calculate the quantum efficiency distribution data;

[0040] Specifically, a three-dimensional PN junction model is constructed based on the optimal PN junction structural parameters using a multi-physics field coupling method. The multi-physics field coupling method combines the electromagnetic field and the physical properties of the semiconductor to more realistically reflect the behavior of the PN junction under different photoelectric conditions. After the construction is completed, the three-dimensional PN junction model is meshed to form a finite element mesh. The meshing discretizes the continuous three-dimensional model so that each unit can be analyzed separately and forms a discrete system for calculation in the entire PN junction area. Based on the finite element mesh, an electromagnetic field-semiconductor coupling equation group is constructed, which includes the propagation behavior of light in semiconductor materials, the change characteristics of the electric field, and the migration and distribution of carriers. In this model, by solving the electromagnetic field-semiconductor coupling equation group, the spatial distribution of the electric field inside the PN junction and the concentration distribution of carriers are obtained, and the electric field strength and the accumulation and migration of carriers in each region of the PN junction under specific illumination and bias voltage conditions are described. Based on the obtained electric field and carrier distribution data, the complex refractive index distribution of each region of the PN junction is calculated to characterize the phase shift and attenuation of the light wave propagating inside the material, thereby accurately describing the light absorption behavior. The complex refractive index distribution reveals the absorption characteristics of the PN junction to light and can also reflect the light response behavior of different structural regions. Substitute the complex refractive index distribution into the photogenerated carrier generation equation to calculate the photogenerated carrier generation rate inside the PN junction. The photogenerated carrier generation rate reflects the efficiency of generating electron-hole pairs after photons are absorbed in the PN junction. By applying the complex refractive index distribution to this equation, the absorption of different wavelengths of light in different regions of the PN junction can be simulated. The carrier transport process is simulated to obtain the recombination probability and collection probability of the carriers. The recombination probability describes the possibility that photogenerated carriers meet and recombine into photons in the PN junction, while the collection probability describes the efficiency of carriers reaching the electrode and being collected under the action of the electric field. The calculation of the two can reveal the response efficiency of the PN junction to the incident light under illumination conditions. After multiplying the photogenerated carrier generation rate with the recombination probability and the collection probability respectively, and then integrating the entire PN junction area, the external quantum efficiency function is obtained. The external quantum efficiency function represents the quantum conversion efficiency of the PN junction for incident light at each wavelength. This quantum conversion efficiency not only takes into account the generation of carriers, but also includes the effects of the recombination and collection processes, accurately reflecting the responsiveness of the PN junction to the light signal. The spectral response analysis of the external quantum efficiency function is performed, and the complete quantum efficiency distribution data is obtained by calculating the quantum efficiency values ​​in the wavelength range from 300nm to 14μm.

[0041] 106. The responsivity of the quantum efficiency distribution data is calculated, and the detection probability and false alarm rate under different light intensities and wavelengths are evaluated in combination with the equivalent circuit model to obtain the sensitivity improvement index of the infrared PD chip.

[0042] Specifically, the quantum efficiency distribution data is integrated by wavelength to obtain the average quantum efficiency The average quantum efficiency is a measure of the photoelectric conversion efficiency over the entire wavelength range. The responsivity of the infrared PD chip is calculated according to the formula R=ηqλ / hc ,in represents the elementary charge, is the wavelength of the incident light, is Planck's constant, is the speed of light. The current responsivity of the optical signal is obtained through the formula, and then the current output performance of the device when receiving optical signals of different wavelengths is measured. The equivalent circuit model of the infrared PD chip is constructed based on the optimal PN junction structure parameters. The equivalent circuit model can simulate the influence of different components in the actual circuit on signal transmission. The finite element analysis method is used to solve the values ​​of each equivalent circuit parameter to obtain the key parameters such as resistance, capacitance, and inductance of the infrared PD chip under different frequencies and working conditions. These parameters help to understand the performance changes of the PN junction under different load conditions and provide basic data for noise characteristic analysis. After completing the construction of the circuit model, the noise characteristics of the infrared PD chip are analyzed using the equivalent circuit model. The total noise power spectrum density is obtained by calculating the thermal noise power spectrum density, shot noise power spectrum density, and 1 / f noise power spectrum density of the chip. Thermal noise originates from the resistance elements in the chip, while shot noise is mainly due to current fluctuations caused by random collisions of carriers. 1 / f noise (flicker noise) is common in the low-frequency region of semiconductor devices and has a significant impact on detection performance. By superimposing the power spectral density of the noise, the total noise power spectral density of the chip at each frequency is calculated and expressed in the form of its square root. The normalized detection rate of the infrared PD chip is obtained by performing a ratio operation on the responsivity R and the square root of the total noise power spectral density. The normalized detection rate is an indicator to measure the detection ability of the chip under the influence of noise. The larger the value, the better the detection performance of the chip under given noise conditions. Based on the maximum likelihood ratio criterion, an infrared PD chip signal detection model is constructed. Substituting the responsivity R, the values ​​of each equivalent circuit parameter and the total noise power spectral density into the model, the probability distribution of signal detection can be obtained. The signal detection probability distribution describes the response of the chip under different signal intensities. The threshold analysis of the signal detection probability distribution is performed to calculate the detection probability and false alarm rate under different light intensities and wavelengths, and the receiver operating characteristic curve of the infrared PD chip is obtained. The receiver operating characteristic curve (ROC curve) depicts the relationship between the detection success rate and the false alarm rate under different conditions, and is used to evaluate the detection efficiency of the chip under various working conditions. Based on this curve, the chip's detection performance under different light signals is determined, so that its performance can be comprehensively evaluated in a variety of usage scenarios. The comprehensive performance score of the infrared PD chip is calculated by weighted summing the normalized detection rate, detection probability, and false alarm rate. This score represents the overall detection capability and reliability of the chip under noise conditions. This score is compared with the performance score of the infrared PD chip before optimization to calculate the percentage of performance improvement, which is the chip's sensitivity improvement index. The sensitivity improvement index quantifies the degree of improvement in the performance of the infrared PD chip caused by the optimized design.

[0043] In the embodiment of the present invention, by integrating nano-metal particle arrays, quantum well structures and gradient doped transition layers into PN junctions, comprehensive enhancement of carrier generation, separation and transport is achieved, and the quantum efficiency and responsiveness of infrared PD chips are significantly improved. The deep reinforcement learning model is used for parameter optimization, which breaks through the limitations of traditional empirical design, can quickly find the optimal solution in high-dimensional parameter space, and greatly improves the design efficiency and performance. By constructing an electromagnetic field-semiconductor coupling equation group, the accurate simulation of the complex physical process inside the infrared PD chip is achieved, providing a reliable theoretical basis for performance evaluation. The quantum efficiency calculation in the wavelength range of 300nm to 14μm is achieved by using Fourier transform and discrete sampling methods, providing reliable performance data for wide spectrum applications. By constructing a comprehensive performance scoring system including normalized detection rate, detection probability and false alarm rate, a comprehensive evaluation of the performance of infrared PD chips in practical applications is achieved. By introducing temperature-related quantum efficiency distribution data, the method can adapt to different environmental conditions, improving the applicability and reliability of infrared PD chips. By calculating thermal noise, shot noise and 1 / f noise in detail, an accurate evaluation of the noise characteristics of infrared PD chips is achieved.

[0044] In a specific embodiment, the process of executing step 101 may specifically include the following steps:

[0045] The P-type semiconductor layer, N-type semiconductor layer and depletion layer of the PN junction in the infrared PD chip are divided into three layers to obtain a first PN junction hierarchical model, and a gradient doping transition layer is constructed between the P-type semiconductor layer and the N-type semiconductor layer of the first PN junction hierarchical model to obtain a second PN junction hierarchical model;

[0046] The thickness, doping concentration and band structure of each layer in the second PN junction hierarchical model are parameterized to obtain a PN junction parameter set, and a Poisson equation is constructed based on the PN junction parameter set. The Poisson equation is solved by finite difference to obtain a potential distribution function of the PN junction.

[0047] The first-order derivative of the potential distribution function is calculated to obtain the electric field intensity distribution function of the PN junction, and a metal-dielectric-semiconductor structure is constructed on the surface of the second PN junction hierarchical model to obtain a surface plasmon enhanced PN junction model;

[0048] The local field enhancement factor of the surface plasmon is calculated based on the surface plasmon enhanced PN junction model to obtain the local field distribution function. The electric field intensity distribution function is convolved with the local field distribution function to obtain a multi-structure PN junction model.

[0049] Specifically, the P-type semiconductor layer, N-type semiconductor layer and the depletion layer between them in the PN junction are divided into three layers to obtain a preliminary PN junction structure, namely the first PN junction hierarchical model. In this model, the P-type and N-type semiconductor layers have different doping concentrations, so that a depletion layer is formed near their interface. The depletion layer is the key area for the separation of photogenerated carriers. In order to make the separation of carriers more effective, a gradually doped transition layer is constructed between the P-type semiconductor layer and the N-type semiconductor layer, so that the doping concentration of the PN junction gradually transitions from P-type to N-type. The constructed structure is the second PN junction hierarchical model. The presence of the gradually doped transition layer can reduce the band discontinuity at the interface, thereby reducing the recombination loss and improving the carrier transmission efficiency. The key structural parameters of each layer in the second PN junction hierarchical model are parameterized, and the parameters include the thickness, doping concentration and band structure distribution of each layer. The PN junction parameter set is formed by parameterization, and these parameter sets are used to describe the physical and electrical properties of each region of the PN junction. For example, the thickness parameter represents the geometric dimensions of each layer, the doping concentration parameter reflects the number of impurity atoms in the material, and the band structure describes the energy distribution of electrons in the semiconductor. Based on these parameters, the Poisson equation is constructed to describe the potential distribution inside the PN junction. The basic form of the Poisson equation is:

[0050] ;

[0051] in, Represents the electric potential function The Laplace operator of is the space charge density, is the dielectric constant of the semiconductor material. Poisson's equation describes the relationship between charge distribution and potential in the PN junction. By solving this equation, the potential distribution inside the PN junction is obtained. Since the boundary conditions of this equation are different in each layer of the PN junction and the structure is relatively complex, the finite difference method is used for numerical solution. Through the finite difference solution, the continuous potential distribution is discretized into a numerical solution on the grid points to obtain the discrete distribution function of the potential inside the PN junction. The first-order derivative operation is performed on the potential distribution function to calculate the electric field strength distribution function of the PN junction. The calculation formula for the electric field strength is:

[0052] ;

[0053] in, Indicates location The electric field strength at For location The potential distribution at the junction. By taking the derivative of the potential distribution function, the electric field intensity distribution of the PN junction at each position is obtained. This distribution describes the change of the electric field inside the PN junction. The stronger the electric field is, the higher the carrier separation efficiency is. In order to improve the light absorption efficiency of the PN junction, a metal-dielectric-semiconductor structure is introduced on the surface of the second PN junction hierarchical model. This structure is used to excite the surface plasmon effect. Surface plasmon is a collective oscillation of electrons existing at the interface between metal and dielectric. It can be excited under the irradiation of light of a specific wavelength and generate a strong local electric field. By constructing such a metal-dielectric semiconductor structure on the surface of the PN junction, a surface plasmon enhanced PN junction model is formed, thereby obtaining a stronger local field enhancement effect on the surface of the PN junction. Based on the surface plasmon enhanced PN junction model, the local field enhancement factor of the surface plasmon is calculated. The local field enhancement factor describes the amplification effect of the surface plasmon on the electric field intensity at a fixed position. Assume that the local field enhancement factor is , then it is calculated by electromagnetic simulation or finite difference time domain method to obtain the distribution function of the local field on the PN junction surface. This distribution function can describe how the intensity of the local electric field changes with the change of position near the metal nanoparticles. In order to obtain a complete multi-structure PN junction model, the electric field intensity distribution function is convolved with the local field distribution function to obtain the enhanced total electric field intensity distribution. The mathematical expression of convolution is:

[0054] ;

[0055] in, represents the total electric field intensity after surface plasmon enhancement, is the electric field intensity distribution of the original PN junction, is the local field enhancement factor distribution. Through the convolution operation, the enhancement effect of the local field is superimposed on the original electric field strength to obtain a distribution with a higher electric field strength in the entire PN junction area.

[0056] In a specific embodiment, the process of executing step 102 may specifically include the following steps:

[0057] Applying multiple discrete reverse bias voltages to the multi-structure PN junction model to obtain a reverse bias voltage set, and performing electric field strength analysis on the multi-structure PN junction model based on each voltage value in the reverse bias voltage set to obtain an electric field strength distribution;

[0058] Using the drift-diffusion equation and the electric field intensity distribution, the carrier concentration distribution at each voltage value in the reverse bias voltage set is calculated, and based on the carrier concentration distribution, the mobility and lifetime of the carriers in the PN junction are calculated to obtain the carrier dynamics parameters;

[0059] The absorption process of incident photons in the multi-structure PN junction model is simulated, the generation rate distribution of photogenerated electron-hole pairs is calculated, and the carrier kinetic parameters and generation rate distribution are substituted into the continuity equation to obtain the carrier separation efficiency under steady-state conditions through numerical solution.

[0060] The key structural parameters of the multi-structure PN junction model are parameterized to obtain the PN junction structural parameter vector, and the mapping relationship between the carrier separation efficiency and the PN junction structural parameter vector is established using the multivariate regression analysis model to obtain the mapping relationship between the carrier separation efficiency and the PN junction structural parameters.

[0061] Specifically, multiple discrete reverse bias voltages are applied to the multi-structure PN junction model to obtain a reverse bias voltage set. The discrete voltage values ​​represent the response characteristics of the PN junction under different working conditions. By applying a series of different reverse bias voltages to the model, the electric field changes of the PN junction are observed, and the electric field strength distribution under different bias voltages is obtained. The PN junction at each voltage value is simulated and analyzed by numerical methods to obtain the distribution characteristics of the electric field strength in space. The calculation of the electric field strength distribution can reflect the electric field changes in the PN junction at different voltages and reveal the behavior mode of carriers affected by the electric field in the PN junction. Using the electric field strength distribution data and combining the drift-diffusion equation, the carrier concentration distribution at each voltage value in the reverse bias voltage set is calculated. The drift-diffusion equation describes the movement of carriers in the PN junction, and its general form is:

[0062] ;

[0063] ;

[0064] in, and are the current densities of electrons and holes, respectively, is the basic charge, and are the mobilities of electrons and holes, respectively, and are the concentrations of electrons and holes, respectively, is the electric field strength, and are the diffusion coefficients of electrons and holes, respectively, and is the concentration gradient. By solving the drift-diffusion equation, the concentration distribution of electrons and holes inside the PN junction under different voltage conditions is obtained. Based on the carrier concentration distribution, the mobility and lifetime of carriers in the PN junction are calculated to obtain the kinetic parameters of the carriers. The mobility describes the movement speed of carriers under the action of the electric field, which affects the efficiency of the carriers reaching the electrode; while the lifetime describes the average time that the carriers can exist before recombination, which is related to the quality of the PN junction material and the distribution of defects. These kinetic parameters can effectively describe the behavior of carriers in the PN junction, and thus affect the efficiency of the entire photoelectric response process. In order to analyze the photoelectric characteristics of the PN junction, the absorption process of the incident photon in the multi-structure PN junction model is simulated. By simulating the absorption of photons, the generation rate distribution of photogenerated electron-hole pairs is calculated, which represents the number of carriers generated after the photon is absorbed at different positions and different wavelengths. The generation rate of photogenerated electron-hole pairs is closely related to the wavelength of the incident light, the absorption coefficient of the material, and the light field distribution inside the PN junction. The generation rate distribution of photogenerated electron-hole pairs is combined with the kinetic parameters of the carriers and substituted into the continuity equation to solve the carrier separation efficiency under steady-state conditions. The continuity equation is in the form of:

[0065] ;

[0066] in, represents the rate of change of the net current density of electrons, is the carrier generation rate, is the recombination rate of carriers. By numerically solving this equation, the efficiency of effective separation of photogenerated carriers under steady-state conditions is obtained. The higher the separation efficiency, the more effective the PN junction can generate and separate electron-hole pairs under illumination, which directly affects the sensitivity of the photodetector. The key structural parameters of the multi-structure PN junction model are parameterized. The key structural parameters obtained by parameterization, such as the thickness of each layer, doping concentration, band spacing, etc., are integrated into a PN junction structural parameter vector, which can effectively describe the geometry, material and electrical properties of the PN junction. By constructing a multivariate regression analysis model, the mapping relationship between the parameter vector and the carrier separation efficiency data is established. The goal of the multivariate regression analysis model is to establish a quantitative correlation between the structural parameters of the PN junction and the carrier separation efficiency by fitting, and to obtain the mapping relationship between the carrier separation efficiency and the PN junction structural parameters.

[0067] In a specific embodiment, the process of executing step 103 may specifically include the following steps:

[0068] The size and spacing of the nano-metal particle array are optimized and calculated to obtain the optimal nano-metal particle array parameters. The surface plasmon resonance characteristics of the optimal nano-metal particle array are simulated using the finite time-domain difference method to obtain the local field enhancement factor distribution.

[0069] According to the distribution of local field enhancement factor, the effective light absorption cross section of the PN junction surface is calculated, and the light field distribution in the multilayer structure is solved based on the transfer matrix method to obtain the target light absorption model;

[0070] Using the target light absorption model, the energy band layout of the quantum well structure is designed, and the energy levels and wave functions in the quantum well structure are solved. Based on the energy levels and wave functions, the carrier transition probability and lifetime in the quantum well structure are calculated, and a quantum well carrier dynamics model is constructed.

[0071] The quantum well carrier dynamics model is combined with the mapping relationship between carrier separation efficiency and PN junction structural parameters to optimize the doping concentration distribution function of the gradient doped transition layer and obtain the optimal gradient doped transition layer structure.

[0072] The electric field distribution of the optimal gradient-doped transition layer structure is simulated, and the carrier transport characteristics are calculated in combination with the drift-diffusion model to obtain the optimized carrier separation efficiency. The optimal nanometal particle array parameters, the energy band layout of the quantum well structure and the optimal gradient-doped transition layer structure are integrated to construct a carrier enhancement structure model.

[0073] Specifically, the size and spacing of the nano-metal particle array are optimized and calculated to obtain the optimal nano-metal particle array parameters. The size and spacing of the nano-metal particle array directly affect its surface plasmon resonance characteristics, and the surface plasmon is a resonance effect caused by the interaction between light and the collective oscillation of metal free electrons. By systematically optimizing the size and spacing of the particles, the resonance intensity within a specific wavelength range is maximized. In order to achieve this optimization, the nano-metal particle array is simulated using the finite time-difference method, which is a numerical method that can accurately solve Maxwell's equations and is used to analyze the propagation behavior of electromagnetic waves in different materials and structures. Through the finite time-difference method, the surface plasmon resonance characteristics of the optimal nano-metal particle array are obtained, and the local field enhancement factor distribution is obtained from it. This distribution describes the degree of enhancement of the local electric field near the nanoparticles. Based on the distribution of the local field enhancement factor, the effective light absorption cross section of the PN junction surface is calculated. The light absorption cross section is the effective area that describes the interaction between nanoparticles and incident light, and its size directly affects the efficiency of photon capture and absorption. The distribution of the light field in the multilayer structure is solved by the transfer matrix method. The transfer matrix method obtains the distribution characteristics of the light field in the entire structure by solving the transmission and reflection behavior of light in each layer of the medium, and obtains the target light absorption model. In the light absorption model, the enhancement and distribution of the light field will affect the absorption efficiency of photons in various regions of the PN junction. The energy band layout of the quantum well structure is designed based on the target light absorption model. A quantum well is a semiconductor heterojunction composed of different materials. Through reasonable energy band engineering, electrons and holes are confined to a specific area, thereby increasing their recombination probability and controlling their energy level distribution. By solving the Schrödinger equation, the energy levels and wave functions in the quantum well are obtained. The general form of the Schrödinger equation is:

[0074] ;

[0075] in, is the reduced Planck constant, is the effective mass of the electron, is the wave function, is the potential energy distribution, is the energy eigenvalue. By numerically solving the equation, the energy levels and wave functions in the quantum well structure are obtained, and these solutions can be used to analyze the transition characteristics of carriers. Based on the energy levels and wave functions, the transition probability and lifetime of carriers in the quantum well are calculated. The transition probability is related to the relative energy difference between electrons and holes in the quantum well, while the carrier lifetime describes the average existence time of electrons and holes before recombination, which has a direct impact on the photoelectric conversion efficiency. By combining the quantum well carrier dynamics model with the previously established mapping relationship between carrier separation efficiency and PN junction structure parameters, the doping concentration distribution function of the gradient doped transition layer in the PN junction structure is optimized. Gradient doping is a means of reducing the recombination of carriers inside the PN junction by gradually changing the doping concentration, thereby achieving a smooth transition of the electric field, which helps to improve the mobility and separation efficiency of carriers. By optimizing the gradient doping concentration distribution, the optimal doping distribution is found to maximize the carrier separation efficiency. The electric field distribution of the optimal gradient doped transition layer structure is simulated, and the carrier transport characteristics are calculated in combination with the drift-diffusion model. The basic form of the drift-diffusion model is:

[0076] ;

[0077] in, is the current density, is the basic charge, is the carrier concentration, is the carrier mobility, is the electric field strength, is the diffusion coefficient, is the concentration gradient. The model simulates the movement of carriers in the PN junction, calculates the transport characteristics of carriers, and finally obtains the optimized carrier separation efficiency. The improvement in efficiency means that under illumination conditions, the PN junction can more efficiently separate photogenerated electrons and holes, thereby increasing current output and improving photoelectric responsiveness. The optimal nanometal particle array parameters, the energy band layout of the quantum well structure, and the optimal gradient doped transition layer structure are integrated together to construct a complete carrier enhancement structure model. The model can optimize the photon absorption and carrier separation process at different levels, thereby significantly improving the overall performance of the infrared PD chip.

[0078] In a specific embodiment, the process of executing step 104 may specifically include the following steps:

[0079] N1: Encode the geometric parameters of the carrier enhancement structure model, construct the state vector, input the state vector into the deep Q network in the deep reinforcement learning model, and obtain the initial strategy function;

[0080] N2: Generate a set of PN junction structure parameters based on the initial strategy function, input the PN junction structure parameters into the photoelectric response model, calculate the photoelectric response value, and construct the reward function;

[0081] N3: Sample the initial strategy function, generate multiple sets of PN junction structure parameters, and calculate the corresponding reward values ​​to obtain a state-action pair sequence;

[0082] N4: Input the state-action pair sequence into the experience replay buffer, randomly extract a batch of samples for training the deep Q network, and obtain the updated Q value function;

[0083] N5: Based on the updated Q-value function, use the ε-greedy strategy to select the optimal action, update the PN junction structure parameters, and obtain a new state vector;

[0084] N6: Input the new state vector into the target network, calculate the target Q value, and compare it with the output of the current deep Q network to obtain the temporal difference error;

[0085] N7: Use the temporal difference error to perform gradient descent optimization on the current deep Q network, update the network parameters, and obtain the optimized policy function;

[0086] N8: Repeat the above steps N1-N7 until the preset number of iterations or convergence conditions are reached, output the target strategy function, and determine the optimal PN junction structure parameters according to the target strategy function.

[0087] Specifically, the geometric parameters of the carrier enhancement structure model are encoded. The geometric parameters include the size of the nanometal particles, the spacing, the thickness of the quantum well layer, the doping concentration of the gradient doped transition layer, etc. These parameters can describe the characteristics of the carrier enhancement structure. The geometric parameters are vectorized to form a state vector so that they can be input into the deep Q network in the deep reinforcement learning model. The goal of the deep Q network is to learn the relationship between state and action, so as to select an action for each state that maximizes the future cumulative reward. By inputting the state vector into the deep Q network, an initial policy function is obtained. Based on the initial policy function, a set of PN junction structure parameters are generated. The structural parameters are input into the photoelectric response model, and the photoelectric responsivity value is calculated by the model. Photoelectric responsivity value ( ) is a measure of the responsiveness of the PN junction to the incident light signal under a specific structure, and its calculation formula is:

[0088] ;

[0089] in, is the quantum efficiency, which indicates the proportion of photons that are effectively converted into electrical signals; is the basic charge, is the wavelength of the incident light, is Planck's constant, is the speed of light. The higher the photoelectric responsivity value, the greater the photocurrent generated by the chip under the same incident light. A reward function is constructed based on the photoelectric responsivity value, and the reward value is proportional to the responsivity, so as to motivate the deep Q network to optimize the structure in the direction that can maximize the responsivity. The initial policy function is sampled to generate multiple sets of PN junction structure parameters, and the corresponding reward value is calculated for each set of parameters to obtain a set of state-action pair sequences. Each state-action pair describes a specific operation performed under a specific structure and its corresponding photoelectric response performance. The state-action pair sequence is stored in the experience replay buffer for subsequent network training. In the experience replay buffer, a batch of samples are randomly selected for training the deep Q network. Through random sampling, the correlation between training data is effectively reduced and the generalization ability of the model is improved. By training the deep Q network, an updated Q value function is obtained. The Q value function ( ) describes the state of Next action The future cumulative reward obtained. Based on the updated Q value function, use -The greedy strategy selects the optimal action to update the PN junction structure parameters. - Greedy strategy is a method to balance exploration and exploitation in reinforcement learning. Under this strategy, the probability Randomly select an action to explore, with probability Select the action with the largest current Q value for use. This ensures that the model can find a potential better solution while avoiding falling into a local optimum. After selecting the optimal action through this strategy, the new PN junction structure parameters are obtained and encoded as a new state vector. The new state vector is input into the target network, which is used to estimate the true Q value of the optimal action in the current state. The target Q value is compared with the output of the current deep Q network to obtain the time difference error (TD error). The calculation formula of the TD error is:

[0090] ;

[0091] in, is the timing difference error, For instant rewards, is a discount factor used to measure the importance of future rewards, The Q value of the next state calculated for the target network, The Q value calculated for the current deep Q network. The temporal difference error describes the difference between the Q value predicted by the current network and the target Q value. By minimizing this error, the deep Q network is continuously optimized. The deep Q network is optimized by gradient descent using the temporal difference error, and the parameters of the network are updated by the back propagation algorithm to obtain the optimized policy function. The goal of gradient descent is to minimize the loss function (usually the square of the TD error) and update the network parameters to make the output of the network closer to the target Q value. This process is repeated from N1 to N7, that is, the geometric structure of the PN junction is improved by continuously generating state vectors, updating the policy function, and optimizing the network parameters. After many iterations, the deep Q network gradually learns how to select the optimal PN junction structural parameters to maximize the photoelectric response performance. When the preset number of iterations is reached or the convergence condition is met, the target policy function is output, which represents the best PN junction structure design in the current model. According to this target policy function, the optimal PN junction structural parameters are determined, so that the photoelectric response of the chip is maximized.

[0092] In a specific embodiment, the process of executing step 105 may specifically include the following steps:

[0093] Perform multi-physics field coupling on the optimal PN junction structural parameters to obtain a three-dimensional PN junction model, and mesh the three-dimensional PN junction model to obtain a finite element mesh, and construct an electromagnetic field-semiconductor coupling equation group based on the finite element mesh;

[0094] Solve the electromagnetic field-semiconductor coupling equations to obtain the electric field distribution and carrier distribution inside the PN junction, and calculate the complex refractive index distribution of each region of the PN junction based on the electric field distribution and carrier distribution;

[0095] Substitute the complex refractive index distribution into the photogenerated carrier generation equation to calculate the photogenerated carrier generation rate inside the PN junction, simulate the carrier transport process inside the PN junction, and calculate the carrier recombination probability and collection probability;

[0096] The carrier generation rate is multiplied by the recombination probability and collection probability of the carriers, respectively, and integrated over the entire PN junction area to obtain the external quantum efficiency function;

[0097] The spectral response analysis of the external quantum efficiency function is performed, the quantum efficiency values ​​in the wavelength range of 300nm to 14μm are calculated, and the quantum efficiency distribution data are obtained.

[0098] Specifically, multi-physics field coupling is performed on the optimal PN junction structural parameters to obtain a more accurate three-dimensional model. When multi-physics field coupling is performed on the optimal PN junction structural parameters, the influence of various aspects such as electric field, carrier motion, electromagnetic field and thermal effect of materials is considered. Through the interaction of these physical fields, the physical behavior and mutual coupling characteristics inside the PN junction are described. Based on the coupling results of multi-physics fields, a three-dimensional PN junction model is obtained, which more realistically reflects the performance and behavior of the PN junction under actual working conditions. The three-dimensional PN junction model is meshed to discretize the continuous geometric structure into many small finite element meshes. Each mesh element is an independent calculation unit, which can accurately calculate the electric field, carrier distribution, etc. in the model. Based on the finite element mesh, an electromagnetic field-semiconductor coupling equation group is constructed. This equation group describes the interaction between the electric field and carriers in semiconductor materials, and its general form is the coupling of Poisson's equation, carrier continuity equation and electromagnetic field equation. These equations make the calculation more accurate by reflecting the interaction between different physical fields. Poisson's equation is used to describe the potential distribution in the PN junction, and its form is:

[0099] ;

[0100] in, is the dielectric constant of the material, is the electric potential distribution, is the space charge density. Poisson's equation is used to describe the electric field generated by the carrier distribution inside the PN junction. By solving the Poisson's equation, the potential distribution inside the PN junction is obtained. In order to obtain the electric field strength, the potential distribution is gradient calculated:

[0101] ;

[0102] in, is the electric field strength, Represents the gradient of the electric potential, and the negative sign indicates that the direction of the electric field is opposite to the direction of the electric potential gradient. Through the electric field distribution, the movement characteristics of carriers in the PN junction are analyzed. Solve the carrier continuity equation to obtain the carrier distribution. The general form of the carrier continuity equation is:

[0103] ;

[0104] in, is the carrier concentration, is the carrier generation rate, is the recombination rate of carriers, is the current density of the carrier flow. This equation describes the generation, recombination and migration of carriers. By solving this equation, the carrier concentration distribution inside the PN junction is obtained. Based on the electric field and carrier distribution, the complex refractive index distribution of each region of the PN junction is calculated. Complex refractive index ( ) is an important parameter that describes the refraction and absorption characteristics of a material to light, among which, is the refractive index, which indicates the speed of light in a material relative to its speed in a vacuum. is the extinction coefficient, which indicates the degree of absorption of light in the material. The complex refractive index of each region is calculated through the electric field and carrier distribution. The complex refractive index distribution is substituted into the photogenerated carrier generation equation to calculate the photogenerated carrier generation rate inside the PN junction. The photogenerated carrier generation rate describes the efficiency of the incident light inducing electron-hole pairs inside the PN junction, and its calculation formula is:

[0105] ;

[0106] in, For location Wavelength The photogenerated carrier generation rate, is the intensity of the incident light, The material's wavelength The absorption coefficient of is Planck's constant, is the frequency of photons. By calculating the generation rate of photogenerated carriers, the efficiency of photons inside the PN junction being converted into electron-hole pairs is obtained. After obtaining the generation rate of photogenerated carriers, the transport process of carriers in the PN junction is simulated to calculate the recombination probability and collection probability of carriers. The recombination probability of carriers describes the possibility that electrons and holes meet and recombine into photons in the PN junction, while the collection probability describes the efficiency of carriers successfully reaching the electrode and being collected under the action of the electric field. These two probabilities have an important influence on the final photoelectric conversion efficiency. Multiply the photogenerated carrier generation rate by the recombination probability and the collection probability respectively, and integrate them over the entire area of ​​the PN junction to obtain the external quantum efficiency function. The external quantum efficiency function (EQE) indicates how many photogenerated carriers are successfully collected and converted into current under different wavelengths of light. Its integral expression is:

[0107] ;

[0108] in, is the external quantum efficiency, is the carrier generation rate, is the composite probability, To collect the probability, is the depth of the PN junction. Through this integration, we can get the efficiency of photons of various wavelengths being successfully converted into current in the entire PN junction area. We perform spectral response analysis on the external quantum efficiency function, calculate the quantum efficiency values ​​in the wavelength range from 300nm to 14μm, and get the quantum efficiency distribution data. Spectral response analysis can show the conversion efficiency of the PN junction under different wavelengths of light. For infrared PD chips, these data are used to evaluate their detection capabilities in different bands.

[0109] In a specific embodiment, the process of executing step 106 may specifically include the following steps:

[0110] The quantum efficiency distribution data is integrated with wavelength to obtain the average quantum efficiency η, and the responsivity R of the infrared PD chip is calculated according to the formula R=ηqλ / hc, where q is the basic charge, λ is the wavelength of the incident light, h is the Planck constant, and c is the speed of light;

[0111] Based on the optimal PN junction structure parameters, the equivalent circuit model of the infrared PD chip is constructed, and the values ​​of each equivalent circuit parameter are obtained through finite element analysis;

[0112] Based on the equivalent circuit model, the thermal noise power spectrum density, shot noise power spectrum density and 1 / f noise power spectrum density of the infrared PD chip are calculated to obtain the total noise power spectrum density.

[0113] The normalized detection rate of the infrared PD chip is obtained by calculating the ratio of the responsivity R to the square root of the total noise power spectrum density;

[0114] An infrared PD chip signal detection model based on the maximum likelihood ratio criterion is constructed. The responsivity R, the values ​​of each equivalent circuit parameter and the total noise power spectrum density are substituted into the infrared PD chip signal detection model to obtain the signal detection probability distribution.

[0115] The signal detection probability distribution is analyzed by threshold value, and the detection probability and false alarm rate under different light intensities and wavelengths are calculated to obtain the receiver operating characteristic curve of the infrared PD chip.

[0116] The normalized detection rate, detection probability and false alarm rate are weighted and summed to obtain the comprehensive performance score of the infrared PD chip. The comprehensive performance score is compared with the performance score of the infrared PD chip before optimization, and the performance improvement percentage is calculated to obtain the sensitivity improvement index of the infrared PD chip.

[0117] Specifically, for quantum efficiency distribution data, the average quantum efficiency is obtained by integrating the wavelength . Quantum efficiency Describes the fraction of photons successfully converted into electron-hole pairs at a specific wavelength. arrive The average quantum efficiency is obtained by integrating:

[0118] ;

[0119] The average quantum efficiency describes the photoelectric conversion capability over the entire wavelength range. The higher the value, the better the overall efficiency of the chip over a wide spectral range. , calculate the responsivity of the infrared PD chip Responsivity is a measure of the chip's ability to generate photocurrent in response to incident light, and its calculation formula is:

[0120] ;

[0121] in, is the responsivity, the unit is A / W; is the average quantum efficiency; is the basic charge (approximately coulomb); is the wavelength of the incident light; is Planck's constant (approximately J.s); is the speed of light (approximately m / s). The responsivity reflects the efficiency of the chip in converting optical signals into electrical signals. The higher the value, the stronger the chip's ability to detect optical signals. Based on the optimal PN junction structure parameters, an equivalent circuit model of the infrared PD chip is constructed. The equivalent circuit model represents the physical properties of the chip with electrical components, such as resistance, capacitance, and inductance. These equivalent components are used to simulate the behavior of the chip under different working conditions. Through finite element analysis, the geometric structure of the chip is meshed, and the equivalent circuit model is numerically solved to obtain the parameter values ​​of each equivalent circuit component. Finite element analysis discretizes the continuous chip structure into a finite number of small units, accurately simulates the distribution of electric fields, currents, and carriers, and calculates each equivalent circuit parameter, such as junction resistance, junction capacitance, and leakage current. The noise characteristics of the infrared PD chip are calculated based on the equivalent circuit model. The main noise sources include thermal noise, shot noise, and 1 / f noise. Thermal noise (also known as Johnson noise) is caused by the thermal motion of electrons in the resistor, and its power spectrum density is:

[0122] ;

[0123] in, is the thermal noise power spectral density; is the Boltzmann constant (approximately J / K); is the absolute temperature (unit K); is the resistance. Shot noise is caused by the irregular motion of carriers, and its power spectrum density is:

[0124] ;

[0125] in, is the shot noise power spectral density, is the average current. 1 / f noise (flicker noise) usually plays a leading role in the low frequency range, and its power spectrum density is expressed as being proportional to the inverse of the frequency. By superimposing these three noise power spectrum densities, the total noise power spectrum density of the infrared PD chip is obtained. The normalized detection rate (D , the formula is:

[0126] ;

[0127] in, is the normalized detection rate; is responsiveness; is the total noise power spectral density. Normalized detectivity is a key parameter to measure the ability of photodetectors to detect weak light in a noisy background. The larger the value, the better the detection performance of the detector. In order to evaluate the signal detection capability of the chip, an infrared PD chip signal detection model based on the maximum likelihood ratio criterion is constructed. The maximum likelihood ratio criterion is a classic signal detection method used to determine the presence or absence of a signal in a noise background. , the values ​​of each equivalent circuit parameter and the total noise power spectrum density are substituted into the signal detection model to obtain the probability distribution of signal detection. The signal detection probability distribution reflects the detection capability of the infrared PD chip under different noise levels and signal strengths. Threshold analysis is performed on the signal detection probability distribution to calculate the detection probability and false alarm rate under different light intensities and wavelengths. The detection probability refers to the probability that the chip correctly detects a signal under a given signal strength, while the false alarm rate refers to the probability that the chip incorrectly detects a signal when there is no signal. By performing statistical analysis on these probabilities, the receiver operating characteristic (ROC) curve of the infrared PD chip is obtained. The ROC curve describes the detection performance of the chip under different conditions. The normalized detection rate, detection probability and false alarm rate are weighted and summed to obtain the comprehensive performance score of the infrared PD chip. The comprehensive performance score is used to quantitatively evaluate the overall detection performance of the chip. The score is compared with the performance score of the infrared PD chip before optimization, the percentage of performance improvement is calculated, and the sensitivity improvement index of the infrared PD chip is obtained.

[0128] The above describes the infrared PD chip diode junction sensitivity enhancement method in the embodiment of the present invention. The following describes the infrared PD chip diode junction sensitivity enhancement device in the embodiment of the present invention. Figure 2In one embodiment of the present invention, an infrared PD chip diode junction sensitivity enhancement device comprises:

[0129] A modeling module 201 is used to perform multi-structure modeling on the PN junction of the infrared PD chip and construct a multi-structure PN junction model;

[0130] A calculation module 202 is used to calculate the electric field intensity distribution and carrier concentration distribution under different reverse bias voltages based on a multi-structure PN junction model, and obtain a mapping relationship between carrier separation efficiency and PN junction structural parameters;

[0131] A creation module 203 is used to create a carrier enhancement structure model composed of a nano-metal particle array, a quantum well structure and a gradient doped transition layer according to a mapping relationship between carrier separation efficiency and PN junction structure parameters;

[0132] An optimization module 204 is used to input the geometric parameters of the carrier enhancement structure model as state variables into the deep reinforcement learning model, and optimize the parameters using the photoelectric response model as a reward function to obtain the optimal PN junction structure parameters;

[0133] A construction module 205 is used to construct a quantum efficiency evaluation model based on the optimal PN junction structure parameters and calculate quantum efficiency distribution data;

[0134] The evaluation module 206 is used to calculate the responsivity of the quantum efficiency distribution data, and evaluate the detection probability and false alarm rate under different light intensities and wavelengths in combination with the equivalent circuit model to obtain the sensitivity improvement index of the infrared PD chip.

[0135] Through the synergy of the above components, by integrating nano-metal particle arrays, quantum well structures and gradient doped transition layers into PN junctions, the comprehensive enhancement of carrier generation, separation and transport is achieved, and the quantum efficiency and responsiveness of infrared PD chips are significantly improved. The deep reinforcement learning model is used for parameter optimization, which breaks through the limitations of traditional empirical design and can quickly find the optimal solution in high-dimensional parameter space, greatly improving the design efficiency and performance. By constructing the electromagnetic field-semiconductor coupling equation group, the accurate simulation of the complex physical process inside the infrared PD chip is achieved, providing a reliable theoretical basis for performance evaluation. The quantum efficiency calculation in the wavelength range of 300nm to 14μm is realized by Fourier transform and discrete sampling methods, providing reliable performance data for wide spectrum applications. By constructing a comprehensive performance scoring system including normalized detection rate, detection probability and false alarm rate, a comprehensive evaluation of the performance of infrared PD chips in practical applications is achieved. By introducing temperature-related quantum efficiency distribution data, the method can adapt to different environmental conditions, improving the applicability and reliability of infrared PD chips. By calculating thermal noise, shot noise and 1 / f noise in detail, an accurate evaluation of the noise characteristics of infrared PD chips is achieved.

[0136] above Figure 2 The mid-infrared PD chip diode junction sensitivity enhancement device in the embodiment of the present invention is described in detail from the perspective of modular functional entities. The infrared PD chip diode junction sensitivity enhancement device in the embodiment of the present invention is described in detail from the perspective of hardware processing.

[0137] Figure 3 : is a structural schematic diagram of an infrared PD chip diode sensitivity enhancement device provided by an embodiment of the present invention. The infrared PD chip diode sensitivity enhancement device 300 may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) 310 (for example, one or more processors) and a memory 320, and one or more storage media 330 (for example, one or more mass storage device terminals) storing application programs 333 or data 332. Among them, the memory 320 and the storage medium 330 can be short-term storage or permanent storage. The program stored in the storage medium 330 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations in the infrared PD chip diode sensitivity enhancement device 300. Furthermore, the processor 310 may be configured to communicate with the storage medium 330, and execute a series of instruction operations in the storage medium 330 on the infrared PD chip diode sensitivity enhancement device 300 to implement the steps of the above-mentioned infrared PD chip diode sensitivity enhancement method.

[0138] The infrared PD chip diode junction sensitivity enhancement device 300 may also include one or more power supplies 340, one or more wired or wireless network interfaces 350, one or more input and output interfaces 360, and / or one or more operating systems 331, such as Windows Serve, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art will appreciate that Figure 3 The structure of the infrared PD chip diode junction sensitivity enhancement device shown does not constitute a limitation of the infrared PD chip diode junction sensitivity enhancement device provided by the present invention, and may include more or fewer components than shown in the figure, or a combination of certain components, or a different arrangement of components.

[0139] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are executed on a computer, the computer executes the steps of the infrared PD chip diode junction sensitivity enhancement method.

[0140] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0141] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the whole or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program code.

[0142] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for improving the sensitivity of an infrared PD chip diode junction, characterized in that: The method comprises: Carry out multi-structure modeling on the PN junction of the infrared PD chip and construct a multi-structure PN junction model; Based on the multi-structure PN junction model, the electric field intensity distribution and the carrier concentration distribution under different reverse bias voltages are calculated to obtain a mapping relationship between the carrier separation efficiency and the PN junction structure parameters; According to the mapping relationship between the carrier separation efficiency and the PN junction structure parameters, a carrier enhancement structure model consisting of a nano-metal particle array, a quantum well structure and a gradient doped transition layer is created; The geometric parameters of the carrier enhancement structure model are input into the deep reinforcement learning model as state variables, and the photoelectric response model is used as the reward function to perform parameter optimization to obtain the optimal PN junction structure parameters; Constructing a quantum efficiency evaluation model based on the optimal PN junction structure parameters and calculating quantum efficiency distribution data; The responsivity of the quantum efficiency distribution data is calculated, and the detection probability and false alarm rate under different light intensities and wavelengths are evaluated in combination with an equivalent circuit model to obtain the sensitivity improvement index of the infrared PD chip.

2. The method for improving the sensitivity of the infrared PD chip diode junction according to claim 1, characterized in that: The multi-structure modeling of the PN junction of the infrared PD chip is performed to construct a multi-structure PN junction model, including: Divide the P-type semiconductor layer, N-type semiconductor layer and depletion layer of the PN junction in the infrared PD chip into three layers to obtain a first PN junction hierarchical model, and construct a gradient doping transition layer between the P-type semiconductor layer and the N-type semiconductor layer of the first PN junction hierarchical model to obtain a second PN junction hierarchical model; Parameterizing the thickness, doping concentration and band structure of each layer in the second PN junction hierarchical model to obtain a PN junction parameter set, constructing a Poisson equation based on the PN junction parameter set, performing finite difference solution on the Poisson equation, and obtaining a potential distribution function of the PN junction; Performing a first-order derivative calculation on the potential distribution function to obtain an electric field intensity distribution function of the PN junction, and constructing a metal-dielectric-semiconductor structure on the surface of the second PN junction hierarchical model to obtain a surface plasmon enhanced PN junction model; The local field enhancement factor of the surface plasmon is calculated based on the surface plasmon enhanced PN junction model to obtain a local field distribution function, and the electric field intensity distribution function is convolved with the local field distribution function to obtain a multi-structure PN junction model.

3. The method for improving the sensitivity of the infrared PD chip diode junction according to claim 1, characterized in that: The method of calculating the electric field intensity distribution and carrier concentration distribution under different reverse bias voltages based on the multi-structure PN junction model to obtain a mapping relationship between carrier separation efficiency and PN junction structural parameters includes: Applying a plurality of discrete reverse bias voltages to the multi-structure PN junction model to obtain a reverse bias voltage set, and performing an electric field strength analysis on the multi-structure PN junction model based on each voltage value in the reverse bias voltage set to obtain an electric field strength distribution; Calculating the carrier concentration distribution at each voltage value in the reverse bias voltage set using the drift-diffusion equation and the electric field intensity distribution, and calculating the mobility and lifetime of the carriers in the PN junction based on the carrier concentration distribution to obtain carrier kinetic parameters; Simulating the absorption process of incident photons in a multi-structure PN junction model, calculating the generation rate distribution of photogenerated electron-hole pairs, substituting the carrier kinetic parameters and the generation rate distribution into the continuity equation, and obtaining the carrier separation efficiency under steady-state conditions by numerical solution; The key structural parameters of the multi-structure PN junction model are parameterized to obtain a PN junction structural parameter vector, and a multivariate regression analysis model is used to establish a mapping relationship between the carrier separation efficiency and the PN junction structural parameter vector to obtain a mapping relationship between the carrier separation efficiency and the PN junction structural parameters.

4. The method for improving the sensitivity of the infrared PD chip diode junction according to claim 1, characterized in that: The method of creating a carrier enhancement structure model composed of a nano-metal particle array, a quantum well structure and a gradient doped transition layer according to the mapping relationship between the carrier separation efficiency and the PN junction structure parameters includes: Optimizing and calculating the size and spacing of the nano-metal particle array to obtain the optimal nano-metal particle array parameters, and using the finite time-domain difference method to simulate the surface plasmon resonance characteristics of the optimal nano-metal particle array to obtain the local field enhancement factor distribution; According to the local field enhancement factor distribution, the effective light absorption cross section of the PN junction surface is calculated, and the light field distribution in the multilayer structure is solved based on the transfer matrix method to obtain the target light absorption model; Using the target light absorption model, designing the energy band layout of the quantum well structure, and solving the energy level and wave function in the quantum well structure, calculating the carrier transition probability and lifetime in the quantum well structure based on the energy level and wave function, and constructing a quantum well carrier dynamics model; The quantum well carrier dynamics model is combined with the mapping relationship between the carrier separation efficiency and the PN junction structure parameters to optimize the doping concentration distribution function of the gradient doped transition layer to obtain the optimal gradient doped transition layer structure; The electric field distribution of the optimal gradient-doped transition layer structure is simulated, and the carrier transport characteristics are calculated in combination with the drift-diffusion model to obtain the optimized carrier separation efficiency. The optimal nanometal particle array parameters, the energy band layout of the quantum well structure and the optimal gradient-doped transition layer structure are integrated to construct a carrier enhancement structure model.

5. The method for improving the sensitivity of the infrared PD chip diode junction according to claim 1, characterized in that: The method of inputting the geometric parameters of the carrier enhancement structure model as state variables into the deep reinforcement learning model, optimizing the parameters using the photoelectric response model as the reward function, and obtaining the optimal PN junction structure parameters includes: N1: Encode the geometric parameters of the carrier enhancement structure model, construct a state vector, input the state vector into the deep Q network in the deep reinforcement learning model, and obtain an initial strategy function; N2: Generate a set of PN junction structure parameters based on the initial strategy function, input the PN junction structure parameters into the photoelectric response model, calculate the photoelectric responsivity value, and construct a reward function; N3: Sampling the initial strategy function, generating multiple sets of PN junction structure parameters, and calculating corresponding reward values ​​to obtain a state-action pair sequence; N4: Input the state-action pair sequence into the experience replay buffer, randomly extract a batch of samples for training the deep Q network, and obtain an updated Q value function; N5: Based on the updated Q value function, use the ε-greedy strategy to select the optimal action, update the PN junction structure parameters, and obtain a new state vector; N6: Input the new state vector into the target network, calculate the target Q value, and compare it with the output of the current deep Q network to obtain the timing difference error; N7: Use the temporal difference error to perform gradient descent optimization on the current deep Q network, update the network parameters, and obtain an optimized policy function; N8: Repeat the above steps N1-N7 until a preset number of iterations or convergence conditions are reached, output a target strategy function, and determine the optimal PN junction structure parameters according to the target strategy function.

6. The method for improving the sensitivity of the infrared PD chip diode junction according to claim 1, characterized in that: The step of constructing a quantum efficiency evaluation model based on the optimal PN junction structure parameters and calculating quantum efficiency distribution data includes: Performing multi-physical field coupling on the optimal PN junction structural parameters to obtain a three-dimensional PN junction model, and meshing the three-dimensional PN junction model to obtain a finite element mesh, and constructing an electromagnetic field-semiconductor coupling equation group based on the finite element mesh; Solving the electromagnetic field-semiconductor coupling equations to obtain the electric field distribution and carrier distribution inside the PN junction, and calculating the complex refractive index distribution of each region of the PN junction based on the electric field distribution and carrier distribution; Substituting the complex refractive index distribution into the photogenerated carrier generation equation, calculating the photogenerated carrier generation rate inside the PN junction, and simulating the carrier transport process in the PN junction to calculate the carrier recombination probability and collection probability; The carrier generation rate is multiplied by the recombination probability and collection probability of the carriers respectively, and integrated over the entire PN junction area to obtain an external quantum efficiency function; The spectral response analysis of the external quantum efficiency function is performed to calculate the quantum efficiency value within the wavelength range of 300 nm to 14 μm to obtain quantum efficiency distribution data.

7. The method for improving the sensitivity of the infrared PD chip diode junction according to claim 1, characterized in that: The responsivity calculation of the quantum efficiency distribution data is performed, and the detection probability and false alarm rate under different light intensities and wavelengths are evaluated in combination with an equivalent circuit model to obtain the sensitivity improvement index of the infrared PD chip, including: The quantum efficiency distribution data is wavelength-integrated to obtain an average quantum efficiency η, and the responsivity R of the infrared PD chip is calculated according to the formula R=ηqλ / hc, where q is the basic charge, λ is the wavelength of the incident light, h is the Planck constant, and c is the speed of light; Based on the optimal PN junction structure parameters, an equivalent circuit model of the infrared PD chip is constructed, and each equivalent circuit parameter value is obtained by finite element analysis; The thermal noise power spectral density, shot noise power spectral density and 1 / f noise power spectral density of the infrared PD chip are calculated based on the equivalent circuit model to obtain the total noise power spectral density; wherein the 1 / f noise is flicker noise; Calculate the ratio of the responsivity R to the square root of the total noise power spectrum density to obtain the normalized detection rate of the infrared PD chip; An infrared PD chip signal detection model based on the maximum likelihood ratio criterion is constructed, and the responsivity R, each equivalent circuit parameter value and the total noise power spectrum density are substituted into the infrared PD chip signal detection model to obtain a signal detection probability distribution; Performing threshold analysis on the signal detection probability distribution, and calculating the detection probability and false alarm rate under different light intensities and wavelengths, to obtain the receiver operating characteristic curve of the infrared PD chip; The normalized detection rate, the detection probability and the false alarm rate are weightedly summed to obtain a comprehensive performance score of the infrared PD chip, and the comprehensive performance score is compared with the performance score of the infrared PD chip before optimization, the performance improvement percentage is calculated, and the sensitivity improvement index of the infrared PD chip is obtained.

8. A device for improving the sensitivity of an infrared PD chip diode junction, characterized in that: The device is used to perform the method for improving the sensitivity of the infrared PD chip diode junction according to any one of claims 1 to 7, and comprises: A modeling module is used to perform multi-structure modeling on the PN junction of the infrared PD chip and construct a multi-structure PN junction model; A calculation module, for calculating the electric field intensity distribution and carrier concentration distribution under different reverse bias voltages based on the multi-structure PN junction model, and obtaining a mapping relationship between the carrier separation efficiency and the PN junction structure parameters; A creation module is used to create a carrier enhancement structure model composed of a nano-metal particle array, a quantum well structure and a gradient doped transition layer according to the mapping relationship between the carrier separation efficiency and the PN junction structure parameters; An optimization module, used to input the geometric parameters of the carrier enhancement structure model as state variables into a deep reinforcement learning model, perform parameter optimization using a photoelectric response model as a reward function, and obtain optimal PN junction structure parameters; A construction module is used to construct a quantum efficiency evaluation model based on the optimal PN junction structure parameters and calculate quantum efficiency distribution data; The evaluation module is used to calculate the responsiveness of the quantum efficiency distribution data, and evaluate the detection probability and false alarm rate under different light intensities and wavelengths in combination with the equivalent circuit model to obtain the sensitivity improvement index of the infrared PD chip.

9. An infrared PD chip diode junction sensitivity enhancement device, characterized in that: The infrared PD chip diode junction sensitivity enhancement device comprises: a memory and at least one processor, wherein instructions are stored in the memory; The at least one processor calls the instructions in the memory to enable the infrared PD chip diode junction sensitivity enhancement device to execute the infrared PD chip diode junction sensitivity enhancement method as described in any one of claims 1-7.

10. A computer-readable storage medium having instructions stored thereon, characterized in that: When the instructions are executed by the processor, the method for improving the sensitivity of the infrared PD chip diode junction as described in any one of claims 1 to 7 is implemented.

Images