Optical second harmonic intelligent characterization system and adjustment control method thereof

Abstract

The invention relates to an optical second harmonic intelligent characterization system, and the system comprises an incident light path unit which is used for providing a fundamental frequency light beam meeting a test condition; the four-dimensional sample table is used for receiving the fundamental frequency light beam meeting the test condition from the incident light path unit and bearing a sample, and the fundamental frequency light beam meeting the test condition irradiates the sample and generates a reflection light beam; the emergent light path unit is used for screening out a second harmonic light beam from the reflected light beam, and the second harmonic light beam possibly carries defect information of the sample; the FROG detection module is used for receiving the second harmonic light beam, separating an SHG signal from the second harmonic light beam, outputting the SHG signal, obtaining an FROG trace graph based on the SHG signal, and analyzing and processing the FROG trace graph to obtain interface state defect information of the sample; and the control unit is used for receiving the SHG signal and adjusting the incident light path unit, the four-dimensional sample table and the emergent light path unit based on the SHG signal.

Description

Technical Field

[0001] This disclosure relates to the field of optical characterization technology, and more specifically, to an intelligent optical second harmonic characterization system and its adjustment and control method. Background Technology

[0002] The performance of semiconductor devices is highly dependent on their interface quality, especially since semiconductor interface states act as traps or scattering centers for charge carriers, severely restricting the development of high-performance semiconductor devices. Therefore, developing a sensitive, non-destructive, and specific technique for characterizing the physicochemical states of interfaces is crucial. Second harmonic generation (SHG), as a non-destructive and non-invasive detection technique, has been widely used in the semiconductor field. This technique can detect the electric field at the semiconductor interface without direct contact with the sample surface, thereby investigating the distribution and concentration of semiconductor interface states.

[0003] In the field of second harmonic characterization, system stability and signal collection efficiency are key factors determining measurement accuracy. Such systems typically contain multiple degrees of freedom requiring precise adjustment. However, the adjustment of existing second harmonic characterization systems remains at the coarse-tuning stage, heavily relying on the operator's experience and skill level. Finding the optimal state involves manually and repeatedly adjusting various components, often resulting in the following technical shortcomings: (1) Inefficient: Each time the sample is changed, the entire optical path needs to be recalibrated due to differences in sample surface height, reflectivity, tilt angle, etc., which is time-consuming.

[0004] (2) Difficult to achieve the global optimal solution: The system has a high degree of freedom and there is coupling between parameters. Manual adjustment is easy to fall into local optima and cannot guarantee that the signal strength is in the best state.

[0005] (3) Stability cannot be guaranteed: During the Z-scan measurement process, mechanical motion may introduce slight optical path drift, causing the signal to fluctuate during the scan, affecting the accuracy and stability of the data.

[0006] Although there are some existing automated control schemes based on simple feedback in the field of second harmonic characterization, such as the SPGD (Stochastic Parallel Gradient Descent) algorithm, these existing automated control schemes often converge slowly in high-dimensional, continuous control spaces, cannot quickly handle the complex coupling relationships between multiple system actuators, and also suffer from technical defects such as low efficiency, difficulty in achieving the global optimum, inability to guarantee stability, and the need for cumbersome manual intervention.

[0007] Therefore, the market urgently needs a technological solution that can intelligently and efficiently maintain the optical system in its optimal working condition. Summary of the Invention

[0008] This disclosure proposes an intelligent optical second harmonic characterization system and its adjustment and control method to solve at least one of the technical problems mentioned in the background art. The technical solution adopted in this disclosure is as follows: In a first aspect, this disclosure provides an optical second harmonic intelligent characterization system, the system comprising: Incident optical path unit, used to provide a fundamental frequency beam that meets the test conditions; A four-dimensional sample stage is used to receive a fundamental frequency beam that meets the test conditions from the incident optical path unit and to carry the sample. The fundamental frequency beam that meets the test conditions illuminates the sample and generates a reflected beam. An output optical path unit is used to filter out the second harmonic beam from the reflected beam, the second harmonic beam possibly carrying defect information of the sample; The FROG detection module is used to receive the second harmonic beam, separate the SHG signal (second harmonic signal) from the second harmonic beam, output the SHG signal, and obtain the FROG trace map based on the SHG signal. By analyzing and processing the FROG trace map, the interface state defect information of the sample can be obtained. The control unit is used to receive the SHG signal and adjust the incident optical path unit, the four-dimensional sample stage and the outgoing optical path unit based on the SHG signal, thereby improving the efficiency of the FROG detection module in receiving the second harmonic beam and further improving the characterization accuracy and stability of the separated SHG signal.

[0009] In the aforementioned optical second harmonic intelligent characterization system, the incident optical path unit, the four-dimensional sample stage, the exit optical path unit, and the FROG detection module together constitute the SHG measurement platform mechanism. The SHG measurement platform mechanism and the control unit together form the optical second harmonic intelligent characterization system.

[0010] As is well known, the sample can be a semiconductor device or similar object that needs to be tested.

[0011] Preferably, the incident optical path unit includes, in sequence along the incident direction of the fundamental frequency beam, a laser, a half-wave plate, a polarizer, another half-wave plate, at least two first fast-reflecting mirrors, a fundamental frequency filter, a first five-dimensional adjustment frame, and an incident objective lens.

[0012] Furthermore, a Gran-Thompson polarizer can be selected as the polarizer.

[0013] When the incident optical path unit is working, the fundamental frequency beam emitted by the laser passes through a half-wave plate to adjust the polarization direction, a polarizer to polarize, a first fast-reflecting mirror for rapid beam guidance, a fundamental frequency filter to filter out stray light, a first five-dimensional adjustment frame, and a moving incident objective lens for focusing. The first five-dimensional adjustment frame can carry the moving incident objective lens, thereby illuminating the fundamental frequency beam that meets the test conditions onto the sample carried by the four-dimensional sample stage.

[0014] Preferably, the control unit is connected to the first fast-reflecting mirror, the first five-dimensional adjustment frame, and the four-dimensional sample stage.

[0015] In practice, the control unit sends control signals to the first fast-reflecting mirror, the first five-dimensional adjustment frame, and the four-dimensional sample stage based on the SHG signal, thereby controlling and adjusting the first fast-reflecting mirror, the first five-dimensional adjustment frame, and the four-dimensional sample stage.

[0016] Preferably, the outgoing optical path unit includes, in sequence along the reflection direction of the reflected beam, an adjustable detection objective lens, a second five-dimensional adjustment frame, at least two second fast-reflecting mirrors, a polarizer, and a frequency doubling filter.

[0017] When the output optical path unit is in operation, the reflected beam generated after the sample is irradiated by the fundamental frequency beam is collimated by the adjustable detection objective lens and the second five-dimensional adjustment frame, quickly guided by the second fast-reflecting mirror, analyzed by the analyzer, and filtered by the frequency doubling filter to select the second harmonic beam. The second harmonic beam may carry defect information of the sample for subsequent analysis.

[0018] The second harmonic beam is received by the FROG detection module, which separates the SHG signal (second harmonic signal). The FROG detection module synthesizes the FROG trace based on the SHG signal and obtains interface state defect information based on the analysis and processing of the FROG trace diagram. The FROG detection module also transmits the SHG signal to the control unit, which serves as the basis for the control unit to perform control adjustments.

[0019] Preferably, the control unit is connected to the four-dimensional sample stage, the second five-dimensional adjustment frame, and the second fast-reflecting mirror.

[0020] In practice, the control unit sends control signals to the four-dimensional sample stage, the second five-dimensional adjustment frame, and the second fast-reflecting mirror based on the SHG signal, thereby controlling and adjusting the four-dimensional sample stage, the second five-dimensional adjustment frame, and the second fast-reflecting mirror.

[0021] Preferably, the FROG detection module includes: The beam splitter module is used to receive the second harmonic beam, split the second harmonic beam, and obtain and output a synchronized first beam signal (original signal pulse) and second beam signal. The delay module is used to receive the second beam signal, delay the second beam signal by adjusting the optical path, and obtain and output a delayed beam signal (delay time signal pulse); An SHG crystal is used to receive the first beam signal and the delayed beam signal, mix the first beam signal and the delayed beam signal to generate a sum-frequency signal, and output the sum-frequency signal as the SHG signal. A spectrometer is used to receive the SHG signal, acquire and output the spectrum of the SHG signal as a function of delay time—that is, the FROG trace plot. The PC terminal is used to receive the FROG trace map and adjust the delay time of the second beam signal by the delay module.

[0022] Preferably, the PC is communicatively connected to the spectrometer and is also controllably connected to the delay module.

[0023] Secondly, this disclosure provides an adjustment and control method for an optical second harmonic intelligent characterization system provided in the first aspect and any one thereof. The method is configured in the control unit, and the control unit uses the method to precisely adjust the incident light path unit, the four-dimensional sample stage and the outgoing light path unit to adjust the SHG signal intensity of the optical second harmonic intelligent characterization system after changing the sample and / or Z-scanning.

[0024] Preferably, the method is a pre-improved DDPG algorithm.

[0025] Preferably, the method can also be to execute the SPGD (Stochastic Parallel Gradient Descent) algorithm, apply random perturbations to the control parameters included in the performance evaluation function, and complete gradient estimation based on the change in the evaluation function after the perturbation. This enables precise control of the control parameters of the incident optical path unit, the four-dimensional sample stage, and the exit optical path unit, so as to adjust the SHG signal intensity after the optical second harmonic intelligent characterization system replaces the sample and / or performs a Z-scan.

[0026] The beneficial effects of this disclosure are as follows: The optical second harmonic intelligent characterization system and its adjustment and control method proposed in this disclosure have a simple system structure, modular design, convenient operation, and good applicability. The optical second harmonic intelligent characterization system can be divided into an incident optical path unit, a four-dimensional sample stage, an exit optical path unit, an FROG detection module, and a control unit. The control unit can uniformly and accurately adjust the incident optical path unit, the four-dimensional sample stage, and the exit optical path unit, thereby ensuring the coordination of each module, improving the efficiency of the FROG detection module in receiving SHG signals, and further enhancing the accuracy and stability of SHG signal characterization.

[0027] This disclosure allows for a more detailed structural design of the incident optical path unit, the four-dimensional sample stage, and the exit optical path unit, enabling the control unit to control the first fast reflector, the first five-dimensional adjustment frame, the four-dimensional sample stage, the second five-dimensional adjustment frame, and the second fast reflector among the three. This limits the adjustment and control range to the controlled components that actually play a role in adjusting the optical path of the incident optical path unit, the four-dimensional sample stage, and the exit optical path unit, reducing the complexity of the control and adjustment circuit and the control and adjustment logic.

[0028] In this disclosure, control algorithms or control logic can be specifically designed for the control unit based on the control parameters of the first fast-reflecting mirror, the first five-dimensional adjustment frame, the four-dimensional sample stage, the second five-dimensional adjustment frame, and the second fast-reflecting mirror. This allows the optical second harmonic intelligent characterization system to automatically and quickly adjust multiple controlled optical elements to their optimal configuration based on the SHG signal strength after sample replacement. This maximizes the SHG signal strength received by the FROG detection module, restoring it to or even exceeding the optimal level before sample replacement. This eliminates a lot of tedious manual intervention and avoids errors caused by manual adjustment.

[0029] In this disclosure, the control unit innovatively configures a pre-improved DDPG algorithm. By constructing the state space and action space using the SHG signal intensity received by the FROG detection module and the control parameters of the first fast mirror, the first five-dimensional adjustment frame, the four-dimensional sample stage, the second five-dimensional adjustment frame, and the second fast mirror, the control unit can adaptively adjust the detection optical path and other controlled components of the optical second harmonic intelligent characterization system in real time during the dynamic measurement process of Z-scanning. This compensates for changes in the optical path caused by Z-axis movement, ensuring that the signal intensity received by the FROG detection module remains stable and strong throughout the scanning process, and avoiding signal attenuation or distortion due to optical path misalignment.

[0030] In this disclosure, the pre-improved DDPG algorithm is configured with a unique composite reward function, which includes a light intensity increase reward, a light intensity maintenance reward, a light intensity decrease reward, and a target light intensity reward. This comprehensively evaluates the changes in SHG signal intensity, ensuring that the control unit fully considers the increasing trend, maintenance trend, decreasing trend, and target trend of SHG signal intensity. Based on these trends, it achieves precise and coordinated control and adjustment of each controlled component, including the first fast mirror, the first five-dimensional adjustment frame, the four-dimensional sample stage, the second five-dimensional adjustment frame, and the second fast mirror. This ensures the accuracy, automation, and reliability of the adjustment action, eliminates the need for optical path recalibration, has high adjustment efficiency, and can continuously ensure that the control parameters of each controlled component are at the optimal solution, thus guaranteeing stability.

[0031] This disclosure, through the unique design of the control unit and the controlled units in the incident optical path unit, the four-dimensional sample stage and the outgoing optical path unit, achieves a global optimization control strategy that is more efficient, has wider applicability and is more flexible and reliable than traditional existing optimization algorithms in a control environment with high degrees of freedom, continuous parameters and complex coupling.

[0032] In this disclosure, by fully combining second harmonic generation (SHG) characterization technology and DDPG (deep deterministic policy gradient) algorithm, the optical path can be automatically aligned and the SHG signal can be maximized, thereby intelligently and efficiently maintaining the optical system in the best working state. Attached Figure Description

[0033] The accompanying drawings, which form part of this application, are used to provide a further understanding of this disclosure. The illustrative embodiments of this disclosure and their descriptions are used to explain this disclosure and do not constitute an undue limitation of this disclosure.

[0034] Figure 1 This is a schematic diagram of the combination of the incident optical path unit, four-dimensional sample stage, outgoing optical path unit, FROG detection module and control unit as described in Embodiment 1 of this disclosure.

[0035] Figure 2 This is a schematic diagram of the combination of the incident optical path unit, the four-dimensional sample stage, the outgoing optical path unit, and the FROG detection module as described in Embodiment 1 of this disclosure.

[0036] Figure 3 This is a schematic diagram of the structure and optical path of the FROG detection module described in Embodiment 1 of this disclosure.

[0037] Figure 4 This is a simplified flowchart of the preferred steps of the pre-improved DDPG algorithm described in Embodiment 2 of this disclosure. Detailed Implementation

[0038] The present disclosure will now be described in detail with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in the present application can be combined with each other.

[0039] The following detailed descriptions are exemplary and intended to provide further detailed explanation of this disclosure. Unless otherwise specified, all technical terms used in this disclosure have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. The terminology used in this disclosure is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this disclosure.

[0040] Example 1: like Figure 1-3As shown, this embodiment provides an optical second harmonic intelligent characterization system, the system comprising: Incident optical path unit, used to provide a fundamental frequency beam that meets the test conditions; The four-dimensional sample stage 10 is used to receive a fundamental frequency beam that meets the test conditions from the incident optical path unit and to carry the sample 11. The fundamental frequency beam that meets the test conditions illuminates the sample 11 and generates a reflected beam. The outgoing optical path unit is used to filter out the second harmonic beam from the reflected beam, which may carry defect information of the sample 11. FROG detection module 17 is used to receive the second harmonic beam, separate the SHG signal (second harmonic signal) from the second harmonic beam, output the SHG signal, and obtain the FROG trace map based on the SHG signal. By analyzing and processing the FROG trace map, the interface state defect information of the sample 11 can be obtained. The control unit is used to receive the SHG signal and adjust the incident optical path unit, the four-dimensional sample stage 10 and the outgoing optical path unit based on the SHG signal, thereby improving the efficiency of the FROG detection module 17 in receiving the second harmonic beam and further improving the characterization accuracy and stability of the separated SHG signal.

[0041] In the aforementioned optical second harmonic intelligent characterization system, the incident optical path unit, the four-dimensional sample stage, the exit optical path unit, and the FROG detection module together constitute the SHG measurement platform mechanism. The SHG measurement platform mechanism and the control unit together form the optical second harmonic intelligent characterization system.

[0042] As is well known, sample 11 can be a semiconductor device or the like that to be tested.

[0043] In one feasible embodiment, the incident optical path unit, along the incident direction of the fundamental frequency beam, sequentially includes a laser 1, a half-wave plate 2, a polarizer 3, a half-wave plate 4, at least two first fast-reflecting mirrors 5, a fundamental frequency filter 6, a first five-dimensional adjustment frame 7, and an incident objective lens 8.

[0044] Furthermore, polarizer 3 can be a Gran-Thompson polarizer.

[0045] When the incident optical path unit is working, the fundamental frequency beam emitted by the laser 1 passes through the half-wave plate 2 to adjust the polarization direction, the polarizer 3 to polarize, the first fast-reflecting mirror 5 to guide the beam quickly, the fundamental frequency filter 6 to filter out stray light, and the first five-dimensional adjustment frame 7 and the moving incident objective 8 to focus the beam. The first five-dimensional adjustment frame 7 can carry the moving incident objective 8, thereby illuminating the fundamental frequency beam that meets the test conditions onto the sample 11 carried by the four-dimensional sample stage 10.

[0046] Furthermore, the control unit is connected to the first fast-reflecting mirror 5, the first five-dimensional adjustment frame 7, and the four-dimensional sample stage 10 respectively. Based on the SHG signal, the control unit sends control signals to the first fast-reflecting mirror 5, the first five-dimensional adjustment frame 7, and the four-dimensional sample stage 10, thereby controlling and adjusting the first fast-reflecting mirror 5, the first five-dimensional adjustment frame 7, and the four-dimensional sample stage 10.

[0047] In one feasible embodiment, the outgoing optical path unit, along the reflection direction of the reflected beam, sequentially includes an adjustable detection objective lens 12, a second five-dimensional adjustment frame 13, at least two second fast-reflecting mirrors 14, a polarizer 15, and a frequency doubling filter 16.

[0048] When the output optical path unit is working, the reflected beam generated after the sample 11 is irradiated by the fundamental frequency beam is collimated by the adjustable detection objective lens 12 and the second five-dimensional adjustment frame 13, quickly guided by the second fast-reflecting mirror 14, analyzed by the analyzer 15, and filtered out by the frequency doubling filter 16 to obtain the second harmonic beam. The second harmonic beam may carry defect information of the sample 11 for subsequent analysis.

[0049] The second harmonic beam is received by the FROG detection module 17, which separates the SHG signal (second harmonic signal). The FROG detection module 17 synthesizes the FROG trace based on the SHG signal and obtains interface state defect information based on the analysis and processing of the FROG trace diagram. The FROG detection module 17 also transmits the SHG signal to the control unit, which serves as the basis for the control unit to perform control and adjustment.

[0050] Furthermore, the control unit is connected to the four-dimensional sample stage 10, the second five-dimensional adjustment frame 13, and the second fast-reflecting mirror 14. Based on the SHG signal, the control unit sends control signals to the four-dimensional sample stage 10, the second five-dimensional adjustment frame 13, and the second fast-reflecting mirror 14, thereby controlling and adjusting the four-dimensional sample stage 10, the second five-dimensional adjustment frame 13, and the second fast-reflecting mirror 14.

[0051] In one feasible embodiment, the FROG detection module 17 includes: Beam splitter module 1701 is used to receive the second harmonic beam, split the second harmonic beam, and obtain and output a synchronized first beam signal (original signal pulse) and second beam signal. The delay module 1702 is used to receive the second beam signal, delay the second beam signal by adjusting the optical path, and obtain and output a delayed beam signal (delay time signal pulse); The SHG crystal 1705 is used to receive the first beam signal and the delayed beam signal, mix the first beam signal and the delayed beam signal to generate a sum-frequency signal, and output the sum-frequency signal as the SHG signal. Spectrometer 1707 is used to receive the SHG signal, acquire and output the spectrum of the SHG signal as a function of delay time—that is, the FROG trace plot. PC terminal 1708 is used to receive the FROG trace map and adjust the delay time of the second beam signal by the delay module 1702.

[0052] Furthermore, the PC terminal 1708 is communicatively connected to the spectrometer 1707, and the PC terminal 1708 is controllably connected to the delay module 1702.

[0053] The first beam signal (original signal pulse) and the second beam signal are synchronization signals, therefore they can be uniformly represented as... The delay module 1702 sets the delay time of the second beam signal to τ, then the delayed beam signal is expressed as follows: The SHG crystal 1705 receives the first beam signal. With delayed beam signal The sum-frequency signal generated by mixing the two can be expressed as The sum-frequency signal is the obtained SHG signal. A first lens 1704 and a second lens 1706 can be respectively disposed on both sides of the SHG crystal 1705 for auxiliary beam refraction and mixing.

[0054] Example 2: This embodiment provides an adjustment and control method for an optical second harmonic intelligent characterization system provided in Embodiment 1 and any of them. The method is configured in the control unit, and the control unit uses the method to precisely adjust the incident light path unit, the four-dimensional sample stage 10 and the outgoing light path unit to adjust the SHG signal intensity of the optical second harmonic intelligent characterization system after replacing the sample 11 and / or Z-scanning.

[0055] The method is a pre-improved DDPG algorithm, including the following steps S100 to S900.

[0056] Step S100: Construct and initialize the Actor network, Critic network, target Actor network, and target Critic network.

[0057] The construction and initialization of the Actor network, Critic network, target Actor network, and target Critic network may include steps S101 to S103.

[0058] Step S101: Randomly initialize the Actor network. and Critic network Then, set the target Actor network. Target Critic network .

[0059] Step S102: Initialize the attenuation factor Soft update coefficient The maximum number of iterations T, the cumulative number of iterations t, the random noise function N, and the number of samples M; Among them, attenuation factor Used to solve for the objective Q-value, soft update coefficients Used for updating and The maximum number of iterations T represents the total number of times the network is trained and its parameters are updated; the random noise function N represents Gaussian noise, which is used to introduce a certain degree of exploration during the learning process, using noise to explore the action space; the number of samples M represents the number of samples taken from the replay buffer during training.

[0060] Step S103: Define the state space and action space based on the control parameters of the incident optical path unit, the four-dimensional sample stage 10 and the outgoing optical path unit.

[0061] More specifically, the state space and action space can be defined based on the control parameters of the first fast-reflecting mirror 5, the first five-dimensional adjustment frame 7, the four-dimensional sample stage 10, the second five-dimensional adjustment frame 13, and the second fast-reflecting mirror 14. In short, the state space and action space are defined by the control parameters of the components in the incident light path unit, the four-dimensional sample stage 10, and the exit light path unit that have a controllable connection with the control unit.

[0062] The method described above is a pre-improved DDPG algorithm that can be used as follows: Figure 4 Use the flowchart shown to understand.

[0063] It should be noted that the first fast-reflecting mirror 5 has two control parameters: the x-axis angle and the y-axis angle in a coordinate system with its mirror surface as the reference. These can be denoted as... It is used to precisely control the direction of the incident beam. The first five-dimensional adjustment frame 7 has three translational degrees of freedom in the X-axis, Y-axis, and Z-axis directions, and rotation around the X-axis in a coordinate system based on the adjustment frame. Rotation around the Y-axis With two rotational degrees of freedom and a total of five control parameters, it can be denoted as... It is used to locate the focal point of the incident beam.

[0064] The four-dimensional sample stage 10 has three translational degrees of freedom (X-axis, Y-axis, and Z-axis) and one rotational degree of freedom (T-axis) about the Z-axis in a coordinate system based on the sample stage, for a total of four control parameters, which can be denoted as: This is used to achieve three-dimensional positioning and horizontal rotation of the sample. The second five-dimensional adjustment frame 13 has three translational degrees of freedom in the X-axis, Y-axis, and Z-axis directions, and rotation around the X-axis in the coordinate system based on the adjustment frame. Rotation around the Y-axis With two rotational degrees of freedom and a total of five control parameters, it can be denoted as... This is used to collimate the reflected beam. The second fast-reflecting mirror 14 has two control parameters in a coordinate system with its mirror surface as the reference: the x-axis angle and the y-axis angle, which can be denoted as... This is used to precisely control the direction of the probe beam, ensuring efficient coupling of the reflected beam to subsequent detection modules. The control unit can be configured to acquire these control parameters and incorporate them into the pre-improved DDPG algorithm.

[0065] like Figure 2 As shown, the pre-improved DDPG algorithm is executed in the control unit to precisely adjust the control parameters of the first fast-reflecting mirror 5, the first five-dimensional adjustment frame 7, the four-dimensional sample stage 10, the second five-dimensional adjustment frame 13, and the second fast-reflecting mirror 14, respectively, so that the optical second harmonic intelligent characterization system can receive a higher SHG signal intensity after changing sample 11 or Z-scan. The pre-improved DDPG algorithm iterates continuously, collecting empirical tuples... The samples are added to the experience replay pool, and then randomly selected from the experience replay pool to update the AC network, thereby improving the efficiency of the FROG detection module 17 in receiving SHG signals. The state in the tuple This represents the current state of the optical second harmonic intelligent representation system during its t-th iteration of training. This includes the SHG signal intensity received by the detector, and the control parameters of the first fast-reflecting mirror 5, the first five-dimensional adjustment frame 7, the four-dimensional sample stage 10, the second five-dimensional adjustment frame 13, and the second fast-reflecting mirror 14. Indicates the current state The next action to be executed is to process the summarized control parameters. Adjustments; Indicates the current state Next action The rewards received This represents the next state of the optical second harmonic intelligent representation system (i.e., the state in the t-th iteration of training).

[0066] The state of the optical second harmonic intelligent characterization system can include the SHG signal strength I received by the control unit, as well as the parameter states of all adjustable components. These parameters include the x and y axis angle control parameters of the first fast-reflecting mirror 5 and the second fast-reflecting mirror 14, the x, y, and z axis control parameters of the first five-dimensional adjustment frame 7 and the second five-dimensional adjustment frame 13, and the rotation around the X-axis. Rotation around the Y-axis Two rotational degrees of freedom, control parameters for the x, y, and z axes of the four-dimensional sample stage 10, and one rotational degree of freedom for rotation about the Z-axis.

[0067] Therefore, the defined state space can be represented as .in, These represent the x-axis angle and y-axis angle of the first fast-reflecting mirror 5, respectively; These represent the three translational degrees of freedom (X-axis, Y-axis, and Z-axis) and the rotational degree of freedom (around the X-axis) of the first five-dimensional adjustment frame 7. Rotation around the Y-axis Two rotational degrees of freedom; The four-dimensional sample stage 10 has three translational degrees of freedom in the X-axis, Y-axis, and Z-axis directions, and one rotational degree of freedom about the Z-axis, T. These represent the three translational degrees of freedom (X-axis, Y-axis, and Z-axis) and the rotational degree of freedom (around the X-axis) of the five-dimensional adjustment frame 13. Rotation around the Y-axis Two rotational degrees of freedom; These represent the x-axis angle and y-axis angle of the fast-reflecting mirror 14, respectively.

[0068] Furthermore, the Action of the optical second harmonic intelligent characterization system is the adjustment amount—that is, the parameter adjustment amount—of the control parameters of the first fast-reflecting mirror 5, the first five-dimensional adjustment frame 7 and the four-dimensional sample stage 10, the second five-dimensional adjustment frame 13 and the second fast-reflecting mirror 14. The parameter adjustment amount can be expressed as: The control unit, based on the pre-improved DDPG algorithm, adjusts the control parameters involved in the motion space to control signal output, thereby realizing the adjustment and control of the first fast-reflecting mirror 5, the first five-dimensional adjustment frame 7, the four-dimensional sample stage 10, the second five-dimensional adjustment frame 13, and the second fast-reflecting mirror 14.

[0069] For the Actor network, Critic network, target Actor network, and target Critic network, the Actor network input is considered as the current state S. t The Actor network outputs a specific action. , It can be represented as: This can be understood as the current state being... Decisions (actions) generated through the Actor network After the adjustment, The included control parameters are processed through the action. This will bring the signal closer to the desired ideal state, thus increasing the accuracy of the SHG signal received by the control unit. The parameters of the Actor network are the neural network parameters. The better the settings, the better the actions generated by the Actor network. The more accurate the system, the faster the optical second harmonic intelligent characterization system can converge.

[0070] The Critic network input is the current state S. t And the actions taken by the current Actor network The Critic network output is value.

[0071] Furthermore, The value can be represented as: ;in, For the neural network parameters of the Critic network, in the current state The more precise the actions generated by the Actor network, the better the performance of the Critic network. The larger the value.

[0072] The target Actor network and the target Critic network have the same structure as the Actor network and the Critic network, respectively (the target Actor network is the same as the Actor network, and the target Critic network is the same as the Critic network), and each has its own parameters. and The target Actor network input is the next state. The target Actor network outputs the action to be taken in the next state. , It can be represented as: .

[0073] The target Critic network input is the next state. and the action to be taken in the next state The target Critic network output is , It can be represented as: .

[0074] Step S200: Begin iterative training, set the state Input into the Actor network to obtain actions ; where state S t Represents the current state and action in the t-th iteration of training. Representing state S t The following action.

[0075] Step S300: Perform the action , get the state and rewards ; where state S t+1 This represents the next state in the t-th iteration of training.

[0076] Step S400: Obtain the The samples are stored in the experience replay pool.

[0077] Step S500: If the number of samples stored in the experience replay pool is greater than the number of sampled samples M, firstly, M samples are randomly sampled from the experience replay pool, and then the target Q value is calculated. The formula for calculating the target Q value is as follows:

[0078] in, This represents the target Q value.

[0079] Step S600: Obtain using the Critic network The Q-value is calculated using the mean squared error loss function as the Critic loss function, as follows: ; Minimize using gradient descent algorithm To update the parameters of the Citic network .

[0080] Step S700: Set the loss function of the Actor network as follows: ; Optimize the parameters of the Actor network by maximizing the Q-value. .

[0081] Step S800: Soft update the target Actor network and the target Critic network ,as follows: .

[0082] Step S900: Repeat steps S200 to S800 until the cumulative number of iterations t is greater than or equal to the maximum number of iterations T, then end the iterative training.

[0083] Furthermore, the reward The calculation process may include steps S301 to S305.

[0084] Step S301: Calculate the light intensity increase reward ,as follows: ; in, The bonus for increasing light intensity represents a positive reward given when the intensity of the SHG signal increases compared to the previous time step; The strength of the SHG signal at time step t; It is a positive coefficient and can be used to adjust the reward magnitude when the SHG signal strength increases.

[0085] Furthermore, It can be a linear function of the increase in light intensity, or it can be a fixed positive value (if you want to simplify the design). Step S302: Calculate the light intensity maintenance reward ,as follows: ; in, The light intensity maintenance reward means that when the change of the SHG signal compared to the previous time step is less than or equal to the tolerance threshold (considered as no change or a very small change), a small positive reward is given; otherwise, a zero reward is given. This can be used to encourage the algorithm to maintain its current good state. It is a positive coefficient. The set tolerance threshold is used to evaluate the variation amplitude of the SHG signal within adjacent time steps; otherwise, it represents other.

[0086] Step S303: Calculate the light intensity reduction reward ,as follows: ; in, The light intensity reduction reward represents a negative reward given when the SHG signal decreases compared to the previous time step; It is a negative coefficient, which can be used to give a negative reward (penalty) when the SHG signal decreases.

[0087] Step S304: Calculate the reward for achieving the target light intensity. ,as follows: ; in, To achieve the target light intensity reward, the SHG signal must reach or exceed a preset target value. An additional positive reward is given at that time; It is a positive coefficient, which can be used to encourage the algorithm to reach the required SHG signal strength as quickly as possible.

[0088] Step S305: Calculate the reward ,as follows: ; Among them, rewards Increase the reward for light intensity Light intensity retention reward Light intensity reduction reward and target light intensity reward The total reward obtained by accumulating the total amount.

[0089] During the debugging process, the reward and penalty coefficients in the reward function can be adjusted based on the experimental process and convergence results. Adjustments are made to ensure that the network in the pre-improved DDPG algorithm is optimized in the right direction.

[0090] Furthermore, |k4|>|k1|>|k2|>|k1|.

[0091] It should be noted that the adjustment and control method described in Example 2 is merely a preferred method for the optical second harmonic intelligent characterization system described in Example 1, and does not imply that the optical second harmonic intelligent characterization system described in Example 1 must rely on the adjustment and control method described in Example 2. Other existing adjustment and control methods can also be used in the optical second harmonic intelligent characterization system described in Example 1.

[0092] In summary, the optical second harmonic intelligent characterization system and its adjustment and control method provided in Embodiments 1-2 of this disclosure have a simple system structure, modular design, convenient operation, and good applicability. The optical second harmonic intelligent characterization system can be divided into an incident optical path unit, a four-dimensional sample stage, an exit optical path unit, an FROG detection module, and a control unit. The control unit can uniformly and precisely adjust the incident optical path unit, the four-dimensional sample stage, and the exit optical path unit, thereby ensuring the coordination of each module, improving the efficiency of the FROG detection module in receiving SHG signals, and further improving the accuracy and stability of SHG signal characterization. This disclosure allows for a more detailed structural design of the incident optical path unit, the four-dimensional sample stage, and the exit optical path unit, enabling the control unit to control the first fast-reflecting mirror, the first five-dimensional adjustment frame, the four-dimensional sample stage, the second five-dimensional adjustment frame, and the second fast-reflecting mirror among the three. This limits the adjustment and control range to the controlled components that actually play a role in adjusting the optical path of the incident optical path unit, the four-dimensional sample stage, and the exit optical path unit, reducing the complexity of the control adjustment circuit and control adjustment logic.

[0093] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this disclosure and not to limit them. Although this disclosure has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of this disclosure. Any modifications or equivalent substitutions that do not depart from the spirit and scope of this disclosure should be covered within the protection scope of the claims of this disclosure.

Claims

1. An intelligent optical second harmonic characterization system, characterized in that, The system includes: Incident optical path unit, used to provide a fundamental frequency beam that meets the test conditions; A four-dimensional sample stage (10) is used to receive a fundamental frequency beam that meets the test conditions from the incident optical path unit and to carry a sample (11). The fundamental frequency beam that meets the test conditions illuminates the sample (11) and generates a reflected beam. The outgoing optical path unit is used to filter out the second harmonic beam from the reflected beam, which may carry defect information of the sample (11). The FROG detection module (17) is used to receive the second harmonic beam, separate the SHG signal from the second harmonic beam, output the SHG signal, and obtain the FROG trace map based on the SHG signal. The interface state defect information of the sample (11) is obtained by analyzing and processing the FROG trace map. The control unit is used to receive the SHG signal and adjust the incident optical path unit, the four-dimensional sample stage (10) and the outgoing optical path unit based on the SHG signal.

2. The optical second harmonic intelligent characterization system as described in claim 1, characterized in that, The incident optical path unit, along the incident direction of the fundamental frequency beam, sequentially includes a laser (1), a half-wave plate (2), a polarizer (3), a half-wave plate (4), at least two first fast-reflecting mirrors (5), a fundamental frequency filter (6), a first five-dimensional adjustment frame (7), and an incident objective lens (8).

3. The optical second harmonic intelligent characterization system as described in claim 2, characterized in that, The control unit is connected to the first fast-reflecting mirror (5), the first five-dimensional adjustment frame (7), and the four-dimensional sample stage (10) respectively.

4. The optical second harmonic intelligent characterization system as described in claim 1, characterized in that, The outgoing optical path unit includes, in sequence along the reflection direction of the reflected beam, an adjustable detection objective (12), a second five-dimensional adjustment frame (13), at least two second fast-reflecting mirrors (14), a polarizer (15), and a frequency doubling filter (16).

5. The optical second harmonic intelligent characterization system as described in claim 4, characterized in that, The control unit is connected to the four-dimensional sample stage (10), the second five-dimensional adjustment frame (13), and the second fast-reflecting mirror (14) respectively.

6. The optical second harmonic intelligent characterization system as described in claim 1, characterized in that, The FROG detection module (17) includes: The beam splitter module (1701) is used to receive the second harmonic beam, split the second harmonic beam, and obtain and output a synchronized first beam signal and second beam signal. The delay module (1702) is used to receive the second beam signal, delay the second beam signal by adjusting the optical path, and obtain and output the delayed beam signal; An SHG crystal (1705) is used to receive the first beam signal and the delayed beam signal, mix the first beam signal and the delayed beam signal to generate a sum-frequency signal, and output the sum-frequency signal as the SHG signal. A spectrometer (1707) is used to receive the SHG signal, acquire and output the spectrum of the SHG signal as a function of delay time—that is, the FROG trace plot. The PC (1708) is used to receive the FROG trace map and adjust the delay time of the second beam signal by the delay module (1702).

7. The optical second harmonic intelligent characterization system as described in claim 6, characterized in that, The PC terminal (1708) is communicatively connected to the spectrometer (1707), and the PC terminal (1708) is controllably connected to the delay module (1702).

8. A method for adjusting and controlling the optical second harmonic intelligent characterization system according to any one of claims 1-8, characterized in that, The method is configured in the control unit, which uses the method to precisely adjust the incident light path unit, the four-dimensional sample stage (10) and the outgoing light path unit to adjust the intensity of the SHG signal after the optical second harmonic intelligent characterization system replaces the sample (11) and / or after Z-scan.

9. The adjustment and control method as described in claim 8, characterized in that, The method is a pre-improved DDPG algorithm, which includes the following steps: S100: Construct and initialize the Actor network, Critic network, target Actor network, and target Critic network, specifically: S101: Randomly initialize the Actor network and Critic network Then, set the target Actor network. Target Critic network ; S102: Initialize the attenuation factor Soft update coefficient The maximum number of iterations T, the cumulative number of iterations t, the random noise function N, and the number of samples M; Among them, attenuation factor Used to solve for the objective Q-value, soft update coefficients Used for updating and The maximum number of iterations T represents the total number of times the network is trained and its parameters are updated; the random noise function N represents Gaussian noise, which is used to introduce a certain degree of exploration during the learning process, using noise to explore the action space; the number of samples M represents the number of samples taken from the experience replay pool during training. S103: Define the state space and action space based on the control parameters of the incident optical path unit, the four-dimensional sample stage (10) and the outgoing optical path unit; S200: Start iterative training, set the state Input into the Actor network to obtain actions ; where state S t Represents the current state and action in the t-th iteration of training. Representing state S t The following action; S300: Perform action , get the state and rewards ; where state S t+1 This represents the next state in the t-th iteration of training; S400: The obtained The samples are stored in the experience replay pool; S500: If the number of samples stored in the experience replay pool is greater than the number of sampled samples M, firstly, M samples are randomly sampled from the experience replay pool, and then the target Q value is calculated. The formula for calculating the target Q value is as follows: in, Represents the target Q value; S600: Obtained using a Critic network The Q-value is calculated using the mean squared error loss function as the Critic loss function, as follows: ； Minimize using gradient descent algorithm To update the parameters of the Citic network ; S700: The loss function for the Actor network is set as follows: ； Optimize the parameters of the Actor network by maximizing the Q-value. ; S800: Soft update target Actor network and the target Critic network ,as follows: ； S900: Repeat S200 to S800 until the cumulative number of iterations t is greater than or equal to the maximum number of iterations T, then end the iterative training.

10. The adjustment and control method as described in claim 9, characterized in that, The reward The calculation process is as follows: S301, Calculate light intensity to increase reward ,as follows: ； in, The bonus for increasing light intensity represents a positive reward given when the intensity of the SHG signal increases compared to the previous time step; The strength of the SHG signal at time step t; It is a positive coefficient; S302, Calculate the light intensity maintenance reward ,as follows: ； in, The light intensity retention reward means that when the change in the SHG signal compared to the previous time step is less than or equal to the tolerance threshold, a positive reward is given; otherwise, a zero reward is given. It is a positive coefficient. The set tolerance threshold is used to evaluate the variation amplitude of the SHG signal within adjacent time steps; otherwise, it represents other. S303, Calculate the light intensity reduction reward ,as follows: ； in, The light intensity reduction reward represents a negative reward given when the SHG signal decreases compared to the previous time step; It is a negative coefficient, used to provide a negative reward when the SHG signal decreases; S304, Calculate the reward for achieving the target light intensity. ,as follows: ； in, To achieve the target light intensity reward, the SHG signal must reach or exceed a preset target value. An additional positive reward is given at that time; It is a positive coefficient; S305, Calculate Rewards ,as follows: ； Among them, rewards Increase the reward for light intensity Light intensity retention reward Light intensity reduction reward and target light intensity reward The total reward obtained by accumulating the total amount.

Images