Method for inhibiting dislocation defect of VGF method indium phosphide

Abstract

The invention belongs to the technical field of semiconductor processing and preparation, and particularly relates to a method for inhibiting VGF indium phosphide dislocation defect, which comprises the following steps: optimizing a local thermal field and a low concave solid-liquid interface through indium phosphide single crystal thermal field simulation and unsteady single crystal growth simulation; then, real-time monitoring of single crystal growth key parameters, namely actual temperature or cooling rate and model parameters is realized by establishing a low-dislocation indium phosphide single crystal growth model and an adaptive fuzzy PID algorithm, and deviation automatic fitting correction control is combined; the influence of uncontrollable factors such as thermal field attenuation between different single crystal furnaces and different rounds in production scheduling is avoided, the controllability of the thermal field and single crystal growth is improved, and the deviation range of the growth temperature or the cooling rate in the indium phosphide single crystal growth process is controlled.

Description

Technical Field

[0001] This invention belongs to the field of semiconductor processing and preparation technology, and specifically relates to a method for suppressing VGF indium phosphide dislocation defects. Background Technology

[0002] Indium phosphide (InP) is a new generation of microelectronic and optoelectronic functional materials following silicon (Si) and gallium arsenide (GaAs). The dislocation index of InP directly determines the performance and yield of devices. The VGF method, due to its low thermal stress advantage, has become the mainstream technology for preparing large-size InP single crystals. However, controlling the dislocation density in production practice remains a significant challenge. As a lattice defect, the multiplication and movement of dislocations follow the basic laws of plastic deformation. Current technology suggests that the generation of dislocations is due to: 1) During crystal growth, the radial and axial temperature gradients will cause uneven cooling and shrinkage of different parts of the crystal, generating thermoelastic stress. When this thermal stress exceeds the critical shear stress of indium phosphide crystal, it will induce dislocations. 2) Unstable growth interface, with unevenness or local abrupt changes, such as the faceting of the interface and the bending of the interface caused by the growth rate fluctuation in the early stage of growth of the conical part, will cause uneven stress distribution inside the crystal, which will lead to the generation and propagation of dislocations. 3) Lattice hardening or precipitation of a third phase induces dislocations. For example, sulfur-doped indium phosphide has higher elastic constants, Young's modulus, and shear modulus compared to intrinsic indium phosphide. The incorporation of sulfur atoms, after occupying P sites, forms stronger and shorter In-S bonds, which leads to lattice distortion around them and generates a local tensile strain field. This strain field interacts elastically with the stress field of dislocations, effectively pinning dislocation movement. Macroscopically, this manifests as an increase in critical shear stress. The dislocation density of heavily doped S is less than that of lightly doped S, which is less than that of undoped S. For iron-doped indium phosphide, due to the large segregation coefficient of iron atoms and the low solid solubility of iron atoms in InP, uniformly distributed iron atoms precipitate as FeP2 during crystal cooling growth due to supersaturation. These precipitates have a large mismatch with the lattice constant of the InP matrix, generating huge local stress concentrations around them, becoming powerful dislocation sources. 4) Seed crystal quality: The seed crystal itself contains dislocations, impurities or other defects, which will expand and multiply during crystal growth, leading to an increase in dislocation density.

[0003] Therefore, the preparation of indium phosphide by the VGF method requires consideration of dislocation suppression from multiple perspectives. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention proposes a method for suppressing VGF indium phosphide dislocation defects.

[0005] This method includes the following steps: S1, Indium phosphide thermal field simulation and unsteady single crystal growth simulation: By using thermal field simulation and unsteady single crystal growth simulation software, after completing the calibration of the thermal field simulation model through the steps of establishing an indium phosphide single crystal furnace model, finite cell analysis, material property matching and boundary condition setting, the existence of thermal field weak points is identified through unsteady single crystal growth simulation. S2, Optimization of local thermal field and concave solid-liquid interface of indium phosphide: For weak points in the thermal field, materials with excellent thermal insulation coefficient or thermal conductivity are used to insulate or conduct heat to the local area, and the concave solid-liquid interface is adjusted by adjusting the height of the heat shield and the position of the crucible. S3, Establishment of a long-term model to suppress VGF indium phosphide twin defects: Collect temperature data of growth from at least 30 sets of indium phosphide single crystal samples with low dislocations and high uniformity. The indium phosphide single crystal samples should be selected with an average dislocation of ≤100cm. -2 The 4-inch indium phosphide temperature data includes temperature distribution data, expected value and standard deviation statistical calculations at each temperature measurement point, to obtain the indium phosphide single crystal growth temperature and cooling rate model; S4, Adaptive Fuzzy PID Algorithm and Adaptive Fuzzy PID Control: During the growth of indium phosphide single crystals, actual temperature data is collected and combined with the indium phosphide single crystal growth temperature and cooling rate model to form temperature deviation e and temperature deviation change rate ec or cooling rate deviation ve and temperature change rate vec as input quantities; the proportional gain K of the PID is adjusted by the fuzzy control rule base and fuzzy algorithm. P Integral gain K I Differential gain K D Online adjustments are made by continuously monitoring the values ​​of temperature deviation e and temperature deviation change rate ec in real time through the fuzzy relationship between them, thereby improving the control effect of the temperature control system and achieving the control target of temperature deviation e and temperature deviation change rate ec on the parameters.

[0006] Based on the germanium single crystal growth temperature controlled by the adaptive fuzzy PID algorithm, we continuously collect more growth temperature data of indium phosphide samples with no twins, low dislocations, and high uniformity, and continuously iterate the indium phosphide single crystal growth temperature and cooling rate model to improve the accuracy of the model.

[0007] Specifically, the method for identifying whether there are weak points in the thermal field in S1 is to determine the weak points by stress distribution and temperature gradient.

[0008] Specifically, the method for localized heat preservation or heat conduction in S2 is as follows: the quartz tube capping area is located in the high-temperature zone of the thermal field and is insulated with fiber blankets; silicon carbide or graphite is used as a heat-conducting material in the crucible cone and is attached to the crucible cone for uniform heat conduction.

[0009] Specifically, the adaptive fuzzy PID algorithm in S4 is as follows: ; Where u(t) is the output of the control system; The error of the controlled object at time t; The rate of change of the error of the controlled object at time t; k P k I k D These are proportional gain, integral gain, and derivative gain. Adjust K online P K I K D : k P =K´ P +∆K P k I =K´ I +∆K I k D =K´ D +∆K D Where, ∆K P ∆K I ∆K D These correspond to the online changes in proportional gain, integral gain, and derivative gain, respectively.

[0010] The proportional gain enhances the system's responsiveness and rapidly adjusts system errors; the integral gain gradually eliminates steady-state errors; and the derivative gain predicts the trend of system error changes in advance, thereby eliminating errors and suppressing oscillations generated during regulation. Compared to conventional PID algorithms, the adaptive fuzzy PID algorithm incorporates fuzzy control logic, resulting in higher stability, flexibility, and robustness. It is particularly effective for controlled objects with significant time-varying and nonlinear characteristics. Furthermore, incremental adjustments to the PID parameters based on the input parameters enhance the adaptive performance of the PID controller, enabling online adjustment of PID parameters to suit different operating environments and higher control precision requirements.

[0011] Specifically, the adaptive fuzzy PID control in S4 is as follows: Temperature error e and temperature error rate ec are selected as the input variables of the adaptive fuzzy PID controller, with K... P K I K D Online variation of the three parameters ∆K P ∆K I ∆K DAs the output variable of the adaptive fuzzy PID controller; define e and ec as fuzzy linguistic variables E and EC, respectively, with the basic universe of discourse [-6, 6]; ∆K P ∆K I ∆K D The corresponding fuzzy linguistic variable is K. P K I K D The basic universe of discourse is the same as that of its fuzzy linguistic variables, which are [0.64, 1.2], [8, 12], and [2, 3], respectively; the quantization factors of the input variables e(t) and ∆e(t) are obtained as follows: ; Output variable ∆K P ∆K I ∆K D The scaling factors are as follows: ; Based on the characteristics of the single-crystal temperature control system and the above requirements, a high-resolution triangular membership function is selected. Therefore, the input and output variables of the adaptive fuzzy PID algorithm are selected as 7 linguistic variables: NB (negative large), NM (negative medium), NS (negative small), ZO (zero), PS (positive small), PM (positive medium), and PB (positive large).

[0012] PID control is characterized by its simple structure, convenient design, and strong stability. However, its control effect is poor under conditions of nonlinearity, time-varying factors, and numerous disturbances, making it difficult to meet practical control requirements.

[0013] The beneficial effects of this invention are as follows: 1) Based on thermal field simulation and unsteady single crystal growth simulation, the weak points of the thermal field physical structure are identified in a targeted manner, and the thermal field structure is systematically optimized and verified; for the common cone region of PBN crucible, the thermal field can be quickly optimized by adding heat insulation or thermal conductive materials, which is simple to operate and easy to implement; 2) By establishing a model of germanium single crystal growth temperature and cooling rate, key control parameters in the single crystal growth process are identified, and fuzzy control rules for temperature or cooling rate are formulated. Fuzzy algorithms are then used to control the three parameters K of the PID controller. P K I K D Online adjustments were made, enabling real-time monitoring of key parameters for single crystal growth, namely actual temperature or cooling rate, and model parameters. Combined with automatic deviation fitting and correction control, the influence of uncontrollable factors such as thermal field decay between different single crystal furnaces and different batches during production scheduling was avoided, improving the controllability of the thermal field and single crystal growth, and effectively suppressing dislocation defects. Attached Figure Description

[0014] Figure 1 This is a schematic flowchart of the method for suppressing low-dislocation defects in indium phosphide using the VGF method according to the present invention.

[0015] Figure 2 This is a schematic diagram of the thermal field simulation and unsteady single crystal growth simulation of VGF indium phosphide.

[0016] Figure 3 This diagram illustrates the large dislocation density and stress distribution on the shoulder of the VGF indium phosphide thermal field PBN crucible.

[0017] Figure 4 Temperature and cooling rate models for typical temperature measurement points of VGF indium phosphide.

[0018] Figure 5 This is a schematic diagram of adaptive PID fuzzy control.

[0019] Figure 6 A schematic diagram of the fitting and correction interface for adaptive PID fuzzy control.

[0020] Figure 7 The dislocation diagrams of indium phosphide crystals before and after applying this VGF indium phosphide single crystal growth dislocation suppression method are shown. Detailed Implementation

[0021] To better illustrate the purpose, technical solution, and advantages of this invention, specific embodiments will be provided below to further explain the invention. The purpose of this explanation is to provide a more detailed understanding of the invention, not to limit it. Other embodiments obtained by those skilled in the art without inventive effort are all within the scope of protection of this invention.

[0022] S1: Thermal field simulation and unsteady single crystal growth simulation. A transient heat transfer model was established using the CGsim finite element method. Through calibration of the geometric model, process parameters, physical property parameters, boundary conditions, heater power, and other actual and simulated conditions, the thermal field simulation and calibration were completed. Furthermore, combined with the single crystal growth process, unsteady single crystal growth simulation was conducted, obtaining visualized results such as temperature gradient distribution, solid-liquid interface, dislocation distribution, and stress distribution during VGF single crystal growth. Figure 2 As shown.

[0023] S2: Thermal Field Optimization and Solid-Liquid Interface Control at Low-Concave Surfaces: For weak points in the thermal field, materials with excellent thermal insulation or thermal conductivity are used to insulate or conduct heat to the local structure. Simulation results show that a large temperature gradient and stress distribution exist in the conical region of the PBN crucible, such as... Figure 3 As shown.

[0024] S3: Optimization of thermal conductivity of crucible conical surface and control of solid-liquid interface of concave surface: In the conical region of PBN crucible, silicon carbide, diamond or graphite materials are preferably used to bond with the conical surface of PBN crucible to increase the thermal conductivity of the conical surface.

[0025] S4: Establishment of a low-dislocation, high-uniformity indium phosphide single crystal growth model. Multiple sets of temperature change data were collected throughout the entire single crystal growth process of low-dislocation, high-uniformity indium phosphide single crystal growth samples. This process covers data from all temperature points throughout the single crystal growth process. In this example, measurement point TC2 was selected. During data acquisition, large-scale temperature data was first collected at the TC2 temperature measurement point. Abnormal data removal, interpolation to fill missing values, and temperature data differentiation to convert to cooling rate were then performed to form a model. Figure 5 The TC2 continuous spline curve is shown. This spline curve includes the optimal cooling rate curve, the upper and lower limits of the 60% confidence interval, and determines the minimum or maximum threshold of the cooling rate. Integrating the cooling rate curves at all temperature points yields the crystal growth model.

[0026] S5: Adaptive Fuzzy PID Algorithm and Adaptive Fuzzy PID Control. Specifically, the temperature deviation and its rate of change are first divided into fuzzy subsets such as "negative large," "negative medium," "negative small," "zero," "positive small," "positive medium," and "positive large," and a control strategy covering 7×7 rules, totaling 49 rules, is formulated. When the temperature deviation is "positive large" and the rate of change is "positive small," the rule table outputs a large positive control quantity to quickly reduce the temperature; while when the deviation is close to zero, a smaller control quantity is used for fine-tuning to avoid temperature overshoot. In actual operation, the PLC collects temperature data from multiple temperature zones in the furnace in real time through the analog input module, compares it with the optimal value after A / D conversion, and calculates the temperature deviation and rate of change. Subsequently, the PLC performs fuzzification processing on the input variables according to the preset membership function and quickly matches the output value of the corresponding fuzzy rule. Finally, this value is used as a control signal to drive the actuator to adjust the power of the resistance wire. A schematic diagram of the adaptive fuzzy PID control principle is shown below. Figure 6 As shown.

[0027] Throughout the single crystal growth process, the real-time cooling rate at the temperature measurement point is compared with the optimal cooling rate at the corresponding time point. The output value is corrected by inputting the deviation (e) and the rate of change of deviation (ec) and combining the above-mentioned fuzzy control principle. Finally, this value is used as a control signal to drive the actuator to adjust the temperature field power.

[0028] S5: Iteration of the single crystal growth model. The single crystal controlled by the adaptive fuzzy algorithm will form a new temperature data cluster. Continuous iteration of the indium phosphide temperature model or cooling rate model will improve the accuracy of the model.

Claims

1. A method of inhibiting VGF indium phosphide dislocation defects, characterized by The method comprises the following steps: S1, indium phosphide thermal field simulation and non-steady-state single crystal growth simulation: through the help of thermal field simulation and non-steady-state single crystal growth simulation software, through the steps of indium phosphide single crystal furnace model establishment, finite element analysis, material property matching and boundary condition setting, after the calibration of the thermal field simulation model, through non-steady-state single crystal growth simulation, whether there is a weak point in the thermal field is identified; S2, local thermal field and low concave solid-liquid interface optimization of indium phosphide: for the weak point of the thermal field, the local area is heat-insulated or heat-conducted by using materials with excellent heat insulation coefficient or conduction coefficient, and the low concave solid-liquid interface is adjusted by adjusting the height of the heat shield and the position of the crucible; S3, long model establishment of inhibiting VGF InP twin defects: collect at least 30 sets of temperature data of growth of low dislocation and high uniformity InP single crystal samples, and the InP single crystal samples need to be selected with an average dislocation ≤100 cm -2 The temperature data of the 4-inch InP includes temperature distribution data, expected value and standard deviation statistical calculation of each temperature measurement point, and the InP single crystal growth temperature and cooling rate model is obtained; S4, adaptive fuzzy PID algorithm and adaptive fuzzy PID control: in the process of indium phosphide single crystal growth, the actual temperature data is collected, combined with the indium phosphide single crystal growth temperature and cooling rate model, the temperature deviation e, temperature deviation change rate ec or cooling rate deviation ve, temperature deviation change rate vec as input; By fuzzy control rule base and fuzzy algorithm on PID proportion gain K P , integral gain K I , differential gain K D Adjust online, through the fuzzy relationship between temperature deviation e and temperature deviation change rate ec, real-time continuous detection of temperature deviation e and temperature deviation change rate ec value, improve the control effect of temperature control system, so as to realize the control target of temperature deviation e and temperature deviation change rate ec to parameters.

2. The method of claim 1, wherein the VGF InP threading dislocation defects are suppressed by The method for identifying whether there is a weak point in the thermal field in S1 is: to judge the weak point of the thermal field through stress distribution and temperature gradient.

3. The method of claim 1, wherein the VGF InP threading dislocation defects are suppressed by The method for heat-insulating or heat-conducting the local area in S2 is: the quartz tube cap area in the high temperature zone of the thermal field is heat-insulated by using fiber blanket; the crucible cone part is uniformly heat-conducted by using silicon carbide or graphite as heat-conducting material and being attached to the crucible cone part.

4. The method of claim 1, wherein the VGF InP threading dislocation defects are suppressed by The adaptive fuzzy PID algorithm in S4 is: ； Wherein, u(t) is the output of the control system; Error of the controlled object at time t; error rate of the controlled object at the time instant; k P , k I , k D are proportional gain, integral gain, derivative gain; Adjust K online P , K I , K D : k P =K´ P +∆K P k I =K´ I +∆K I k D =K´ D +∆K D where ΔKp, ΔK1, ΔKd P , ΔKp I , ΔK1 D correspond to the online variation of the proportional gain, integral gain, and derivative gain, respectively.

5. The method of claim 1, wherein the VGF InP threading dislocation defects are suppressed by the addition of a dopant to the VGF InP crystal. 5 The adaptive fuzzy PID control in S4 is: Temperature error e and temperature error rate ec are selected as the input variables of the adaptive fuzzy PID controller, with K... P K I K D Online variation of the three parameters ∆K P ∆K I ∆K D As the output variable of the adaptive fuzzy PID controller; define e and ec as fuzzy linguistic variables E and EC, respectively, with the basic universe of discourse [-6, 6]; ∆K P ∆K I ∆K D The corresponding fuzzy linguistic variable is K. P K I K D The basic universe of discourse is the same as that of its fuzzy linguistic variables, which are [0.64, 1.2], [8, 12], and [2, 3], respectively; the quantization factors of the input variables e(t) and ∆e(t) are obtained as follows: ； The output variable ΔK P The scaling factors for the output variables ΔK I The scaling factors for the output variables ΔK D are respectively: ； Based on the characteristics of the single crystal temperature control system and the above requirements, a high-resolution triangular membership function is selected, so the input and output variables of the adaptive fuzzy PID algorithm are selected as 7 language variables: NB (negative big), NM (negative medium), NS (negative small), ZO (zero), PS (positive small), PM (positive medium) and PB (positive big).

Images