Genetic algorithm-based GaN HEMT device model parameter automatic extraction method
By using a dual-modal data and iterative solution method based on genetic algorithms, the electrical and thermal parameters of gallium nitride high electron mobility transistors (GaN HEMT) devices are decoupled, solving the parameter coupling problem in traditional methods and achieving efficient and accurate parameter extraction and model prediction.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-01-16
- Publication Date
- 2026-05-01
AI Technical Summary
Existing technologies struggle to effectively decouple the electrical and thermal parameters of gallium nitride high electron mobility transistors (GaN HEMTs) under high power conditions, resulting in complex and low-precision parameter extraction. Traditional methods are time-consuming and sensitive to initial values, while deep learning methods lack physical interpretability and generalization ability.
A genetic algorithm-based approach was adopted, combining dual-mode data from pulse and DC tests. A model was constructed using ASM-HEMT theory, and iterative solutions were obtained using mobility degradation, channel length modulation, and self-heating effects to achieve physical-level decoupling of electrical and thermal parameters. A two-stage optimization was performed using a genetic algorithm.
It significantly improves the accuracy and efficiency of parameter extraction, ensures the physical interpretability and generalization ability of the model, reduces the dependence on expert experience, and makes the simulation of the electrothermal coupling characteristics of the model under different working conditions more accurate.
Smart Images
Figure CN121960338A_ABST
Abstract
Description
An Automatic Parameter Extraction Method for GaN HEMT Device Models Based on Genetic Algorithm Technical Field
[0001] This invention belongs to the field of semiconductor device modeling technology, and relates to an automatic parameter extraction method for GaN HEMT device models based on genetic algorithms. Background Technology
[0002] Gallium nitride high electron mobility transistors (GaN HEMTs) are core components in fifth-generation mobile communication technology and power electronics, attracting widespread attention due to their high power density and high-frequency characteristics. However, accurate modeling of these devices has always faced significant challenges, especially under high-power operating conditions. The significant self-heating effect leads to a high degree of coupling between the electrical and thermal characteristics of the device, making the parameter extraction process exceptionally complex.
[0003] Current parameter extraction techniques mainly fall into two categories. Traditional physical extraction methods rely on expert experience for manual tuning or use gradient-based local optimization algorithms such as the Levenberg-Marquardt method. These methods are not only time-consuming and inefficient, but also highly sensitive to initial parameter values. When dealing with the strong nonlinear characteristics of gallium nitride HEMT devices, they struggle to overcome the coupling relationships between multiple parameters and are prone to getting trapped in local optima. The other category, black-box modeling methods based on deep learning, directly fit test data using neural networks. While exhibiting high accuracy within a specific data range, these models lack physical interpretability, have poor generalization ability, and cannot reliably predict unseen physical conditions.
[0004] With the rapid development of artificial intelligence technology, integrating integrated circuit design and manufacturing with intelligent algorithms has become an industry trend. However, existing methods have failed to effectively solve the coupling problem between electrical and thermal parameters under DC testing conditions, leading to extraction results that deviate from the true physical values. Therefore, the industry urgently needs a new method that can improve efficiency by leveraging the global search capabilities of artificial intelligence while maintaining the rigor and interpretability of the physical model. This invention is proposed against the backdrop of this technological bottleneck. Summary of the Invention
[0005] In view of this, the purpose of this invention is to provide an automatic parameter extraction method for GaN HEMT device models based on genetic algorithms. By deeply integrating artificial intelligence technology with semiconductor device physics, and utilizing pulse test data and self-heating effect iterative models, the method achieves physical-level decoupling of electrical and thermal parameters, thereby improving extraction accuracy and model generalization ability.
[0006] To achieve the above objectives, the present invention provides the following technical solution:
[0007] An automatic parameter extraction method for GaN HEMT device model based on genetic algorithm includes: Step 1, constructing a GaN HEMT simulation model with multiple physical effects coupled: Based on ASM-HEMT theory, establishing the drain current containing multiple key physical mechanisms. Computational model. The model takes into account the following effects: (1) Mobility degradation effect: introduces a mobility degradation coefficient. (1) Describe the phenomenon of decreased carrier mobility caused by enhanced vertical electric field; (2) Channel length modulation effect: introduce channel length modulation coefficient to describe the phenomenon of non-zero current slope caused by shortening of effective channel length in saturation region; (3) Self-heating effect: introduce thermal resistance to describe the phenomenon of junction temperature rise caused by power dissipation and parameter degradation of device under high power operation.
[0008] The device junction temperature is determined through the following iterative process. and current The equilibrium point:
[0009]
[0010]
[0011] in, For power dissipation, The ambient temperature.
[0012] The model has a mode switch. When the switch is off, it forces... To simulate ideal pulse test conditions without self-heating; when the switch is turned on, the above self-heating iterative calculation is performed to simulate DC test conditions with self-heating.
[0013] Step 2: Obtain pulsed and DC dual-mode test data for the device under test: Perform an IV scan on the device using nanosecond or microsecond-level narrow pulses. Due to the extremely short pulses, the device does not have enough time to accumulate heat, and it can be considered to have no self-heating effect; perform an IV test on the device using a conventional DC scan. The device heats up sufficiently, and the self-heating effect is significant.
[0014] Step 3, First Stage Parameter Extraction: Using only pulse data, set... And a genetic algorithm was used to analyze the electrical parameter set. Global optimization is performed, which utilizes pulse data from the linear region for extraction. and Extracting pulse data from the saturation region , and .
[0015] Step 4, Second Stage Parameter Extraction: Using DC data, activate the self-heating switch of the physical simulation model and lock all electrical parameters extracted in Step 3 to prevent them from changing. Then, use a genetic algorithm to extract only thermal parameters. Perform single-parameter optimization.
[0016] Step 5, Model Validation and Output: Substitute the extracted complete parameter set into the model and predict the pulse validation set and DC validation set data that were not used in training, respectively, to evaluate the model's generalization error. If the error meets the requirements, output the final parameter set.
[0017] Preferably, the constructed model is a dual-mode physical model: a GaN HEMT physical simulation model with an iterative solution mechanism for self-heating effect is established. This model has a self-heating switch and can simulate the pulse mode without self-heating and the DC mode with self-heating respectively.
[0018] Preferably, the dataset is bimodal data, consisting of pulse IV test data and DC IV test data of the device under test.
[0019] Preferably, the decoupling process is a two-stage decoupling. The first stage, based on pulse IV data, disables the model's self-heating effect and uses a genetic algorithm to extract and lock the set of electrical parameters of the device. The second stage, based on DC IV data, enables the model's self-heating effect and, after locking the set of electrical parameters, uses a genetic algorithm to extract the thermal parameters of the device separately.
[0020] Preferably, the GaN HEMT physical simulation model includes correction mechanisms for mobility degradation effects, channel length modulation effects, and self-heating effects, specifically expressed as follows: Parameters in the mobility degradation effect The expression is:
[0021] in, For low field mobility, The first-order mobility degradation coefficient, This is the second-order mobility degradation coefficient. For an effective vertical electric field.
[0022] Drain current in channel length modulation effect The expression is:
[0023] in, This is the uncorrected drain current. The channel length modulation coefficient, This is the drain-source voltage. This is the effective drain-source voltage.
[0024] (3) Device junction temperature in self-heating effect The expression is:
[0025] in, For ambient temperature, For thermal resistance, This indicates the power dissipation of the device.
[0026] Preferably, the self-heating effect is calculated through iterative solution: in each simulation step, the new junction temperature is calculated using the current value from the previous iteration. Then, the temperature-dependent model parameters are updated using the new junction temperature. These temperature-dependent model parameters include at least: temperature-dependent mobility.
[0027] Temperature-dependent saturation rate:
[0028] Repeat the above calculations until the junction temperature is reached. convergence.
[0029] The beneficial effects of this invention are as follows: (1) This invention achieves a fundamental breakthrough in parameter extraction methods by deeply integrating the global optimization capability of genetic algorithms with the rigor of GaN HEMT physical models. This method effectively overcomes the inherent limitations of traditional optimization techniques, such as sensitivity to initial values and susceptibility to local optima, and significantly improves the intelligence level and solution efficiency of the parameter extraction process.
[0030] (2) By adopting a dual-mode data strategy combining pulse testing and DC testing, and with the intelligent switching mechanism of the model's self-heating switch, physical-level decoupling of electrical and thermal parameters was successfully achieved. This innovative design fundamentally solves the technical problem of multi-physics coupling in high-dimensional parameter space, resulting in higher physical authenticity and reliability of the extracted results.
[0031] (3) The iterative solution framework based on ASM-HEMT theory can accurately simulate the electrothermal coupling characteristics of devices under different operating conditions. This model not only retains the interpretability of physical parameters, but also ensures the consistency between parameters through a step-by-step extraction strategy, which greatly enhances the prediction accuracy of the model in practical applications.
[0032] (4) The automated extraction process established in this invention significantly reduces the reliance on expert experience, making the modeling of high-performance GaNHEMT devices more standardized and repeatable. This method provides a novel technical path for the field of semiconductor device modeling and plays an important role in promoting the development of wide-bandgap semiconductor technology.
[0033] (5) Through a two-stage optimization mechanism, the pure extraction of electrical parameters under non-self-heating conditions is ensured, and the accurate calibration of thermal parameters in real working scenarios is achieved. This design enables the final model parameters to accurately characterize both the transient and steady-state characteristics of the device, providing more comprehensive simulation support for circuit design.
[0034] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description
[0035] To make the objectives, technical solutions, and advantages of this invention clearer, the following is a preferred detailed description of the invention in conjunction with the accompanying drawings, wherein: Figure 1 is an overall flowchart of the method of this invention; Figure 2 is a schematic diagram of the GaN HEMT structure; Figure 3 is a flowchart of the two-stage optimization parameters of the genetic algorithm of this invention; Figure 4 is a logical schematic diagram of the iterative solution mechanism for the self-heating effect in this invention; Figure 5 is a comparison diagram of the fitting curve and target data of this invention; Figure 6 is a trend diagram of RMSE changes during the iterative optimization process of this invention. Detailed Implementation
[0036] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Unless otherwise specified, the following embodiments and features can be combined with each other.
[0037] The accompanying drawings are for illustrative purposes only and are schematic diagrams, not actual pictures. They should not be construed as limiting the invention. To better illustrate the embodiments of the invention, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the actual product dimensions. It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.
[0038] In the accompanying drawings of the embodiments of the present invention, the same or similar reference numerals correspond to the same or similar components. In the description of the present invention, it should be understood that if terms such as "upper," "lower," "left," "right," "front," and "rear" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, they are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, the terms used to describe positional relationships in the drawings are only for illustrative purposes and should not be construed as limiting the present invention. For those skilled in the art, the specific meaning of the above terms can be understood according to the specific circumstances.
[0039] First, please refer to Figures 1 to 4 to extract parameters for a GaN HEMT device using a 0.25µm process.
[0040] Step 1: Perform pulse (500ns pulse width) and DC IV tests at temperatures of 300K, 350K, and 400K respectively. Test voltage From -2V to 0.5V, in 0.5V steps; The data ranges from 0V to 5V, with a step size of 0.1V. There are a total of 900 data sets, of which 700 are used as the training set and 200 as the validation set.
[0041] Step 2, First Stage Optimization: Configure genetic algorithm parameters: population size 300, iterations 2000. Optimize for pulse data and extract electrical parameters: , , , .
[0042] At this point, the model can perfectly fit the pulse curve, but the predicted value is too high in the DC high power region.
[0043] Step 3: Lock in the above electrical parameters and optimize based on the DC data, optimizing only the parameters. The thermal resistance was obtained through optimization using a genetic algorithm. .
[0044] Step 4: Result Validation: The extracted parameters are used to predict the validation set and plot the IV characteristic curve, as shown in Figure 5. It can be seen that the curve fitted by this invention (solid line portion) is extremely close to the target data. Furthermore, the RMSE of this invention on the validation set is 0.000713. Meanwhile, compared to traditional single-stage extraction methods (… (The error is typically >20%), but this method reduces the thermal resistance extraction error to 1.15%, verifying the effectiveness and high accuracy of the method. Figure 6 shows the RMSE change trend during the iterative optimization process of this invention.
[0045] This invention proposes a method for automatically extracting high-precision GaN HEMT SPICE model parameters, enabling high-precision simulation of electrothermal coupling characteristics. By acquiring pulsed IV and DC IV dual-mode data, a genetic algorithm is used to perform a two-stage step-by-step extraction of ASM-HEMT physical model parameters, and a self-heating iterative solution mechanism is established to simulate the dynamic influence of junction temperature on device parameters. This model accurately reflects the current degradation characteristics of GaN HEMT under different test modes, effectively decoupling the strong correlation between electrical and thermal parameters. The simulation results of the extracted parameters show good agreement with the measured curves of the device in both pulsed and DC modes. In particular, the thermal resistance extraction error is reduced to within 1.15%, demonstrating the accuracy and physical reliability of the method.
[0046] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.
Claims
1. A method for automatically extracting GaN HEMT device model parameters based on genetic algorithm, characterized in that: The process includes the following steps: Step 1, constructing a GaN HEMT simulation model with multi-physical effect coupling. The model is based on the Advanced Semiconductor Model-High Electron Mobility Transistor (ASM-HEMT) theory and includes correction mechanisms for mobility degradation effects, channel length modulation effects, and self-heating effects. Step 2, acquiring pulse IV test data and DC IV test data of the device under test. Step 3, first-stage parameter extraction: using only the pulse IV test data, disabling the self-heating effect, and using a genetic algorithm to globally optimize the electrical parameter set. Step 4, second-stage parameter extraction: using the DC IV test data, enabling the self-heating effect, and locking the electrical parameter set extracted in Step 3, using a genetic algorithm to optimize the thermal parameters individually. Step 5, model verification and output: substituting the extracted complete parameter set into the model, predicting the validation set data, evaluating the error, and outputting the parameter set.
2. The method for automatic extraction of GaN HEMT device model parameters based on genetic algorithm according to claim 1, characterized in that: The GaN HEMT simulation model is a dual-mode physical model with a self-heating switch; when the self-heating switch is closed, the device junction temperature is forced to a certain level. Equal to ambient temperature To simulate conditions without self-heating; when the self-heating switch is turned on, perform self-heating iterative calculations to simulate conditions with self-heating.
3. The method for automatic extraction of GaN HEMT device model parameters based on genetic algorithm according to claim 1, characterized in that: The pulse IV test data is obtained through nanosecond or microsecond-level narrow pulse IV scanning; the DC IV test data is obtained through conventional DC IV scanning.
4. The method for automatic extraction of GaN HEMT device model parameters based on genetic algorithm according to claim 1, characterized in that: The first stage parameter extraction and the second stage parameter extraction constitute a two-stage decoupled process; The first stage parameter extraction extracts the electrical parameter set based on the pulse IV test data with the self-heating effect disabled; the second stage parameter extraction extracts the thermal parameters based on the DC IV test data with the self-heating effect enabled.
5. The method for automatic extraction of GaN HEMT device model parameters based on genetic algorithm according to claim 1, characterized in that: The effective mobility of the mobility degradation effect in the GaN HEMT simulation model The expression is: in, For low field mobility, The first-order mobility degradation coefficient, This is the second-order mobility degradation coefficient. For an effective vertical electric field; drain current due to channel length modulation effect. The expression is: in Uncorrected drain current The channel length modulation coefficient, This is the drain-source voltage. Effective drain-source voltage; device junction temperature due to self-heating effect. The expression is: in For ambient temperature, This is for thermal resistance.
6. The method for automatic extraction of GaN HEMT device model parameters based on genetic algorithm according to claim 2, characterized in that: The self-heating iterative calculation includes: utilizing dissipated power Calculate junction temperature And update temperature-related parameters; the temperature-related parameters include temperature-dependent mobility. Temperature-dependent saturation rate ,in For the nominal temperature; repeat the above calculations until... convergence.