Semiconductor device electrical parameter prediction method, electronic device and storage medium

Abstract

The present invention provides a method, electronic device, and storage medium for predicting the electrical parameters of semiconductor devices. The method comprises: obtaining historical electrical parameter measurement data for semiconductor devices of different sizes; constructing a first training sample set based on the historical electrical parameter measurement data, wherein the first training sample includes measurement data and size data related to a first electrical parameter of small-sized devices, as well as size data and first electrical parameter measurement data of large-sized devices; training a first neural network model based on the first training sample set to obtain a corresponding first electrical parameter prediction model; and inputting the first electrical parameter measurement data and size data of the measured small-sized devices, as well as the size data of the large-sized device to be predicted, into the first electrical parameter prediction model to obtain first electrical parameter prediction data for the large-sized device to be predicted. The present invention can accurately predict the electrical parameters of a large-sized device based on the electrical parameters of the small-sized device.

Description

Technical Field

[0001] The present invention relates to the field of semiconductor processing and manufacturing technology, and in particular to a method for predicting electrical parameters of a semiconductor device, an electronic device, and a storage medium. Background Art

[0002] Against the backdrop of today's rapidly developing semiconductor industry, the advancement of advanced semiconductor processes has become a key area of ​​global technological competition. However, within this process, the efficiency bottleneck faced by manual data processing has become increasingly prominent, becoming a key constraint on further breakthroughs and upgrades in semiconductor manufacturing. From the real-time monitoring and collection of inline process parameters on the production line to the comprehensive evaluation and analysis of device electrical parameters, the entire data chain involves the flow and processing of massive amounts of information, and each link faces the dual dilemma of low efficiency and high costs.

[0003] Since the performance of semiconductor devices is affected by multiple factors such as material properties, process conditions, and device structure, their electrical parameters show a high degree of nonlinearity and multi-dimensional correlation. Therefore, to comprehensively and accurately evaluate device performance, it is necessary to conduct in-depth mining and correlation analysis of the entire link data from inline process parameters to device electrical parameters. Taking the existing 40HV platform as an example, since the platform needs to measure different sizes (different threshold voltages V t ) of the device (such as the on-state current I on , shutdown current I off , saturation threshold voltage V t,sat , linear region threshold voltage V t,lin Its huge test sample volume and complex parameter system lead to extended measurement and analysis cycles and increased test costs.

[0004] In this context, it is necessary to apply artificial intelligence technology to empower engineering teams. By adopting artificial intelligence technology, it can not only improve work efficiency and reduce wafer fab operating costs, but also promote the evolution of semiconductor manufacturing to a higher-level development stage through quality optimization and intelligent upgrades. Summary of the Invention

[0005] The purpose of the present invention is to provide a method, electronic device and storage medium for predicting the electrical parameters of semiconductor devices, which can accurately predict the electrical parameters of large-size devices based on the electrical parameters of small-size devices. This can not only effectively shorten the analysis cycle of the electrical parameters of large-size devices and reduce the testing cost of large-size devices, but also establish a new prediction paradigm for the development of semiconductor devices.

[0006] To achieve the above-mentioned objectives, the present invention provides a method for predicting electrical parameters of semiconductor devices, comprising: obtaining historical electrical parameter measurement data of semiconductor devices of different sizes; constructing a first training sample set based on the historical electrical parameter measurement data, wherein each first training sample in the first training sample set includes first electrical parameter-related measurement data and size data of small-size devices and size data and first electrical parameter measurement data of large-size devices, wherein the first electrical parameter measurement data of the large-size devices is the label of the first training sample; training a first neural network model based on the first training sample set to obtain a corresponding first electrical parameter prediction model; inputting the first electrical parameter-related measurement data and size data of the measured small-size devices and the size data of the large-size device to be predicted into the first electrical parameter prediction model to obtain the first electrical parameter prediction data of the large-size device to be predicted.

[0007] Optionally, the first electrical parameter includes a saturation region threshold voltage, an on-current and an off-current; and the measurement data related to the first electrical parameter of the small-size device includes the gate voltage and corresponding drain current sequence data of the small-size device measured under the condition that the drain voltage is equal to the supply voltage.

[0008] Optionally, the semiconductor device electrical parameter prediction method provided by the present invention also includes: constructing a second training sample set based on the historical measurement data of the electrical parameters, each second training sample in the second training sample set includes the second electrical parameter related measurement data and size data of the small-size device and the size data and second electrical parameter measurement data of the large-size device, wherein the second electrical parameter measurement data of the large-size device is the label of the second training sample, the second electrical parameter includes the linear region threshold voltage, and the second electrical parameter related measurement data of the small-size device includes the gate voltage and corresponding drain current sequence data of the small-size device measured under the condition that the drain voltage is lower than the power supply voltage; training the second neural network model according to the second training sample set to obtain the corresponding second electrical parameter prediction model; inputting the second electrical parameter related measurement data and size data of the measured small-size device and the size data of the large-size device to be predicted into the second electrical parameter prediction model to obtain the second electrical parameter prediction data of the large-size device to be predicted.

[0009] Optionally, the training of the first neural network model according to the first training sample set includes: training the first neural network model according to the first training sample set using a mixed precision training method; the training of the second neural network model according to the second training sample set includes: training the second neural network model according to the second training sample set using a mixed precision training method.

[0010] Optionally, the semiconductor device electrical parameter prediction method provided by the present invention further includes: obtaining short channel voltage prediction data of the large-size device to be predicted based on the first electrical parameter prediction data and the second electrical parameter prediction data of the large-size device to be predicted.

[0011] Optionally, both the first neural network model and the second neural network model are neural network models based on a self-attention mechanism.

[0012] Optionally, the semiconductor device electrical parameter prediction method provided by the present invention also includes: when a preset update condition is met, updating the first electrical parameter prediction model and / or the second electrical parameter prediction model based on newly acquired electrical parameter measurement data of multiple semiconductor devices.

[0013] Optionally, the semiconductor device electrical parameter prediction method provided by the present invention also includes: obtaining historical measurement data of process parameters of the semiconductor devices of different sizes; each of the first training sample and the second training sample also includes process parameter data of the small-size device and process parameter data of the large-size device; the first electrical parameter-related measurement data and size data of the measured small-size device and the size data of the large-size device to be predicted are input into the first electrical parameter prediction model, including: inputting the first electrical parameter-related measurement data, size data and process parameter data of the measured small-size device and the size data and process parameter data of the large-size device to be predicted into the first electrical parameter prediction model; the second electrical parameter-related measurement data and size data of the measured small-size device and the size data of the large-size device to be predicted into the second electrical parameter prediction model, including: inputting the second electrical parameter-related measurement data, size data and process parameter data of the measured small-size device and the size data and process parameter data of the large-size device to be predicted into the second electrical parameter prediction model.

[0014] To achieve the above-mentioned object, the present invention further provides an electronic device, comprising a processor and a memory, wherein the memory stores a computer program, and when the computer program is executed by the processor, the semiconductor device electrical parameter prediction method described above is implemented.

[0015] To achieve the above-mentioned object, the present invention further provides a readable storage medium, wherein the readable storage medium stores a computer program, and when the computer program is executed by a processor, the semiconductor device electrical parameter prediction method described above is implemented.

[0016] Compared with the prior art, the semiconductor device electrical parameter prediction method, electronic device and storage medium provided by the present invention have the following unexpected technical effects: the semiconductor device electrical parameter prediction method provided by the present invention first obtains historical electrical parameter measurement data of semiconductor devices of different sizes, then constructs a first training sample set based on the historical electrical parameter measurement data, and then trains a first neural network model based on the first training sample set to obtain a corresponding first electrical parameter prediction model. Finally, the first electrical parameter-related measurement data and size data of the measured small-size device and the size data of the large-size device to be predicted are input into the first electrical parameter prediction model to obtain the first electrical parameter prediction data of the large-size device to be predicted. Thus, the electrical parameters of the large-size device can be accurately predicted based on the electrical parameters of the small-size device, thereby effectively shortening the analysis cycle of the electrical parameters of the large-size device and reducing the testing cost of the large-size device, and establishing a new prediction paradigm for the development of semiconductor devices.

[0017] Since the electronic device and storage medium provided by the present invention belong to the same inventive concept as the semiconductor device electrical parameter prediction method provided by the present invention, the electronic device and storage medium provided by the present invention have at least all the beneficial effects of the semiconductor device electrical parameter prediction method provided by the present invention. For details, please refer to the relevant description of the beneficial effects of the semiconductor device electrical parameter prediction method provided by the present invention in the above text. Therefore, the beneficial effects of the electronic device and storage medium provided by the present invention will not be described one by one here. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is a flow chart of a method for predicting electrical parameters of a semiconductor device provided in one embodiment of the present invention.

[0019] Figure 2 This is a characteristic diagram showing the correlation between the threshold voltage change caused by drain-induced barrier lowering and the saturation region threshold voltage of different devices.

[0020] Figure 3 The figure shows the correlation between the on / off current ratio and the saturation threshold voltage of different devices.

[0021] Figure 4 This is a characteristic diagram showing the correlation between the short channel voltage and the saturation region threshold voltage of different devices.

[0022] Figure 5 The figure is a comparison diagram of the short channel voltage prediction result obtained by using the semiconductor device electrical parameter prediction method provided by the present invention and the short channel voltage actual measurement result.

[0023] Figure 6 A schematic block diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0024] The following is a further detailed description of the semiconductor device electrical parameter prediction method, electronic device, and storage medium proposed in the present invention, in conjunction with the accompanying drawings and specific embodiments. The advantages and features of the present invention will become more apparent from the following description. It should be noted that, in this document, relational terms such as first and second, etc., are merely used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations.

[0025] The core idea of ​​the present invention is to provide a method, electronic device and storage medium for predicting the electrical parameters of semiconductor devices, which can accurately predict the electrical parameters of large-size devices based on the electrical parameters of small-size devices. This can not only effectively shorten the analysis cycle of the electrical parameters of large-size devices and reduce the testing cost of large-size devices, but also establish a new prediction paradigm for the development of semiconductor devices.

[0026] It should be noted that the semiconductor device electrical parameter prediction method provided by the present invention can be applied to the electronic device provided by the present invention, wherein the electronic device can be a personal computer, a mobile terminal, etc., and the mobile terminal can be a mobile phone, a tablet computer, and other hardware devices with various operating systems.

[0027] To realize the above idea, the present invention provides a method for predicting the electrical parameters of semiconductor devices. Figure 1 , which is a flow chart of a method for predicting electrical parameters of a semiconductor device provided by one embodiment of the present invention. Figure 1 As shown, the semiconductor device electrical parameter prediction method provided by the present invention includes the following steps: step S100, obtaining historical electrical parameter measurement data of semiconductor devices of different sizes; step S200, constructing a first training sample set based on the historical electrical parameter measurement data, each first training sample in the first training sample set includes first electrical parameter-related measurement data and size data of small-size devices and size data and first electrical parameter measurement data of large-size devices, wherein the first electrical parameter measurement data of the large-size devices is the label of the first training sample; step S300, training a first neural network model based on the first training sample set to obtain a corresponding first electrical parameter prediction model; step S400, inputting the first electrical parameter-related measurement data and size data of the measured small-size devices and the size data of the large-size device to be predicted into the first electrical parameter prediction model to obtain the first electrical parameter prediction data of the large-size device to be predicted.

[0028] It can be seen that by adopting the semiconductor device electrical parameter prediction method provided by the present invention, the electrical parameters of large-size devices can be accurately predicted based on the electrical parameters of small-size devices, thereby not only effectively shortening the analysis cycle of the electrical parameters of large-size devices and reducing the testing cost of large-size devices, but also establishing a new prediction paradigm for the development of semiconductor devices.

[0029] It should be noted that the small-size device can be a Core device (core device, including LVT (low threshold voltage transistor), SVT (standard threshold voltage transistor), HVT (high threshold voltage transistor) and PD (pull-down) / PG (transmission gate) unit of 6T-SRAM (six-transistor static random access memory)), and the large-size device can be an MV device (medium voltage device) or an HV device (high voltage device). It should also be noted that the dimensional parameters of the small-size device include the gate length and gate width of the small-size device, and the dimensional parameters of the large-size device include the gate length and gate width of the large-size device. In addition, it should be noted that the historical measurement data of the electrical parameters of semiconductor devices of different sizes collected are derived from the same process platform (for example, a 40nm HV platform). In actual applications, the electrical parameter-related measurement data of the measured small-size devices from the same process platform can be used to predict the electrical parameters of large-size devices.

[0030] In some exemplary embodiments, the method for predicting electrical parameters of a semiconductor device provided by the present invention further includes: preprocessing the historical measurement data of the electrical parameters.

[0031] Therefore, by preprocessing the historical measurement data of electrical parameters of semiconductor devices of different sizes, data of different scales can be normalized to the same scale, which can facilitate subsequent model training.

[0032] In some exemplary embodiments, the first electrical parameter includes a saturation region threshold voltage, an on-current, and an off-current; and the measurement data related to the first electrical parameter of the small-size device includes a gate voltage and a corresponding drain current sequence data of the small-size device measured under the condition that the drain voltage is equal to the supply voltage.

[0033] Since the saturation threshold voltage V t,sat Reflects the threshold characteristics of the device in the saturation region; the on-state current I on Characterizes the maximum conduction performance of the device and determines the switching speed; the off current I off It reflects the static power consumption level of the device and affects the energy efficiency. Therefore, by predicting these three parameters at the same time, the comprehensive performance of the device under high voltage working state can be fully evaluated. d Equal to the supply voltage V ddThe gate voltage V of the small-size device measured under the conditions of g and drain current I d Sequence data (V g -I d Curve) predicts the saturation threshold voltage V of large-size devices t,sat , conduction current I on , shutdown current I off , which can ensure that the first electrical parameter prediction model can be based on the gate voltage V g and drain current I d Sequence data, intuitively extract high-order physical features (such as transconductance g m , subthreshold swing SS, etc.), which can effectively ensure the predicted saturation threshold voltage V t,sat , conduction current I on , shutdown current I off accuracy.

[0034] In some exemplary embodiments, the training the first neural network model according to the first training sample set includes: training the first neural network model according to the first training sample set using a mixed precision training method.

[0035] Because mixed-precision training combines the advantages of half-precision (FP16) and single-precision (FP32) floating-point numbers, it can accelerate model convergence while ensuring numerical accuracy, significantly improving the training rate with almost no loss in the performance of the first neural network model.

[0036] In some exemplary embodiments, the method for predicting electrical parameters of a semiconductor device provided by the present invention further includes: acquiring historical measurement data of process parameters of the semiconductor devices of different sizes.

[0037] Correspondingly, each of the first training samples also includes the process parameter data of the small-size device and the process parameter data of the large-size device.

[0038] The step of inputting the measurement data and size data related to the first electrical parameter of the measured small-size device and the size data of the large-size device to be predicted into the first electrical parameter prediction model includes: inputting the measurement data related to the first electrical parameter of the measured small-size device (specifically, the gate voltage and corresponding drain current sequence data of the measured small-size device measured under the condition that the drain voltage is equal to the power supply voltage), size data and process parameter data, as well as the size data and process parameter data of the large-size device to be predicted into the first electrical parameter prediction model.

[0039] Therefore, by integrating electrical parameter-related measurement data, dimensional data, and process parameters (such as doping concentration and CMP (chemical mechanical polishing) polishing conditions), the first electrical parameter prediction model can effectively capture the complex interactions between physical, geometric, and process factors that affect device performance, thereby effectively improving the prediction accuracy of the first electrical parameter prediction model. In addition, since device performance is highly sensitive to process fluctuations (such as film thickness deviation and uneven doping distribution), by inputting process parameters, the first electrical parameter prediction model can distinguish between performance changes caused by process deviations and the regular effects of dimensional scaling itself, thereby improving the robustness of the prediction.

[0040] Furthermore, the historical process parameter measurement data includes data on process fluctuation scenarios under abnormal operating conditions. By adding data on process fluctuation scenarios (including abnormal operating conditions such as CMP overshoot and ion implantation dose deviation), the first electrical parameter prediction model can learn how these abnormalities affect the first electrical parameter, thus avoiding prediction failures caused by process deviations.

[0041] In some exemplary embodiments, the first neural network model is a neural network model based on a self-attention mechanism.

[0042] Since the self-attention mechanism can directly calculate the association weight between any two parameters in the input features (regardless of their distance), it can capture the complex interactions between electrical parameters, process parameters and dimensional data. Through multi-level feature interactions, it can automatically extract high-order physical features and reveal the deep association mechanism of electrical parameters between devices with different thresholds, thereby effectively ensuring the prediction accuracy of the first electrical parameter prediction model.

[0043] It should be noted that, as those skilled in the art can understand, the first neural network model can be but is not limited to a Transformer model. For the specific network structure of the Transformer model, reference can be made to relevant content known to those skilled in the art, and will not be elaborated here.

[0044] In some exemplary embodiments, the semiconductor device electrical parameter prediction method provided by the present invention further includes: visualizing the attention weight of the first electrical parameter prediction model.

[0045] Therefore, by visualizing the attention weights of the first electrical parameter prediction model, engineers can intuitively identify the key input features of the first electrical parameter prediction model (such as the doping concentration in the process parameters and the gate length in the dimensional data) to verify whether they are consistent with the known semiconductor physics laws. At the same time, it can also reveal the differences in the attention distribution of the first electrical parameter prediction model when predicting different electrical parameters. If the attention weight shows that a certain process step (such as the ion implantation dose) has a significant impact on the prediction results, engineers can prioritize optimizing this step to improve device performance.

[0046] In some exemplary embodiments, the semiconductor device electrical parameter prediction method provided by the present invention further includes: when a first preset update condition is met, updating the first electrical parameter prediction model based on newly acquired electrical parameter measurement data and process parameter data of multiple semiconductor devices.

[0047] Therefore, by updating the first electrical parameter prediction model based on the newly acquired electrical parameter measurement data and process parameter data of multiple semiconductor devices, the first electrical parameter prediction model can learn the potential rules in the new data in a timely manner, avoid prediction deviations caused by process fluctuations, and thus effectively ensure the accuracy of the first electrical parameter prediction model.

[0048] In some exemplary embodiments, the first preset update condition includes that the number of newly measured wafers exceeds a first preset number, or that data distribution drift occurs in the first electrical parameter prediction data predicted by the first electrical parameter prediction model.

[0049] Thus, by triggering an update of the first electrical parameter prediction model when the number of newly measured wafers exceeds a first preset number (e.g., 500), process fluctuations in semiconductor device manufacturing can be effectively addressed, significantly improving the robustness and generalization capabilities of the first electrical parameter prediction model. By triggering an update of the first electrical parameter prediction model when data distribution drift occurs in the first electrical parameter prediction model, the prediction logic of the first electrical parameter prediction model can be recalibrated to maintain the high prediction accuracy of the first electrical parameter prediction model.

[0050] It should be noted that, as those skilled in the art will appreciate, statistical methods such as KL divergence can be used to monitor data distribution changes, and when a significant offset is detected, the update of the first electrical parameter prediction model is triggered.

[0051] In some exemplary embodiments, the semiconductor device electrical parameter prediction method provided by the present invention further includes: constructing a second training sample set based on the historical measurement data of the electrical parameters, each second training sample in the second training sample set includes the second electrical parameter-related measurement data and size data of the small-size device and the size data and second electrical parameter measurement data of the large-size device, wherein the second electrical parameter measurement data of the large-size device is the label of the second training sample, the second electrical parameter includes the linear region threshold voltage, and the second electrical parameter-related measurement data of the small-size device includes the gate voltage of the small-size device and the corresponding drain current sequence data measured under the condition that the drain voltage is lower than the power supply voltage; training a second neural network model based on the second training sample set to obtain a corresponding second electrical parameter prediction model; inputting the second electrical parameter-related measurement data and size data of the measured small-size device and the size data of the large-size device to be predicted into the second electrical parameter prediction model to obtain the second electrical parameter prediction data of the large-size device to be predicted.

[0052] Since the linear region (V d <V dd ) and saturation region (V d =V dd ) is dominated by different physical mechanisms. Therefore, separate modeling of the first electrical parameter and the second electrical parameter can avoid the accuracy loss caused by mixing multiple region characteristics in a single model. In addition, since the gate voltage V measured under linear region conditions (low drain voltage) g and drain current I d Sequence data (V g -I d The curve) can more completely characterize the subthreshold characteristics (such as the subthreshold swing SS). Therefore, the linear region threshold voltage V of the large-scale device predicted by the second electrical parameter prediction model can be effectively guaranteed. t,lin accuracy.

[0053] In some exemplary embodiments, the training the second neural network model according to the second training sample set includes: training the second neural network model according to the second training sample set using a mixed precision training method.

[0054] Because mixed-precision training combines the advantages of half-precision (FP16) and single-precision (FP32) floating-point numbers, it can accelerate model convergence while ensuring numerical accuracy, significantly improving the training rate with almost no loss in the performance of the second neural network model.

[0055] In some exemplary embodiments, each of the second training samples further includes process parameter data of the small-size device and process parameter data of the large-size device.

[0056] Correspondingly, the inputting of the measurement data and size data related to the second electrical parameter of the measured small-size device and the size data of the large-size device to be predicted into the second electrical parameter prediction model includes: inputting the measurement data related to the second electrical parameter of the measured small-size device (specifically, the gate voltage and corresponding drain current sequence data of the measured small-size device measured under the condition that the drain voltage is lower than the power supply voltage), size data and process parameter data, as well as the size data and process parameter data of the large-size device to be predicted into the second electrical parameter prediction model.

[0057] Therefore, by inputting process parameters, the second electrical parameter prediction model can distinguish between performance changes caused by process deviations and the regular effects of size scaling itself, thereby improving the robustness of the prediction.

[0058] In some exemplary embodiments, the second neural network model is a neural network model based on a self-attention mechanism.

[0059] Because the self-attention mechanism can directly calculate the association weight between any two parameters in the input features (regardless of their proximity), it can capture the complex interactions between electrical parameters, process parameters, and dimensional data. Through multi-level feature interactions, it can automatically extract high-order physical features, revealing the deep association mechanism of electrical parameters between devices with different thresholds, and thus effectively ensuring the prediction accuracy of the second electrical parameter prediction model. It should be noted that, as those skilled in the art will appreciate, the second neural network model can be, but is not limited to, a Transformer model.

[0060] In some exemplary embodiments, the semiconductor device electrical parameter prediction method provided by the present invention further includes: visualizing the attention weight of the second electrical parameter prediction model.

[0061] Therefore, by visualizing the attention weights of the second electrical parameter prediction model, engineers can intuitively identify the key input features of the second electrical parameter prediction model (such as doping concentration in process parameters and gate length in dimensional data) to verify whether they are consistent with known semiconductor physics laws. It can also reveal the differences in attention distribution when predicting different electrical parameters, so that engineers can determine the optimized process steps for the device.

[0062] In some exemplary embodiments, the semiconductor device electrical parameter prediction method provided by the present invention further includes: when a second preset update condition is met, updating the second electrical parameter prediction model based on newly acquired electrical parameter measurement data and process parameter data of multiple semiconductor devices.

[0063] Therefore, by updating the second electrical parameter prediction model based on the newly acquired electrical parameter measurement data and process parameter data of multiple semiconductor devices, the second electrical parameter prediction model can learn the potential rules in the new data in a timely manner, avoid prediction deviations caused by process fluctuations, and thus effectively ensure the accuracy of the second electrical parameter prediction model.

[0064] In some exemplary embodiments, the second preset update condition includes that the number of newly measured wafers exceeds a second preset number, or that data distribution drift occurs in the second electrical parameter prediction data predicted by the second electrical parameter prediction model.

[0065] Thus, by triggering an update of the second electrical parameter prediction model when the number of newly measured wafers exceeds a second preset number (e.g., 500), process fluctuations in semiconductor device manufacturing can be effectively addressed, significantly improving the robustness and generalization capabilities of the second electrical parameter prediction model. By triggering an update of the second electrical parameter prediction model when data distribution drift occurs in the second electrical parameter prediction model, the prediction logic of the second electrical parameter prediction model can be recalibrated to maintain the high-precision predictions of the second electrical parameter prediction model.

[0066] It should be noted that, as those skilled in the art will appreciate, statistical methods such as KL divergence can be used to monitor data distribution changes, and when a significant offset is detected, the update of the second electrical parameter prediction model is triggered.

[0067] In some exemplary embodiments, the semiconductor device electrical parameter prediction method provided by the present invention further includes: obtaining short channel voltage prediction data of the large-size device to be predicted based on the first electrical parameter prediction data and the second electrical parameter prediction data of the large-size device to be predicted.

[0068] Because the short-channel effect is the result of the combined effects of saturation and linear region parameters, the first and second electrical parameter prediction data for the large-scale device being predicted can more completely capture the physical drivers of the short-channel voltage, effectively ensuring the accuracy of the acquired short-channel voltage prediction data. Furthermore, obtaining this short-channel voltage prediction data facilitates reverse analysis of the impact of process parameters (such as ion implantation dose) on saturation and linear region characteristics, thereby optimizing process design.

[0069] Specifically, the short channel voltage prediction data can be calculated according to the following formula:

[0070]

[0071]

[0072] Among them, V sce represents the short channel voltage, It represents the change in threshold voltage caused by drain-induced barrier lowering.

[0073] Please continue to refer to Figures 2 to 4 ,in, Figure 2 The graph shows the correlation between the threshold voltage change caused by drain-induced barrier lowering and the saturation threshold voltage for different devices. Figure 3 The correlation characteristic diagram between the on / off current ratio and the saturation threshold voltage of different devices; Figure 4 The figure is the correlation characteristic diagram between the short channel voltage and the saturation threshold voltage of different devices. Figure 2 As shown in the figure, the threshold voltage change caused by drain-induced barrier lowering is and the saturation threshold voltage V t,sat There is a negative correlation between them, that is, as the saturation threshold voltage V t,sat The increase in the threshold voltage caused by the lowering of the drain barrier gradually decreases, which indicates that the higher the saturation threshold voltage V t,sat Corresponding to stronger gate control capability, the influence of drain voltage on threshold voltage is effectively reduced, that is, by increasing the saturation threshold voltage V t,sat Can reduce the short channel effect. Figure 3 As shown, the conduction current I on and shutdown current I off The ratio (i.e. I on / I off ) with the saturation threshold voltage V t,sat This is because the saturation threshold voltage V t,sat The increase of will increase the on-state current I on , which results in the switching efficiency (I on / I off ) increases. Figure 4 As shown, the short channel voltage V sce and the saturation threshold voltage V t,sat It shows a negative correlation trend, which shows that the high saturation threshold voltage V t,sat Not only reduced (short channel effect), and also by improving the switching efficiency (I on / I off ) indirectly optimizes the short channel voltage V sce, indicating the saturation threshold voltage V t,sat It is a key parameter to improve the overall performance of the device.

[0074] Please continue to refer to Figure 5 , which is a comparison chart of the predicted results and the measured results of the semiconductor device electrical parameter prediction method provided by the present invention. Figure 5 As shown, the short channel voltage prediction result obtained by the semiconductor device electrical parameter prediction method provided by the present invention is highly consistent with the short channel voltage data obtained by actual measurement.

[0075] Based on the same inventive concept, the present invention also provides an electronic device, please refer to Figure 6 , which is a block diagram of an electronic device provided by one embodiment of the present invention. Figure 6 As shown, the electronic device includes a processor 101 and a memory 103. The memory 103 stores a computer program. When the computer program is executed by the processor 101, the method for predicting the electrical parameters of a semiconductor device described above is implemented. Since the electronic device provided by the present invention and the method for predicting the electrical parameters of a semiconductor device provided by the present invention belong to the same inventive concept, the electronic device provided by the present invention has at least all the beneficial effects of the method for predicting the electrical parameters of a semiconductor device provided by the present invention. Therefore, for the beneficial effects of the electronic device provided by the present invention, reference can be made to the relevant description of the beneficial effects of the method for predicting the electrical parameters of a semiconductor device provided by the present invention, and no further description will be given here.

[0076] Please continue to refer to Figure 6 ,like Figure 6 As shown, the electronic device further includes a communication interface 102 and a communication bus 104. The processor 101, the communication interface 102, and the memory 103 communicate with each other via the communication bus 104. The communication bus 104 can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, for example. The communication bus 104 can be divided into an address bus, a data bus, a control bus, and the like. For ease of illustration, only one thick line is used in the figure, but this does not mean that there is only one bus or only one type of bus. The communication interface 102 is used for communication between the electronic device and other devices.

[0077] It should be noted that the processor 101 referred to in the present invention may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor 101 is the control center of the electronic device, and connects various parts of the entire electronic device using various interfaces and lines.

[0078] It should also be noted that the memory 103 may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable memory (PROM), electrically programmable memory (EPROM), electrically erasable programmable memory (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, random access memory is available in various forms, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous random access memory (SDRAM), double data rate synchronous random access memory (DDRSDRAM), enhanced synchronous random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), RAMbus direct random access memory (RDRAM), direct memory bus dynamic random access memory (DRDRAM), and RAMbus dynamic random access memory (RDRAM).

[0079] The present invention also provides a readable storage medium having a computer program stored therein. When the computer program is executed by a processor, the computer program can implement the semiconductor device electrical parameter prediction method described above. Since the readable storage medium provided by the present invention and the semiconductor device electrical parameter prediction method provided by the present invention belong to the same inventive concept, the readable storage medium provided by the present invention has at least all the beneficial effects of the semiconductor device electrical parameter prediction method provided by the present invention. Therefore, for the beneficial effects of the readable storage medium provided by the present invention, reference can be made to the relevant description of the beneficial effects of the semiconductor device electrical parameter prediction method provided by the present invention above, and no further details will be given here.

[0080] To sum up, compared with the existing technology, the semiconductor device electrical parameter prediction method, electronic device and storage medium provided by the present invention have the following unexpected technical effects: the present invention can accurately predict the electrical parameters of large-size devices based on the electrical parameters of small-size devices, thereby not only effectively shortening the analysis cycle of the electrical parameters of large-size devices and reducing the testing cost of large-size devices, but also establishing a new prediction paradigm for the development of semiconductor devices.

[0081] It should be noted that the above description is merely a description of preferred embodiments of the present invention and does not limit the scope of the present invention. Any changes or modifications made by persons skilled in the art based on the above disclosure are within the scope of protection of the present invention. Obviously, various modifications and variations may be made by persons skilled in the art without departing from the spirit and scope of the present invention. Thus, provided such modifications and variations fall within the scope of the present invention and its equivalents, the present invention is intended to include such modifications and variations.

Claims

1. A method for predicting electrical parameters of a semiconductor device, characterized in that: include: Obtain historical measurement data of electrical parameters of semiconductor devices of different sizes; Constructing a first training sample set based on the historical electrical parameter measurement data, wherein each first training sample in the first training sample set includes first electrical parameter-related measurement data and size data of a small-sized device, and size data and first electrical parameter measurement data of a large-sized device, wherein the first electrical parameter measurement data of the large-sized device is a label of the first training sample, and the first electrical parameter includes a saturation region threshold voltage, an on-current, and an off-current; and the first electrical parameter-related measurement data of the small-sized device includes gate voltage and corresponding drain current sequence data of the small-sized device measured under the condition that the drain voltage is equal to the supply voltage; Training a first neural network model according to the first training sample set to obtain a corresponding first electrical parameter prediction model; Inputting the first electrical parameter related measurement data and size data of the measured small-size device and the size data of the large-size device to be predicted into the first electrical parameter prediction model to obtain the first electrical parameter prediction data of the large-size device to be predicted; The method further comprises: Constructing a second training sample set based on the historical electrical parameter measurement data, wherein each second training sample in the second training sample set includes measurement data and size data related to the second electrical parameter of a small-sized device, and size data and second electrical parameter measurement data of a large-sized device, wherein the second electrical parameter measurement data of the large-sized device is a label of the second training sample, the second electrical parameter includes a linear region threshold voltage, and the measurement data related to the second electrical parameter of the small-sized device includes gate voltage and corresponding drain current sequence data of the small-sized device measured under a condition where the drain voltage is lower than the supply voltage; Training a second neural network model according to the second training sample set to obtain a corresponding second electrical parameter prediction model; The second electrical parameter related measurement data and size data of the measured small-size device and the size data of the large-size device to be predicted are input into the second electrical parameter prediction model to obtain the second electrical parameter prediction data of the large-size device to be predicted.

2. The method for predicting electrical parameters of a semiconductor device according to claim 1, wherein: The training of the first neural network model according to the first training sample set includes: Training the first neural network model using a mixed precision training method according to the first training sample set; The training of the second neural network model according to the second training sample set includes: The second neural network model is trained using a mixed precision training method based on the second training sample set.

3. The method for predicting electrical parameters of a semiconductor device according to claim 1, wherein: The method further comprises: According to the first electrical parameter prediction data and the second electrical parameter prediction data of the large-scale device to be predicted, the short channel voltage prediction data of the large-scale device to be predicted is obtained.

4. The method for predicting electrical parameters of a semiconductor device according to claim 1, wherein: Both the first neural network model and the second neural network model are neural network models based on the self-attention mechanism.

5. The method for predicting electrical parameters of a semiconductor device according to claim 1, wherein: The method further comprises: When a preset update condition is met, the first electrical parameter prediction model and / or the second electrical parameter prediction model is updated according to the newly acquired electrical parameter measurement data of the plurality of semiconductor devices.

6. The method for predicting electrical parameters of a semiconductor device according to claim 1, wherein: The method further comprises: Acquiring historical measurement data of process parameters of the semiconductor devices of different sizes; Each of the first training sample and the second training sample further includes process parameter data of the small-size device and process parameter data of the large-size device; The step of inputting the first electrical parameter related measurement data and size data of the measured small-sized device and the size data of the large-sized device to be predicted into the first electrical parameter prediction model includes: Inputting the first electrical parameter related measurement data, size data and process parameter data of the measured small-size device and the size data and process parameter data of the large-size device to be predicted into the first electrical parameter prediction model; The step of inputting the measurement data and size data related to the second electrical parameter of the measured small-sized device and the size data of the large-sized device to be predicted into the second electrical parameter prediction model includes: The second electrical parameter related measurement data, size data and process parameter data of the measured small-size device and the size data and process parameter data of the large-size device to be predicted are input into the second electrical parameter prediction model.

7. An electronic device, characterized in that: The method comprises a processor and a memory, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, the method for predicting electrical parameters of a semiconductor device according to any one of claims 1 to 6 is implemented.

8. A readable storage medium, characterized in that: The readable storage medium stores a computer program, and when the computer program is executed by a processor, the method for predicting electrical parameters of a semiconductor device according to any one of claims 1 to 6 is implemented.

Images