Method and system for establishing complementary metal oxide semiconductor tube model at extremely low temperature
By combining the dual-temperature power law method with modeling software, the problem of large prediction errors in CMOS transistor models at extremely low temperatures was solved. This method enables efficient acquisition and accurate modeling of CMOS transistor electrical characteristic data, and is applicable to multi-size and multi-process devices, thereby improving the accuracy and stability of low-temperature circuit design.
Patent Information
- Application Number
- CN202511926448.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-19
- Publication Date
- 2026-05-08
AI Technical Summary
Existing technologies have prediction errors of up to 15% to 20% in CMOS transistor simulation models at extremely low temperatures, which cannot meet the requirements of circuit design and are limited to devices of specific sizes, making it impossible to optimize circuit performance through simulation.
By combining dual-temperature power-law prediction with modeling software, the electrical characteristic curves of CMOS transistors at multiple low-temperature points are obtained. Parameter correction and curve fitting are then performed to establish a CMOS transistor model suitable for extremely low-temperature environments.
It enables efficient and accurate acquisition of CMOS transistor electrical characteristic data, reduces modeling costs, shortens the cycle, and improves the accuracy and stability of circuit design. It is applicable to multi-size and multi-process devices.
Smart Images

Figure CN121997866A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of semiconductor technology, and particularly relates to a method and system for establishing a complementary metal-oxide-semiconductor (CMOS) transistor model at extremely low temperatures. Background Technology
[0002] Since the pioneering research on various radioactive phenomena by scientists such as Röntgen, Becquerel, and Curie in the late 19th century, people have continuously explored and expanded the applications of radioactivity in production and daily life. How to achieve accurate detection of radioactive rays and high-energy particles has always been a hot research topic in nuclear physics and nuclear electronics. In recent years, novel semiconductor radiation detectors, represented by cadmium zinc telluride detectors and high-purity germanium detectors, have received widespread attention due to their high radiation absorption and energy conversion rates. Because they can rapidly detect ultra-trace amounts of X-rays / gamma rays, they have broad application prospects in radiation detection and imaging fields such as environmental radioactivity monitoring, radiation source identification, metal composition analysis, geological mapping and research, mineral and raw material exploration, security inspection, emergency monitoring, biomedical imaging, space exploration, astrophysics, and high-energy physics research.
[0003] To fully leverage the superior performance of these radiation detectors at low temperatures, their front-end readout circuits typically operate in a cryogenic environment, usually housed in liquid nitrogen tanks (77K). However, existing CMOS transistor simulation models for this temperature in integrated circuit manufacturing processes have significant shortcomings. The conventional method uses "4K+218K linear interpolation" to predict electrical data at 77K, neglecting key physical effects such as carrier freezing at extremely low temperatures, a sharp drop in impurity ionization rate, and temperature dependence of mobility. This results in prediction errors as high as 15%–20%, failing to meet circuit design requirements. Furthermore, this linear interpolation method relies solely on temperature, neglecting CMOS transistor structure and process parameters, and its applicability is limited to devices of specific sizes. This technological gap severely restricts researchers' ability to predict and optimize CMOS transistor performance in cryogenic environments. Designers cannot use effective and accurate models to improve circuit performance through simulation, and must rely on extensive experimental testing, which is not only time-consuming and labor-intensive but also carries high costs and risks. Summary of the Invention
[0004] Purpose of the invention: In order to solve the problems existing in the prior art, the present invention provides a method and system for establishing a complementary metal-oxide-semiconductor (CMOS) transistor model at extremely low temperatures. By using the electrical characteristic curves of CMOS transistors at several low temperatures of known models, the electrical characteristic curve of CMOS transistors at target low temperatures is predicted based on a combination of theoretical and modeling software curve fitting methods, thereby establishing a reliable CMOS transistor model at the target low temperature.
[0005] Technical solution: This invention provides a method for establishing a complementary metal-oxide-semiconductor (CMOS) transistor model at extremely low temperatures, specifically as follows:
[0006] At extremely low temperatures The electrical characteristic curve of the CMOS transistor is obtained, or curve fitting is performed on a CMOS transistor with a known model; the extremely low temperature is a temperature below the temperature threshold; wherein... I represents the total number of extremely low temperature points. ;
[0007] Based on the obtained CMOS transistor electrical characteristic curves at at least two low temperature points, the dual-temperature power law method is used to predict the CMOS transistor electrical characteristic curve at the target low temperature T_target.
[0008] Modeling software was used to extract parameters from the electrical characteristic curves of the CMOS transistor under T_target, and important parameters were corrected. These important parameters include those that affect the characteristics of the CMOS transistor and are related to temperature.
[0009] Based on the extracted and corrected parameters, the electrical characteristic curves of the CMOS transistor under T_target are reconstructed;
[0010] The reconstructed electrical curves are compared with the predicted electrical characteristic curves to determine whether the extracted parameters need to be corrected.
[0011] After completing the parameter extraction modeling of the CMOS transistor under T_target using modeling software, the complete modeling of the low-temperature electrical behavior of the CMOS transistor is further completed.
[0012] Furthermore, the method for obtaining the electrical characteristic curves of CMOS tubes at at least two extremely low temperature points includes testing CMOS tubes of different sizes in an extremely low temperature environment. When testing CMOS tubes in an extremely low temperature environment, multiple CMOS tubes of the same size are placed at one time, the temperature gradient is controlled to be ≤5K / min during the test, and dry nitrogen gas is introduced into the test chamber.
[0013] Furthermore, the target temperature T_target is 77K. The electrical characteristic curve at 77K is obtained using the dual-temperature-zone power-law low-temperature prediction method, specifically as follows:
[0014] Each acquired electrical characteristic curve is used as a prediction unit to predict the electrical characteristic curve corresponding to the next prediction unit at the target temperature T_targe.
[0015] Furthermore, the actual electrical characteristic curves of the CMOS transistor at the first and second temperatures are as follows:
[0016] The first set of curves: With the drain-source voltage Vds1 fixed, the CMOS transistor is made to work in the linear region. The gate-source voltage Vgs is scanned, and the body-source voltage Vbs is changed synchronously to obtain the family of linear region transfer characteristic curves.
[0017] The second set of curves: fix the first body source voltage Vbs1 to eliminate the body effect of the CMOS transistor, scan the drain source voltage Vds, and synchronously change the gate source voltage Vgs to obtain the family of output characteristic curves.
[0018] The third set of curves: With the drain-source voltage Vds2 fixed, the CMOS transistor is in the deep saturation region. The gate-source voltage Vgs is scanned, and the body-source voltage Vbs is changed synchronously to obtain the threshold voltage characteristic curve family.
[0019] The fourth set of curves: With the second body source voltage Vbs2 fixed, the CMOS transistor is subjected to the limiting body effect. The drain-source voltage Vds is scanned, and the gate-source voltage Vgs is changed synchronously to obtain the family of mobility characteristic curves.
[0020] Furthermore, the electrical characteristics of a prediction unit corresponding to T_target are predicted using the power-law method, specifically as follows:
[0021] Set the power-law scaling relationship between current and temperature: ;
[0022] Where I(.) represents the current at the corresponding temperature, To solve for parameters C and D, let T = All Substitute the data of the electrical characteristic curve under the corresponding prediction unit into the power-law scaling relationship between current and temperature, and obtain the values of C and D simultaneously; then let T=T_target, substitute it into the relationship, and obtain the electrical characteristic curve corresponding to the prediction unit at temperature T_target.
[0023] Furthermore, important parameters include carrier mobility μ, saturation velocity vsat, and threshold voltage V. th .
[0024] Furthermore, after parameter extraction via curve fitting in the modeling software, a correction term based on the physical scattering mechanism is used to correct important parameters, specifically:
[0025] Calculate the baseline temperature index: ;
[0026] Where X represents any important parameter, X( ) indicates temperature. The value of the important parameter, X( (The temperature is) The value of the important parameter, , For all The set that constitutes;
[0027] Adjust the base temperature index: ;
[0028] Where k_process is a process-related correction factor;
[0029] Calculate the baseline values of important parameters at T_target without considering the special effects of extremely low temperatures:
[0030] ;
[0031] Calculate the correction values for important parameters under T_target:
[0032] ;
[0033] in, The Coulomb scattering coefficient is... is the surface roughness scattering coefficient.
[0034] Furthermore, it is determined whether the extracted parameters need to be corrected. Specifically, the deviation between the reconstructed electrical curve and the predicted electrical characteristic curve is quantified using the maximum error and the average error. If the quantized result is greater than the preset deviation threshold, then parameters other than important parameters are selected from the extracted parameters for adjustment so that the quantized result is less than or equal to the preset deviation threshold.
[0035] Furthermore, when using modeling software for curve fitting, the following extreme low-temperature patterns are set:
[0036] μ(77K)>μ(218K), V th (77K)>V th (218K);
[0037] Where μ represents carrier mobility, V th This represents the threshold voltage.
[0038] A system for establishing a model of a complementary metal-oxide-semiconductor (CMOS) transistor at extremely low temperatures, including...
[0039] Data acquisition module: used to acquire the actual electrical characteristic curves of CMOS transistors under extremely low temperature environments;
[0040] Curve prediction module: Used to predict the electrical characteristic curve of CMOS transistor at target temperature T_target using the power law method based on the acquired CMOS transistor characteristic curve;
[0041] Parameter correction module: Used to extract parameters from the electrical characteristic curves of CMOS transistors at T_target temperature using parameter extraction software, and to correct important parameters in the extracted parameters;
[0042] Curve reconstruction module: Used to reconstruct the CMOS transistor electrical characteristic curve under T_target using parameter extraction software based on the extracted and corrected parameters;
[0043] Comparison module: Used to compare the reconstructed electrical characteristic curve with the electrical characteristic curve predicted by the power law method, and to determine whether the extracted parameters need to be corrected.
[0044] Modeling module: Used to complete the parameter extraction modeling at T_target temperature through parameter extraction software, and further complete the full modeling of the low-temperature electrical behavior of MOSFET.
[0045] Beneficial Effects: This invention enables efficient and accurate construction of CMOS models suitable for extremely low temperature environments, solving the problems of difficulty in acquiring electrical characteristic data of CMOS transistors and insufficient model construction accuracy at extremely low temperatures. It provides reliable model support for the design and analysis of ultra-low temperature integrated circuits, contributing to technological development and performance optimization in related fields. This invention constructs a complete solution encompassing "low-temperature data acquisition → power-law prediction → parameter extraction and correction → error analysis and curve reconstruction," filling a systematic gap in the physical mechanisms lacking in ultra-low temperature modeling. It innovatively adopts a data filling and correction mechanism of "electrical characteristic prediction + low-temperature parameter correction," compatible with full-size, multi-process devices of BSIM 4.5 and above. In practical applications, this method can acquire a parameter system covering the entire temperature range (4K~218K), and a scientific deviation evaluation system provides a basis for performance prediction, reducing the number of low-temperature tests, lowering modeling costs, and shortening the modeling cycle. From an industry perspective, this method promotes the transformation of low-temperature integrated circuit design from "experience-based trial and error" to "precise simulation," providing core tools for cutting-edge fields such as ultra-low temperature environment detection, establishing replicable technical specifications, and contributing to industrial upgrading and technology popularization.
[0046] This invention addresses the challenges of modeling CMOS transistors at extremely low temperatures such as 77K, including high model inaccuracies. It proposes a method combining physical mechanism-driven exponential (power-law) prediction with software parameter correction, aiming to overcome the limitations of traditional linear interpolation methods. Traditional linear interpolation predicts CMOS transistor characteristic curves at extremely low temperatures like 77K using only two data points (e.g., 4K and 218K), relying solely on temperature as a single parameter and ignoring other physical effects at extremely low temperatures. This results in prediction errors of ±15%~20% and poor adaptability. The current characteristics of CMOS transistors are fundamentally determined by the Boltzmann statistical law of charge carriers. The subthreshold current exhibits an exponential relationship with the gate voltage, and temperature directly affects the kT factor in the exponential term. This physical mechanism dictates that the change in current with temperature is inherently nonlinear. Therefore, linear interpolation can produce catastrophic errors in the subthreshold region, potentially mispredicting nanoamp-level currents as microamp-level currents, resulting in errors of up to three orders of magnitude. Power-law prediction transforms the problem into a linear space through logarithmic transformation, preserving the exponential characteristics of the subthreshold region while reflecting the power-law temperature dependence of mobility, and maintaining a smooth transition from the subthreshold region to the strong inversion region. Mathematically, this method strictly satisfies boundary conditions, accurately reproducing measurements at 4K and 218K, and physically conforms to the temperature scaling equation used in the BSIM model, ensuring the physical rationality and numerical accuracy of the prediction results across the entire operating voltage range. Furthermore, by incorporating device structure and process parameters such as channel length and doping concentration during modeling, rather than relying solely on temperature interpolation, the prediction error for 77K characteristics is reduced to ±5%~8%, and it is compatible with multi-size and multi-process CMOS devices, eliminating the need for repeated testing and achieving precise matching between the 77K characteristic curve and the actual physical state of the device. This invention, by constructing a CMOS transistor simulation model at 77K, greatly assists researchers in deeply studying the working mechanism of semiconductor chips in extremely low-temperature environments, optimizing low-temperature circuit design, and improving the stability and reliability of circuits operating under extreme conditions. Attached Figure Description
[0047] Figure 1 This is the low-temperature modeling process of the present invention.
[0048] Figure 2 To predict the 218K output characteristic curve using the power law interpolation method based on 4K and 300K data, Figure (a) shows the output characteristic curves at 4K and 300K temperatures obtained from actual measurements and the process library model, while Figure (b) compares the output characteristics predicted by the power law at 218K temperature with those provided by the process library model.
[0049] Figure 3The 4K transfer characteristic curves are predicted by extrapolation using the power law method based on 218K and 300K data. Figure (a) shows the transfer characteristic curves at 218K and 300K temperatures provided by the process library model, while Figure (b) compares the power law prediction at 4K temperature with the actual measured transfer characteristics.
[0050] Figure 4 The 4K mobility characteristic curves are predicted by extrapolation using the power law method based on 218K and 300K data. Figure (a) shows the mobility characteristic curves at 218K and 300K temperatures provided by the process library model, and Figure (b) compares the mobility characteristics predicted by the power law at 4K temperature with the actual measured mobility characteristics.
[0051] Figure 5 The figure shows the IV characteristic curve obtained at 77K using the method of this invention. Figure 4 (a) is the transfer characteristic curve. Figure 4 (b) is the output characteristic curve. Figure 4 (c) is the threshold voltage characteristic curve. Figure 4 (d) is the mobility characteristic curve. Detailed Implementation Plan
[0052] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.
[0053] like Figure 1 As shown, to address the difficulty of directly measuring CMOS electrical parameters at the extremely low temperature of 77K, a software-based parameter extraction method combining dual-temperature-zone power-law prediction and parameter correction is adopted. The specific method is as follows:
[0054] S1: Obtain basic data under multiple bias conditions
[0055] At least two low-temperature points T1 (e.g., 4K) and T2 (e.g., 218K) are established. Multiple characteristic curves of the CMOS transistor at these two low-temperature points are obtained using a testing method. For low-temperature points where the model is known, curve fitting can also be used to obtain data. The characteristic curves must include at least the following:
[0056] First set of curves: Fixed drain-source voltage V ds1 This allows the CMOS transistor to operate in the linear region, scanning the gate-source voltage V. gs And synchronously change the body source voltage V bs Obtain the family of linear region transfer characteristic curves;
[0057] The second set of curves: With the first source voltage Vbs1 fixed, ensuring no body effect in the CMOS transistor, the drain-source voltage V is scanned. ds Simultaneously, the gate-source voltage Vgs is changed to obtain a family of output characteristic curves;
[0058] Third set of curves: Fixed drain-source voltage V ds2 This causes the CMOS transistor to be in the deep saturation region, scanning the gate-source voltage Vgs and synchronously changing the body-source voltage V. bs Obtain the family of threshold voltage characteristic curves (V in this embodiment) ds2 =1.2V);
[0059] Fourth set of curves: Fixed second source voltage V bs2 This causes the CMOS transistor to operate under limiting body effect, scanning the drain-source voltage V. ds And synchronously change the gate-source voltage V gs Obtain the family of mobility characteristic curves.
[0060] Taking the acquisition of CMOS transistor characteristic curves at 4K temperature as an example: During testing, test samples of various sizes covering BSIM 4.5 and above are prepared to ensure the model's applicability under different process parameters. A test circuit board is then built to provide the hardware foundation for subsequent electrical characteristic testing. To obtain the basic data for the CMOS transistor low-temperature simulation model, the CMOS transistors are placed in a low-temperature test chamber. Three CMOS transistors of different sizes are placed in each test to reduce the impact of temperature fluctuations during a single test. Data acquisition for all tested sizes is completed sequentially (after each set of size tests, the test chamber is heated to room temperature before placing the next set to avoid the impact of residual low temperature on test accuracy). A Keithley 4200 semiconductor parameter analyzer and chip test probe station are used to acquire transfer characteristics (Vt) at 4K temperature. g The range is 0V~5V, with a step size of 0.1V), and the output characteristics (V d The range is 0V~5V, with a step size of 0.2V. Threshold voltage characteristics (calculated using linear extrapolation) and mobility characteristics (calculated using the transconductance method) are also included. During testing, the temperature gradient is controlled to ≤5K / min, and dry nitrogen gas is purged into the test chamber to prevent condensation. After data acquisition, the data is cleaned, outliers with current fluctuations exceeding 5% are removed, and the threshold voltage V is extracted. th Mobility μ, Drain-source saturation voltage V dsat The characteristic parameters were standardized and organized into CSV format.
[0061] S2: Grouped Power-Law Prediction Based on Physical Scattering Mechanism
[0062] Use a curve as a prediction unit;
[0063] For each prediction unit, a dual-temperature-zone power-law prediction method is applied. The core of this method lies in establishing a power-law scaling relationship between current and temperature: I(T) = I(T1) × C × T^D, where I(.) is the current at the corresponding temperature, and C and D are scaling parameters. To solve for C and D, T = 4K and T = 218K are set respectively. The 4K and 218K measurement data of one prediction unit are substituted into the power-law scaling relationship between current and temperature to obtain the values of C and D. Then, T = T_target is set and substituted into the relationship to obtain the electrical characteristic curve corresponding to the prediction unit at the target temperature T_target.
[0064] To verify the physical rationality and prediction accuracy of the dual-temperature-zone power-law low-temperature prediction method used in this invention in the extremely low-temperature region, the embodiments of this invention provide the interpolation prediction of the transfer characteristic curve at 218K using extremely low temperature (4K) and normal temperature (300K) as T1 and T2, respectively. Figure 2 As shown, the power-law prediction and linear interpolation prediction results were compared with the electrical data obtained from the commercial model provided in the process library. The results showed that the power-law prediction results were better than the linear prediction results, with an average relative error (RMSE) of 5.7%. The mobility, subthreshold slope, and threshold voltage drift trend of the two methods were consistent, and the boundary conditions were well preserved.
[0065] Furthermore, this invention uses low temperature (218K) and normal temperature (300K) as T1 and T2 respectively, and employs the same power-law extrapolation method as described above to predict the output characteristic curve at 4K. The prediction results are as follows: Figure 3 As shown, the power-law prediction and linear interpolation prediction results were compared with the actual electrical data measured in a 4K environment. The results also showed that the power-law prediction results were better than the linear prediction results. The average relative error (RMSE) of its output characteristic curve was 4.8%, and the mobility, subthreshold slope, and threshold voltage drift trend were consistent with the measured results, and the boundary conditions were well preserved.
[0066] Similarly, Figure 4 The power-law prediction results for mobility at 4K are presented. The above interpolation and extrapolation power-law prediction results show that the power-law method proposed in this invention can effectively capture the nonlinear temperature-dependent characteristics of carrier transport behavior from room temperature to extremely low temperature (4K), and its mathematical form is highly compatible with physical mechanisms (such as power-law temperature scaling of mobility and Boltzmann statistics in the subthreshold region).
[0067] S3: BISM4.5 modeling software curve fitting for parameter extraction and correction. While modeling software can automatically extract parameters, it may overlook special physical effects at extremely low temperatures. Therefore, some important temperature-related parameters of the CMOS model need to be corrected during parameter extraction. Important parameters considered in this embodiment include carrier mobility (μ) and threshold voltage (V). th ) and saturation velocity (vsat ).
[0068] Important parameters are corrected using a correction term based on the physical scattering mechanism, specifically as follows:
[0069] Calculate the baseline temperature index: ;
[0070] Where X represents any important parameter, X( ) indicates temperature. The value of the important parameter, X( (The temperature is) The value of the important parameter, , For all The set that constitutes;
[0071] Adjust the base temperature index: ;
[0072] Where k_process is a process-related correction factor;
[0073] Calculate the baseline values of important parameters at T_target without considering the special effects of extremely low temperatures:
[0074] ;
[0075] Calculate the correction values for important parameters under T_target:
[0076] ;
[0077] in, The Coulomb scattering coefficient is... is the surface roughness scattering coefficient.
[0078] This embodiment takes carrier mobility as an example: First, the values of low-field mobility μ(218K) and μ(300K) in the linear region are extracted from the known 218K and 300K device models; then, the basic temperature index is calculated through the power law relationship: γ=-ln[μ(300K) / μ(218K)] / ln(300K / 218K). This temperature index γ reflects the average scattering characteristics in the high-temperature range and serves as the basic parameter for subsequent extrapolation.
[0079] Then, based on the device process characteristics, the temperature index is adjusted accordingly: γ_adj=γ×k_process, where the process correction factor k_process is determined according to the device type: bulk silicon planar devices: k_process=0.9~1.0, SOI devices: k_process=0.8~0.9, FinFET three-dimensional devices: k_process=0.7~0.8.
[0080] Based on the adjusted temperature index, the base mobility value without considering the special effects of extremely low temperatures is calculated as: μ_base(4K)=μ(218K)×(4K / 218K)^(-γ_adj), which represents the theoretical prediction result under the simplified model that only considers lattice scattering.
[0081] Finally, to address the dominant effects of Coulomb scattering and surface roughness scattering at the extremely low temperature of 4K, a composite correction term is introduced, resulting in the following mobility: Among them, the correction parameters and Determined based on device doping level and interface quality: Light doping uses 0.10, medium doping uses 0.15, and heavy doping uses 0.20~0.25. : 0.10 for standard interfaces, and 0.05~0.08 for high-quality interfaces.
[0082] The corrected 4K temperature-downloaded carrier mobility can be obtained by following the above steps, such as... Figure 3 As shown, this model can then be used in modeling software for parameter fitting to obtain a more accurate ultra-low temperature simulation model. A similar approach is used for other parameters, and based on the specific parameters and different trends in temperature change, linear interpolation, exponential prediction, and other methods are selected to obtain the required parameter values.
[0083] In this embodiment, when using parameter extraction software for curve fitting, the mobility and threshold voltage are set to reflect the extremely low temperature characteristics.
[0084] μ(77K)>μ(218K), ;
[0085] S4: Characteristic Curve Reconstruction. Based on the corrected parameters and extracted parameters, BSIM4.5 was used to perform curve fitting and parameter extraction for a low-threshold PMOS transistor (L=0.06μm, W=0.2μm) at 77K. Compared with the curve predicted by the power-law method, the RMSE of the 77K transfer characteristic curve was 3.8%, and the RMSE of the output characteristic curve was 4.5%.
[0086] S5: Error Calculation and Judgment. The deviation between the characteristic curve obtained from parameter extraction by the computational modeling software and the characteristic curve predicted by the power-law method is calculated, and the maximum error is evaluated. and average error When the error (average error) exceeds 8%, the number of adjustment parameters is increased, and other parameters besides the critical ones are selected for correction until the error is less than 8%, at which point the modeling result is considered reliable. The final characteristic curve at 77K is shown below. Figure 5 As shown.
[0087] S6: Replace the default model in the simulation software with the BSIM4.5 or later version model file (.model file) created at 77K temperature to perform circuit design simulation (such as operational amplifier) at 77K temperature.
[0088] This invention addresses key electrical parameters, such as μ and V, under low-temperature conditions. th and V dsat The fitting terms in the device model card have been modified, allowing designers to directly measure the circuit's performance indicators such as noise and bandwidth through simulation, thus reducing the trial-and-error costs of low-temperature circuit design.
[0089] It should also be noted that the various specific technical features described in the above embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, the present invention will not describe the various possible combinations separately.
Claims
1. A method for establishing a complementary metal-oxide-semiconductor (CMOS) transistor model at extremely low temperatures, characterized in that, Specifically: At extremely low temperatures The electrical characteristic curve of the CMOS transistor is obtained, or curve fitting is performed on a CMOS transistor with a known model; the extremely low temperature is a temperature below the temperature threshold; wherein... I represents the total number of extremely low temperature points. ; Based on the obtained CMOS transistor electrical characteristic curves at at least two low temperature points, the dual-temperature power law method is used to predict the CMOS transistor electrical characteristic curve at the target low temperature T_target. Modeling software was used to extract parameters from the electrical characteristic curves of the CMOS transistor under T_target, and important parameters were corrected. These important parameters include those that affect the characteristics of the CMOS transistor and are related to temperature. Based on the extracted and corrected parameters, the electrical characteristic curves of the CMOS transistor under T_target are reconstructed; The reconstructed electrical curves are compared with the predicted electrical characteristic curves to determine whether the extracted parameters need to be corrected. After completing the parameter extraction modeling of the CMOS transistor under T_target using modeling software, the complete modeling of the low-temperature electrical behavior of the CMOS transistor is further completed.
2. The method for establishing a complementary metal-oxide-semiconductor (CMOS) transistor model at extremely low temperatures according to claim 1, characterized in that, The method for obtaining the electrical characteristic curves of CMOS tubes at at least two extremely low temperature points includes testing CMOS tubes of different sizes in an extremely low temperature environment. When testing CMOS tubes in an extremely low temperature environment, multiple CMOS tubes of the same size are placed at one time. During the test, the temperature gradient is controlled to be ≤5K / min, and dry nitrogen gas is introduced into the test chamber.
3. The method for establishing a complementary metal-oxide-semiconductor (CMOS) transistor model at extremely low temperatures according to claim 1, characterized in that, The target temperature T_target is 77K. The electrical characteristic curve at 77K is obtained using the dual-temperature-zone power-law low-temperature prediction method, as follows: Each acquired electrical characteristic curve is used as a prediction unit to predict the electrical characteristic curve corresponding to the next prediction unit at the target temperature T_targe.
4. The method for establishing a complementary metal-oxide-semiconductor (CMOS) transistor model at extremely low temperatures according to claim 3, characterized in that, The actual electrical characteristic curves of the CMOS transistor at the first and second temperatures are as follows: The first set of curves: With the drain-source voltage Vds1 fixed, the CMOS transistor is made to work in the linear region. The gate-source voltage Vgs is scanned, and the body-source voltage Vbs is changed synchronously to obtain the family of linear region transfer characteristic curves. The second set of curves: fix the first body source voltage Vbs1 to eliminate the body effect of the CMOS transistor, scan the drain source voltage Vds, and synchronously change the gate source voltage Vgs to obtain the family of output characteristic curves. The third set of curves: With the drain-source voltage Vds2 fixed, the CMOS transistor is in the deep saturation region. The gate-source voltage Vgs is scanned, and the body-source voltage Vbs is changed synchronously to obtain the threshold voltage characteristic curve family. The fourth set of curves: With the second body source voltage Vbs2 fixed, the CMOS transistor is subjected to the limiting body effect. The drain-source voltage Vds is scanned, and the gate-source voltage Vgs is changed synchronously to obtain the family of mobility characteristic curves.
5. The method for establishing a complementary metal-oxide-semiconductor (CMOS) transistor model at extremely low temperatures according to claim 3, characterized in that, The electrical characteristics of a prediction unit corresponding to T_target are predicted using the power-law method, specifically as follows: Set the power-law scaling relationship between current and temperature: ; Where I(.) represents the current at the corresponding temperature, To solve for parameters C and D, let T = All Substitute the data of the electrical characteristic curve under the corresponding prediction unit into the power-law scaling relationship between current and temperature, and obtain the values of C and D simultaneously; then let T=T_target, substitute it into the relationship, and obtain the electrical characteristic curve corresponding to the prediction unit at temperature T_target.
6. The method for establishing a complementary metal-oxide-semiconductor (CMOS) transistor model at extremely low temperatures according to claim 1, characterized in that, Key parameters include carrier mobility μ, saturation velocity vsat, and threshold voltage V. th .
7. The method for establishing a complementary metal-oxide-semiconductor (CMOS) transistor model at extremely low temperatures according to claim 1, characterized in that, After curve fitting and parameter extraction in the modeling software, a correction term based on the physical scattering mechanism is used to correct important parameters, specifically: Calculate the baseline temperature index: ; Where X represents any important parameter, X( ) indicates temperature. The value of the important parameter, X( (The temperature is) The value of the important parameter, , For all The set that constitutes; Adjust the base temperature index: ; Where k_process is a process-related correction factor; Calculate the baseline values of important parameters at T_target without considering the special effects of extremely low temperatures: ; Calculate the correction values for important parameters under T_target: ; in, The Coulomb scattering coefficient is... is the surface roughness scattering coefficient.
8. The method for establishing a complementary metal-oxide-semiconductor (CMOS) transistor model at extremely low temperatures according to claim 1, characterized in that, To determine whether the extracted parameters need to be corrected, the deviation between the reconstructed electrical curve and the predicted electrical characteristic curve is quantified using the maximum error and average error. If the quantized result is greater than the preset deviation threshold, parameters other than important parameters are selected from the extracted parameters for adjustment so that the quantized result is less than or equal to the preset deviation threshold.
9. The method for establishing a complementary metal-oxide-semiconductor (CMOS) transistor model at extremely low temperatures according to claim 1, characterized in that, When using modeling software for curve fitting, set the extreme low temperature rules: μ(77K)>μ(218K)、V th (77K)>V th (218K); Where μ represents the carrier mobility, V th This represents the threshold voltage.
10. A system for establishing a model of a complementary metal-oxide-semiconductor (CMOS) transistor at extremely low temperatures, characterized in that, include Data acquisition module: used to acquire the actual electrical characteristic curves of CMOS transistors under extremely low temperature environments; Curve prediction module: Used to predict the electrical characteristic curve of CMOS transistor at target temperature T_target using the power law method based on the acquired CMOS transistor characteristic curve; Parameter correction module: Used to extract parameters from the electrical characteristic curves of CMOS transistors at T_target temperature using parameter extraction software, and to correct important parameters in the extracted parameters; Curve reconstruction module: Used to reconstruct the CMOS transistor electrical characteristic curve under T_target using parameter extraction software based on the extracted and corrected parameters; Comparison module: Used to compare the reconstructed electrical characteristic curve with the electrical characteristic curve predicted by the power law method, and to determine whether the extracted parameters need to be corrected. Modeling module: Used to complete the parameter extraction modeling at T_target temperature through parameter extraction software, and further complete the full modeling of the low-temperature electrical behavior of MOSFET.
