A method for quickly extracting parameter degradation of semiconductor devices after aging

Abstract

The present invention discloses a method for rapidly extracting parameter degradation after aging of semiconductor devices. First, the characteristic parameters in the pdk model are modified, and the characteristic parameters are the characteristic parameters of the degradation to be measured later. Then, the model lcard and netlist after the characteristic parameters are modified are called to simulate the IV characteristics of the device. Then, a data set is constructed by combining the modified characteristic parameters in the pdk model with the experimental parameters extracted from the simulated IV curve, and a neural network is built to learn the mapping relationship between the two sets of parameters. Finally, the IV curve of the semiconductor device before and after aging is actually measured, and the characteristic parameter set P0 before aging and the characteristic parameter set P after aging of the semiconductor device are obtained according to the mapping relationship of machine learning. The degradation amount of the characteristic parameters after aging of the semiconductor device is ΔP=P0-P. The present invention realizes the rapid extraction of device aging parameters through the early learning of the device simulation data set and a small amount of testing in the later stage, and has the advantages of high accuracy and high efficiency.

Description

Technical Field

[0001] The present invention relates to the field of integrated circuits, and in particular to a method for rapidly extracting parameter degradation amounts of semiconductor devices after aging, including rapid extraction of characteristic parameter degradation amounts such as threshold voltage, mobility, and saturation speed. Background Art

[0002] As integrated circuits enter the nanometer node, the size of MOS devices continues to shrink, and new structures with low heat dissipation capabilities (such as FinFET, GAAFET, etc.) are introduced, resulting in the continuous enhancement of the electric field of the gate oxide layer and the increase of the local operating temperature of the device, making the reliability of devices and circuits more and more important. In addition, the continuous reduction of chip operating voltage has also made the chip design margin smaller and smaller. The reliability issue has become a key bottleneck hindering the further development of integrated circuits.

[0003] Reliability issues mainly stem from device aging, which manifests itself as the shift of device characteristic parameters. Digital circuits focus on the switching speed of devices, so they often pay attention to the device threshold voltage V th The analog circuit focuses on the current characteristics of the entire voltage domain, and the parameters that affect it include the threshold voltage V th , mobility μ, saturation velocity v sat How to quickly and accurately extract the degradation amount of characteristic parameters is of great significance for establishing a reliable and accurate aging model.

[0004] Current threshold voltage V th The extraction methods of μ include constant current method, maximum transconductance method, etc. The extraction methods of mobility μ include Y function method, separation CV method, etc., and the saturation velocity v sat Parameters such as φ and φ are difficult to extract through experiments, which restricts the establishment of an accurate reliability model.

[0005] Although many current device parameter extraction methods are proposed and verified based on the traditional MOS model, they are basically simplified derivations of complex models without considering the influence of high-order attenuation of other parameters, such as the threshold voltage V th The extraction method does not consider the impact of mobility attenuation and source-drain series resistance. In addition, different characteristic parameters require different experiments to extract. If multiple parameters are to be extracted, it will take more time. Moreover, the extraction of each characteristic parameter is independent of each other, and the experimental conditions are not the same. The extracted parameters are not the same working conditions. The correlation between the parameters is low. Finally, the extraction of some parameters requires complex test structures, which makes it difficult to quickly extract parameters in practical applications.

[0006] Therefore, it is necessary for us to improve such existing methods to overcome the above-mentioned defects. Summary of the invention

[0007] The purpose of the present invention is to provide a method for quickly extracting the parameter degradation amount of semiconductor devices after aging. While extracting the target characteristic parameters, the influence of high-order attenuation of other characteristic parameters can be taken into account, and more characteristic parameters can be extracted quickly and efficiently from limited measurement data, thereby improving the parameter extraction efficiency and ensuring the accuracy of the device model after aging.

[0008] The above technical objectives of the present invention are achieved through the following technical solutions:

[0009] A method for quickly extracting the parameter degradation amount of a semiconductor device after aging, comprising the following steps:

[0010] 1) First, modify the characteristic parameters in the pdk model, which are the characteristic parameters of the degradation to be measured later, and then call the modelcard and netlist after the modified characteristic parameters to simulate the IV characteristics of the device;

[0011] 2) Then construct a data set by combining the modified characteristic parameters in the pdk model with the experimental parameters extracted from the simulated IV curve, and build a neural network to learn the mapping relationship between the two sets of parameters;

[0012] 3) Finally, the IV curve of the semiconductor device before and after aging is actually measured, and the characteristic parameter set P0 of the semiconductor device before aging and the characteristic parameter set P after aging are obtained based on the mapping relationship of machine learning. The degradation amount of the characteristic parameters of the semiconductor device after aging is ΔP=P0-P.

[0013] Furthermore, the process of constructing the data set in step 2) is as follows:

[0014] First, the characteristic parameters of the transistor in the pdk model are modified through the central control program, and then the simulator is controlled by the central control program to run the circuit simulation. The simulator calls the modified pdk modelcard and the pre-tested circuit netlist through instructions, and finally sends the simulator simulation results back to the central control program for data processing. The above method is repeated N times to finally obtain an N-dimensional data set.

[0015] Furthermore, the process of obtaining the degradation amount of the device characteristic parameter in step 3) is as follows:

[0016] 3.1) Test the IV curve of the unaged semiconductor device and extract the experimental parameter I0;

[0017] 3.2) After the semiconductor device is aged, the IV curve of the aged semiconductor device is tested and the experimental parameter I is extracted;

[0018] 3.3) The experimental parameters I0 and I are respectively introduced into the trained neural network to obtain the characteristic parameters P0 and P to be measured. The degradation amount of the characteristic parameters after aging of the semiconductor device is ΔP = P0-P.

[0019] In summary, the present invention has the following beneficial effects:

[0020] 1. Consider the impact of high-order attenuation of other parameters

[0021] In the data set construction stage, the present invention modifies the pdk model adapted to the actual device process, which accurately models the device (covering various second-order effects). Therefore, in the data set creation and neural network training stage, various second-order effects have been automatically taken into consideration, and the degradation amount of the final target parameter is also obtained based on the consideration of the influence of high-order attenuation of other parameters.

[0022] 2. Efficiency

[0023] The parameter extraction scheme proposed in the present invention is different from the traditional method, which requires different parameters to be extracted one by one through experiments. After the parameter extraction network is trained in the early stage, the present invention only needs to perform two IV measurements on the device before and after aging in the later stage, that is, the required characteristic parameter degradation amount can be extracted. In addition, the early data set creation and neural network training are all performed on the software, without consuming a lot of time and labor costs, and the efficiency can be improved exponentially. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is a flow chart of the rapid extraction method of the present invention.

[0025] Figure 2 It is a schematic diagram of the data set construction described in the present invention.

[0026] Figure 3 It is a training diagram of the fully connected BP neural network described in the present invention.

[0027] Figure 4a-4c This is a training result diagram of the fully connected BP neural network described in the present invention.

[0028] Figure 5 It is a flow chart of the method for extracting the degradation amount of parameters described in the present invention.

[0029] Figure 6a-6c This is a diagram comparing the prediction and measurement results described in the present invention. DETAILED DESCRIPTION

[0030] In order to make the technical means, creative features, objectives and effects achieved by the present invention easy to understand, the present invention is further described below in conjunction with illustrations and specific embodiments.

[0031] The present invention proposes a method for quickly extracting the parameter degradation of semiconductor devices after aging, which extracts some simple parameters (such as V th (lin), V th (sat), etc.), and through the mapping relationship, more other complex parameters can be quickly extracted (u0_fac, vsat_fac, etc.).

[0032] 1. Technical process framework

[0033] The specific process framework of the rapid extraction method of device characteristic parameter degradation is as follows: Figure 1 As shown in the figure, first, by modifying the characteristic parameters in the pdk model (the parameters are the characteristic parameters of the degradation to be measured later), such as dvtshift, u0_fac, vsat_fac, etc. Then call the modelcard and netlist with modified parameters to perform the IV characteristics of the device (including the linear current Id lin and saturation current Id sat ) simulation. Extract relevant parameters from the simulated IV curve, including the linear region threshold voltage V th (lin), saturation threshold voltage V th (sat), linear region V g =V dd The corresponding current I d (lin), saturation region V g =V dd The corresponding current I d (sat), the maximum transconductance g in the linear region m (lin) and the maximum transconductance g in the saturation region m (sat).

[0034] Then, a data set is constructed by combining the characteristic parameters modified in the pdk model with the experimental parameters extracted from the simulated IV curve, and a neural network is built to learn the mapping relationship between the two sets of parameters.

[0035] Finally, the IV curves of the semiconductor devices before and after aging are actually measured, and the characteristic parameter set P0 of the semiconductor devices before aging and the characteristic parameter set P after aging are obtained based on the mapping relationship of machine learning. The degradation amount of the characteristic parameters of the semiconductor devices after aging is ΔP=P0-P.

[0036] 2. Dataset Construction

[0037] The data set construction scheme of the present invention is as follows Figure 2As shown, firstly, the characteristic parameters of the transistor in the pdk model are modified through the central control program, and then the simulator is controlled by the central control program to run the circuit simulation. The simulator calls the modified pdk modelcard and the pre-tested circuit netlist through instructions, and finally the simulator simulation results are sent back to the central control program for data processing. The above method is repeated N times to finally obtain an N-dimensional data set.

[0038] 3. Machine Learning

[0039] After constructing the N-dimensional data set by the above-mentioned loop simulation in the simulator, the data set is brought into Figure 3 The training is performed in the fully connected BP neural network shown in the figure. The training results are as follows Figure 4a-4c The results show that the neural network has a strong learning ability for the mapping relationship between input and output parameters.

[0040] 4. Prediction of parameter degradation

[0041] When the actual device ages, the method for extracting the degradation of its parameters is as follows: Figure 5 shown.

[0042] The comparison between prediction and measurement is shown in Figure 6a-6c As shown in the figure, the results show that the degradation of characteristic parameters predicted by the neural network is basically consistent with the actual measured degradation, but the test method based on the neural network can significantly improve the test efficiency.

[0043] In this document, the directions or positional relationships indicated by terms such as "up", "down", "front", "back", "left", "right", "top", "bottom", "inside", "outside", "vertical", and "horizontal" are based on the directions or positional relationships shown in the accompanying drawings and are only for the clarity of expressing the technical solutions and the convenience of description, and therefore should not be understood as limitations of the present invention.

[0044] In this document, the terms "comprises," "comprising," or any other variations thereof, are intended to cover a non-exclusive inclusion of elements other than those listed and may also include additional elements not expressly listed.

[0045] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. The present invention may be subject to various changes and improvements without departing from the spirit and scope of the present invention. These changes and improvements fall within the scope of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A method for quickly extracting the parameter degradation amount of a semiconductor device after aging, characterized in that: The steps include: 1) First, modify the characteristic parameters in the pdk model, which are the characteristic parameters of the degradation to be measured later, and then call the modelcard and netlist after the modified characteristic parameters to simulate the IV characteristics of the device; 2) Then construct a data set by combining the modified characteristic parameters in the pdk model with the experimental parameters extracted from the simulated IV curve, and build a neural network to learn the mapping relationship between the two sets of parameters; 3) Finally, the IV curve of the semiconductor device before and after aging is actually measured, and the characteristic parameter set P0 of the semiconductor device before aging and the characteristic parameter set P after aging are obtained according to the mapping relationship of machine learning. The degradation amount of the characteristic parameters of the semiconductor device after aging is ΔP=P0-P; 3.1) Test the IV curve of the unaged semiconductor device and extract the experimental parameter I0; 3.2) After the semiconductor device is aged, the IV curve of the aged semiconductor device is tested and the experimental parameter I is extracted; 3.3) The experimental parameters I0 and I are respectively introduced into the trained neural network to obtain the characteristic parameters P0 and P to be measured. The degradation amount of the characteristic parameters after aging of the semiconductor device is ΔP = P0-P.

2. The method for rapidly extracting the parameter degradation amount of a semiconductor device after aging according to claim 1, characterized in that: The process of constructing the data set in step 2) is as follows: First, the characteristic parameters of the transistor in the pdk model are modified through the central control program, and then the simulator is controlled by the central control program to run the circuit simulation. The simulator calls the modified pdk modelcard and the pre-tested circuit netlist through instructions, and finally sends the simulator simulation results back to the central control program for data processing. The above method is repeated N times to finally obtain an N-dimensional data set.

Images