A method for processing semiconductor test data

Abstract

The present invention provides a method for processing semiconductor test data, which relates to the technical field of data processing. The method includes: acquiring semiconductor test data; selecting device target test data and epitaxial wafer target test data; constructing a test data matrix from the epitaxial wafer test data and the device target test data; performing a first matrix transformation and a second matrix transformation on the matrix; extracting a sub-matrix, performing linear fitting on the column vectors to obtain a corresponding relationship, and using this corresponding relationship as the corresponding relationship between the device target test data and the epitaxial wafer target test data. By constructing a test data matrix and performing matrix transformations, then extracting a sub-matrix and performing linear fitting on the two column vectors of the sub-matrix, the corresponding relationship between the device target test data and the epitaxial wafer target test data is obtained, solving the problem of establishing the corresponding relationship between device test parameters and epitaxial test parameters, and helping to realize the optimization and control of the epitaxial process using device test parameters.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and particularly to a method for processing semiconductor test data. Background Art

[0002] In the large-scale mass production of molecular beam epitaxy (MBE), the uniformity between multiple epitaxial wafers grown on the same substrate pallet and the stability of the parameters of epitaxial wafers grown in different rounds or batches seriously affect the yield of semiconductor device products. In order to improve the yield of semiconductor device products, it is necessary to stably control the epitaxial process of semiconductor epitaxial wafers grown in batches.

[0003] For the grown semiconductor epitaxial wafers, some epitaxial test parameters of the epitaxial wafers (such as some structural test parameters and electrical test parameters) can be obtained through conventional epitaxial wafer test means (for example, XRD test, Hall test, etc.), but the device test parameters of the final device products prepared from the epitaxial wafers cannot be accurately predicted, and the device test parameters are the direct determining factors for the yield of semiconductor device products. Therefore, in order to achieve stable control of the epitaxial process of semiconductor epitaxial wafers grown in batches, in addition to timely monitoring the conventional epitaxial test parameters of the epitaxial wafers, it is also necessary to timely obtain the device test parameters of the final device products and optimize the control of the epitaxial process based on the device test parameters. For epitaxial wafer manufacturers, the relationship between the conventional epitaxial test parameters of epitaxial wafers and the epitaxial process is known. In order to optimize the control of the epitaxial process based on the device test parameters, it is also necessary to know the relationship between the device test parameters and the conventional epitaxial test parameters.

[0004] There is a certain correlation between the device test parameters of device products and each conventional epitaxial test parameter of semiconductor epitaxial wafers. However, due to the complexity of the epitaxial wafer structure and the influence of factors such as device preparation processes, it is usually difficult to directly deduce this correlation based on physical principles, which brings challenges to using device test parameters to optimize the control of the epitaxial process. Therefore, a method needs to be proposed to establish the corresponding relationship between device test parameters and conventional epitaxial test parameters through this method. Summary of the Invention

[0005] The purpose of the present invention is to provide a method for processing semiconductor test data to solve the problem of establishing the corresponding relationship between device test parameters and epitaxial test parameters in view of the deficiencies of the above-mentioned prior art.

[0006] To achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0007] The present invention provides a method for processing semiconductor test data, which includes test data of semiconductor epitaxial wafers with the same structure in mass production and test data of semiconductor devices prepared using the semiconductor epitaxial wafers. The number of types of test data of the semiconductor epitaxial wafers is m, where m is an integer greater than or equal to 2, and the number of types of test data of the semiconductor devices is s, where s is an integer greater than or equal to 1. The processing method is used to obtain the correspondence between device target test data and epitaxial wafer target test data. The device target test data is a type of test data selected from the test data of the semiconductor devices; the epitaxial wafer target test data is a type of test data selected from the test data of the semiconductor epitaxial wafers. The method includes:

[0008] Obtain the semiconductor test data;

[0009] Select a type of test data from the s types of test data of the semiconductor devices as the device target test data, and select a type of test data from the m types of test data of the semiconductor epitaxial wafers as the epitaxial wafer target test data;

[0010] Construct a test data matrix T from the m types of test data of the semiconductor epitaxial wafers and the device target test data,

[0011]

[0012] where w ij represents the jth type of test data of the ith semiconductor epitaxial wafer in mass production, i = 1, 2, 3,..., n, j = 1, 2, 3,..., m, q i represents the device target test data of the semiconductor device corresponding to the ith semiconductor epitaxial wafer, and the selected epitaxial wafer target test data is used as the mth type of test data. n represents the number of semiconductor epitaxial wafers, and n > 100·m;

[0013] Perform a first matrix transformation on the test data matrix T to form a data matrix T',

[0014]

[0015] The first matrix transformation means that the element value of any row in the data matrix T' is obtained by subtracting the element value of another row in the test data matrix T except for the corresponding row from the element value of the corresponding row in the test data matrix T;

[0016] Perform a second matrix transformation on the data matrix T' to form a data matrix T”,

[0017]

[0018] The second matrix transformation represents performing elementary row operations on T' so that the elements w” in the data matrix T” satisfy the following conditions kt all equal 0: k < m and t < k; and k ≥ m and t < m;

[0019] Extract the submatrix t from the data matrix T”.

[0020]

[0021] Perform linear fitting on the column vector t2 and the column vector t1 to obtain the correspondence between t2 and t1, and use this correspondence as the correspondence between the device target test data and the epitaxial wafer target test data, where the column vector t1 represents the column vector composed of all the elements in the first column of the submatrix t, and the column vector t2 represents the column vector composed of all the elements in the second column of the submatrix t.

[0022] Optionally, the first matrix transformation represents that the element value of the g-th row in the data matrix T' is obtained by subtracting the element value of the (g + 1)-th row in the test data matrix T from the element value of the g-th row in the test data matrix T, where g = 1, 2, 3,..., n - 1, and the element value of the n-th row in the data matrix T' is obtained by subtracting the element value of the first row in the test data matrix T from the element value of the n-th row in the test data matrix T.

[0023] Optionally, after obtaining the semiconductor test data, the method further includes: preprocessing the semiconductor test data to remove abnormal data.

[0024] Optionally, abnormal data means that for any type of test data in the semiconductor test data, calculate the average value of this type of test data, and mark the test data whose absolute value of the difference from the average value is greater than the preset threshold as abnormal data. The preset threshold is less than or equal to 3% of the absolute value of the average value, and mark the test data of the semiconductor epitaxial wafer and the test data of the semiconductor device associated with this abnormal data as abnormal data.

[0025] Optionally, the preset threshold is equal to 2% of the absolute value of the average value.

[0026] Optionally, the semiconductor epitaxial wafer is an epitaxial wafer with a pHEMT structure, the semiconductor device is a pHEMT device prepared using the epitaxial wafer with a pHEMT structure, and the test data of the semiconductor epitaxial wafer includes barrier layer composition data, epitaxial layer thickness data, channel layer doping concentration data, channel layer carrier mobility data, and epitaxial wafer surface sheet resistance data.

[0027] Optionally, the test data of the semiconductor device includes pinch-off voltage data, source-drain current data when the gate-source voltage is zero, and transconductance data.

[0028] Optionally, n > 500·m.

[0029] The beneficial effects of the present invention include:

[0030] The method for processing semiconductor test data provided by the present invention includes: obtaining semiconductor test data; selecting one type of test data from the s types of test data of a semiconductor device as the device target test data, and selecting one type of test data from the m types of test data of a semiconductor epitaxial wafer as the epitaxial wafer target test data; constructing a test data matrix T from the m types of test data of the semiconductor epitaxial wafer and the device target test data; performing a first matrix transformation on the test data matrix T to form a data matrix T'; performing a second matrix transformation on the data matrix T' to form a data matrix T''; extracting a sub-matrix t from the data matrix T'', and performing a linear fit on the column vector t2 and the column vector t1, so as to obtain the corresponding relationship between t2 and t1, and taking this corresponding relationship as the corresponding relationship between the device target test data and the epitaxial wafer target test data. By constructing a test data matrix and performing a first matrix transformation and a second matrix transformation on the matrix, then extracting a sub-matrix from the transformed data matrix and performing a linear fit on two column vectors of the sub-matrix, the corresponding relationship between the two column vectors is obtained, thereby obtaining the corresponding relationship between the device target test data and the epitaxial wafer target test data, solving the problem of establishing the corresponding relationship between the device test parameters and the epitaxial test parameters, and helping to realize the optimization and control of the epitaxial process by using the device test parameters. Description of the Drawings

[0031] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0032] Figure 1 It shows a schematic flow chart of the method for processing semiconductor test data provided by an embodiment of the present invention;

[0033] Figure 2 It shows a schematic flow chart of the method for processing semiconductor test data provided by another embodiment of the present invention. Detailed Embodiments

[0034] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0035] To improve the yield of semiconductor device products, it is necessary to stably control the epitaxial process of batch-grown semiconductor epitaxial wafers. For the grown semiconductor epitaxial wafers, some epitaxial test parameters of the epitaxial wafers (such as some structure test parameters and electrical test parameters) can be obtained through conventional epitaxial wafer test means (such as XRD test, Hall test, etc.). These conventional epitaxial test parameters can reflect the structure and performance of the epitaxial wafers to a certain extent, but they cannot comprehensively characterize the epitaxial wafers. Therefore, the device test parameters of the final device products prepared from the epitaxial wafers cannot be accurately predicted based on the conventional epitaxial test parameters, and the device test parameters are the direct determining factors for the yield of semiconductor device products. Therefore, to achieve stable control of the epitaxial process of batch-grown semiconductor epitaxial wafers, in addition to timely monitoring the conventional epitaxial test parameters of the epitaxial wafers, it is also necessary to timely obtain the device test parameters of the final device products and optimize the control of the epitaxial process based on the device test parameters. For epitaxial wafer manufacturers, the relationship between the conventional epitaxial test parameters of the epitaxial wafers and the epitaxial process is known. That is to say, for epitaxial wafer manufacturers, it is known how to quantitatively adjust the conventional epitaxial test parameters by adjusting the epitaxial process. This known relationship is usually associated with specific epitaxial equipment and is usually obtained in advance through multiple experiments. After obtaining the epitaxial wafers, device manufacturers process them into semiconductor devices with specific functions using specific device preparation processes, and the device test parameters are used to characterize the performance of the semiconductor devices. To achieve optimized control of the epitaxial process based on the device test parameters, it is also necessary to know the relationship between the device test parameters and the epitaxial test parameters.

[0036] There is a certain correlation between the device test parameters of device products and each epitaxial test parameter of semiconductor epitaxial wafers. However, due to the complexity of the epitaxial wafer structure and the influence of factors such as device preparation processes, it is usually difficult to directly derive this correlation based on physical principles, which poses a challenge to using the device test parameters to optimize the control of the epitaxial process. Therefore, a method needs to be proposed to establish the corresponding relationship between the device test parameters and the epitaxial test parameters through this method.

[0037] Figure 1 The flowchart of the method for processing semiconductor test data provided by an embodiment of the present invention is shown, as Figure 1As shown in the figure. The present invention provides a method for processing semiconductor test data, which includes test data of semiconductor epitaxial wafers with the same structure produced in batch and test data of semiconductor devices prepared using the semiconductor epitaxial wafers. Batch production can be understood as hundreds or even thousands of semiconductor epitaxial wafers with the same structure being produced.

[0038] The test data of semiconductor epitaxial wafers are usually test data obtained by epitaxial wafer manufacturers using conventional epitaxial wafer testing means. These test data can include, for example, but are not limited to: epitaxial layer thickness data, surface sheet resistance data, and doping concentration data obtained by Hall testing, etc. The number of types of test data of semiconductor epitaxial wafers is m, and m is an integer greater than or equal to 2. In practical applications, the types of test data of semiconductor epitaxial wafers are selected by epitaxial wafer manufacturers based on the types of device test parameters. Optionally, when the semiconductor epitaxial wafer is, for example, an epitaxial wafer with a pHEMT structure, the test data of the semiconductor epitaxial wafer includes barrier layer composition data, epitaxial layer thickness data, channel layer doping concentration data, channel layer carrier mobility data, and epitaxial wafer surface sheet resistance data. In this case, m is equal to 5.

[0039] The test data of semiconductor devices are usually data obtained by semiconductor device manufacturers using device preparation processes to fabricate semiconductor epitaxial wafers into semiconductor devices with specific functions and then using device testing means to test the semiconductor devices. These data can be, for example, electrical and / or optical test data related to device performance. The number of types of test data of semiconductor devices is s, and s is an integer greater than or equal to 1. In practical applications, the types of test data of semiconductor devices are selected according to the requirements of device manufacturers for device product specifications. Optionally, when the semiconductor epitaxial wafer is, for example, an epitaxial wafer with a pHEMT structure, the semiconductor device is a pHEMT device prepared using the epitaxial wafer with a pHEMT structure. At this time, the test data of the semiconductor device can include pinch-off voltage data, source-drain current data when the gate-source voltage is zero, and transconductance data. In this case, s is equal to 3.

[0040] The processing method provided by the embodiments of the present invention is used to obtain the correspondence between device target test data and epitaxial wafer target test data. The device target test data is a type of test data selected from the test data of semiconductor devices; the epitaxial wafer target test data is a type of test data selected from the test data of semiconductor epitaxial wafers. The method includes:

[0041] Step 101, obtain semiconductor test data.

[0042] For an epitaxial wafer manufacturer, the test data of semiconductor epitaxial wafers are the data obtained and saved in daily batch production tests; the test data of semiconductor devices can be obtained from device manufacturers.

[0043] Step 102: Select a type of test data from the s types of test data of semiconductor devices as the device target test data, and select a type of test data from the m types of test data of semiconductor epitaxial wafers as the epitaxial wafer target test data.

[0044] The purpose of this step is to select two types of test data for establishing a correspondence relationship. That is to say, the subsequent steps are used to establish the correspondence relationship between the device target test data and the epitaxial wafer target test data selected in this step. It should be understood that according to needs, any type can be selected from the s types of test data of semiconductor devices as the device target test data, and any type can be selected from the m types of test data of semiconductor epitaxial wafers as the epitaxial wafer target test data. In order to know the correspondence relationship of other types of test data, after completing the subsequent steps to establish the correspondence relationship, the selection in Step 102 can be made again, and then the subsequent steps can be repeated.

[0045] Step 103: Construct a test data matrix T from the m types of test data of semiconductor epitaxial wafers and the device target test data.

[0046]

[0047] where w ij represents the j-th type of test data of the i-th semiconductor epitaxial wafer in batch production, i = 1, 2, 3,..., n, j = 1, 2, 3,..., m, q iDenote the device target test data of the semiconductor device corresponding to the i-th semiconductor epitaxial wafer, and take the selected epitaxial wafer target test data as the m-th type of test data. n represents the number of semiconductor epitaxial wafers. Generally speaking, the larger the value of n, the larger the amount of data obtained, and the higher the accuracy of the finally established corresponding relationship. Here, it is set that n > 100·m. For example, when m = 3, n is greater than 300. It should be understood that according to the actual amount of data, n can be selected as large as possible. Optionally, n > 500·m. In the test data matrix T constructed here, there are m + 1 data in each row, and these m + 1 data are associated with the same semiconductor epitaxial wafer. The data in different rows correspond to different semiconductor epitaxial wafers. Generally speaking, the same type of test data of semiconductor epitaxial wafers of the same structure grown in batches should theoretically be exactly the same, but due to issues such as the stability of the epitaxial process and the test accuracy, the same type of test data is approximately equal, but it is difficult to be exactly equal; similarly, due to the influence of factors such as the stability of the epitaxial wafer itself, the stability of the device process, and the device test accuracy, the same type of test data of semiconductor devices prepared from semiconductor epitaxial wafers of the same structure grown in batches is difficult to be exactly equal, but is approximately equal. In other words, in the test data matrix T, for the test data w ij , when j is fixed, for i = 1, 2, 3,..., n, the w ij is approximately equal for different values of i, and in addition, the q i is approximately equal for different values of i.

[0048] Step 104: Perform a first matrix transformation on the test data matrix T to form a data matrix T'.

[0049] Wherein,

[0050] The first matrix transformation means that the element value of any row in the data matrix T' is obtained by subtracting the element value of another row in the test data matrix T except for the corresponding row from the element value of the corresponding row in the test data matrix T. Through the first matrix transformation, the element value of each row of the formed data matrix T' represents the difference between the test data associated with two epitaxial wafers and their corresponding semiconductor devices. Optionally, the first matrix transformation means that the element value of the g-th row in the data matrix T' is obtained by subtracting the element value of the (g + 1)-th row in the test data matrix T from the element value of the g-th row in the test data matrix T, where g = 1, 2, 3,..., n - 1, and the element value of the n-th row in the data matrix T' is obtained by subtracting the element value of the 1-st row in the test data matrix T from the element value of the n-th row in the test data matrix T.

[0051] In the test data matrix T, for each epitaxial wafer, the m types of epitaxial wafer test data correspond to a device target test data. The device target test data is used as the dependent variable, and the functional relationship between the m types of test data as independent variables is complex and difficult to directly express. Since for any continuous function, within a very small change interval of the independent variable, the continuous function can be approximated as a linear function. From the description in step 103, it can be seen that the values of the same type of test data are approximately equal. Therefore, the above first matrix transformation can be performed on the test data matrix T to obtain the data matrix T'. The first m elements in each row of the data matrix T' respectively represent the change amounts of the m types of epitaxial wafer test data, and the (m + 1)-th element in each row of the data matrix T' represents the change amount of the device target test data caused by the aforementioned changes in the m types of epitaxial wafer test data.

[0052] Step 105: Perform a second matrix transformation on the data matrix T' to form the data matrix T”.

[0053] Wherein,

[0054] The second matrix transformation means performing elementary row operations on T' so that the element w” in the data matrix T” kt all equal 0: when k < m and t < k; and when k ≥ m and t < m.

[0055] In other words, the second matrix transformation means performing quasi-diagonalization on the data matrix T' so that the w” corresponding to k < m and t < k kt all equal 0, and the w” corresponding to k ≥ m and t < m kt also all equal 0. The transformed data matrix T” has the following form:

[0056]

[0057] Similarly, based on the description in step 104, it can be known that after the second matrix transformation, the first m elements in each row of the data matrix T” respectively represent the change amounts of the m types of epitaxial wafer test data, and the (m + 1)-th element in each row of the data matrix T” represents the change amount of the device target test data caused by the aforementioned changes in the m types of epitaxial wafer test data. By performing the above second matrix transformation, among the elements from the m-th row to the n-th row of the data matrix T”, the data associated with the epitaxial wafer test data only remains the test data change amount in the m-th column (that is, the change amount of the epitaxial wafer target test data).

[0058] Step 106: Extract a sub-matrix t from the data matrix T”. Perform linear fitting on the column vector t2 and the column vector t1 to obtain the corresponding relationship between t2 and t1, and use this corresponding relationship as the corresponding relationship between the device target test data and the epitaxial wafer target test data, where the column vector t1 represents the column vector composed of all elements of the first column of the sub-matrix t, and the column vector t2 represents the column vector composed of all elements of the second column of the sub-matrix t.

[0059] Among them,

[0060] The first column of the sub-matrix t represents the change amount of the epitaxial wafer target test data, and the second column of the sub-matrix t represents the change amount of the device target test data. From the description of step 105, it can be seen that the change amount of the device target test data in the second column of the sub-matrix t is only related to the change amount of the epitaxial wafer target test data and has nothing to do with other types of test data of the epitaxial wafer. Therefore, through linear fitting, it is easy to obtain the corresponding relationship between the column vector t2 and the column vector t1, and use this corresponding relationship as the corresponding relationship between the device target test data and the epitaxial wafer target test data. From the above method, it can be seen that this corresponding relationship is an approximate corresponding relationship, and this corresponding relationship can be used to guide the optimization control of the process in mass production. In practical applications, for example, if the value of the device target test data is too large and Δq needs to be reduced, then through the corresponding relationship obtained above, the corresponding epitaxial test data Δw that needs to change can be deduced m , and then based on Δw m , the epitaxial process is correspondingly optimized and adjusted to realize the optimization control of the epitaxial process using the device test parameters.

[0061] In summary, by constructing a test data matrix, performing the first matrix transformation and the second matrix transformation on this matrix, then extracting a sub-matrix from the transformed data matrix and performing linear fitting on the two column vectors of the sub-matrix, the corresponding relationship between the two column vectors is obtained, and thus the corresponding relationship between the device target test data and the epitaxial wafer target test data is obtained, solving the problem of establishing the corresponding relationship between the device test parameters and the epitaxial test parameters, which helps to realize the optimization control of the epitaxial process using the device test parameters.

[0062] Figure 2 shows a schematic flow chart of a method for processing semiconductor test data provided by another embodiment of the present invention, as Figure 2As shown, after obtaining the semiconductor test data, that is, after step 101, the method further includes step 1011: preprocessing the semiconductor test data to remove abnormal data. The abnormal data means that for any type of test data in the semiconductor test data, the average value of this type of test data is calculated, and the test data whose absolute value of the difference from the average value is greater than a preset threshold is marked as abnormal data. The preset threshold is less than or equal to 3% of the absolute value of the average value, and the test data of the semiconductor epitaxial wafer and the test data of the semiconductor device associated with the abnormal data are both marked as abnormal data. Optionally, the preset threshold is equal to 2% of the absolute value of the average value. By removing the abnormal data, the adverse effects of the abnormal data on subsequent data processing and fitting can be avoided, and the accuracy of the correspondence between the established device target test data and the epitaxial wafer target test data can be improved.

[0063] The above embodiments are only for illustrating the technical concept and features of the present invention, and their purpose is to enable those of ordinary skill in the art to understand the content of the present invention and implement it, and it cannot be used to limit the protection scope of the present invention. Any equivalent changes or modifications made according to the spirit and essence of the present invention should be covered within the protection scope of the present invention.

Claims

1. A method for processing semiconductor test data, characterized in that The semiconductor test data includes the test data of semiconductor epitaxial wafers of the same structure in mass production and the test data of semiconductor devices prepared using the semiconductor epitaxial wafers. The number of types of the test data of the semiconductor epitaxial wafers is m, where m is an integer greater than or equal to 2. The number of types of the test data of the semiconductor devices is s, where s is an integer greater than or equal to 1. The processing method is used to obtain the correspondence between the device target test data and the epitaxial wafer target test data. The device target test data is a type of test data selected from the test data of the semiconductor devices. The epitaxial wafer target test data is a type of test data selected from the test data of the semiconductor epitaxial wafers. The method includes: Obtaining the semiconductor test data; Selecting a type of test data from the s types of test data of the semiconductor devices as the device target test data, and selecting a type of test data from the m types of test data of the semiconductor epitaxial wafers as the epitaxial wafer target test data; Constructing a test data matrix T from the m types of test data of the semiconductor epitaxial wafers and the device target test data; where w ij represents the j-th type of test data of the i-th semiconductor epitaxial wafer in mass production, i = 1, 2, 3,..., n, j = 1, 2, 3,..., m, q i represents the device target test data of the semiconductor device corresponding to the i-th semiconductor epitaxial wafer, and the selected epitaxial wafer target test data is used as the m-th type of test data, n represents the number of semiconductor epitaxial wafers, and n > 100·m; Performing a first matrix transformation on the test data matrix T to form a data matrix T'; The first matrix transformation means that the element value of any row in the data matrix T' is obtained by subtracting the element value of another row in the test data matrix T except for the corresponding row from the element value of the corresponding row in the test data matrix T; Performing a second matrix transformation on the data matrix T' to form a data matrix T''; Said second matrix transformation represents performing elementary row transformations on T' such that the elements w'' in the data matrix T'' kt all equal 0: k < m and t < k; and k ≥ m and t < m; Extracting a sub-matrix t from the data matrix T''; Performing a linear fitting on the column vector t2 and the column vector t1 to obtain the correspondence between t2 and t1, and taking this correspondence as the correspondence between the device target test data and the epitaxial wafer target test data, where the column vector t1 represents the column vector composed of all the elements in the first column of the sub-matrix t, and the column vector t2 represents the column vector composed of all the elements in the second column of the sub-matrix t.

2. The method for processing semiconductor test data according to claim 1, wherein The first matrix transformation means that the element value of the g-th row in the data matrix T' is obtained by subtracting the element value of the (g + 1)-th row in the test data matrix T from the element value of the g-th row in the test data matrix T, where g = 1, 2, 3,..., n - 1. The element value of the n-th row in the data matrix T' is obtained by subtracting the element value of the first row in the test data matrix T from the element value of the n-th row in the test data matrix T.

3. The method for processing semiconductor test data according to claim 1, wherein After obtaining the semiconductor test data, the method further includes: preprocessing the semiconductor test data to remove abnormal data.

4. The method for processing semiconductor test data according to claim 3, wherein The abnormal data means that for any type of test data in the semiconductor test data, calculating the average value of this type of test data, and marking the test data whose absolute value of the difference from the average value is greater than a preset threshold as abnormal data. The preset threshold is less than or equal to 3% of the absolute value of the average value, and marking both the test data of the semiconductor epitaxial wafers and the test data of the semiconductor devices associated with the abnormal data as abnormal data.

5. The method for processing semiconductor test data according to claim 4, wherein The preset threshold is equal to 2% of the absolute value of the average value.

6. The method for processing semiconductor test data according to claim 1, wherein The semiconductor epitaxial wafer is an epitaxial wafer with a pHEMT structure, the semiconductor device is a pHEMT device prepared by using the epitaxial wafer with the pHEMT structure, and the test data of the semiconductor epitaxial wafer includes barrier layer composition data, epitaxial layer thickness data, channel layer doping concentration data, channel layer carrier mobility data, and epitaxial wafer surface sheet resistance data.

7. The method for processing semiconductor test data according to claim 6, wherein The test data of the semiconductor device includes pinch-off voltage data, source-drain current data when the gate-source voltage is zero, and transconductance data.

8. The method for processing semiconductor test data according to claim 1, wherein n > 500·m.

Images