Method and device for the re-inspection of a wafer classified as abnormal in semiconductor production

Abstract

Computer-implemented method (100) for the re-inspection of a wafer classified as anomalous to distinguish a wafer defect from a defect of a test measuring device in the production quality inspection of semiconductor production based on a comparison of S data distributions f i , which have been determined from sensor measurements on chips on a wafer by S test measuring devices. For this purpose, an anomaly value of this chip is determined from the sensor measurements on this chip, whereby a data distribution f i the frequency distribution of anomaly values ​​in measurements from exactly one test measuring device each. If the frequency distribution of the anomaly values ​​of all chips of a wafer deviates significantly from a reference distribution of anomaly values ​​of a wafer, the pairwise deviation values ​​of the S data distributions are compared by applying a normalized metric to each of the two data distributions f idetermined (101). The deviation values ​​are compared with a predefined deviation threshold (102), and the total number of deviation values ​​that exceed the predefined deviation threshold is determined (103). If the total number of deviation values ​​exceeds a total threshold (104), a test measuring device is flagged for further testing (105).

Description

[0001] The invention relates to a method for the subsequent inspection of a wafer classified as abnormal in semiconductor production, a computer program and a machine-readable storage medium, as well as a device for data processing for the subsequent inspection of a wafer classified as abnormal in semiconductor production. State of the art

[0002] In multi-stage manufacturing processes, it can be important, particularly for cost reasons, to be able to identify and sort out defective components that were produced in one process step and are to be used in later stages of the manufacturing process at an early stage. This is particularly true for the production of semiconductor chips. In the latter case, wafer-level tests (WLT) are carried out after a wafer has been produced, during which all chips on the wafer are individually subjected to several tests that indicate the final performance of the chip. Chips that have passed the WLT test can then be assembled into a package together with chips of the same or a different type as the end result of the at least two-stage manufacturing process. The costs of disposing of chips in later process stages are higher than in earlier stages. Therefore, it is important to identify defective or abnormal chips early in WLT.Univariate measurements in WLT often fail to identify all defective or abnormal chips. DE102023200852.1 proposes a method for determining an anomaly value for chips within the WLT. The described anomaly value can be determined from multivariate WLT sensor measurements and can provide a measure of whether an individual chip that has passed previous WLT still has an anomaly.

[0003] The area under the receiver operating characteristic curve (AUC) is an aggregate measure of the performance of a binary classification model across all possible classification thresholds. In arXiv:2107.02990, a weighted AUC is applied to a model predicting deviations between two distributions.

[0004] From https: / / doi.org / 10.2307 / 2347084 a method for determining a kernel density estimator using the Fast Fourier Transform (FFT) is known. Disclosure of the invention

[0005] According to one aspect of the present invention, a computer-implemented method for the post-inspection of a wafer classified as anomalous for distinguishing a wafer defect from a defect of a test measuring device in the production quality inspection of semiconductor manufacturing based on a comparison of S data distributions f idisclosed. The test measuring device can be a test measuring device for performing wafer-level tests on wafers manufactured in semiconductor production. Corresponding wafer-level tests can be performed on wafers before the chips of the wafer are separated from the corresponding wafer by cutting. The S data distributions have been determined from sensor measurements by S test measuring devices or test modules of a test device, wherein the sensor measurements were performed on chips on a wafer. Sensor measurements were performed on each chip on the wafer by exactly one of the S test measuring devices. An anomaly value for this chip is determined from the sensor measurements on a chip. An anomaly value can be a value that indicates whether one or more sensor measurements on a chip or a quantity derived from the sensor measurement(s) on the chip lies outside predefined / specified limits.An anomaly value of a chip can be a value that indicates a measure or probability that the chip is abnormal, i.e., may have a defect. An anomaly value of a chip can also indicate whether or not a measured value from a measurement on the chip lies outside a predetermined interval for that measurement. For example, a chip may have shown no abnormalities in previously performed, e.g., univariate, test measurements on the chip, and an anomaly value determined for that chip may nevertheless assume a value that may indicate an increased probability that the chip is abnormal. For example, an anomaly value can be determined by a machine learning system, where the machine learning system can receive multivariate sensor measured values ​​from test measurements of a chip as input and provide an anomaly value as output.It is also possible for anomaly values ​​to be determined from one or more sensor measurements on a chip using further analytical or numerical methods. An anomaly value of a chip can, for example, provide a measure of the extent to which derived and / or combined physical or chemical properties of the chip deviate from a reference component. A reference component can be a chip that meets the properties specified in a specification. The specification can, for example, be a document in which the electrical, chemical, mechanical, etc. properties of a non-defective chip are precisely defined. A data distribution f. icomprises the frequency distribution of anomaly values ​​in measurements from exactly one test measuring device out of a total of S test measuring devices. The method comprises at least the following steps, which are carried out if the frequency distribution of the anomaly values ​​of all chips on a wafer deviates significantly from a reference distribution of anomaly values ​​for a wafer. A reference distribution of anomaly values ​​for a wafer is a frequency distribution of anomaly values ​​for the chips on at least one reference wafer or a plurality of reference wafers, a reference wafer being a wafer whose chips meet the properties specified in a specification, and the frequency distribution of the anomaly values ​​for the chips on the reference wafer further meeting specified properties. For example, specified properties can consist in a center of gravity ora moment of the reference distribution lies within a specified interval, and / or the ends of the reference distribution have a specified shape and / or lie within a specified interval. In one process step, deviation values ​​are determined that quantify the pairwise deviations of the S data distributions from each other. A deviation value is determined by applying a normalized metric to each of two data distributions f. i In other words, the normalized metric can measure the agreement of any two data distributions f i measure. A deviation value of two data distributions f ifrom each other, which is obtained by applying the normalized metric to the corresponding two data distributions, can thus quantify how well the two data distributions agree with each other. In a further step of the method, each of the deviation values ​​from the pairwise comparisons of the S data distributions is compared with a predetermined deviation threshold. In a further step, the total number of deviation values ​​that exceed the predetermined deviation threshold in the comparison is determined. The total number of deviation values ​​is then compared with a predetermined total count threshold. If the total number of deviation values ​​exceeds the total count threshold, the test device is flagged for further testing.

[0006] Optionally, in a further step, a warning message can be issued, for example, to a system controlling a production test system and / or to a user of the production test system. The warning message can include a visual (lamp, display on the control element / computer screen) and / or acoustic alarm.

[0007] Furthermore, optionally in response to receiving a warning message or in response to a test measuring device being marked for further testing, the test measuring device can be stopped for further testing and the test measuring device can be serviced and / or cleaned or defective parts of the test measuring device can be identified and then replaced. For example, a test measuring device for carrying out WLT on chips on a wafer in semiconductor production can have contacts, e.g. needles or pins. The contacts can be used to carry out the sensor measurements of the individual chips on the wafer, which are included, among other things, in the calculation of an anomaly value for a chip. One or more contacts can be bent, for example, as a result of use or wear. It is also possible that one or more contacts are contaminated, for example by dust. As part of the further testing of the test measuring device, for example,an inspection and, in case of defects or contamination, a replacement or cleaning of the contacts (e.g. the needles or pins) of the test measuring device must be carried out.

[0008] Preferably, the individual chips on the wafer have been classified as non-defective by previous measurements on the individual chips as part of the quality inspections of the individual chips. However, the wafer has been classified as conspicuous, i.e., anomalous, in a further test in which the distribution of the anomaly values ​​determined from sensor measurements on the individual chips is considered. Using a method proposed here, it can be determined whether the wafer's classification as anomalous originates from a (manufacturing) defect in the wafer or from a defect / contamination in the test measuring device.

[0009] The S test measuring devices can be S test modules of a test device, ie the S test measuring devices can be provided by the S test modules of a test device, e.g. for carrying out WLT in semiconductor production.

[0010] Preferably, the normalized metric is the Kolmogorov-Smirnov metric, the Hellinger distance, the Wasserstein metric or the normalized area under the curve (AUC).

[0011] The normalized AUC can be used, for example, as a measure of the overlap of two data distributions f i and f j The AUC can be determined by calculating the integral ∫−∞∞Fj(x)fi(x)dx, where F j (x) the cumulative distribution function of the distribution f j (x) denotes, Fj(x)=∫−∞xfj(y)dy.

[0012] Preferably, S reference data distributions are fi' from sensor measurements of S reference test instruments, whereby the deviation threshold can then be determined by the following steps. The S reference data distributions fi' are compared pairwise in one step by calculating the normalized metric. The normalized metric determines a deviation value between two reference data distributions for each of these two reference data distributions. The deviation value can provide a measure of the agreement between the two reference data distributions. In a further step, the deviation threshold can then be determined depending on the maximum deviation value from the comparison of the reference data distributions. Preferably, the deviation threshold can be set as the maximum deviation value. However, it is also possible for the deviation threshold to be set as the value that exceeds the maximum deviation value by a specified factor, e.g., a specified percentage or a multiple of the interquartile range.

[0013] Preferably, the total count threshold depends on the number S of test measuring devices. For example, the total count threshold can be given by a value S / 2 or a comparable value of the same order of magnitude S / 2. For example, the total count threshold can be given by (S-1) / 2 or (S+1) / 2.

[0014] The total number threshold can thus advantageously be given by a value above which deviations can no longer be achieved by individual, statistical fluctuations, but above which deviations are based on deviations of a data distribution f i from the other data distributions.

[0015] According to an alternative embodiment of a method described here, the deviation values ​​can first be determined, as before, by applying a normalized metric to two data distributions f ifrom the S data distributions. In this case, a total of S(S-1) / 2 deviation values ​​can be obtained. In a next step, a test instrument-specific deviation value can be calculated for each test instrument s of the total S test instruments from the determined S(S-1) / 2 deviation values, in which - in a first sub-step, all those deviation values ​​are determined in whose calculation a data distribution f determined by the test measuring device s i has been received, and - in a second sub-step, a test device-specific deviation value for the test device s is calculated as the mean of the deviation values ​​determined in the first sub-step. In a further step, each of the total S calculated test device-specific deviation values ​​can be compared with a threshold value. The threshold value can, for example, be the deviation threshold value determined from the reference distributions. If a test device-specific deviation value from the S calculated test device-specific deviation values ​​exceeds the threshold value, the test device to which the test device-specific deviation value belongs can be marked for further testing. In this case, the wafer anomaly can be attributed to a defect or contamination of the test measuring device whose test device-specific deviation value exceeds the threshold value.This makes it possible to advantageously distinguish a defect in the wafer from a defect in a test measuring device, both of which can manifest themselves in deviations of the frequency distributions of the anomaly values ​​from reference distributions of the anomaly values.

[0016] According to a preferred embodiment, the normalized metric is the normalized area under the curve. The deviation value of a first and a second data distribution is determined from the S data distributions f iThe area under the curve is determined by integrating the product of the first data distribution with the cumulative distribution function of the second data distribution over the range from the smallest occurring anomaly value to the largest occurring anomaly value. The normalized AUC assumes a value of 0.5 if the data distributions match. The deviation value can further be compared with the deviation threshold by considering, for example, the following inequality: |θ - 0.5| > θ th , where θ in this case is the deviation value and θ th in this case denotes the deviation threshold and |·| indicates the absolute value of the argument.

[0017] According to a preferred embodiment, a method proposed here may further comprise the following steps. In a further step, the test measuring device may be flagged for further testing if the total number of deviation values ​​exceeds the total threshold. Flagging may be performed, for example, by placing a corresponding note or reference in a document, file, or database containing the maintenance status of test measuring devices. Optionally, an alarm may also be triggered if the total number of deviation values ​​exceeds the total threshold.The alarm may be the display of a warning message on a screen for a user or operator in production quality control, the sending of a corresponding email to a user or operator in production quality control, a visual and / or audible alarm signal, and / or an alarm signal sent to a control unit of a test device in quality control. In the latter case, the alarm signal sent may result in the operation of the marked test device being stopped.

[0018] In a further step, optional maintenance of the marked test measuring device can be performed. This may include, for example, cleaning the test measuring device or, for example, the needles / pins of the test measuring device, whereby the needles / pins may be brought into contact with the chips when performing sensor measurements on the chips of a wafer. Maintenance of the marked test measuring device may also include replacing needles of the test measuring device that have been identified as, for example, bent or particularly dirty during the testing or maintenance of the device.

[0019] Furthermore, the invention relates to a computer program with machine-readable instructions which, when executed on one or more computers, cause the computer(s) to execute one of the methods described above and below. The invention also encompasses a machine-readable data carrier on which the above computer program is stored, as well as a computer or data processing device equipped with the aforementioned computer program and / or the aforementioned machine-readable data carrier.

[0020] Embodiments of the invention are explained in more detail below with reference to the accompanying drawings. In the drawings: Fig. 1 schematically shows an information flow overview of a method described here; Fig. 2 schematically shows a further information flow overview of a method described here; Fig. 3 a data processing device comprising means for carrying out a method described here. Description of the embodiments

[0021] Fig. 1 schematically shows an information flow overview of a computer-implemented method 100 for the re-inspection of a wafer classified as anomalous in order to distinguish a wafer defect from a defect of a test measuring device in the production quality inspection of semiconductor production based on a comparison of S data distributions f i . The S data distributions f i from sensor measurements of S test instruments. Furthermore, the sensor measurements were performed on chips on a wafer, and from the sensor measurements on a chip, an anomaly value for that chip is determined. A data distribution f iThis includes the frequency distribution of anomaly values ​​in measurements from exactly one test measuring device each. If the frequency distribution of the anomaly values ​​of all chips of a wafer deviates significantly from a reference distribution of anomaly values ​​of a wafer, the following method steps of method 100 are executed. In step 101, deviation values ​​are determined that quantify the pairwise deviations of the S data distributions from each other, with a deviation value being determined by applying a normalized metric to each of two data distributions f iis obtained. The normalized metric can be, for example, the Kolmogorov-Smirnov metric, the Hellinger distance, or the normalized area under the curve. In step 102, each of the deviation values ​​from the pairwise comparisons of the S data distributions is compared with a predetermined deviation threshold. Then, in step 103, the total number of deviation values ​​that exceed the predetermined deviation threshold in the comparison of the previous step 102 is determined. The total number of deviation values ​​is compared with a predetermined total number threshold in step 104. If the total number of deviation values ​​exceeds the total number threshold, the test measuring device is marked for further testing in step 105.

[0022] Fig. Figure 2 schematically shows a further information flow overview of a computer-implemented method 100 for the re-inspection of a wafer classified as anomalous. The method steps 101, 102, 103, 104 and 105 have already been described in connection with the Fig. 1 has been described. Fig. 2 shows further method steps 201 and 202, which relate to the determination of the deviation threshold. These steps can preferably be carried out before the deviation values ​​from the pairwise comparisons of the S data distributions are compared with the predetermined deviation threshold in method step 102. Before performing step 201 or during the execution of this step, S reference data distributions are fi' from sensor measurements of S reference test instruments. Step 201 then involves the pairwise comparison of the S reference data distributions fi' by calculating the normalized metric for each of the two reference data distributions, whereby the normalized metric determines a deviation value between these two reference data distributions. In step 202, the deviation threshold is then determined depending on the maximum deviation value from the comparison of the reference data distributions. For example, the deviation threshold can be given by the maximum deviation value. In optional step 106, the test measuring device can be marked for further testing if the total number of deviation values ​​exceeds the total number threshold. Furthermore, in optional step 107, maintenance of the marked test measuring device can be performed.

[0023] Fig.3 shows an embodiment of a data processing device 10 comprising at least one processor 30 and at least one machine-readable storage medium 20, wherein the machine-readable storage medium 20 contains instructions which, when executed by the processor 30, cause the data processing device 10 to carry out a method according to one of the aspects of the invention.

[0024] The term "computer" encompasses any device capable of executing specific computational instructions. These computational instructions can be in the form of software, hardware, or a combination of both software and hardware.

[0025] In general, a plurality can be understood as indexed, meaning that each element of the plurality is assigned a unique index, preferably by assigning consecutive integers to the elements included in the plurality. Preferably, when a plurality comprises N elements, where N is the number of elements in the plurality, the elements are assigned integers from 1 to N. QUOTES CONTAINED IN THE DESCRIPTION

[0000] This list of documents submitted by the applicant was generated automatically and is included solely for the convenience of the reader. This list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions. Cited patent literature

[0000] DE 102023200852.1

[0002] Cited non-patent literature

[0000] https: / / doi.org / 10.2307 / 2347084

[0004]

Claims

[1] Computer-implemented method (100) for the re-inspection of a wafer classified as anomalous to distinguish a wafer defect from a defect of a test measuring device in the production quality inspection of semiconductor production based on a comparison of S data distributions f i , which were determined from sensor measurements of S test measuring instruments, where the sensor measurements were carried out on chips on a wafer, where sensor measurements have been carried out on each chip by exactly one of the S test measuring devices, where an anomaly value of this chip is determined from the sensor measurements on this chip, where a data distribution f i the frequency distribution of anomaly values ​​in measurements from exactly one test measuring device each, the method comprising the steps of, if the frequency distribution of the anomaly values ​​of all chips of a wafer deviates significantly from a reference distribution of anomaly values ​​of a wafer: - Determining (101) deviation values ​​that quantify the pairwise deviations of the S data distributions from each other, whereby a deviation value is determined by applying a normalized metric to each of two data distributions f i is received, - comparing (102) each of the deviation values ​​from the pairwise comparisons of the S data distributions with a predetermined deviation threshold, - determining (103) the total number of deviation values ​​which, in comparison, exceed the specified deviation threshold, - comparing (104) the total number of deviation values ​​with a predetermined total threshold value, - Marking (105) of the test measuring instrument for further verification if the total number of deviation values ​​exceeds the total number threshold. [2] The method (100) of claim 1, wherein the normalized metric is the Kolmogorov-Smirnov metric, the Hellinger distance, or the normalized area under the curve. [3] Method (100) according to one of the preceding claims, wherein S reference data distributions fi' from sensor measurements of S reference test instruments, where the deviation threshold was determined by the following steps: - Pairwise comparison (201) of the S reference data distributions fi' by calculating the normalized metric, which determines a deviation value between two reference data distributions, - Determining (202) the deviation threshold value depending on the maximum deviation value from the comparison of the reference data distributions. [4] Method (100) according to one of the preceding claims, wherein the total number threshold depends on the number S of test measuring devices. [5] Method (100) according to one of the preceding claims, wherein the normalized metric is the normalized area under the curve, wherein the deviation value of a first and a second data distribution from the S data distributions f i is determined in each case by the area under the curve in which the product of the first data distribution with the cumulative distribution function of the second data distribution is integrated over the range from the smallest occurring anomaly value to the largest occurring anomaly value. [6] Method (100) according to one of the preceding claims, further comprising the steps: - Marking (106) of the test measuring instrument for further verification if the total number of deviation values ​​exceeds the total number threshold. - Carrying out (107) maintenance of the marked test measuring instrument. [7] Data processing device (10) comprising means for carrying out the method (100) according to one of claims 1 to 6. [8] A computer program comprising instructions which, when executed by a computer, cause the computer to carry out the method (100) according to any one of claims 1 to 6. [9] A computer-readable storage medium (20) comprising instructions which, when executed by a computer, cause the computer to carry out the method (100) according to any one of claims 1 to 6.

Images