Wafer preceding quantity detection method and semiconductor process equipment

Abstract

The invention provides a front quantity detection method of a wafer and semiconductor process equipment. The front quantity detection method of the wafer comprises the following steps: firstly, acquiring a feature vector set of the wafer; inputting the feature vector set into a pre-trained prediction model to enable the prediction model to output a prediction information set according to the feature vector set; global feature vectors of the wafer are generated according to the prediction information set, and the global feature vectors comprise a film thickness prediction mean value, a variable coefficient, a center-to-edge ratio, an orientation non-uniformity parameter and an abnormal point coefficient; and finally, generating a pre-quantity detection result of the wafer according to the global feature vector and a preset feature threshold value, thereby realizing rapid and lossless pre-quantity detection of the wafer, avoiding a situation of wrong operation of a subsequent process caused by wafer abnormality, and further improving process stability.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of semiconductor technology, and in particular to a wafer pre-quantity detection method and a semiconductor process equipment. BACKGROUND

[0002] CMP (Chemical Mechanical Planarization) is a process widely used in the manufacturing process of IC (Intergrated Circuit), mainly used for global planarization of metal materials such as tungsten (W), copper (Cu), cobalt (Co), aluminum (Al), etc. The current CMP process is mainly divided into a main polishing stage and a fine polishing stage, and is equipped with an in-situ endpoint detection system to monitor the material removal rate in real time. In addition, in order to ensure the effect of the CMP process, a pre-quantity detection needs to be performed before the CMP process, that is, the film thickness of the wafer metal layer before the CMP process is detected, so as to control the CMP process parameters and the CMP process operation according to the film thickness detected by the pre-quantity detection.

[0003] In the related technical solutions, the pre-quantity detection often needs to be combined with XRF (X-ray Fluorescence Spectrometer) technology or RS (Resistance), however, the XRF technology has limited measurement penetration ability, and it is difficult to reflect the change of the film thickness under a multi-layer structure, and at the same time, due to the time-consuming measurement, it is difficult to be applied to mass production; the RS is a contact test, which will cause unexpected damage or pollution, therefore, how to realize fast and non-destructive pre-quantity detection is a problem to be solved. SUMMARY

[0004] Therefore, the purpose of the present application is to provide a wafer pre-quantity detection method and a semiconductor process equipment to alleviate the above technical problems, so as to realize fast and non-destructive pre-quantity detection of the wafer, avoid the situation that the subsequent process is incorrectly operated due to the wafer abnormality, and further improve the process stability.

[0005] In a first aspect, an embodiment of the present application provides a wafer pre-quantity detection method, which comprises:

[0006] obtaining a feature vector set of a wafer; wherein the feature vector set comprises feature vectors corresponding to a plurality of sampling points on the wafer respectively, each feature vector is determined according to a detection signal of a corresponding sampling point, and comprises at least one of a time domain feature parameter, a frequency domain feature parameter and a statistical feature parameter;

[0007] inputting the feature vector set into a pre-trained prediction model, so that the prediction model outputs a prediction information set according to the feature vector set; wherein the prediction information set comprises prediction information corresponding to each of the sampling points, and the prediction information comprises a film thickness prediction value and an uncertainty;

[0008] generating a global feature vector of the wafer according to the prediction information set, wherein the global feature vector comprises a film thickness prediction mean value, a coefficient of variation, a center-to-edge ratio, an azimuthal non-uniformity parameter, and an outlier coefficient;

[0009] generating a front quantity detection result of the wafer according to the global feature vector and a preset feature threshold; wherein the front quantity detection result comprises a front quantity detection pass result, a front quantity detection fail result, and a front quantity detection conditional pass result.

[0010] Optionally, the step of generating a global feature vector of the wafer according to the prediction information set comprises: determining a plurality of effective sampling points according to the prediction information of the sampling points; and generating the global feature vector according to the film thickness prediction values of the plurality of effective sampling points.

[0011] Optionally, the plurality of effective sampling points comprises a first effective sampling point located at the center of the wafer and a plurality of second effective sampling points located at the edge of the wafer; and the step of generating the global feature vector according to the film thickness prediction values of the plurality of effective sampling points comprises: calculating the film thickness prediction mean value and the film thickness prediction standard deviation according to the film thickness prediction values of the plurality of effective sampling points, and determining the coefficient of variation according to the film thickness prediction mean value and the film thickness prediction standard deviation; calculating the center-to-edge ratio and the radial gradient according to the film thickness prediction value of the first effective sampling point and the film thickness prediction values of the plurality of second effective sampling points, and determining the azimuthal non-uniformity parameter according to the center-to-edge ratio and the radial gradient; determining an abnormal sampling point in the plurality of effective sampling points, and determining the outlier coefficient according to the number of abnormal sampling points and the number of effective sampling points.

[0012] Optionally, the step of determining an abnormal sampling point in the plurality of effective sampling points comprises: calculating an outlier value of each effective sampling point according to the film thickness prediction value of the effective sampling point; and regarding an effective sampling point with an outlier value greater than a preset outlier threshold as the abnormal sampling point.

[0013] Optionally, the step of determining a plurality of effective sampling points according to the prediction information of the sampling points comprises: calculating a confidence degree of each sampling point according to the film thickness prediction value and the uncertainty of the sampling point; and regarding a sampling point with a confidence degree not less than a preset confidence threshold as the effective sampling point.

[0014] Optionally, the preset feature threshold comprises a preset target film thickness and a preset error; the step of generating the pre-quantity detection result of the wafer according to the global feature vector and the preset feature threshold comprises: generating a film thickness detection result according to the film thickness prediction mean value, the preset target film thickness and the preset error; wherein the film thickness detection result comprises a film thickness detection qualified result and a film thickness detection unqualified result; if the film thickness detection result is the film thickness detection unqualified result, the pre-quantity detection result is determined as the pre-quantity detection unpass result.

[0015] Optionally, the preset feature threshold further comprises a preset coefficient of variation threshold, a preset center-to-edge ratio threshold, a preset azimuthal non-uniformity threshold and a preset abnormal point coefficient threshold; the step of generating the pre-quantity detection result of the wafer according to the global feature vector and the preset feature threshold comprises: if the film thickness detection result is the film thickness detection qualified result, generating a uniformity detection result according to the coefficient of variation and the preset coefficient of variation threshold; generating a radial distribution detection result according to the center-to-edge ratio and the preset center-to-edge ratio threshold; generating a circumferential distribution detection result according to the azimuthal non-uniformity parameter and the preset azimuthal non-uniformity threshold; generating an abnormal point detection result according to the abnormal point coefficient and the preset abnormal point coefficient threshold; and determining the pre-quantity detection result according to the uniformity detection result, the radial distribution detection result, the circumferential distribution detection result and the abnormal point detection result.

[0016] Preferably, the uniformity detection result, the radial distribution detection result, the circumferential distribution detection result and the abnormal point detection result each comprise a detection qualified result and a detection unqualified result; the step of determining the pre-quantity detection result according to the uniformity detection result, the radial distribution detection result, the circumferential distribution detection result and the abnormal point detection result comprises: if the number of the detection unqualified results in the uniformity detection result, the radial distribution detection result, the circumferential distribution detection result and the abnormal point detection result is only one, determining the pre-quantity detection result as the pre-quantity detection conditional pass result; or, if the number of the detection unqualified results in the uniformity detection result, the radial distribution detection result, the circumferential distribution detection result and the abnormal point detection result is greater than one, determining the pre-quantity detection result as the pre-quantity detection unpass result; or, if none of the uniformity detection result, the radial distribution detection result, the circumferential distribution detection result and the abnormal point detection result contains the detection unqualified result, determining the pre-quantity detection result as the pre-quantity detection pass result.

[0017] Preferably, the method further comprises: if the pre-quantity detection result is the pre-quantity detection unpass result, generating an alarm information and controlling the wafer to stop entering a next process for process processing.

[0018] In a second aspect, the embodiments of the present application further provide a semiconductor process equipment, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method of the first aspect when executing the computer program.

[0019] The embodiments of the present application have the following beneficial effects:

[0020] The embodiments of the present application provide a wafer front quantity detection method and a semiconductor process equipment. Firstly, a feature vector set of a wafer is obtained. The feature vector set comprises feature vectors corresponding to a plurality of sampling points on the wafer. Then, the feature vector set is input into a pre-trained prediction model, so that the prediction model outputs a prediction information set according to the feature vector set. The prediction information set comprises prediction information corresponding to the plurality of sampling points. The prediction information comprises a film thickness prediction value and an uncertainty. A global feature vector of the wafer is generated according to the prediction information set. The global feature vector comprises a film thickness prediction mean value, a coefficient of variation, a center-to-edge ratio, an azimuthal non-uniformity parameter and an abnormal point coefficient. Finally, a front quantity detection result of the wafer is generated according to the global feature vector and a preset feature threshold. Thus, the wafer rapid non-destructive front quantity detection is realized, the situation that the subsequent process is incorrectly operated due to wafer abnormality is avoided, and the process stability is improved.

[0021] Other features and advantages of the present application will be further described in the following description, and will be apparent to those skilled in the art from the following description and the embodiments. The objects and other advantages of the present application will be achieved and obtained by the structure particularly pointed out in the description and the drawings.

[0022] In order to make the above objects, features and advantages of the present application more apparent, the following will specifically describe preferred embodiments, and the accompanying drawings will be described in detail as follows. BRIEF DESCRIPTION OF DRAWINGS

[0023] In order to more clearly illustrate the specific embodiments of the present application or the technical solutions in the prior art, the following will briefly introduce the drawings needed to be used in the specific embodiments or the related technical description. Obviously, the drawings in the following description are some embodiments of the present application, and those skilled in the art can also obtain other drawings according to these drawings without any creative labor.

[0024] Figure 1 A flowchart of a wafer front quantity detection method provided by the embodiments of the present application is shown in the figure.

[0025] Figure 2 A schematic structural diagram of an eddy current system provided by the embodiments of the present application is shown in the figure. DETAILED DESCRIPTION

[0026] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions of the present application will be described clearly and completely below with reference to the drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments of the present application, all the other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0027] With the semiconductor process evolving to 28 nm, 14 nm, and even 7 nm / 5 nm, CMP technology gradually becomes crucial, especially in BEOL (Backend of Line), CMP process has become an indispensable key step, and the core principle is to eliminate the step difference, recess and metal redundant layer on the wafer surface caused by the previous process through the synergistic effect of chemical corrosion and mechanical grinding. In practical application, CMP process needs to be precisely matched with lithography, etching and other processes to ensure that the flatness error of tens of layers of metal interconnection structure is controlled in the sub-nanometer level.

[0028] In addition, in order to ensure the effect of CMP process, pre-amount detection needs to be performed before CMP process; wherein, the pre-amount detection is used to detect the wafer film thickness (also can be called as the thickness of the front film layer) before CMP process, so as to prevent the previous process of CMP from being run multiple times or not running due to special reasons (such as thin film deposition). In actual detection, the pre-amount detection mainly detects the film thickness of the wafer metal layer, and often needs to be combined with XRF technology or RS, etc. However, the XRF technology is difficult to reflect the change of the film thickness under the multi-layer structure due to the limited measurement penetration ability, and is difficult to be applied to mass production due to the time-consuming measurement. RS is a contact test, which will bring unexpected damage or pollution.

[0029] And, the related technical solutions also adopt an eddy current system to detect the end point of the CMP process, wherein the eddy current system is a non-contact measurement technology suitable for multi-layer metal structures, and the system can accurately measure the thickness of a metal layer such as a Cu film by using the principle of electromagnetic induction, and the detection range is from 1500 nm to less than 30 nm. However, the eddy current system cannot provide accurate film thickness information before polishing to guide the setting of the CMP process parameters, that is, it cannot be applied to the pre-quantity detection scenario before CMP. In addition, since the eddy current system can only provide film thickness measurement data, it cannot automatically determine whether it can enter the CMP process or which parameters need to be adjusted according to the pre-quantity detection. At the same time, it cannot predict possible problems in the CMP process, such as incorrect process recipe running, excessive polishing or insufficient polishing, etc. In addition, for different product types and different film thickness ranges, there is a lack of calibration, resulting in errors in the film thickness measurement data, thereby affecting the process effect of CMP.

[0030] Therefore, the embodiment of the present application provides a wafer pre-quantity detection method and a semiconductor process equipment. First, a feature vector set of a wafer is obtained. The feature vector set includes feature vectors corresponding to a plurality of sampling points on the wafer. Then, the feature vector set is input into a pre-trained prediction model, so that the prediction model outputs a prediction information set according to the feature vector set. The prediction information set includes prediction information corresponding to the plurality of sampling points. The prediction information includes a film thickness prediction value and an uncertainty. A global feature vector of the wafer is generated according to the prediction information set. The global feature vector includes a film thickness prediction mean, a coefficient of variation, a center-to-edge ratio, an azimuthal non-uniformity parameter, and an outlier coefficient. Finally, a pre-quantity detection result of the wafer is generated according to the global feature vector and a preset feature threshold. Thus, the wafer pre-quantity detection is realized quickly and non-destructively, and the situation that the subsequent process is incorrectly run due to wafer abnormalities is avoided, thereby improving the process stability.

[0031] In order to facilitate the understanding of the present embodiment, the present embodiment will be described in detail below.

[0032] Embodiment one

[0033] The present embodiment provides a wafer pre-quantity detection method. The pre-quantity detection refers to the process of detecting the wafer after the previous process. As shown in the following formula (1), the method includes the following steps: Figure 1

[0034] Step S102, obtaining a feature vector set of a wafer.

[0035] ​The feature vector set includes feature vectors corresponding to a plurality of sampling points on the wafer, each feature vector being determined according to a detection signal of the corresponding sampling point, and including at least one of a time-domain feature parameter, a frequency-domain feature parameter, and a statistical feature parameter. In actual application, the time-domain feature parameter includes an amplitude feature and a shape feature, where the amplitude feature includes but is not limited to a maximum value, a mean value, a peak value, a root mean square (RMS) value, etc., and the shape feature includes kurtosis, skewness, and a waveform factor, etc.; the frequency-domain feature parameter includes but is not limited to a main frequency amplitude, a phase, a harmonic ratio, and a power spectral density, etc.; and the statistical feature parameter includes but is not limited to signal stability, a signal-to-noise ratio, and phase fluctuation, etc. The specific time-domain feature parameter, frequency-domain feature parameter, and statistical feature parameter can be set according to actual conditions.

[0036] In addition, the detection signal of each sampling point can be an eddy current signal, that is, each sampling point on the wafer is detected by an eddy current system. Specifically, as shown in Figure 2 The eddy current system emits a high-frequency electromagnetic field through a drive electromagnetic 22 to excite the metal film layer 11 of the wafer 10 to generate an eddy current 23. At this time, the secondary magnetic field generated by the eddy current 23 is received by an eddy current sensor 21, and the eddy current signal can be obtained after signal processing. The film thickness of the wafer can be obtained according to the eddy current signal. Therefore, by detecting each sampling point on the wafer through the eddy current system, the eddy current signal of each sampling point can be obtained, thereby realizing rapid and non-destructive detection of the wafer film layer. In addition, this detection method shortens the detection time to 12 seconds per wafer, which is 5-10 times faster than traditional offline detection, further improving the pre-quantity detection efficiency of the wafer.

[0037] Therefore, when the wafer is loaded onto the CMP polishing disc, a plurality of sampling points are first set at the center and the edge of the wafer; where the edge can be a region at a preset distance from the center of the wafer, which can be set according to actual conditions. Then, each sampling point on the wafer is detected by an eddy current system, such as collecting 200 ms of eddy current signal for each sampling point, realizing rapid coverage of the metal layer; finally, the feature signal of each sampling point is determined according to the eddy current signal of the sampling point.

[0038] For example, for the eddy current signal S(i, t) of the i-th sampling point on the wafer within t time, t∈[0, T], where T represents the detection period, first apply digital filtering to eliminate high-frequency noise to obtain the processed eddy current signal S f (i,t)=αS f (i,t)+(1-α)S f (i,t), where a represents a smoothing factor, then according to the processed eddy current signal S f(i, t) determines the feature vector of the i-th sampling point. Here, the feature vector includes the average amplitude, phase information and waveform factor, wherein the average amplitude A(i) is calculated by the mean function, that is, A(i) = mean(|S f (i, t)|), the phase information is determined by the formula , and the waveform factor K(i) is determined by the formula K(i) = max(|S f (i, t)|) / mean(|S f (i, t)|), so that the feature vector of the i-th sampling point is determined.

[0039] It should be noted that in the above process of determining each feature vector, the detection signal can also be other signals except the eddy current signal, and the feature vector can be generated according to the other signals. The specific adaptation can be adjusted according to the actual situation, and the embodiment of the present application will not be described in detail.

[0040] In step S104, the feature vector set is input into the pre-trained prediction model, so that the prediction model outputs a prediction information set according to the feature vector set.

[0041] After the feature vector set of the wafer is obtained, the feature vector set is input into the pre-trained prediction model, so that the prediction model outputs a prediction information set according to the feature vector set; wherein the prediction information set includes prediction information corresponding to each sampling point, and each prediction information includes a film thickness prediction value and an uncertainty; for example, for any sampling point i, the prediction information of the sampling point can be obtained, wherein the prediction information includes the film thickness prediction value T(i) and the uncertainty U(i) of the point, so that the prediction information of each sampling point can be calculated through the prediction model and the feature vector of each sampling point, so as to detect the wafer according to the prediction information of each sampling point. Compared with the detection mode of the whole wafer prediction information, the detection precision of the wafer is further improved.

[0042] In step S106, a global feature vector of the wafer is generated according to the prediction information set.

[0043] The global feature vector includes: film thickness prediction mean, coefficient of variation, center-to-edge ratio, azimuthal non-uniformity parameter and abnormal point coefficient, so that the wafer is detected according to the global feature vector. Compared with the wafer detection mode of a single detection information, the wafer detection mode of the multi-dimensional feature not only further improves the wafer detection precision, but also realizes the accurate process control of the wafer, and ensures the process quality of the wafer.

[0044] In step S108, the pre-quantity detection result of the wafer is generated according to the global feature vector and the preset feature threshold.

[0045] The pre-quantity detection result includes a pre-quantity detection pass result, a pre-quantity detection fail result and a pre-quantity detection conditional pass result. In actual application, the pre-quantity detection pass result indicates that the wafer is normal and can enter the next CMP process. The pre-quantity detection fail result indicates that the wafer is abnormal and cannot enter the next CMP process, thereby avoiding the situation that the wafer abnormally enters the CMP process and increases the equipment maintenance cost. The pre-quantity detection conditional pass result indicates that the wafer is abnormal but can be restored to normal by adjustment and enter the next CMP process. Therefore, the pre-quantity detection of the wafer avoids the situation that the wafer abnormally enters the next CMP process, thereby improving the process control precision of the wafer and ensuring the process quality of the wafer.

[0046] The pre-quantity detection method of the wafer provided by the embodiment of the application realizes the fast and lossless pre-quantity detection of the wafer, avoids the situation that the subsequent process is incorrectly operated due to the wafer abnormality, and further improves the process stability.

[0047] In one embodiment, the step of generating the global feature vector of the wafer according to the prediction information set includes the following sub-steps: (1) determining a plurality of effective sampling points according to the prediction information of the sampling points; and (2) generating the global feature vector according to the film thickness prediction values of the plurality of effective sampling points.

[0048] In the sub-step (1), the plurality of effective sampling points are determined according to the prediction information of the sampling points as follows: first, the confidence of each sampling point is calculated according to the film thickness prediction value and the uncertainty of the sampling point. For example, for any sampling point i, the confidence C(i) of the sampling point is calculated according to the film thickness prediction value T(i) and the uncertainty U(i) of the sampling point as C(i)=exp(-U(i) 2 / T(i) 2 ). Then, the sampling points with the confidence not less than a preset confidence threshold C threshold are taken as the effective sampling points, that is, when C(i)≥C threshold , the corresponding sampling point is taken as the effective sampling point; and similarly, when C(i)<C threshold , the corresponding sampling point is taken as the ineffective sampling point.

[0049] Therefore, the plurality of effective sampling points are obtained by screening the plurality of sampling points, and the global feature vector is generated according to the film thickness prediction values of the plurality of effective sampling points, thereby avoiding the situation that the global feature vector precision is reduced due to the ineffective sampling points, and further improving the pre-quantity detection precision of the wafer.

[0050] In addition, for the plurality of effective sampling points, in order to facilitate the description, it is assumed that the total number of effective sampling points is n, then the corresponding film thickness prediction value set T is T={T(1), T(2),..., T(n)}; at the same time, the plurality of effective sampling points are divided into a first effective sampling point located at the center of the wafer and a plurality of second effective sampling points located at the edge of the wafer, that is, the second effective sampling points are a plurality of effective sampling points in the edge region.

[0051] Specifically, in sub-step (2), the step of generating a global feature vector according to the film thickness prediction values of the plurality of effective sampling points comprises: ① calculating the film thickness prediction mean and the film thickness prediction standard deviation according to the film thickness prediction values of the plurality of effective sampling points, and determining the coefficient of variation according to the film thickness prediction mean and the film thickness prediction standard deviation; ② calculating the center-to-edge ratio and the radial gradient according to the film thickness prediction value of the first effective sampling point and the film thickness prediction value of the plurality of second effective sampling points, and determining the azimuthal non-uniformity parameter according to the center-to-edge ratio and the radial gradient; ③ determining the abnormal sampling points in the plurality of effective sampling points, and determining the abnormal point coefficient according to the number of abnormal sampling points and the number of effective sampling points.

[0052] In ①, for the film thickness prediction mean μ, it is calculated according to μ=mean(T); similarly, for the film thickness prediction standard deviation σ, it is calculated according to the std (standard deviation) function, that is, σ=std(T); and the coefficient of variation CV is calculated according to CV=σ / μ. In particular, in some scenarios, the film thickness prediction maximum Tmax=max(T), the film thickness prediction minimum Tmin=min(T), and the film thickness prediction range R=Tmax-Tmin can also be determined according to the film thickness prediction value set T, which can be set according to actual conditions.

[0053] In addition, in ②, in order to facilitate the description, the film thickness prediction value of the first effective sampling point is referred to as Tcenter, and the film thickness prediction value of the second effective sampling point is referred to as Tedge, then for the center-to-edge ratio CER, it can be calculated according to CER=Tcenter / mean(Tedge), where mean(Tedge) represents the film thickness prediction mean of the plurality of second effective sampling points; similarly, for the radial gradient Gr, it can be calculated according to Gr=(mean(Tedge)-Tcenter) / D, where D represents the average radius of the plurality of second effective sampling points, that is, the total radius is obtained according to the distance from each second effective sampling point to the center, and D is determined according to the total radius and the number of second effective sampling points; for the azimuthal non-uniformity parameter U θ , it can be calculated according to U θ =std(Tedge) / mean(Tedge).

[0054] In ③, firstly, the abnormal sampling points in the plurality of effective sampling points are determined, and the specific determination process is as follows: the outlier value of each effective sampling point is calculated according to the film thickness prediction value of the effective sampling point; and the effective sampling point with an outlier value greater than a preset outlier threshold is taken as an abnormal sampling point. Wherein, the outlier value can be calculated by using an improved GESD (Generalized Extreme Studentized Deviate) algorithm, and the outlier value of each effective sampling point is calculated according to Z(i) = 0.6745 (T(i) - median (T)) / MAD, where median (T) represents the median of the film thickness prediction value set T, MAD (Median absolute deviation) represents the median absolute deviation, and |Z(i)| is taken as the outlier value of each effective sampling point. At this time, if the outlier value is greater than the preset outlier threshold (preferably 3.5, which can be adaptively adjusted according to actual conditions), that is, |Z(i)|>3.5, the corresponding effective sampling point is taken as an abnormal sampling point, otherwise, if the outlier value is not greater than the preset outlier threshold, that is, |Z(i)|≤3.5, the corresponding effective sampling point is taken as a non-abnormal sampling point. Therefore, the abnormal sampling points in the effective sampling points can be determined according to the outlier value of each effective sampling point.

[0055] After the above determination of the abnormal sampling points, the number of the abnormal sampling points, that is, the abnormal sampling point number m, can be determined, and then the abnormal point coefficient Q = m / n can be determined according to the abnormal sampling point number m and the effective sampling point number n (that is, the total number of the effective sampling points).

[0056] Therefore, the global feature vector G of the wafer can be generated, where G = [μ, CV, CER, U θ , Q], and in some scenarios, the global feature vector G can also be G = [μ, σ, CV, CER, Gr, U θ , Q], so that comprehensive and detailed wafer film thickness distribution information is provided through the global feature vector G. In addition, in order to facilitate the description, the embodiment of the present application takes G = [μ, CV, CER, U θ , Q] as an example to illustrate the process of generating the front quantity detection result of the wafer, and the global feature vector G of the remaining cases can be adaptively adjusted according to the case, which will not be described in detail herein.

[0057] In one embodiment, the preset feature threshold includes: a preset target film thickness T target , a preset error δ, a preset coefficient of variation threshold CV limit , a preset center-to-edge ratio threshold CER limit , a preset azimuthal non-uniformity threshold U θlimit , and a preset abnormal point coefficient threshold Q limit; generating a pre-quantity detection result of the wafer according to the global feature vector G and a preset feature threshold value, including the following cases:

[0058] (1) generating a film thickness detection result according to the film thickness prediction mean value, the preset target film thickness and the preset error; wherein the film thickness detection result includes a film thickness detection qualified result and a film thickness detection unqualified result; if the film thickness detection result is the film thickness detection unqualified result, determining that the pre-quantity detection result is a pre-quantity detection failure result.

[0059] Specifically, calculating an absolute value of a difference between the film thickness prediction mean value and the preset target film thickness, and generating the film thickness detection result according to the absolute value of the difference and the preset error; wherein when the absolute value of the difference is not greater than the preset error, i.e. | μ - T t arg et |≤δ, the film thickness detection result is the film thickness detection qualified result; otherwise, when the absolute value of the difference is greater than the preset error, i.e. | μ - T t arg et |>δ, the film thickness detection result is the film thickness detection unqualified result, thereby realizing the average film thickness detection of the wafer according to the film thickness prediction mean value, the preset target film thickness and the preset error.

[0060] In addition, when the film thickness detection result is the film thickness detection unqualified result, it is directly determined that the pre-quantity detection result is the pre-quantity detection failure result, i.e. the wafer abnormality cannot enter the next CMP process at this time, and the abnormality reason is recorded, and an alarm information is generated, etc., thereby avoiding the CMP process failure caused by the wafer abnormality, and reducing the waste of CMP consumables (polishing pad, polishing liquid).

[0061] (2) if the film thickness detection result is the film thickness detection qualified result, generating a uniformity detection result according to the coefficient of variation and a preset coefficient of variation threshold value; generating a radial distribution detection result according to the center-to-edge ratio and a preset center-to-edge ratio threshold value; generating a circumferential distribution detection result according to the azimuthal non-uniformity parameter and a preset azimuthal non-uniformity threshold value; generating an abnormal point detection result according to the abnormal point coefficient and a preset abnormal point coefficient threshold value; wherein the uniformity detection result, the radial distribution detection result, the circumferential distribution detection result and the abnormal point detection result all include a detection qualified result and a detection unqualified result; determining the pre-quantity detection result according to the uniformity detection result, the radial distribution detection result, the circumferential distribution detection result and the abnormal point detection result.

[0062] Specifically, when the film thickness detection result is the film thickness detection qualified result, at this time, the remaining feature parameters in the global feature vector G are detected respectively. For example, if the coefficient of variation is not greater than the preset coefficient of variation threshold value, i.e. CV limit , the uniformity detection result is the detection qualified result; otherwise, if the coefficient of variation is greater than the preset coefficient of variation threshold value, i.e. CV limit , the uniformity detection result is the detection unqualified result.

[0063] Similarly, if the center-to-edge ratio CER and the preset center-to-edge ratio threshold CER limit satisfy |CER-1|≤CER limit , the radial distribution detection result is a detection qualified result, otherwise, if the center-to-edge ratio CER and the preset center-to-edge ratio threshold CER limit satisfy |CER-1|>CER limit , the radial distribution detection result is a detection unqualified result. If the azimuthal non-uniformity parameter is not greater than the preset azimuthal non-uniformity threshold, i.e., U θ ≤U θlim it , the circumferential distribution detection result is a detection qualified result, otherwise, if the azimuthal non-uniformity parameter is greater than the preset azimuthal non-uniformity threshold, i.e., U θ >U θlimit , the circumferential distribution detection result is a detection unqualified result. If the outlier coefficient is not greater than the preset outlier coefficient threshold, i.e., Q limit ≤Q limit , the outlier detection result is a detection qualified result, otherwise, if the outlier coefficient is greater than the preset outlier coefficient threshold, i.e., Q >Q .

[0064] When the film thickness detection result is a film thickness detection qualified result, at this time, it is determined whether the number of detection unqualified results in the uniformity detection result, the radial distribution detection result, the circumferential distribution detection result and the outlier detection result is only one, if yes, it is determined that the pre-quantity detection result is a pre-quantity detection conditional passing result, i.e., the average film thickness detection is qualified but there is one item of distribution characteristic unqualified, which needs to be adjusted after the parameter corresponding to the unqualified distribution characteristic enters the next CMP process, for example, if the radial distribution detection result is a detection unqualified result, the chamber pressure distribution is adjusted to improve the wafer radial uniformity; or, if the circumferential distribution detection result is a detection unqualified result, the rotation speed is adjusted to improve the wafer circumferential uniformity, so as to realize automatic adjustment of the parameter of certain distribution characteristics according to the pre-quantity detection result, so as to ensure the safe execution of the CMP process.

[0065] (3) When the film thickness detection result is a film thickness detection qualified result, if the number of detection unqualified results in the uniformity detection result, the radial distribution detection result, the circumferential distribution detection result and the outlier detection result is greater than one, at this time, it is determined that the pre-quantity detection result is a pre-quantity detection unqualified result, the wafer is controlled not to enter the next CMP process, and the abnormal reason is recorded, and alarm information is generated, etc., to prompt the operator to manually check the semiconductor process equipment; at the same time, the CMP process failure caused by wafer abnormality is avoided, so as to reduce the waste of CMP consumables (polishing pad, polishing liquid) and the pollution caused by the wafer without pre-quantity detection in error operation.

[0066] (4) When the film thickness detection result is a film thickness detection qualified result, if none of the uniformity detection result, the radial distribution detection result, the circumferential distribution detection result and the abnormal point detection result contains a detection unqualified result, that is, each feature parameter in the global feature vector G is detected qualified, at this time the unqualified item number is 0, then the pre-quantity detection result is a pre-quantity detection passing result, and the wafer is controlled to proceed to the next CMP process.

[0067] Therefore, the wafer is detected by the global feature vector G, and the semiconductor process equipment is controlled according to the pre-quantity detection result. If the pre-quantity detection result is a pre-quantity detection passing result, the wafer is directly controlled to enter the CMP process; if the pre-quantity detection result is a pre-quantity detection conditional passing result, the wafer is adjusted according to the unqualified distribution feature parameter and then enters the CMP process, which not only avoids the waste of CMP consumables (polishing pad, polishing liquid) caused by wafer abnormalities, but also reduces the equipment maintenance cost caused by wafer abnormalities, such as reducing the CMP process cost by about 10% to 12%, and avoids running an incorrect process recipe caused by abnormal feature parameters when directly entering the CMP, thereby improving the stability of the CMP process; if the pre-quantity detection result is a pre-quantity detection unpassing result, the wafer is directly controlled not to enter the next CMP process, and the abnormal reason is recorded, and an alarm information is generated, etc., to prompt the operator to manually check the semiconductor process equipment, so as to not only realize the comprehensive detection of the wafer and improve the quality control ability, but also save the logistics time and labor cost without transferring the wafer to a special measurement equipment, thereby significantly improving the detection efficiency.

[0068] In an embodiment, for the prediction model pre-trained in the embodiment of the application, the training process is as follows:

[0069] (S1) Construct a feature vector training set.

[0070] Specifically, a plurality of baseline film layer wafers with standard film thickness are selected, the metal layer of each baseline film layer wafer is different, and the thickness range of each baseline film layer wafer covers ±30% of the previous metal film layer thickness, and the film thickness distribution should include at least 5 sampling points of different thicknesses, which are uniformly distributed at the center and edge of the baseline film layer wafer. In addition, the real film thickness of the baseline film layer wafer can also be pre-verified by a high-precision offline measurement tool (such as XRF).

[0071] For any baseline film layer wafer, multiple sampling points thereof are detected, and it is ensured that the center, the median and the edge area thereof are detected; meanwhile, the signal set S(x, j, k) of the baseline film layer wafer is obtained by repeating the detection k times for each sampling point, where x represents the identification information such as the sequence information of the baseline film layer wafer in multiple baseline film layer wafers, j represents the jth sampling point of the baseline film layer wafer, and j ∈ [1, y], y being the total number of sampling points, for example, for the 2nd sampling point of 17 sampling points on the first baseline film layer wafer, the signal set S(1, 2, 3) is obtained by sampling 3 times.

[0072] Then, the average signal value AvgS(x, j) = ∑S(x, j, k) / k of each sampling point is calculated for the signal set S(x, j, k) of the baseline film layer wafer, and the signal standard deviation stdS(x, j) = sqrt(∑S(x, j, k) - AvgS(x, j) 2 ) / k of each sampling point is also calculated, if stdS(x, j) / AvgS(x, j) > 0.5, the sampling point j is marked as an unstable point, and needs to be detected again until each sampling point satisfies stdS(x, j) / AvgS(x, j) ≤ 0.5, and finally the data pairs of multiple stable points are obtained, each data pair including the average signal value and the standard film thickness of the stable point, that is, {AvgS(x, j), T(x, j)}, where AvgS(x, j) represents the eddy current signal of the jth sampling point in the xth baseline film layer wafer, and T(x, j) represents the standard film thickness of the jth sampling point in the xth baseline film layer wafer.

[0073] Therefore, the training data pairs of the baseline film layer wafer can be obtained according to the data pairs of multiple sampling points on the same baseline film layer wafer, and the training data pair set can be composed according to the training data pairs of multiple baseline film layer wafers, so as to generate the feature vector training set according to the training data pair set.

[0074] Wherein, the eddy current signal of any sampling point can be subjected to feature extraction, which includes but is not limited to time domain feature extraction, frequency domain feature extraction and statistical feature extraction; specifically, the time domain feature parameter F time extracted by the time domain feature extraction includes but is not limited to the maximum value, the mean value, the peak value, the RMS value and the like, and the shape feature includes the kurtosis, the skewness and the waveform factor and the like; similarly, the frequency domain feature parameter F freq extracted by the frequency domain feature extraction (such as FFT (Fast Fourier Transform)) includes but is not limited to the main frequency amplitude, the phase, the harmonic ratio and the power spectral density and the like; and the statistical feature parameter F fastThe signal stability, signal-to-noise ratio, and phase fluctuation, etc. are included but not limited to, and thus the feature vector F = [F time , F freq , F fast ] of the sampling point can be generated.

[0075] In addition, the generated feature vector F can be reduced in dimension by using the PCA (Principal Components Analysis) technology, for example, the characteristic covariance matrix of the feature vector F is calculated, and the eigenvalues and eigenvectors of the characteristic covariance matrix are calculated, the eigenvectors corresponding to the L largest eigenvalues are selected to form a projection matrix P, and thus the reduced feature vector F reduced = P * F is calculated.

[0076] Therefore, by reducing the dimension of the feature vector of each sampling point, the feature vector training set can be finally generated, and the feature vector training set includes a data training pair corresponding to each sampling point, and each data training pair includes the reduced feature vector F reduced of the sampling point and the standard film thickness. It should be noted that the above dimension reduction processing and feature extraction can refer to the prior art, and the embodiments of the present application will not be described in detail here.

[0077] After the feature vector training set is generated, a prediction model is constructed, and the prediction model uses an enhanced multi-feature fusion calibration model (EMFCM) to solve the complex nonlinear mapping relationship between the eddy current signal and the film thickness. The EMFCM model is a segmented model, and different segmented models are determined for different film thicknesses W, for example, when W < 100 nm, the segmented model is an SVR (Support Vector Regression) model; when 100 nm ≤ W < 500 nm, the segmented model is an XGBoost (eXtreme Gradient Boosting) model; and when W ≥ 500 nm, the segmented model is a polynomial regression model. The specific segmented model and film thickness range can be adaptively adjusted according to actual conditions.

[0078] After the segmented model is determined, each segmented model is trained by using the feature vector training set, and in the training process, the optimal hyperparameters can be determined by using the grid search, and the K-fold cross-validation is used to avoid overfitting. For details, please refer to the prior art, and the embodiments of the present application will not be described in detail here.

[0079] In addition, for the trained segmented model, in order to ensure the accuracy of the model, an error compensation value C(F) is also pre-set, and the error compensation value C(F) is a fixed value which can be set according to actual conditions. Therefore, for the trained prediction model, the prediction function is T final (F) = Ensemble(F) + C(F), where C(F) represents the error compensation value, and Ensemble(F) represents the segmented model, for example, when the film thickness W < 100 nm, Ensemble(F) is the SVR model, when 100 nm ≤ W < 500 nm, Ensemble(F) is the XGBoost model, and when W ≥ 500 nm, Ensemble(F) is the polynomial regression model, and the specific segmented model can refer to the prior art, and the embodiment of the present application will not be described in detail here. In addition, actual production data can be collected regularly and the model can be updated, and finally a trained prediction model meeting the requirements is obtained. It should be noted that the specific model training process can refer to the prior art, and the embodiment of the present application will not be described in detail here.

[0080] Therefore, according to the prediction model, the wafer prediction information set can be quickly output according to the wafer feature vector set, and the wafer pre-quantity detection efficiency is further improved.

[0081] Embodiment two

[0082] In order to facilitate understanding, the wafer pre-quantity detection process is described by taking a 300 mm wafer and a preset Cu film thickness of 300 nm as an example. First, the feature vector set is generated according to the eddy current signals of the plurality of sampling points on the wafer, and the feature vector set is input into the pre-trained prediction model, so as to output the prediction information set, and the global feature vector G = [μ, σ, CV, CER, Gr, U θ , of the wafer is generated according to the prediction information set, where μ is 585 nm, and |μ-T target | > δ, indicating that the film thickness detection is significantly high, that is, the film thickness detection result is a film thickness detection unqualified result.

[0083] Similarly, σ is 35.1 nm, which is greater than the corresponding threshold value, indicating that the standard deviation is large, and the detection result is a detection unqualified result; CV = 6.0%, which is greater than CV limit (3%, for example), indicating that the coefficient of variation is high, and the uniformity detection result is a detection unqualified result; CER = 1.15, |CER-1| > 0.1, indicating that the center is significantly thicker than the edge, and the radial distribution detection result is a detection unqualified result; Gr = 0.05 nm / mm, indicating that the radial gradient is large; U θ is 3.5%, which is greater than U θlimit(2%), which indicates that the orientation non-uniformity is high, the circumferential distribution detection result is a detection failure result; the abnormal point coefficient Q = 0.22, which is greater than Q limit (5%), the abnormal point detection result is a detection failure result, therefore, it is determined that the front quantity detection result is a front quantity detection failure result, at this time, the semiconductor process equipment needs to control the wafer not to enter the next CMP process, and record the abnormal reason, and generate alarm information and the like to prompt the operator to manually check the semiconductor process equipment, for example, according to the abnormal data feedback, the upstream deposition process finds that the air flow control valve is faulty, so that timely maintenance is realized.

[0084] Further, the embodiment of the present application also provides a semiconductor process equipment, including a memory, a processor and a computer program stored on the memory and executable on the processor, and the processor executes the computer program to realize the above-mentioned method embodiment.

[0085] The semiconductor process equipment provided by the embodiment of the present application has the same technical features as the wafer front quantity detection method provided by the above-mentioned embodiment, so it can also solve the same technical problems and achieve the same technical effects.

[0086] The embodiment also provides a machine readable storage medium, which stores machine executable instructions, and the machine executable instructions make the processor realize the wafer front quantity detection method when the machine executable instructions are called and executed by the processor.

[0087] The wafer front quantity detection method and the computer program product of the semiconductor process equipment provided by the embodiment of the present application include a computer readable storage medium storing program codes, and the instructions included in the program codes can be used to execute the method described in the above-mentioned method embodiment, and the specific implementation can be referred to the method embodiment, which will not be described here.

[0088] Those skilled in the art can clearly understand that, for the convenience and brevity of the description, the specific working process of the above-mentioned system and device can refer to the corresponding process in the above-mentioned method embodiment, which will not be described here.

[0089] In addition, in the description of the embodiment of the present application, unless otherwise explicitly specified and limited, the terms "mounting", "connection" and "connection" should be understood in a broad sense, for example, it can be fixed connection, or detachable connection, or integral connection; it can be mechanical connection, or electrical connection; it can be direct connection, or indirect connection through intermediate medium, or internal communication of two elements. For those skilled in the art, the specific meaning of the above-mentioned terms in the present application can be understood according to the specific circumstances.

[0090] If the functions are realized in the form of software function units and sold or used as independent products, they can be stored in a nonvolatile computer readable storage medium executable by a processor. Based on this understanding, the technical solutions of the present application essentially or the parts that contribute to the prior art or parts of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0091] In the description of the present application, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as limiting the present application. In addition, the terms "first", "second", "third" are only for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0092] Finally, it should be noted that: the above-described embodiments are only specific embodiments of the present application, used to illustrate the technical solutions of the present application, and are not limited thereto, the protection scope of the present application is not limited thereto, although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art within the technical scope disclosed by the present application can still modify or easily think of changes to the technical solutions recorded in the foregoing embodiments, or make equivalent replacements to part of the technical features; and these modifications, changes or replacements do not make the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A wafer pre-measurement detection method, characterized in that: The method comprises: Obtaining a feature vector set of the wafer; wherein the feature vector set includes feature vectors corresponding to a plurality of sampling points on the wafer, each feature vector being determined based on a detection signal of the corresponding sampling point and including at least one of the following: a time domain feature parameter, a frequency domain feature parameter, and a statistical feature parameter; Inputting the feature vector set into a pre-trained prediction model so that the prediction model outputs a prediction information set based on the feature vector set; wherein the prediction information set includes prediction information corresponding to a plurality of sampling points, and the prediction information includes a film thickness prediction value and uncertainty; Generating a global feature vector of the wafer according to the prediction information set, wherein the global feature vector includes: a film thickness prediction mean, a coefficient of variation, a center-to-edge ratio, an azimuthal non-uniformity parameter, and an outlier coefficient; The pre-quantity detection result of the wafer is generated according to the global feature vector and the preset feature threshold; wherein the pre-quantity detection result includes: a pre-quantity detection pass result, a pre-quantity detection fail result and a pre-quantity detection conditional pass result.

2. The method according to claim 1, characterized in that The step of generating a global feature vector of the wafer according to the prediction information set includes: determining a plurality of valid sampling points according to the prediction information of the sampling points; The global feature vector is generated according to the film thickness prediction values ​​of the plurality of effective sampling points.

3. The method according to claim 2, characterized in that The plurality of effective sampling points include a first effective sampling point located at the center of the wafer and a plurality of second effective sampling points located at the edge of the wafer; The step of generating the global feature vector according to the film thickness prediction values ​​of the plurality of effective sampling points comprises: Calculating the film thickness prediction mean and the film thickness prediction standard deviation based on the film thickness prediction values ​​of the plurality of valid sampling points, and determining the coefficient of variation based on the film thickness prediction mean and the film thickness prediction standard deviation; Calculating the center-to-edge ratio and the radial gradient according to the film thickness prediction value of the first effective sampling point and the film thickness prediction values ​​of the plurality of second effective sampling points, and determining the azimuthal nonuniformity parameter according to the center-to-edge ratio and the radial gradient; Abnormal sampling points among the plurality of valid sampling points are determined, and the abnormal point coefficient is determined according to the number of abnormal sampling points and the number of valid sampling points.

4. The method according to claim 3, characterized in that The step of determining abnormal sampling points among the plurality of valid sampling points comprises: Calculate the outlier value of each effective sampling point according to the film thickness prediction value of the effective sampling point; The valid sampling point whose outlier value is greater than the preset outlier threshold is regarded as the abnormal sampling point.

5. The method according to claim 2, characterized in that The step of determining a plurality of valid sampling points according to the prediction information of the sampling points comprises: Calculating the confidence of each sampling point based on the film thickness prediction value and uncertainty of the sampling point; The sampling points whose confidence levels are not less than a preset confidence threshold are taken as the valid sampling points.

6. The method according to claim 1, characterized in that The preset characteristic threshold includes: a preset target film thickness and a preset error; The step of generating a pre-measurement detection result of the wafer according to the global feature vector and a preset feature threshold comprises: Generate a film thickness detection result according to the film thickness prediction mean, the preset target film thickness and the preset error; wherein the film thickness detection result includes: a film thickness detection qualified result and a film thickness detection unqualified result; If the film thickness detection result is an unqualified film thickness detection result, the pre-quantity detection result is determined to be a failed pre-quantity detection result.

7. The method according to claim 6, characterized in that The preset feature thresholds further include: a preset coefficient of variation threshold, a preset center-to-edge ratio threshold, a preset orientation non-uniformity threshold, and a preset outlier coefficient threshold; and the step of generating the pre-measurement detection result of the wafer according to the global feature vector and the preset feature thresholds includes: If the film thickness test result is a qualified film thickness test result, a uniformity test result is generated according to the coefficient of variation and the preset coefficient of variation threshold; a radial distribution test result is generated according to the center-to-edge ratio and the preset center-to-edge ratio threshold; a circumferential distribution test result is generated according to the azimuthal nonuniformity parameter and the preset azimuthal nonuniformity threshold; and an outlier detection result is generated according to the outlier coefficient and the preset outlier coefficient threshold. The front quantity detection result is determined according to the uniformity detection result, the radial distribution detection result, the circumferential distribution detection result and the abnormal point detection result.

8. The method according to claim 7, characterized in that The uniformity test result, the radial distribution test result, the circumferential distribution test result and the abnormal point test result all include a qualified test result and a failed test result; The step of determining the front quantity detection result according to the uniformity detection result, the radial distribution detection result, the circumferential distribution detection result and the abnormal point detection result comprises: If there is only one unqualified test result among the uniformity test result, the radial distribution test result, the circumferential distribution test result, and the abnormal point test result, the front quantity test result is determined to be a conditionally passed result of the front quantity test; or, If the number of unqualified test results in the uniformity test result, the radial distribution test result, the circumferential distribution test result, and the abnormal point test result is greater than one, the front quantity test result is determined to be a failed front quantity test result; or If none of the uniformity detection result, the radial distribution detection result, the circumferential distribution detection result and the abnormal point detection result include the unqualified detection result, the front quantity detection result is determined to be the passed front quantity detection result.

9. The method according to claim 1, characterized in that The method further comprises: If the pre-quantity detection result is a failure result of the pre-quantity detection, an alarm message is generated, and the wafer is controlled to stop entering the next process for processing.

10. A semiconductor process device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 9 are implemented.

Images