Analysis and evaluation method for wafer height detection signal

Through the combination of wavelet transform and Fourier transform, the problem that the existing technology cannot obtain the time and frequency domain characteristics of the highly detectable signal at the same time is solved, and the accurate analysis and evaluation of non-stationary signals is realized, and the signal stability evaluation is improved. accuracy.

CN113177534BActive Publication Date: 2024-12-06INST OF MICROELECTRONICS CHINESE ACAD OF SCI LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202110600845.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-05-31
Publication Date
2024-12-06
Estimated Expiration
2041-05-31

AI Technical Summary

Technical Problem

The existing time-domain and frequency-domain analysis methods cannot obtain the time-domain and frequency-domain characteristics of highly probed signals at the same time, and due to the non-stationarity of the signal, direct Fourier transform will lead to a loss of analysis accuracy.

Method used

The wavelet transform method is used to decompose the height detection signal, and combine one-dimensional discrete wavelet transform and fast Fourier transform to obtain the time and frequency domain characteristics of the signal, and improve the accuracy of the analysis through block analysis and wavelet packet transformation.

Benefits of technology

Simultaneous time and frequency domain analysis of non-stationary height detection signals is realized, which can evaluate the stability and interference frequency band of the signal, suppress the impact of the environment and drift, and improve the accuracy and effect of the analysis.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure BDA0003092647220000051
    Figure BDA0003092647220000051
  • Figure BDA0003092647220000052
    Figure BDA0003092647220000052
  • Figure BDA0003092647220000053
    Figure BDA0003092647220000053
Patent Text Reader

Abstract

The present invention provides an analysis and evaluation method for wafer height detection signals, comprising: using a wavelet transform method, utilizing the characteristics of multi-resolution, performing wavelet decomposition on the height detection signal, and simultaneously obtaining the frequency domain characteristics and time domain characteristics of the height detection signal. The mathematical method of wavelet transform is used to simultaneously obtain the time domain and frequency domain characteristics of the height detection signal, solving the problem that the detection signal belongs to a non-stationary model.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of semiconductor manufacturing and detection technology, and in particular to an analysis and evaluation method for wafer height detection signals. Background Art

[0002] In the field of semiconductor manufacturing and testing technology, it is generally necessary to measure the surface height morphology of wafer samples to be processed or tested for further exposure processing or direct wafer characteristic analysis.

[0003] At present, the industry often uses detection technology based on image methods or optical methods, and the second detection method is often used for nano-level measurement. This method generally requires an independent wafer height measurement device to obtain the height detection signal of the wafer surface through triangulation and other methods.

[0004] The signal stability of the height detection device has an important impact on the accuracy of the wafer topography results and is one of the key technical indicators of the height detection device. The evaluation method for this indicator also needs to be strictly defined. The commonly used method is to use the time domain or frequency domain analysis method. When the time domain method is used, the detection signal of a certain time period is generally obtained at a fixed sampling frequency, and then the statistical results such as the pv value or average value of the detection signal are calculated. When the frequency domain method is used, the time domain signal is generally converted to the frequency domain through Fourier transform, and then the frequency characteristics of each frequency band or certain characteristic frequency points are analyzed. However, both methods have certain limitations. First, time domain analysis and frequency domain analysis cannot be obtained at the same time, so that neither the frequency band where the abnormal signal value is located in the time domain analysis nor the time when the signal value with certain characteristic frequencies occurs can be obtained. At the same time, since the height detection signal is not a stationary signal, the direct use of Fourier transform will result in a loss of analysis accuracy. Summary of the invention

[0005] In view of this, the main purpose of the present invention is to provide a method for analyzing and evaluating wafer height detection signals, in order to partially solve at least one of the above-mentioned technical problems.

[0006] In order to achieve the above object, as one aspect of the present invention, a method for analyzing and evaluating a wafer height detection signal is provided, comprising:

[0007] The wavelet transform method is adopted and the multi-resolution characteristic is utilized to perform wavelet decomposition on the height detection signal, and the frequency domain characteristics and time domain characteristics of the height detection signal are obtained at the same time.

[0008] Wherein, the wavelet transform uses one-dimensional discrete wavelet transform to analyze the position detection signal, and uses haar wavelet basis or db wavelet basis for analysis.

[0009] The selected wavelet basis is used to analyze the position detection signal, obtain the corresponding decomposed wavelet coefficient sequence, and draw a discrete wavelet decomposition diagram.

[0010] Among them, if the value of the vanishing moment is too large, it will affect the compact support of the wavelet function. The wavelet transform improves the accuracy of height detection signal analysis and evaluation by selecting different vanishing moments and numbers of layers and performing repeated iterative analysis.

[0011] The wavelet transform selects a method based on continuous wavelet transform or wavelet packet wavelet transform.

[0012] The method further comprises performing frequency domain analysis on the height detection signal to preliminarily obtain characteristic analysis of the position detection signal in the frequency domain; the frequency domain analysis method adopts fast Fourier transform.

[0013] Among them, the method of performing block analysis on the height detection signal is processed.

[0014] The block analysis method comprises the following steps:

[0015] Format the acquired height detection signal and determine the data sampling frequency F;

[0016] Split the data to obtain a certain number of data blocks;

[0017] For each data block, calculate the statistical value corresponding to the data block using all the sample values ​​contained therein;

[0018] By analyzing the changes in the mean values ​​and mean square errors corresponding to all data blocks, the overall stability of the entire detection model can be further evaluated.

[0019] Wherein, dividing the data to obtain a certain number of data blocks includes:

[0020] When the total number of height detection signal samples is S, and the number of samples of each data block is N, the number of data blocks obtained is M=S / N, and the remainder is regarded as discarded.

[0021] The number of samples of each data block is the minimum action time of height detection, and the method for determining the number of samples of each data block needs to be determined according to the specific application scenario;

[0022] During the wafer height detection process, assuming that the length of a Die is L and the uniform scanning speed is V, the measurement time for a Die is T=L / V, where T is considered to be the minimum action time for height detection. The number of samples for each data block is N=T*F; if the total number of Dies is K, then S=N*K.

[0023] Based on the above technical solution, it can be seen that the analysis and evaluation method for wafer height detection signals of the present invention has at least one of the following beneficial effects compared with the prior art:

[0024] 1. The mathematical method of wavelet transform is used to simultaneously obtain the time domain and frequency domain characteristics of the height detection signal to solve the problem that the detection signal belongs to a non-stationary model.

[0025] 2. Use the time domain information of wavelet transform to analyze the signal and time-related characteristics during the detection process, such as drift and transient jitter; use the frequency domain information of wavelet transform to analyze the frequency band of the interference signal used to detect the signal.

[0026] 3. The time domain segmented analysis method is used to evaluate the overall time domain characteristics of the altitude detection signal, thereby suppressing the influence of slow-changing factors such as the environment and drift; the frequency domain segmented analysis method is used to calculate the components and cumulative amounts of the altitude detection signal in each frequency band, and evaluate the distribution of the altitude detection signal in the frequency domain.

[0027] 4. The method proposed in the present invention can be applied to semiconductor manufacturing and testing equipment, especially to the wafer height detection system used in chip manufacturing lithography equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 A schematic diagram of wafer height profile detection provided by an embodiment of the present invention;

[0029] Figure 2 A time domain curve of a height signal provided by an embodiment of the present invention;

[0030] Figure 3 The Fourier transform-based height signal analysis result provided by the embodiment of the present invention;

[0031] Figure 4 A height signal stability curve for data block analysis provided by an embodiment of the present invention;

[0032] Figure 5 The components of each frequency band and the accumulated amount of the altitude signal provided by the embodiment of the present invention;

[0033] Figure 6 A time domain curve of a height signal provided by an embodiment of the present invention (with airflow disturbance and mechanical vibration);

[0034] Figure 7 The Fourier transform-based height signal analysis results (with airflow disturbance and mechanical vibration) provided in the embodiment of the present invention;

[0035] Figure 8 The height signal analysis result based on discrete wavelet transform provided by the embodiment of the present invention;

[0036] Fig. 9 The height signal analysis result based on continuous wavelet transform provided by the embodiment of the present invention;

[0037] Fig.10 The height signal analysis result based on wavelet packet provided by the embodiment of the present invention (the low-frequency part of the third layer);

[0038] Fig.11 The wavelet packet-based height signal analysis result (the third layer high frequency part) provided in the embodiment of the present invention. DETAILED DESCRIPTION

[0039] The detection signal analysis and evaluation method proposed in the present invention first adopts the segmented time domain and segmented frequency domain analysis methods to evaluate the local and overall characteristics of the height detection signal. Then, the mathematical method of wavelet transform is adopted to perform wavelet decomposition on the obtained height detection signal by utilizing the characteristics of multi-resolution, and the frequency domain characteristics and time domain characteristics of the detection signal are obtained at the same time. For the analysis of the frequency domain, similar to the conventional Fourier transform, the distribution of the detection signal at different resolutions can be obtained, and the frequency characteristics and time information of the occurrence of the system or external interference can be obtained, which is particularly suitable for instantaneous interference and signal mutation.

[0040] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments and with reference to the accompanying drawings.

[0041] The present invention proposes a method for analyzing and evaluating position detection signals, which can analyze the height detection signals in the frequency domain and the spatial domain. The implementation steps adopted by the present invention are:

[0042] The acquired height detection signal is formatted and parameters such as the data sampling frequency F are determined.

[0043] Perform frequency domain analysis on the height detection signal to obtain preliminary characteristic analysis of the position detection signal in the frequency domain. The frequency domain analysis method mainly uses Fourier transform, and the fast Fourier analysis method is generally used in practical applications.

[0044] by Figure 2 Take the height detection signal as an example for analysis, the corresponding frequency domain characteristics are as follows Figure 3 shown.

[0045] In order to suppress the influence of environment or drift, the obtained height detection signal can be processed by block analysis.

[0046] When the total number of height detection signal samples is S, and the number of samples in each data block is N, the number of data blocks that can be obtained is M=S / N, and the remainder can be discarded.

[0047] The method for determining the number of samples N for each data block can be determined according to the specific application scenario, that is, the minimum action time of height detection. Typical application scenarios are as follows: Figure 1 As shown, in the process of wafer height detection, assuming that the length of a Die is L and the speed of uniform scanning is V, the measurement time for a Die is T=L / V, and T can be considered as the minimum evaluation time for height detection.

[0048] The number of samples of each data block is N = T * F. If the total number of Dies is K, then S = N * K.

[0049] For each data block, all the sample values ​​contained in it are used to calculate the statistical value corresponding to the data block. For the i-th data block, i∈[1,K], where each sample is s i,j ,j∈[1,N], then

[0050]

[0051]

[0052] Furthermore, the following formula can be used to represent the comprehensive situation of the above statistical values, where n is used to represent the size of the confidence interval.

[0053] ε i =μ i +nσ i or ε i =nσ i

[0054] By analyzing the changes in the mean and mean square error corresponding to all data blocks, the overall stability of the entire detection model can be further evaluated.

[0055]

[0056] By analogy, we can get the following calculation formula

[0057]

[0058]

[0059] Figure 4 The 3σ variation curve for each data block can describe the stability of the height detection signal in the local time period and the entire time range.

[0060] Furthermore, a frequency segmentation method is adopted to divide the half sampling frequency F / 2 into multiple frequency segments. Without loss of generality, F / 2 is equally divided with a frequency interval of Fs.

[0061] Perform Fourier transform on the height detection signal, calculate the sum of the components in all frequency bands, and draw curves of the components of the height detection signal in each frequency band and the total amount accumulated over the frequency band, such as Figure 5 shown.

[0062] Figure 5 It is the distribution of the height detection signal at each frequency, and can intuitively obtain the frequency band that has the greatest impact on the height detection model.

[0063] The actual height detection signal will be affected by external interference, self-noise, environment and drift, etc. It is not a strictly stable signal. Frequency analysis using Fourier transform alone has great limitations.

[0064] Without loss of generality, Figure 2 The height detection signal is based on the interference signal as an analysis example. The interference of external airflow disturbance and mechanical vibration is added. There is no interference in the first 1 / 3 period, the middle 1 / 3 period is the height detection signal under airflow disturbance, and the last 1 / 3 period is the height detection signal under mechanical vibration, such as Figure 6 shown.

[0065] Figure 6 Two different sources of external interference are listed, but the method in this paper is not limited to the above two types of external interference. For example, random noise, signal drift, etc. can all be analyzed using the wavelet transform method described in this paper.

[0066] like Figure 7 Although the frequency characteristic diagram shown clearly shows that the height detection signal is subject to strong interference, it is impossible to determine the specific frequency distribution and obtain the corresponding time information.

[0067] The wavelet transform mathematical method is used to analyze the height detection signal, and its multi-resolution characteristics are utilized to simultaneously obtain the time domain and frequency domain information of the height detection signal.

[0068] One-dimensional discrete wavelet transform is used to analyze the position detection signal. Various wavelet bases can be used for analysis, such as haar wavelet base, db wavelet base, etc.

[0069] According to the resolution to be analyzed and the data sampling frequency F, select the appropriate decomposition level L and vanishing moment N. When using dyadic wavelet, the frequency detail resolution that can be achieved is F / 2^(L+1).

[0070] If the value of the vanishing moment N is too large, it will affect the compact support of the wavelet function. Generally, a compromise should be made based on the effect after analyzing the detection signal. Typically, 5 to 9 can be selected.

[0071] The low-frequency graph of the signal after wavelet transform processing reflects the overall overview of the signal, covering the signal frequency range of 0 ~ F / 2^(L+1).

[0072] The high-frequency graph of the signal after wavelet transform processing reflects the detailed characteristics of the signal in different frequency bands. When the number of decomposition layers is j, the frequency range corresponding to this layer is F / 2^(L+1)~F / 2^L.

[0073] Take db wavelet as an example for analysis, N takes 5th order, L takes 7th layer. Analyze the detection signal, obtain the corresponding decomposed wavelet coefficient sequence, and draw the discrete wavelet decomposition diagram, such as Figure 8 It can be clearly seen that the height detection signal is in three different time periods.

[0074] Airflow disturbances are mainly concentrated in the low-frequency and medium-frequency bands, and the wavelet systems in each frequency band are of similar magnitude, indicating that the effects on these frequency bands are similar. Mechanical vibrations are mainly concentrated in the high-frequency band, and the frequency bands in which they are located are relatively concentrated. The low-frequency band also shows the drift characteristics of the height detection signal.

[0075] Further, continuous wavelet transform is used for analysis. Taking db wavelet as an example, N is 5th order, L is 7th layer, and the scale factor is 64. The results are as follows: Fig. 9 Through the coefficient distribution diagram, we can intuitively get the size of the wavelet coefficients at different scales in a specific time period, and then get the impact of the corresponding interference signal on the detection signal.

[0076] The use of wavelet packet transform for analysis can make up for the shortcomings of wavelet transform in the high-frequency part of the signal. Taking db wavelet as an example, N takes 5 orders and L takes 7 layers. Fig.10 and Fig.11 shown.

[0077] Based on the above wavelet transform analysis method, different wavelet bases, vanishing moments, number of layers and other parameters can be selected, and the analysis can be iterated repeatedly to improve the accuracy of height detection signal analysis and evaluation.

[0078] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for analyzing and evaluating wafer height detection signals, characterized in that: include: The height detection signal is processed by a block analysis method to obtain the overall stability of the height detection signal, wherein the block analysis method for processing the height detection signal includes: formatting the obtained height detection signal to determine the data sampling frequency F; according to the length of a Die L and the speed of uniform scanning V, the measurement time of the Die and the minimum action time of the height detection are obtained to be T=L / V, and the number of samples of each data block is N=T*F; in the case of K Dies, the total number of samples of the height detection signal is calculated to be S=N*K; according to the total number of samples of the height detection signal and the number of samples of each data block, the number of data blocks is M=S / N; for each data block, using all the sampled values ​​contained therein, the statistical value corresponding to each data block is calculated; by analyzing the changes in the mean value and the mean square error corresponding to all data blocks, the overall stability of the height detection signal is evaluated; The wavelet transform method is adopted and the multi-resolution characteristic is utilized to perform wavelet decomposition on the height detection signal, and the frequency domain characteristics and time domain characteristics of the height detection signal are obtained at the same time.

2. The method according to claim 1, characterized in that The wavelet transform uses one-dimensional discrete wavelet transform to analyze the position detection signal, and uses haar wavelet basis or db wavelet basis for analysis.

3. The method according to claim 2, characterized in that The position detection signal is analyzed using the selected wavelet basis to obtain the corresponding decomposed wavelet coefficient sequence and draw a discrete wavelet decomposition diagram.

4. The method according to claim 2, characterized in that: A vanishing moment that is too large will affect the compact support of the wavelet function. The wavelet transform improves the accuracy of height detection signal analysis and evaluation by selecting different vanishing moments and numbers of layers and performing repeated iterative analysis.

5. The method according to claim 1, characterized in that The wavelet transform selects a continuous wavelet transform or a wavelet packet based wavelet transform method.

6. The method according to claim 1, characterized in that The method also includes performing frequency domain analysis on the height detection signal to preliminarily obtain characteristic analysis of the position detection signal in the frequency domain; the frequency domain analysis method adopts fast Fourier transform.