Synthetic Wavelength for Endpoint Detection in Plasma Etching

Abstract

A method for determining an endpoint of an etching process using optical emission spectroscopy (OES) data as input is described. The OES data is acquired by a spectrometer in a plasma etching process chamber. First, the acquired spectral data evolving over time is filtered and detrended, and subsequently the data is transformed into transformed spectral data or trends using multivariate analysis such as principal component analysis, in which the transformation is accomplished using previously calculated principal component weights. Separate signed trends (synthetic wavelengths) are created by grouping the principal component weights into two separate groups corresponding to positive and negative natural wavelengths.

Description

[0001] Cross - reference to related applications

[0002] This application is related to and claims the priority benefit of U.S. Patent Application No. 16 / 548,333, filed on August 22, 2019, entitled "Synthetic Wavelengths for Endpoint Detection in Plasma Etching", the entire content of which is incorporated herein by reference. Background of the Invention Field of the Invention

[0004] This application relates to a method and system for controlling an etching process of a structure on a substrate, such as in semiconductor manufacturing. More specifically, this application relates to a method for determining an endpoint of an etching process of a substrate.

[0005] Related Applications

[0006] This application is related to U.S. Patent No. 9,330,990 ('990) entitled "Method of endpoint detection of plasma etching process using multivariate analysis", and U.S. Patent No. 10,002,804 entitled "Method of endpoint detection of plasma etching process using multivariate analysis".

[0007] Description of Related Art

[0008] In the process of manufacturing semiconductor devices, liquid crystal displays (LCDs), light emitting diodes (LEDs), and some photovoltaic devices (PVs), plasma etching processes are typically used in combination with photolithography. Generally, a layer of radiation-sensitive material such as photoresist is first coated on a substrate and exposed to patterned light to impart a latent image to it. Subsequently, the exposed radiation-sensitive material is developed to remove the exposed (or, if a negative-tone photoresist is used, the unexposed) radiation-sensitive material, leaving a pattern of radiation-sensitive material that exposes the areas to be etched subsequently and covers the areas that are not desired to be etched. During the etching process (e.g., a plasma etching process), the substrate and the radiation-sensitive material pattern are exposed to high-energy ions in a plasma processing chamber to remove the material beneath the radiation-sensitive material in order to form etched features such as vias, trenches, etc. After the features are etched in the underlying material, the remaining portion of the radiation-sensitive material is removed from the substrate using a lift-off process to expose the formed etched structure for further processing.

[0009] In many types of devices such as semiconductor devices, a plasma etching process is performed in a first material layer covering a second material layer, and importantly, once the etching process has formed an opening or pattern in the first material layer, the etching process is accurately stopped without continuing to etch the underlying second material layer.

[0010] To control the etching process, various types of endpoint controls are used, some of which rely on analyzing the chemical composition of the gas in the plasma processing chamber in order to infer whether the etching process has progressed to, for example, an underlying layer having a chemical composition different from that of the layer being etched. Other processes may rely on making in-situ measurements directly on the etched structure. In the former group, optical emission spectroscopy (OES) is often used to monitor the chemical composition of the gas in the plasma processing chamber. The chemical species in the gas in the plasma processing chamber are excited by the plasma excitation mechanism used, and the excited chemical species produce different spectral features in the optical emission spectrum of the plasma. Changes in the optical emission spectrum due to, for example, clearing the etched layer and exposing the underlying layer on the substrate can be monitored and used to precisely end the etching process (i.e., reach the endpoint) to avoid etching the underlying layer or forming other yield-impacting defects such as undercuts.

[0011] Depending on the type of the etched structure and the etching process parameters, the change in the optical emission spectrum of the plasma at the etching process endpoint may be very obvious and easy to detect, or conversely, subtle and very difficult to detect. For example, etching a structure with a very low opening ratio may result in difficult endpoint detection using current algorithms for processing OES data. Therefore, improvements are needed to make the etching endpoint detection based on OES data more robust under such challenging etching process conditions. SUMMARY OF THE INVENTION

[0012] The features of the present application relate to a method for determining an endpoint of an etching process, where, at this endpoint, once the etching process has formed an opening or pattern in a first material layer, the etching process is accurately stopped without continuing to etch the underlying second material layer.

[0013] In a non-limiting embodiment, optical emission spectroscopy (OES) data for different etching runs are acquired to obtain an OES data matrix, an average OES data matrix, and a mean OES data matrix. This data is used such that a multivariate model of the acquired OES data can be established. Once the multivariate model of the OES data is established, it is subsequently used for in-situ etching endpoint detection.

[0014] Analysis of grouping wavelengths with similar behavior is used to determine a weight vector P to transform the OES data vector into a trend domain. Preferably, by grouping the principal component weights into two separate groups corresponding to positive and negative natural wavelengths, separate signed trends (synthetic wavelengths) are created.

[0015] After determining the synthetic wavelength, during in-situ etching endpoint detection, the functional form of the time evolution value of the synthetic wavelength is plotted against time to determine the endpoint of the etching process.

[0016] For example, in one embodiment, the time evolution of the synthetic wavelength ratio or the time evolution of the time derivative of the synthetic wavelength ratio is calculated. However, in other embodiments, any other functional form can be calculated, such as the square of the synthetic wavelength ratio or just a single signed synthetic wavelength or just the natural wavelength trend.

[0017] In a further non-limiting embodiment, to compensate for OES drift between different wafers, normalized OES spectra are used in the principal component analysis (PCA) method.

[0018] After calculating the time evolution trend variable, it is determined whether the endpoint has been reached. If the endpoint has indeed been reached, the etching process ends; otherwise, the etching process continues, and the etching endpoint is continuously monitored.

[0019] The generation of the synthetic wavelength makes the trend of endpoint detection similar to the natural wavelength, but with an endpoint signal having a higher signal-to-noise ratio (SNR). BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The present application will be better understood in view of the description given in a non-limiting manner, in conjunction with the accompanying drawings, in which:

[0021] Figure 1is a schematic diagram of an exemplary plasma etching processing system having an optical detection device (which includes a spectrometer for acquiring OES data), and a controller for implementing the etch endpoint detection method described herein.

[0022] Figure 2 is an exemplary flow chart of a method for using multivariate analysis to prepare etch endpoint data for later in-situ etch point detection.

[0023] Figure 3 is an exemplary flow chart of a method for using PCA analysis to prepare etch endpoint data for later in-situ etch point detection.

[0024] Figure 4 is an exemplary flow chart of an in-situ etch endpoint detection method.

[0025] Figure 5 Shows an exemplary graph of the time evolution of the time derivative of a trend variable function form involving the synthetic wavelength trend ratio and the single wavelength trend.

[0026] Figure 6A Shows an exemplary graph of the time evolution of the trend variable function form involving the synthetic wavelength trend ratio.

[0027] Figure 6B Shows the Figure 5 exemplary graph of the time evolution of the time derivative of the trend variable function form of A involving the synthetic wavelength trend ratio.

[0028] Figure 7A Shows an exemplary graph of the time evolution of the trend variable function form involving a single wavelength.

[0029] Figure 7B Shows the Figure 6A exemplary graph of the time evolution of the time derivative of the trend variable function form of

[0030] Figure 8A Shows an exemplary graph of the time evolution of the time derivative of the trend variable function form involving a single wavelength.

[0031] Figure 8B Shows an exemplary graph of the time evolution of the time derivative of the trend variable function form involving the synthetic wavelength trend ratio.

[0032] Figure 8C Shows an exemplary graph of the time evolution of the time derivative of the trend variable function form involving a single synthetic wavelength, and is normalized. Detailed Description

[0033] References to "an embodiment" or "embodiments" in the course of this specification mean that a particular feature, structure, material, or characteristic described in connection with the embodiment is included in at least one embodiment of the present application, but does not imply that it is present in every embodiment. Thus, the phrases "in an embodiment" or "in embodiments" appearing throughout this specification do not necessarily refer to the same embodiment of the present application. Moreover, the particular features, structures, materials, or characteristics may be combined in any suitable manner in one or more embodiments.

[0034] According to an embodiment of the present application, Figure 1 a plasma etch processing system 10 and a controller 55 are depicted, wherein the controller 55 is coupled to the plasma etch processing system 10. The controller 55 is configured to monitor the performance of the plasma etch processing system 10 using data obtained from various sensors disposed in the plasma etch processing system 10. For example, the controller 55 can be used to control various components of the plasma etch processing system 10 to detect faults and to detect the endpoint of an etch process.

[0035] According to Figure 1 the illustrated embodiment of the present application depicted in, the plasma etch processing system 10 includes a processing chamber 15, a substrate holder 20 (to which a substrate 25 to be processed is attached), a gas injection system 40, and a vacuum pumping system 58. For example, the substrate 25 can be a semiconductor substrate, a wafer, or an LCD. For example, the plasma etch processing system 10 can be configured to facilitate the generation of a plasma in a processing region 45 adjacent to the surface of the substrate 25, wherein the plasma is formed via collisions between heated electrons and an ionizable gas. An ionizable gas or gas mixture is introduced via the gas injection system 40, and the process pressure is adjusted. It is desirable to use the plasma to produce a material specific to a predetermined material process and to assist in removing material from the exposed surface of the substrate 25. For example, the controller 55 can be used to control the vacuum pumping system 58 and the gas injection system 40.

[0036] For example, the substrate 25 can be moved into and out of the plasma etch processing system 10 via a mechanical substrate transfer system through a gate valve (not shown) and a chamber feedthrough (not shown), where the substrate is received by substrate lift pins (not shown) within the substrate holder 20 and is mechanically translated by equipment housed within the substrate holder. Once the substrate 25 is received from the substrate transfer system, the substrate is lowered to the upper surface of the substrate holder 20.

[0037] For example, the substrate 25 can be attached to the substrate holder 20 via an electrostatic chucking system 28. Additionally, the substrate holder 20 can further include a cooling system that includes a recirculating coolant flow that receives heat from the substrate holder 20 and transfers the heat to a heat exchanger system (not shown), or transfers heat from the heat exchanger system when heating. Further, gas can be delivered to the back side of the substrate via a backside gas delivery system 26 to improve the thermal conductivity of the air gap between the substrate 25 and the substrate holder 20. Such a system can be utilized when temperature control of the substrate is required at elevated or reduced temperatures. For example, at temperatures above the steady-state temperature, which is achieved by the balance between the heat flux delivered to the substrate 25 from the plasma and the heat flux removed from the substrate 25 by conduction to the substrate holder 20, temperature control of the substrate can be useful. In other embodiments, heating elements such as resistive heating elements or thermoelectric heaters / coolers can be included.

[0038] Continuing to refer Figure 1 , for example, process gas can be introduced into the processing region 45 through a gas injection system 40. For example, the process gas can include gas mixtures for oxide etch applications such as argon, CF4, and O2 or Ar, C4F8, and O2, or other chemical substances such as, for example, O2 / CO / Ar / C4F8, O2 / CO / Ar / C5F8, O2 / CO / Ar / C4F6, O2 / Ar / C4F6, N2 / H2. The gas injection system 40 includes a showerhead, where the process gas is supplied from a gas delivery system (not shown) to the processing region 45 through a gas injection plenum (not shown) and a porous showerhead gas injection plate (not shown).

[0039] As Figure 1 Further shown, the plasma etch processing system 10 includes a plasma source 80. For example, RF or microwave power can be coupled from a generator 82 to the plasma source 80 through an impedance matching network or tuner 84. For capacitively coupled (CCP), inductively coupled (ICP), and transformer coupled (TCP) plasma sources, the frequency range for applying RF power to the plasma source is from 10 MHz to 200 MHz, and preferably 60 MHz. For microwave plasma sources 80 such as electron cyclotron resonance (ECR) and surface wave plasma (SWP) sources, the typical operating frequency of the generator 82 is between 1 and 5 GHz, and preferably about 2.45 GHz. An example of an SWP source 80 is a radial line slot antenna (RLSA) plasma source. Additionally, the controller 55 can be coupled to the generator 82 and the impedance matching network or tuner 84 to control the application of RF or microwave power to the plasma source 80.

[0040] As Figure 1As shown, the substrate holder 20 can be electrically biased with an RF voltage by transmitting RF power from an RF generator 30 through an impedance matching network 32 to the substrate holder 20. The RF bias can be used to attract ions from the plasma formed in the processing region 45 to facilitate an etching process. The frequency range for applying power to the substrate holder 20 can be from 0.1 MHz to 30 MHz, and preferably 2 MHz. Alternatively, RF power can be applied to the substrate holder 20 at multiple frequencies. In addition, the impedance matching network 32 is used to maximize the transfer of RF power to the plasma in the processing chamber 15 by minimizing the reflected power. Various matching network topologies (e.g., L-type, π-type, T-type, etc.) and automatic control methods can be utilized.

[0041] Various sensors are configured to receive tool data from the plasma etching processing system 10. The sensors can include both sensors inherent to the plasma etching processing system 10 and sensors external to the plasma etching processing system 10. Inherent sensors can include those related to the functions of the plasma etching processing system 10, such as measuring helium backside gas pressure, helium backside flow rate, electrostatic chuck (ESC) voltage, ESC current, substrate holder 20 temperature (or lower electrode (LEL) temperature), coolant temperature, upper electrode (UEL) temperature, forward RF power, reflected RF power, RF self-inductive DC bias, RF peak-to-peak voltage, chamber wall temperature, process gas flow rate, process gas partial pressure, chamber pressure, capacitor settings (i.e., C1 and C2 positions), focus ring thickness, RF hours, focus ring RF hours, and any statistics thereof. Alternatively, external sensors can include those not directly related to the functions of the plasma etching processing system 10, such as Figure 1 the optical detection device 34 shown for monitoring the light emitted from the plasma in the processing region 45.

[0042] The light detection device 34 may include detectors such as (silicon) photodiodes or photomultiplier tubes (PMTs) for measuring the total light intensity emitted from the plasma. The light detection device 34 may further include filters such as narrowband interference filters. In an alternative embodiment, the light detection device 34 may include a linear array CCD (charge-coupled device) or CID (charge injection device) array and a light dispersion device such as a grating or a prism. Additionally, the light detection device 34 may include a monochromator (e.g., a grating / detector system) for measuring light at a given wavelength, or a spectrometer (e.g., having a rotating grating or a fixed grating) for measuring a spectrum. The light detection device 34 may include a high-resolution OES sensor from a peak sensor system. Such an OES sensor has a broad spectrum spanning the ultraviolet (UV), visible (VIS), and near-infrared (NIR) spectra. In the peak sensor system, the resolution is approximately 1.4 angstroms, i.e., the sensor is capable of collecting 5550 wavelengths from 240 to 1000 nm. In the peak system sensor, the sensor is equipped with highly sensitive miniature fiber optic UV-VIS-NIR spectrometers, which are in turn integrated with a 2048 pixel linear CCD array.

[0043] In one embodiment of the present application, the spectrometer receives light transmitted through a single optical fiber and a fiber bundle, wherein the light output from the optical fiber is dispersed onto a linear array CCD array using a fixed grating. Similar to the above configuration, the light transmitted through the optical vacuum window is focused onto the input end of the optical fiber via a lens or a mirror. Different spectrometers each specifically tuned for a given spectral range (UV, VIS, and NIR) or a broadband spectrometer covering UV, VIS, and NIR form the sensors of the processing chamber. Each spectrometer includes an independent analog-to-digital (A / D) converter. Finally, depending on the use of the sensor, a complete emission spectrum may be recorded every 0.01 to 1.0 seconds or faster.

[0044] Alternatively, in an embodiment, the light detection device 34 may employ a spectrometer with all-reflective optics. Additionally, in an embodiment, a single spectrometer involving a single grating and a single detector may be used to detect the entire light wavelength range. The design and use of the optical emission spectroscopy hardware for obtaining optical OES data using, for example, the light detection device 34 are well known to those skilled in the art of optical plasma diagnostics.

[0045] The controller 55 includes a microprocessor, a memory, and digital I / O ports (possibly including D / A and / or A / D converters) that are capable of generating control voltages sufficient to transmit and activate inputs to the plasma etching processing system 10 and to monitor outputs from the plasma etching processing system 10. As Figure 1As shown, the controller 55 can be coupled to the RF generator 30, the impedance matching network 32, the gas injection system 40, the vacuum pumping system 58, the backside gas delivery system 26, the electrostatic chucking system 28, and the optical detection device 34, and exchange information with them. Programs stored in the memory are used to interact with the aforementioned components of the plasma etching processing system 10 according to the stored process instructions. An example of the controller 55 is the DELL PRECISION WORKSTATION 530 available from Dell Inc. in Austin, Texas TM The controller 55 can be locally located relative to the plasma etching processing system 10, or it can be remotely located relative to the plasma etching processing system 10. For example, the controller 55 can use at least one of a direct connection, an intranet, and the Internet to exchange data with the plasma etching processing system 10. The controller 55 can be coupled to an intranet, for example, at a customer site (i.e., equipment manufacturer, etc.), or the controller can be coupled to an intranet, for example, at a supplier site (i.e., equipment manufacturer). Additionally, for example, the controller 55 can be coupled to the Internet. Further, another computer (i.e., a controller, a server, etc.) can access the controller 55, for example, via at least one of a direct connection, an intranet, and the Internet to exchange data. As further described herein, the controller 55 also implements an algorithm for detecting the endpoint of an etching process performed in the plasma etching processing system 10 based on input data provided by the optical detection device 34

[0046] In plasma etching processes, endpoint detection (EPD) using optical emission spectroscopy is an important technique for controlling etch uniformity between wafers. Monitoring the time-varying trends generated according to one or two selected optical emission wavelengths reveals the endpoint to pause or stop the etching process. Multivariate data analysis using synthetic wavelengths helps improve the SNR and the robustness of the EPD. However, the synthetic wavelengths generated according to multivariate data analysis typically do not retain some of the inherent characteristics of natural wavelengths, such as having physical meaning

[0047] Grouping natural wavelengths using a multivariate model, a non-limiting example of which is PCA, can generate synthetic wavelengths such that the trends of the EPD are similar to those of the natural wavelengths, but with an endpoint signal having a higher SNR. In one example non-limiting embodiment of the present application, the grouping includes selecting natural wavelengths that exhibit constructive or destructive contributions, and using separate positive and negative weights of the wavelengths when grouping the wavelengths to transform the OES data into the PCA domain. However, other wavelength grouping methods can be used to generate synthetic OES data

[0048] The endpoint determination process according to an embodiment is carried out in two stages. In the first stage, a plasma etching process run is performed in the plasma processing chamber 15( Figure 2Step 110) in, and during one or more etching runs performed in the plasma etching processing system 10, OES data is acquired using the optical detection device 34 (step 120), such that a multivariate model of the acquired OES data can be established (step 130).

[0049] Once a multivariate model of the OES data is established, as long as the etching process run during the second phase is substantially similar to those used in the one or more etching runs performed in the first phase in terms of the structure being etched, the etching process conditions, the etching processing system used, etc., the multivariate model can be used for in-situ etching endpoint detection during the second phase (step 140). This is to ensure the effectiveness of the multivariate model.

[0050] In a non-limiting embodiment of endpoint determination ( Figure 3 as shown), wherein PCA analysis is used to group the natural wavelengths of the multivariate model (i.e., having positive and negative weights), the endpoint detection 200 starts from performing the etching process run and acquiring OES data using, for example, the optical detection device 34. During each plasma etching process run, n spectra are acquired ( Figure 3 step 210) in, where n is an integer greater than 1. The sampling interval between consecutive OES data acquisitions (i.e., spectrum acquisitions) may vary between 0.01 and 1.0 seconds, or faster. Each acquired OES data set (i.e., spectrum) contains m measured light intensities corresponding to m pixels of the CCD detector, and each pixel corresponds to a specific light wavelength projected onto the pixel by a diffraction grating, which is commonly used as a light dispersion device in the optical detection device 34. Depending on the required spectral resolution, the CCD detector can have 256 to 8192 pixels, but the most commonly used number of pixels is 2048 or 4096. For example, a two-dimensional detector with 4k×4k pixels can also be used.

[0051] Next, an OES data matrix [X] is established for all plasma etching process runs i = 1, 2,... k [i] (step 215). Each matrix [X] [i] is an n×m matrix, where the acquired spectra are arranged in the rows of the matrix such that the rows correspond to the n instants when the OES data is acquired, and the columns correspond to the number of pixels m. Subsequently, an n×m average OES data matrix [X] is optionally calculated by averaging each element of the acquired matrix [X] [i] within all plasma etching process runs i = 1, 2,... k, avg(Step 220). Optional OES spectral normalization can be performed before calculating the average value. When k = 1, there is only a single wafer OES measurement, and in this case, the average OES matrix is not calculated.

[0052] In one embodiment, the OES data matrix [X] [i] can be optionally normalized as follows. The OES data matrix [X] [i] is an n×m matrix with components x ij , where i = 1, 2, … n and each row corresponds to an OES snapshot at time t; and j = 1, 2 … m and each column corresponds to a trend at wavelength λ, so each column is a single wavelength trend. OES data normalization can be applied in two ways. In the first way, the method selects a reference snapshot S R = x R,j at time R (i.e., the Rth row), and then divides each OES data by this reference snapshot, x i,j = x i,j / x R,j . It can be a single-time snapshot or a snapshot averaged over a period of time. In the second way, the method selects a reference wavelength λ R (i.e., the Rth column), and then divides the intensity of each wavelength by the intensity of the reference wavelength, x i,j = x i,j / x i,R . Similarly, it can be a single wavelength or an average of wavelengths in a specific band. The inventors have found that normalization solves the intensity drift that occurs during OES operation between different wafers.

[0053] Subsequently, as detailed in the '990 patent, noise is filtered from the average OES data matrix [X] avg (step 225), the matrices [X] [i] and [X] avg are truncated (step 230) to remove the spectra acquired during plasma startup and optionally after the actual etching process endpoint, and the mean OES data matrix [S avg is calculated (step 235), where all elements of each column are set to the average of the elements of the entire column (i.e., all instants) of the average OES data matrix [X] avg , and this mean OES data matrix is subtracted from each acquired OES data matrix [X] [i] i = 1, 2, … k (step 240) to perform a de-averaging step before constructing a multivariate model of the acquired OES data, i.e., subtracting the average value.

[0054] Next, in a non - limiting example, the method PCA for determining the principal component weights [P] used in multivariate analysis (see step 242 between step 240 and step 245 in Figure 3 to transform the OES data) is described in the following steps. Other multivariate data analysis methods can also be used, such as the independent component analysis (ICA) method. PCA is an example of an unsupervised training method. As long as the target values of each or some of the OES spectra are available, other supervised methods can also be used, such as partial least squares (PLS), support vector machine (SVM) regression or classification methods. The target values can be obtained by xSEM, transmission electron microscopy (TEM), optical critical dimension (OCD) spectrometry, critical dimension scanning electron microscopy (CDSEM) or other tools.

[0055] During step 1, the average spectrum of [X] is subtracted from each row ( Figure 3 step 240 in

[0056] ), but optionally the data is not normalized using the standard deviation of [X]. 2 kj During step 2, the covariance matrix cov(λ)=[σ is calculated. The covariance matrix is m×m. For each column (each wavelength), the average value is calculated

[0057]

[0058] The covariance of the k - th row and j - th column is:

[0059]

[0060] During step 3, the eigenvectors and eigenvalues of the covariance matrix that satisfy the equation [covariance matrix]·[eigenvector]=[eigenvalue]·[eigenvector] are calculated. This is achieved by performing a singular value decomposition on the covariance matrix cov(λ):

[0061] P’cov(λ)P = L (3)

[0062] where L is a diagonal matrix of the eigenvalues of cov(λ), and P is a matrix of the eigenvectors of cov(λ). The eigenvalues are sorted in descending order so that the method can find the principal component weights in order of importance. For example, in a specific software, the first three (at most five) eigenvectors are used.

[0063] Then the de - averaged OES data [X] [i] -[S avgUsed as input for multivariate analysis (step 245), which multivariate analysis is, for example, PCA that transforms the OES data vector into the PCA domain using the derived weight vector P derived above.

[0064] The inventors have found that by grouping the principal component weights

[0065] Pj(λj) into two separate groups corresponding to positively weighted wavelengths and negatively weighted wavelengths, separate trends Tj are created, where T + j + T - j = Tj. Each T + j or T - j is a single positive trend. Thus, all conventional trend operations (such as snapshot normalization and taking the ratio between any of the trends in the trend) can be easily applied to T + j and T - j.

[0066] In one embodiment, the vector [P] is calculated and subsequently the positive vector [P + and the negative vector [P - are formed. For example, [P + is formed by setting all negative values in [P] to zero, while [P - is formed by setting all positive values in [P] to zero and then taking the absolute value (i.e., converting to a positive number).

[0067] In step 245, the mean - removed OES data [X] [i] - [S avg together with the determined vector [P] are used to derive the OES data transformed into the PCA domain

[0068] [T + = ([X] - [S avg )[P + , and [T - = ([X] - [S avg )[P - (4)

[0069] The method described herein generates synthetic wavelengths (corresponding to positively and negatively weighted natural wavelengths) to create a single signed trend (the transformed OES vector). For example, positive and negative synthetic wavelengths are created:

[0070] Λ + 1 = ∑ j＝1 n1 w + j S j Λ - 1 = ∑ k＝1n2 | w - k| S k (5)

[0071] where n1 is the number of positive weights, and n2 is the number of negative weights, and

[0072] T1 = ∑ i＝1 n w i S i = ∑ j＝1 n1 w + j S j - ∑ k＝1 n2 | w - k| S k = Λ + 1 - Λ - 1, (6)

[0073] where S i is the intensity of λ i at time t i and T1, Λ + 1 and Λ - 1 all vary with time.

[0074] The synthetic wavelength and the generated trend [T + = [Λ + and [T - = [Λ - have been determined, and the second stage of the endpoint detection method is performed by using a functional form of the time-evolution values of [T + and [T - . Since the trends [T + and [T - are already positive signals, they can be separated from each other to obtain an enhanced signal without any bias to shift the trends upward to all be positive and then apply these offsets to the new wafer in real time. For example, in one embodiment, the ratio T + 1(t) / T + 3(t) is calculated. However, in other embodiments, any other functional form can be calculated, such as the square of the synthetic wavelength ratio or just a single synthetic wavelength.

[0075] Since the goal of the first stage is to precompute useful multivariate model parameters for in-situ etch endpoint detection later, various parameters will be saved for later use. In step 250, the mean OES data matrix [Savg Save it to a volatile or non-volatile storage medium to facilitate the de-averaging of in-situ measured OES data. Also in this step, save the principal component (PC) weight vector [P] to a volatile or non-volatile storage medium to facilitate the rapid transformation of in-situ measured OES data into a transformed OES data vector [T].

[0076] In some cases, the inventors have found that shifting the calculated values of the elements of the transformed OES data vector [T] (i.e., the principal components) as they evolve over time so that they are centered around zero rather than growing to large positive or negative values is useful for endpoint detection reliability. This shift is done in step 255, where at each instant during the measurement in the etching process, estimate at least one element T of the transformed OES data vector [T] i , and find the minimum value min(T i ) of such one or more elements. For this purpose, time-evolving data or other data from the averaged OES data matrix [X i can be used. Then store this minimum value in a volatile or non-volatile storage medium in step 260 for later use in in-situ endpoint detection. Thus, the minimum value min(T avg ) of the elements T of the transformed OES data vector [T] can be used to shift the time-evolving values of the same element T of the transformed OES data vector [T] i , which are calculated from in-situ measured optical emission spectrometry (OES) data. i i avg

[0077] The data values stored on the volatile or non-volatile storage medium are now ready for the second stage, i.e., for in-situ etching endpoint detection.

[0078] Figure 4 FIG. 300 shows an exemplary flowchart of an in-situ endpoint detection process in a plasma etching processing system 100 equipped with a light detection device 34. This process can use the data saved in steps 250 and 260 of flowchart 200.

[0079] Figure 1 In steps 310 and 315, retrieve the previously determined mean OES data matrix [S avg and the principal component (PC) weight vector [P] from a volatile or non-volatile storage medium and load them into Figure 1 the memory of the controller 55 of the plasma etching processing system 10. The controller 55 will perform all the in-situ calculations required to determine the plasma process endpoint. Additionally, if the elements T of the transformed OES data vector [T] are to be usedi at least one minimum value min(T i ), it can be loaded from the volatile or non-volatile medium into the memory of the controller 55 in step 320.

[0080] In step 325, the substrate 25 is loaded into the plasma etching processing system 10, and a plasma is formed in the processing region 45.

[0081] In step 330, the optical detection device 34 is used to acquire time-evolving OES data in-situ (i.e., during the etching process).

[0082] In step 335, the retrieved mean OES data matrix [S avg elements are subtracted from each acquired OES data set (i.e., spectrum) to de-mean the acquired spectra before transformation using the developed multivariate model.

[0083] In step 340, using the developed PCA multivariate model, the de-meaned OES data is transformed into a transformed OES data vector [T] (i.e., principal components) using Equation 4 and the retrieved principal component (PC) weight vector [P]. This process is very fast because it only involves simple multiplication and is thus easy to calculate in-situ in real-time. The calculated element T i (e.g., the natural wavelengths Λ + i and Λ - i ) of the transformed OES data vector [T] over time can be used for endpoint detection (step 345).

[0084] In step 350, each time-evolving element T of the transformed OES data vector [T] can optionally be i differentiated to further facilitate endpoint detection using the slope data of the trend variable.

[0085] After calculating the time-evolving trend variable, the controller 55 of the plasma etching processing system 10 determines whether the endpoint has been reached (step 355). If the endpoint has been reached, the etching process ends at step 360, otherwise the etching process continues and each endpoint is continuously monitored via steps 330 to 355 of the flowchart 300.

[0086] Figure 5 shows the time evolution of the time derivative of the trend variable of the etching process. A deep and thus easily recognizable minimum experienced by the differentiated trend variable at the etching endpoint is seen. The bottom set of traces corresponds to using the trend Λ + 1(t) / Λ +The trend obtained for 3(t), where a synthetic wavelength is applied. Various traces corresponding to different wafers used in different etch runs are shown. Figure 5 Also shown is the time evolution of other types of trends (for different wafers), including the time evolution of the time derivative of a single wavelength trend at λ = 656 nm, and also the time evolution of the time derivative of the ratio of two single wavelength trends at λ = 656 nm and λ = 777 nm. As Figure 5 seen, an endpoint occurs near the 32 - second mark of the etch process.

[0087] Figure 6A Shows the trend Λ for different wafers in another etch process run + 1(t) / Λ + The time evolution of 3(t), while Figure 6B shows the trend Λ for different wafers in this other etch process run + 1(t) / Λ + The time evolution of the time derivative of 3(t).

[0088] Figure 7A Shows the time evolution of a single wavelength trend at λ = 656 nm for different wafers in another etch process run, while Figure 7B shows the time evolution of the time derivative of the single wavelength at λ = 656 nm for different wafers in this other etch process run.

[0089] Figure 8A Shows the time evolution of the time derivative of a single wavelength trend at λ = 260 nm for different wafers in another etch process run, while Figure 8B shows the trend obtained using Λ + 1(t) / Λ + The time evolution of the time derivative of the trend obtained for 3(t).

[0090] Figure 8C Shows the time evolution of the time derivative of the normalized synthetic wavelength trend discussed above. This plot refers only to the synthetic wavelength, not the ratio, since normalization has been applied beforehand.

[0091] After computing the time - evolution trend variables, the controller 55 of the plasma etch processing system 10 needs to determine in step 355 whether an endpoint has been reached. If an endpoint has indeed been reached, the etch process ends at step 360; otherwise, the etch process continues and the etch endpoint is continuously monitored via steps 330 - 355 of the flowchart 300.

[0092] In view of the foregoing teachings, many modifications and variations of this application are possible. Accordingly, it is to be understood that within the scope of the appended claims, this application may be practiced in a manner different from that specifically described herein.

Claims

1. A method for determining etch process endpoint data in a plasma processing system, the method comprising: Performing a plasma etch process run in a plasma processing chamber of an etch processing system; Obtaining optical emission spectroscopy (OES) data from the plasma processing chamber during one or more etch processes; Performing multivariate data analysis on the OES data to generate synthetic OES data based on the OES data by grouping wavelengths, wherein generating the synthetic OES data includes grouping wavelengths corresponding to positive and negative weights associated with natural wavelengths; And Using the synthetic OES data for later in-situ determination of the etch process endpoint.

2. The method according to claim 1, wherein Generating the synthetic OES data includes obtaining a transformed OES data vector [T], wherein, [T] = ([X] - [S avg )[P], where, [X] is an OES data matrix, [P] is a weight vector, and [S avg is an n×m mean OES data matrix, and each element of the mean OES data matrix is calculated as the average of the n elements of the corresponding column of the n×m average OES data matrix [X] avg . Each element of the average OES data matrix is calculated as the average of the corresponding element of the OES data matrix [X] within k etching process runs. n corresponds to the instant when the OES data is acquired, and m corresponds to the number of light intensities measured by the detector in the plasma processing chamber.

3. The method according to claim 2, wherein, The weight vector [P] is determined by: Calculating the eigenvectors and eigenvalues of the covariance matrix associated with the matrix [X]; Sorting the eigenvalues in descending order, the eigenvalues representing the weight vector [P]; And Set the positive weight vector [P] by setting all negative components in [P] to zero + , and set the negative weight vector [P] by setting all positive components in [P] to zero and taking the absolute value of it - .

4. The method according to claim 3, further comprising: Obtain the transformed OES data vector [T+] or [T - , where, [T + = ([X] - [S avg )[P + , [T - = ([X] - [S avg )[P - .

5. The method according to claim 4, further comprising: Select a functional form that involves elements of the transformed OES data vector [T + or [T - , and compute the time evolution of the selected functional form.

6. The method according to claim 5, further comprising: Calculating the time derivative of a selected functional form and calculating the time evolution of the time derivative of the selected functional form.

7. The method according to claim 6, wherein, The function form includes + , - , + / - , and the ratio [T + / [T - to the power of, or the transformed OES data vector [T + or [T - of a single element, or using [T + and / or [T - in any mathematical form.

8. The method according to claim 1, further comprising Perform k plasma etching process runs in the plasma processing chamber, where, k is an integer greater than zero, and each of the k plasma etch process runs includes: Loading a substrate to be processed into the plasma processing chamber, the plasma processing chamber including a spectrometer having a detector including m pixels, each pixel corresponding to a different optical wavelength; Forming a plasma in the plasma etch processing chamber; and Obtaining OES data from the plasma processing chamber during one or more etch processes and forming an OES data matrix [X] for each of the k plasma etch process runs.

9. The method according to claim 8, further comprising Calculate the n×m average OES data matrix [X] avg , where, Each element is calculated as the average of the corresponding element of the OES matrix [X] within the k etch process runs; Filter noise from the average OES data matrix [X] avg ; Truncate each OES data matrix [X] and [X] avg , wherein data acquired during plasma startup and during times beyond the etch process endpoint is discarded; Calculate the n×m mean OES data matrix [S avg , where each element is calculated as the average of the n elements of the corresponding column of [X] avg ; For each k, subtract [S avg from the matrix [X] to demean the OES data.

10. The method according to claim 2, wherein, After forming the OES data matrix [X] for each of these plasma etching processes during operation, and before calculating the n×m average OES data matrix [X] avg the method normalizes the OES data matrix [X].

11. The method according to claim 10, wherein, This OES data matrix normalization involves selecting a reference snapshot x at time R R,j , and then dividing each OES data by this reference snapshot, x i,j = x i,j / x R,j .

12. The method according to claim 11, wherein, The reference snapshot is a single moment snapshot or a snapshot averaged over a period of time.

13. The method according to claim 10, wherein, The OES data matrix normalization includes selecting a reference wavelength λ R , and then dividing each OES data by the intensity at that reference wavelength, x i,j = x i,j / x i,R .

14. The method according to claim 13, wherein, The reference wavelength is a single wavelength or the average of a band of wavelengths.

15. A method for determining etch process endpoint data in a plasma processing system, the method comprising: Performing a plasma etch process run in a plasma processing chamber of an etch processing system; Obtaining optical emission spectroscopy (OES) data from the plasma processing chamber during one or more etch processes; Perform multivariate data analysis on the OES data to generate synthetic OES data from the OES data by grouping wavelengths corresponding to positive and negative weights associated with natural wavelengths; and Use the synthetic OES data for later in-situ determination of the endpoint of the etching process.

16. The method according to claim 15, wherein, The multivariate data analysis is performed using independent component analysis.

17. The method according to claim 15, wherein, The multivariate data analysis is performed using a supervised multivariate data analysis method, the supervised multivariate data analysis method including support vector machine regression.

18. The method according to claim 15, further comprising: Obtain a transformed OES data vector [T+] or [T - , wherein, [T + = ([X] - [S avg )[P + , [T - = ([X] - [S avg )[P - , where, [X] is an OES data matrix, [P + is a positive weight vector, [P - is a negative weight vector, and [S avg is an n×m mean OES data matrix, each element of which is calculated as the average of the n elements of the corresponding column of the n×m average OES data matrix [X] avg where each element of the average OES data matrix is calculated as the average of the corresponding elements of the OES data matrix [X] over k etch process runs, n corresponds to the instant at which the OES data is acquired, and m corresponds to the number of light intensities measured by the detector in the plasma processing chamber.

19. The method according to claim 18, further comprising: Select a functional form for elements of the transformed OES data vector [T + or [T - , and compute the time evolution of the selected functional form.

20. The method according to claim 19, further comprising: Calculate the time derivative of a selected functional form and calculate the time evolution of the time derivative of the selected functional form.

21. The method according to claim 20, wherein, The function form includes + , - , + / - , this ratio [T + / [T - power, or the transformed OES data vector [T + or a single element of [T - , or any mathematical form using [T + and / or [T - .

Images