Method for rapidly determining photoetching process conditions according to photoresist topography
By performing multiple exposures on the wafer and using machine learning models to train the lithography process conditions, the problems of long cycles and high costs caused by lithography engineers' blind guessing have been solved. This has enabled the rapid determination of photoresist morphology and the selection of optimal process parameters, thereby improving R&D and mass production efficiency.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHANGHAI HUALI INTEGRATED CIRCUIT CORP
- Filing Date
- 2023-04-19
- Publication Date
- 2026-05-19
AI Technical Summary
In existing technologies, lithography engineers rely on experience to guess the morphology of the photoresist, which requires repeated iterations and adjustments. This process is time-consuming, labor-intensive, and external factors affecting the morphology of the lithography machine cause difficulties in research and development and mass production.
By performing multiple exposures on a wafer to obtain photoresist morphology images, a machine learning model is used to train the photolithography process conditions, calculate the morphology deviation value, and select the optimal process parameters to quickly determine the photolithography process conditions.
It reduces reliance on human judgment, improves the efficiency of quantitative description of photoresist morphology, saves time and wafer costs, and significantly improves R&D and mass production efficiency.
Smart Images
Figure CN116610005B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of semiconductor technology, and in particular to a method for rapidly determining photolithography process conditions based on photoresist morphology. Background Technology
[0002] In the photolithography process, the morphology of the photoresist after development directly affects the quality of the pattern on the silicon wafer, and thus the performance of the final device. As pattern sizes continue to shrink, maintaining a good morphology becomes particularly important. In even smaller nodes of specialty processes (CIS, CMOS image sensors), the application of thicker photoresist also places higher demands on the morphology of the photoresist.
[0003] Some processes in CIS devices employ a double-exposure process. Currently, lithography engineers need to adjust the exposure dose and focal length (Dose1 / Focus1, Dose2 / Focus2) to modify the photoresist morphology. These four variables drastically increase the difficulty of process development and mass production.
[0004] The current method involves engineers guessing based on experience, confirming the morphology through ADI (Advanced Difference Inspection) measurements, and requiring repeated feedback and iterations (n times) for morphology correction. This process is time-consuming, labor-intensive, and wastes a significant amount of wafers. In addition to increasing R&D costs, during mass production, external factors such as changes in photoresist batches and lithography machine PM (re-run) can cause morphology variations, requiring readjustment.
[0005] To address the aforementioned issues, a novel method is needed to rapidly determine photolithography process conditions based on photoresist morphology. Summary of the Invention
[0006] In view of the shortcomings of the prior art described above, the purpose of this invention is to provide a method for rapidly determining photolithography process conditions based on photoresist morphology. This method addresses the shortcomings of existing technologies where engineers rely on experience and guesswork, and morphology is confirmed through ADI (Advanced Difference-Inspection) measurements. This requires repeated feedback and iterations (n times) for morphology correction, resulting in long cycles, high labor costs, and significant wafer waste. Besides increasing R&D costs, during mass production, external factors such as photoresist batch changes and lithography machine PM (rework) can cause morphology variations, necessitating readjustment.
[0007] To achieve the above and other related objectives, the present invention provides a method for rapidly determining photolithography process conditions based on photoresist morphology, comprising:
[0008] Step 1: Provide at least one wafer, form a photoresist layer on the wafer, and perform multiple exposures in different exposure areas using different photolithography process conditions, including exposure dose and focal length; then obtain a photoresist morphology image on each exposure area.
[0009] Step 2: Obtain the photoresist outline based on the photoresist morphology image, cut the photoresist outline to form multiple cutting segments, and then calculate the morphology deviation value of each cutting segment. The morphology deviation value is the difference between the cutting segment and the target pattern.
[0010] Step 3: Train the model using the parameters of the photolithography process conditions and the corresponding morphology deviation values to obtain a machine learning model;
[0011] Step 4: Randomly generate multiple sets of test parameters for the photolithography process conditions, and use the machine learning model to obtain the morphology deviation value for each set of test parameters;
[0012] Step 5: Select the test parameters corresponding to the morphological deviation values that meet the judgment values as the final photolithography process conditions.
[0013] Preferably, the number of times the exposure is performed in different exposure areas under different photolithography conditions in step one is two.
[0014] Preferably, in step one, a scanning electron microscope is used to obtain an image of the photoresist morphology.
[0015] Preferably, the photoresist morphology images on each of the exposure areas in step one have the same specifications.
[0016] Preferably, in step one, the field of view, pixel ratio, and high voltage of the photoresist morphology image on each of the exposure areas are consistent.
[0017] Preferably, in step two, before cutting the photoresist contour to form multiple cutting segments, the photoresist contour is further smoothed.
[0018] Preferably, the method for training the model to obtain the machine learning model in step three includes fully connected neural networks and random forests.
[0019] Preferably, the model training in step three further includes model testing: using the parameters of the lithography process conditions as the input values of the machine learning model, using the morphology deviation value as the output value of the machine learning model, and using the residual standard deviation between the output value and the true value of the machine learning model to define the accuracy of the machine learning model, wherein the true value is the measurement data of the parameters of the lithography process conditions in the actual process.
[0020] Preferably, in step five, the standard deviation is used to define the dispersion of the output value, and the judgment value includes the set residual standard deviation and the standard deviation.
[0021] Preferably, step five further includes exposure verification on the test wafer using the parameters of the final photolithography process conditions.
[0022] As described above, the method for rapidly determining photolithography process conditions based on photoresist morphology of the present invention has the following beneficial effects:
[0023] This invention uses a machine learning-trained model to predict the optimal lithography process. Compared with the traditional method where lithography engineers rely on experience to judge lithography process parameters, this invention eliminates the need for human judgment. By generating photoresist morphology deviation (PPE), it can quantitatively describe the photoresist morphology. Furthermore, by predicting the optimal process conditions through the model, it saves time, manpower, and wafer costs, and significantly improves R&D and mass production efficiency. Attached Figure Description
[0024] Figure 1 The diagram shown is a schematic representation of the process flow of the present invention.
[0025] Figure 2 The diagram shows different photolithography process conditions used for different shots in this invention.
[0026] Figure 3 The diagram shown is a top view of the structure of the photoresist after exposure according to the present invention.
[0027] Figure 4 The diagram shows a cross-sectional structure of the photoresist after exposure according to the present invention.
[0028] Figure 5 The diagram shows the outline of the photoresist acquisition method according to the present invention.
[0029] Figure 6 The diagram shown illustrates the calculation of morphological deviation values according to the present invention. Detailed Implementation
[0030] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention.
[0031] Please see Figure 1 This invention provides a method for rapidly determining photolithography process conditions based on photoresist morphology, comprising:
[0032] Step 1, please refer to Figure 2The process involves providing at least one wafer, forming a photoresist layer on the wafer, and performing multiple exposures in different exposure areas (shots) using different photolithography conditions, including exposure dose and focus. Subsequently, an image of the photoresist morphology on each exposure area is acquired (e.g., ...). Figure 3 , 4 (as shown);
[0033] In an embodiment of the present invention, the number of exposures performed in step one on different exposure areas using different photolithography process conditions is twice. That is, different photolithography process conditions are used for different exposure areas: Dose1, Focus1, Dose2, Focus2 (first exposure energy / focus: Dose1 / Focus1; second exposure energy / focus: Dose2 / Focus2). It should be noted that the number of exposures can also be greater than 2, and is not specifically limited here.
[0034] In the embodiments of the present invention, a scanning electron microscope is used to obtain a photoresist morphology image in step one. It should be noted that other methods known to those skilled in the art can also be used to obtain the photoresist morphology image.
[0035] In an embodiment of the present invention, the photoresist morphology images on each exposure area in step one are of the same specifications, so that the photoresist outlines obtained later are of the same specifications.
[0036] In an embodiment of the present invention, the field of view, pixel ratio, and high voltage of the photoresist morphology image on each exposure area in step one are consistent.
[0037] Step Two, please refer to Figure 5 To obtain the photoresist outline from the photoresist morphology image, please refer to [link / reference]. Figure 6 The photoresist outline is cut to form multiple fragments. Then, the profile error (PPE) value of each fragment is calculated. The profile error value is the difference between the fragment and the target pattern.
[0038] In an embodiment of the present invention, step two, before cutting the photoresist contour to form multiple cutting segments, further includes smoothing the photoresist contour.
[0039] Step 3: Train the model using the parameters of the photolithography process and their corresponding morphological deviation values to obtain the machine learning model;
[0040] In embodiments of the present invention, the method for training the machine learning model in step three includes fully connected neural networks and random forests. A fully connected neural network model is a type of multilayer perceptron (MLP). The principle of a perceptron is to find the most reasonable and robust hyperplane between categories; the most representative perceptron is the Support Vector Machine (SVM) algorithm. Neural networks draw inspiration from both perceptrons and bionics. Typically, when an animal's nervous system receives a signal, it sends signals to its neurons. Each neuron, upon receiving input, makes its own judgment, activates, and generates output signals, which are then aggregated to achieve the identification and classification of the information source. In machine learning, a random forest is a classifier containing multiple decision trees, and its output category is determined by the mode of the categories output by individual trees. It should be noted that other methods can also be used to train the machine learning model; no specific limitations are specified here.
[0041] In an embodiment of the present invention, model training in step three further includes model testing: using the parameters of the photolithography process conditions as the input values of the machine learning model, for example, 75% of the data can be randomly selected for model training, and 25% of the data can be used for model testing. The morphology deviation value is used as the output value of the machine learning model. The accuracy of the machine learning model is defined by the residual standard deviation (RMSE) between the output value and the true value, where the true value is the measurement data of the parameters of the photolithography process conditions in the actual process. The statistical symbol for residual standard deviation is RMSE (Root mean squared error), also known as root mean square error, standard error, or the standard deviation of a regression system. It is related to other statistical indicators such as standard deviation and is used to reflect the degree of data discretization. It is mainly used in distributed regression analysis as a standard for introducing and removing variables. RMSE is the square root of the ratio of the sum of the squares of the deviations of all observed values from the true value to the number of observations n.
[0042] Step 4: Randomly generate multiple sets of test parameters for photolithography process conditions, and use a machine learning model to obtain the morphological deviation value of each set of test parameters;
[0043] Step 5: Select the test parameters corresponding to the morphological deviation values that meet the judgment values as the final photolithography process conditions.
[0044] In an embodiment of the invention, step five utilizes the standard deviation (MES) to define the dispersion of the output values. The judgment values include the set residual standard deviation and the standard deviation (or a set number of predictions, after which the minimum value is retrieved). Standard deviation, a mathematical term, is the square root of the arithmetic mean (i.e., variance) of the squared deviations from the mean, denoted by σ. Standard deviation is also called the standard deviation or experimental standard deviation, and is most commonly used in probability and statistics as a measure of the degree of statistical distribution. Standard deviation is the square root of the variance. Standard deviation reflects the dispersion of a dataset. Two sets of data with the same mean may not have the same standard deviation.
[0045] In an embodiment of the present invention, step five further includes exposure verification on a test wafer using the parameters of the final photolithography process conditions.
[0046] It should be noted that the illustrations provided in this embodiment are only schematic representations of the basic concept of the present invention. Therefore, the drawings only show the components related to the present invention and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.
[0047] In summary, this invention uses a machine learning-trained model to predict the optimal lithography process. Compared to traditional lithography engineers who rely on experience to judge process parameters, this invention eliminates the need for human judgment. By generating photoresist morphology deviations (PPE), it can quantitatively describe the photoresist morphology. Furthermore, by predicting optimal process conditions through a model, it saves time, manpower, and wafer costs, significantly improving R&D and mass production efficiency. Therefore, this invention effectively overcomes the various shortcomings of existing technologies and has high industrial application value.
[0048] The above embodiments are merely illustrative of the principles and effects of the present invention and are not intended to limit the invention. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in the present invention should still be covered by the claims of the present invention.
Claims
1. A method for rapidly determining photolithography process conditions based on photoresist morphology, characterized in that, At least including: Step 1: Provide at least one wafer, form a photoresist layer on the wafer, and perform multiple exposures in different exposure areas using different photolithography process conditions, including exposure dose and focal length; then obtain a photoresist morphology image on each exposure area. Step 2: Obtain the photoresist outline based on the photoresist morphology image, cut the photoresist outline to form multiple cutting segments, and then calculate the morphology deviation value of each cutting segment. The morphology deviation value is the difference between the cutting segment and the target pattern. Step 3: Train the model using the parameters of the photolithography process conditions and the corresponding morphology deviation values to obtain a machine learning model; Step 4: Randomly generate multiple sets of test parameters for the photolithography process conditions, and use the machine learning model to obtain the morphology deviation value for each set of test parameters; Step 5: Select the test parameters corresponding to the morphological deviation values that meet the judgment values as the final photolithography process conditions.
2. The method for rapidly determining photolithography process conditions based on photoresist morphology according to claim 1, characterized in that: The number of times the exposure is performed in step one, using different photolithography conditions in different exposure areas, is two.
3. The method for rapidly determining photolithography process conditions based on photoresist morphology according to claim 1, characterized in that: In step one, a scanning electron microscope is used to obtain an image of the morphology of the photoresist.
4. The method for rapidly determining photolithography process conditions based on photoresist morphology according to claim 1, characterized in that: In step one, the photoresist morphology images on each of the exposure areas have the same specifications.
5. The method for rapidly determining photolithography process conditions based on photoresist morphology according to claim 4, characterized in that: In step one, the field of view, pixel ratio, and high voltage of the photoresist morphology image on each of the exposure areas are consistent.
6. The method for rapidly determining photolithography process conditions based on photoresist morphology according to claim 1, characterized in that: Step two, before cutting the photoresist contour to form multiple cutting segments, also includes smoothing the photoresist contour.
7. The method for rapidly determining photolithography process conditions based on photoresist morphology according to claim 1, characterized in that: The methods for training the model to obtain the machine learning model described in step three include fully connected neural networks or random forests.
8. The method for rapidly determining photolithography process conditions based on photoresist morphology according to claim 1, characterized in that: The model training in step three also includes model testing: using the parameters of the lithography process conditions as the input values of the machine learning model, using the morphology deviation value as the output value of the machine learning model, and using the residual standard deviation between the output value and the true value of the machine learning model to define the accuracy of the machine learning model, wherein the true value is the measurement data of the parameters of the lithography process conditions in the actual process.
9. The method for rapidly determining photolithography process conditions based on photoresist morphology according to claim 8, characterized in that: In step five, the standard deviation is used to define the dispersion of the output value, and the judgment value includes the set residual standard deviation and the standard deviation.
10. The method for rapidly determining photolithography process conditions based on photoresist morphology according to claim 1, characterized in that: Step five also includes exposure verification on the test wafer using the parameters of the final photolithography process conditions.