DETAILED ACTION
Second Non-final Rejection
The Status of the Claims
Claims 1-14 and 16-21 are pending for examination.
Claims 1, 8 and 16 are independent Claims.
Claims 1-14 and 16-21 are rejected under 35 U.S.C. §103.
Claims 1-14 and 16-21 are rejected under 35 U.S.C. §112(b).
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 1-14 and 16-21 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
The Claims recite the limitation "the " in the “applying a domain decomposition electromagnetic simulation …” (in Claim 1 and 16) or “apply a quasi-rigorous electromagnetic simulation …” (in Claim 8) limitation(s). There is insufficient antecedent basis for this limitation in the claim.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim(s) 1-9, 13-14, 16 and 21 is/are rejected under 35 U.S.C. 103 as being unpatentable over Ye et al. (U.S. 20070061773 hereinafter Ye) in view of Adam (U.S. 2004/0122636 hereinafter Adam) in further view of Li et al. ("Predicting Scattering From Complex Nano-Structures via Deep Learning).
As Claim 1, Ye teaches a method comprising:
accessing a description of a lithographic mask design (Ye (¶0036 line 5-7, fig. 2A step 212), mask is inspected to generate mask inspection data);
simulating a remainder of a lithography process based on the improved prediction of the output field (Ye (¶0037), system tunes the lithographic process based on the simulated pattern); and
modifying the lithographic mask design based on the simulation of the lithography process (Ye (¶0036, fig. 2A step 226), “the mask is repaired or reworked based on the simulated patterns produced using the individual mask error model”. “The mask” is construed as a “lithographic mask design” because the (physical) mask is a blueprint or design for the masked integrated circuitry. Modifying the mask will change the integrated circuit).
Ye may not explicitly disclose:
applying a domain decomposition electromagnetic simulation of Maxwell's equations to the lithographic mask design to produce an approximate prediction of an output field resulting from the lithographic mask;
Adam teaches:
applying a domain decomposition electromagnetic simulation (Adam (¶0052 last 3 lines), domain decomposition method) of Maxwell's equations (Adam (¶0070, ¶0076 line 11-17, ¶0078 line 1-4), solving Maxwell’s equations is not practical while decomposition method provides a good approximation) to the lithographic mask design to produce an approximate prediction of an output field resulting from the lithographic mask (Adam (¶0070 line 1-7), a mask simulation is effectively provided by the domain decomposition method); [[and]]
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify prediction of Ye instead be a diffraction modulation taught by Adam, with a reasonable expectation of success. The motivation would be to provide a solution for “solving the problem of the rapid and accurate evaluation of the benchmark signal from non-defective masks for the inspection system” (Adam (¶0078 line 1-4)).
Ye in view of Adam does not explicitly disclose:
applying, by a processor, the approximate prediction of the output field as input to a machine learning model to produce an improved prediction of the output field, wherein the machine learning model accounts for higher order effects that are approximated by the domain decomposition electromagnetic simulation of Maxwell's equations;
Li teaches:
applying, by a processor (Li ((figure 3), architecture of the deep learning framework and input-output scheme), the approximate prediction of the output field (Li (II. Methodology and Formulation; A. Finite Difference Frequency Domain (FDFD); last 4 lines), “then we apply a direct solver or iterative solver [37], [38] to obtain the EM field solutions for the unknown h. It is worth noting that the DL framework is employed to approximate the matrix system for every input”. Maxwell equation is approximated by FDFD method. Deep learning framework is trained to approximate the result for each Maxwell equation input) as input to a machine learning model to produce an improved prediction of the output field (Li (II. Methodology and Formulation; B. 2D Dataset Generation); last 4 lines 23-29), “The overall dataset contains 36,000 sets of data, with 6,000 arbitrary samples for each category of different complex shapes, in which 90 percent of total samples serve as the training set and the rest are used to test the predicted results after training. It is worth mentioning that all sets of data is obtained from FDFD strictly solving process under certain illumination settings”, FDFD solver (Maxwell equation) provides data set for Deep Learning frame work. Li (II. Methodology and Formulation; C. Deep Learning Framework); column 2, last 5 lines), “it is an iterative process to adjust the aforementioned hyperparameters of our model in order to train the network properly, such as steps of the down sampling, the number of layers, the number of kernels, etc”), wherein the machine learning model accounts for higher order effects that are approximated by the domain decomposition electromagnetic simulation of Maxwell's equations (Li (III. Numerical Results, A. Computational Acceleration), “To this end, the DL framework utilizes the parallel computing ability of GPUs, which makes it possible to simultaneously predict large batches of inputs. As a result, 100 testing samples take only 1.76 s to generate their corresponding EM field distributions, which is three orders of magnitude faster than the 3626.2 s required by the FDFD solver”);
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify prediction of Ye in view of Adam instead be a machine learning taught by Li, with a reasonable expectation of success. The motivation would be to allow “enables faster predictions of the EM fields for random scatter configurations than traditional full-wave solvers.” (Li (III. Numerical Results; A. Computational Acceleration; line 1-4)).
As Claim 2, besides Claim 1, Ye in view of Adam in further view of Li teaches wherein the input applied to the machine learning model comprises three-dimensional data, in which two of the dimensions represent spatial dimensions of the lithographic mask (Ye (¶0036 line 5-7, fig. 2A step 212), mask is inspected to generate mask inspection data. The mask is 2-dimensional mask) and the third dimension represents polarization components for the output field (Adam (¶0071 line 4-9), figure 11 shows evaluation for TE and TM polarization).
As Claim 3, besides Claim 1, Ye in view of Adam in further view of Li teaches wherein the approximate prediction produced by the domain decomposition electromagnetic simulation of Maxwell’s equations comprises coupling between individual components of the output field (Adam (¶0071 line 4-9), figure 11 shows evaluation for TE and TM polarization), and the machine learning model improves the prediction of the coupling (Ye (¶0037), system tunes the lithographic process based on the simulated pattern).
As Claim 4, besides Claim 1, Ye in view of Adam in further view of Li teaches wherein the approximate prediction produced by the domain decomposition electromagnetic simulation of Maxwell’s equations comprises higher diffraction orders in k-space (Adam (¶0071 line 4-9), figure 11 shows evaluation for TE and TM polarization), and the machine learning model improves the prediction of the higher diffraction orders (Ye (¶0037), system tunes the lithographic process based on the simulated pattern).
As Claim 5, besides Claim 1, Ye in view of Adam in further view of Li teaches further comprising:
partitioning the lithographic mask design into a plurality of tiles (Adam (¶0021 line 1-5, ¶0059 line 28-30), domain is decomposed into edges. Scatter field end-point is ended in a perfect square shape);
applying the domain decomposition electromagnetic simulation and machine learning model to the tiles to produce improved predictions for the tiles (Adam (¶0021 line 1-5), electromagnetic field from the diffraction is calculated); and
combining the improved predictions for the plurality of tiles to produce the improved prediction for the lithographic mask (Adam (¶0021 line 5-8), sum of the complex fields is taken).
As Claim 6, besides Claim 1, Ye in view of Adam in further view of Li teaches wherein the lithographic mask design is for an entire chip (Ye (¶0004 line 1-5, ¶0006 bottom portion, full chip modelling), photo mask is for chip production).
As Claim 7, besides Claim 1, Ye in view of Adam in further view of Li teaches further comprising:
partitioning a source illumination into multiple components (Adam (¶0071 line 4-9), figure 11 shows evaluation for TE and TM polarization);
for each component, applying the domain decomposition electromagnetic simulation and machine learning model to produce improved prediction for that component (Li (II. Methodology and Formulation; B. 2D Dataset Generation); last 4 lines 23-29), “The overall dataset contains 36,000 sets of data, with 6,000 arbitrary samples for each category of different complex shapes, in which 90 percent of total samples serve as the training set and the rest are used to test the predicted results after training. It is worth mentioning that all sets of data is obtained from FDFD strictly solving process under certain illumination settings”, FDFD solver (Maxwell equation) provides data set for Deep Learning frame work. Li (II. Methodology and Formulation; C. Deep Learning Framework); column 2, last 5 lines), “it is an iterative process to adjust the aforementioned hyperparameters of our model in order to train the network properly, such as steps of the down sampling, the number of layers, the number of kernels, etc”), wherein different machine learning models are used for different components (Ye (¶0066 last 7 lines), different models are used for different exposure tools); and
combining the improved predictions for the multiple components to produce the improved prediction for the lithographic mask (Li (III. Numerical Results; B. Accuracy; line 1-2), “The proposed framework exhibits an excellent acceleration rate with uncompromising accuracy.”).
As Claim 8, Ye teaches a system comprising a memory storing instructions; and a processor (Ye (¶0055 line 3-4), computing platform), coupled with the memory and to execute the instructions, the instructions when executed cause the processor (Ye (¶0055 line 3-4), software programs) to:
The rest of the limitation(s) are rejected for the same reasons as Claim 1.
As Claim 9, besides Claim 1, Ye in view of Adam in further view of Li teaches wherein the instructions further cause the processor to:
balance the input applied to the machine learning model and/or scale the input applied to the machine learning model (Ye (¶0039, fig. 2A step 252), system balances the input by going through a loop).
As Claim 13, besides Claim 8, Ye in view of Adam in further view of Li teaches wherein the lithographic mask contains features that are smaller than a wavelength of an illuminating source (Ye (¶0036 line 4), OPC process produces features smaller than the light wavelength).
As Claim 14, besides Claim 8, Ye in view of Sezginer in further view of Li teaches wherein source illumination for the lithographic mask is an extreme ultraviolet (EUV) or deep ultraviolet (DUV) illumination (Adam (¶0004 line 12), Deep Ultra Violet).
As Claim 16, Ye teaches A non-transitory computer readable medium comprising stored instructions, which when executed by a processor (Ye (¶0055 line 3-4), computing platform and software program) cause the processor to:
The rest of limitation(s) are rejected for the same reasons as Claim 1.
As Claim 21, besides Claim 1, Ye in view of Adam in further view of Jin teaches wherein the description of the lithographic mask design includes a description of a layout and of a stack material of the lithographic mask (Ye (¶0041, fig. 2 step 276, step 278), description of the mask design include simulated pattern (description of a layout) and ideal simulated patterns (a stack material) as simulated by shinning the light through the ideal post-OPC material).
Claim(s) 10-12 and 17-20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Ye and Adam in view of Li in further view of Wantanabe et al. (U.S. 2018/0121592 hereinafter Watanabe).
As Claim 10, besides Claim 1, Ye in view of Adam in further view of Li does not explicitly disclose:
teaches wherein the machine learning model comprises a residual- learning type layer.
Watanabe teaches:
teaches wherein the machine learning model comprises a residual- learning type layer (Watanabe (¶0073, fig. 8), initial condition includes plurality of layers).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify tuning module of Ye in view of Adam in further view of Li instead be a neural network of Watanabe, with a reasonable expectation of success. The motivation would be to allow “the post-patterning pattern can be simulated directly based on the mask” (Watanabe (¶0121 line 2-4)).
As Claim 11, besides Claim 10, Ye and Adam in view of Li in further view of Watanabe teaches wherein the machine learning model further comprises an auto-encoder or GAN type model (Watanabe (¶0073, fig. 8), each layer includes a number of filters).
As Claim 12, besides Claim 10, Ye in view of Adam in further view of Li does not explicitly disclose:
wherein the machine learning model comprises at least 20 layers.
Watanabe teaches:
wherein the machine learning model comprises at least 20 layers (Watanabe (¶0073, fig. 8), number of layers is any number from 10-100).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify tuning module of Ye in view of Adam in further view of Li instead be a neural network of Watanabe, with a reasonable expectation of success. The motivation would be to allow “the post-patterning pattern can be simulated directly based on the mask” (Watanabe (¶0121 line 2-4)).
As Claim 17, besides Claim 16, Ye in view of Adam in further view of Li teaches:
wherein the machine learning model has been trained using a training set of training tiles (Ye (¶0037), system tunes the lithographic process based on the simulated pattern), produced by a fully rigorous Maxwell solver for the individual training tiles (Adam (¶0070 line 1-4), rigorous mask simulation that was performed with TEMPEST).
Ye in view of Adam in further view of Li does not explicitly disclose:
and ground-truth for the training is based on output fields.
Watanabe teaches:
and ground-truth for the training is based on output fields (Watanabe (¶0077, ¶0078), calculated value and actual value are compared for reducing error).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify tuning module of Ye in view of Adam in further view of Li instead be a neural network of Watanabe, with a reasonable expectation of success. The motivation would be to allow “the post-patterning pattern can be simulated directly based on the mask” (Watanabe (¶0121 line 2-4)).
As Claim 18, besides Claim 17, Ye and Adam in view of Li in further view of Watanabe teaches wherein the training set contains not more than 1000 different training tiles (Watanabe (¶0080), learning data includes 100-1000 sets).
As Claim 19, besides Claim 17, Ye and Adam in view of Li in further view of Watanabe teaches wherein the training set includes training tiles with known symmetry (Adam (¶0070 line 8-11), diffraction results are calculated only once and recycles as many times as necessary) and training of the machine learning model enforces the known symmetry (Adam (¶0070 line 8-11), diffraction results are calculated only once and recycles as many times as necessary).
As Claim 20, besides Claim 17, Ye and Adam in view of Li in further view of Watanabe teaches wherein the training is further based on a loss function comparing images predicted (Wantabe (¶0077, ¶0078), calculated value and actual value are compared for reducing error) by (a) the fully rigorous Maxwell solver (Adam (¶0070 line 1-4), rigorous mask simulation that was performed with TEMPEST) and (b) the domain decomposition electromagnetic simulation of Maxwell’s equations (Adam (¶0070 line 1-7), a mask simulation is effectively provided by the domain decomposition method) and the machine learning model (Li (III. Numerical Results; B. Accuracy; line 1-2), “The proposed framework exhibits an excellent acceleration rate with uncompromising accuracy.”).
Response to Arguments
Rejections under 35 U.S.C. §112(b):
Applicants’ arguments are persuasive; therefore, current 35 U.S.C. §112(b) rejections are respectfully withdrawn for Claim 21.
Rejections under 35 U.S.C. §101:
Applicants’ arguments are persuasive; therefore, current 35 U.S.C. §101 rejections are respectfully withdrawn.
Rejections under 35 U.S.C. §103:
Ye does not disclose “modify a lithographic mask design” because Ye does not concern with the design of the lithographic mask design (first paragraph of page 8 in the appeal brief).
PNG
media_image1.png
213
640
media_image1.png
Greyscale
Applicants’ arguments are not persuasive because each lithographic mask is a mask design of its own. Ye discloses a lithographic mask or a lithographic mask design. Ye (¶0036, fig. 2A step 226) teaches “the mask is repaired or reworked based on the simulated patterns produced using the individual mask error model”. “The mask” is construed as a “lithographic mask design” because the (physical) mask is a blueprint or design for the masked integrated circuitry. Modifying the mask will change the integrated circuit.
As Claim 1, Applicants argue that Ye does not teaches “the approximate prediction of the output field as input” and from that “produces an improved prediction of the output field” (second paragraph of page 9 in the remarks).
PNG
media_image2.png
183
627
media_image2.png
Greyscale
Applicants’ arguments are moot because new reference Li teaches the limitation(s).
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to NHAT HUY T NGUYEN whose telephone number is (571)270-7333. The examiner can normally be reached M-F: 12:00-8:00 EST.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Viker Lamardo can be reached at 571-270-5871. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
In view of the appeal brief filed on 8/28/2025, PROSECUTION IS HEREBY REOPENED. A new ground of rejection is set forth below.
To avoid abandonment of the application, appellant must exercise one of the following two options:
(1) file a reply under 37 CFR 1.111 (if this Office action is non-final) or a reply under 37 CFR 1.113 (if this Office action is final); or,
(2) initiate a new appeal by filing a notice of appeal under 37 CFR 41.31 followed by an appeal brief under 37 CFR 41.37. The previously paid notice of appeal fee and appeal brief fee can be applied to the new appeal. If, however, the appeal fees set forth in 37 CFR 41.20 have been increased since they were previously paid, then appellant must pay the difference between the increased fees and the amount previously paid.
A Supervisory Patent Examiner (SPE) has approved of reopening prosecution by signing below:
/NHAT HUY T NGUYEN/Primary Examiner, Art Unit 2147
/VIKER A LAMARDO/Supervisory Patent Examiner, Art Unit 2147