# US 11,264,206 B2 · Methods and systems for forming a pattern on a surface using multi-beam charged particle beam lithography > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 11,264,206 B2 - Document ID: US11264206B2 - Status: Granted - Kind: B2 - Application number: 16/655,582 - Filed: 2019-10-17 - Published: 2020-02-13 - Granted: 2022-03-01 - Earliest claimed: 2014-03-10 - Assignees: D2S, Inc. - Applicants: D2S, Inc. - Inventors: Akira Fujimura; Thang Nguyen; Ajay Baranwal; Michael J. Meyer; Suhas Pillai - Web page: https://patentgrep.com/patent/US11264206B2 ## CPC - H01J 37/3175: {Projection methods, i.e. transfer substantially complete pattern to substrate} - G06N 3/008: based on physical entities controlled by simulated intelligence so as to replicate intelligent life forms, e.g. based on robots replicating pets or humans in their appearance or behaviour - G06N 3/04: Architecture, e.g. interconnection topology - G06N 3/0455: Auto-encoder networks; Encoder-decoder networks - G06N 3/0464: Convolutional networks [CNN, ConvNet] - G06N 3/08: Learning methods - G06N 3/09: Supervised learning - G06N 3/10: Interfaces, programming languages or software development kits, e.g. for simulating neural networks - G06T 2207/20081: Training; Learning - G06T 2207/20084: Artificial neural networks [ANN] - H01J 37/3177: {Multi-beam, e.g. fly's eye, comb probe} - H01J 2237/30416: Handling of data - H01J 2237/31761: Patterning strategy - H01J: ELECTRIC DISCHARGE TUBES OR DISCHARGE LAMPS (spark-gaps H01T; electric arc lamps H05B31/00; particle accelerators H05H) - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS - G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL ## Abstract Methods for fracturing or mask data preparation are disclosed in which a set of single-beam charged particle beam shots is input; a calculated image is calculated using a neural network, from the set of single-beam charged particle beam shots; and a set of multi-beam shots is generated based on the calculated image, to convert the set of single-beam charged particle beam shots to the set of multi-beam shots which will produce a surface image on the surface. Methods for training a neural network include inputting a set of single-beam charged particle beam shots; calculating a set of calculated images using the set of single-beam charged particle beam shots; and training the neural network with the set of calculated images. ## Claims ### Claim 1 A method for forming a pattern on a surface using multi-beam charged particle beam lithography, the method comprising: inputting a set of single-beam charged particle beam shots; calculating a calculated image using the set of single-beam charged particle beam shots, wherein the calculating comprises pixel-to-pixel conversion using a neural network, wherein the neural network comprises a U-Net; generating a set of multi-beam shots based on the calculated image, to convert the set of single-beam charged particle beam shots to the set of multi-beam shots which will produce a surface image on the surface, wherein the surface image matches the calculated image, within a pre-determined tolerance, and wherein the generating is performed using a computing hardware processor; and forming the pattern on the surface using the set of multi-beam shots. ### Claim 2 (depends on 1) The method of claim 1 wherein the pattern comprises one or more patterns, and wherein the calculated image comprises contour information for the one or more patterns. ### Claim 3 (depends on 1) The method of claim 1 wherein the calculated image comprises a predicted image calculated by the neural network. ### Claim 4 (depends on 3) The method of claim 3 wherein the predicted image comprises an aerial image. ### Claim 5 (depends on 1) The method of claim 1 wherein each multi-beam shot comprises a plurality of beamlets. ### Claim 6 (depends on 1) The method of claim 1 wherein shots in the set of single-beam charged particle beam shots comprise assigned dosages. ### Claim 7 (depends on 1) The method of claim 1 wherein shots in the set of single-beam charged particle beam shots have no assigned dosages. ### Claim 8 (depends on 1) The method of claim 1 wherein shots in the set of single-beam charged particle beam shots overlap. ### Claim 9 (depends on 1) The method of claim 1 wherein shots in the set of single-beam charged particle beam shots comprise variable shaped beam (VSB) shots. ### Claim 10 (depends on 1) The method of claim 1 wherein the pattern comprises one or more patterns, wherein the multi-beam charged particle beam lithography comprises exposing a plurality of pixels or beamlets, and wherein the step of generating comprises: determining pixel or beamlet dosage information for each of the patterns in the one or more patterns; and generating the set of multi-beam shots from the pixel or beamlet dosage information for the one or more patterns. ### Claim 11 (depends on 1) The method of claim 1 wherein the calculating further comprises calculating an identity image from the set of single-beam charged particle beam shots. ### Claim 12 (depends on 11) The method of claim 11 wherein the identity image is divided into tiles. ### Claim 13 (depends on 1) The method of claim 1 wherein the U-Net further comprises four convolution layers in a contracting network and four transposed convolution layers in an expansion network. ### Claim 14 (depends on 13) The method of claim 13 wherein a stride of 2 and a kernel size of 4×4 is used in each of the four convolution layers in the contracting network and in each of the four transposed convolution layers in the expansion network.