# US 2026/0090336 A1 · AI-Optimized Semiconductor Manufacturing Process Using Machine Learning Models Trained on Mask Work Datasets

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2026/0090336 A1
- Document ID: US20260090336A1
- Status: Published application
- Kind: A1
- Application number: 18/890,840
- Filed: 2024-09-20
- Published: 2026-03-26
- Earliest claimed: 2024-09-20
- Inventors: Alexander Davis
- Web page: https://patentgrep.com/patent/US20260090336A1

## CPC

- H10P 74/238: {comprising acting in response to an ongoing measurement without interruption of processing, e.g. endpoint detection or in-situ thickness measurement}
- G06F 30/27: using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
- H10P 74/203: {Structural properties, e.g. testing or measuring thicknesses, line widths, warpage, bond strengths or physical defects}
- G06F 2119/18: Manufacturability analysis or optimisation for manufacturability
- H10P 74/23: {characterised by multiple measurements, corrections, marking or sorting processes}
- H10P: GENERIC PROCESSES OR APPARATUS FOR THE MANUFACTURE OR TREATMENT OF DEVICES COVERED BY CLASS H10
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

A method and system for optimizing a semiconductor manufacturing process using an artificial intelligence (AI) system comprising machine learning (ML) models trained on mask work datasets. The AI system generates optimized process parameters for a multi-step semiconductor manufacturing process based on an input mask work. The manufacturing process includes photolithography, etching, ion implantation, chemical vapor deposition (CVD), physical vapor deposition (PVD), atomic layer deposition (ALD), thermal oxidation, and/or chemical-mechanical polishing (CMP). During manufacturing, metrology data is collected and input into the ML models to predict end-of-line electrical performance parameters. If the predicted parameters deviate from target values, the AI system adjusts process parameters to optimize performance. The ML models are retrained using the collected metrology data to improve AI system performance over time. The system includes a semiconductor manufacturing apparatus configured to perform the optimized manufacturing process.

## Claims

### Claim 1

A method for optimizing a semiconductor device manufacturing process using an artificial intelligence (AI) system, comprising:
a. providing a mask work defining a pattern for a semiconductor device;
b. inputting the mask work into an AI system comprising one or more machine learning (ML) models trained on a dataset of mask works and corresponding optimized manufacturing process parameters;
c. generating, by the AI system, a set of optimized process parameters for a multi-step semiconductor manufacturing process based on the input mask work, the multi-step semiconductor manufacturing process including at least one of:
i. photolithography,
ii. etching,
iii. ion implantation,
iv. chemical vapor deposition (CVD),
v. physical vapor deposition (PVD),
vi. atomic layer deposition (ALD),
vii. thermal oxidation, or
viii. chemical-mechanical polishing (CMP);
d. manufacturing the semiconductor device using the generated set of optimized process parameters;
e. obtaining, during the manufacturing of the semiconductor device, metrology data from the semiconductor device;
f. predicting, by inputting the metrology data into the one or more ML models, an end-of-line electrical performance parameter of the semiconductor device;
g. determining that the predicted end-of-line electrical performance parameter deviates from a target value;
h. adjusting, by the AI system, a process parameter of the multi-step semiconductor manufacturing process based on the deviation to optimize the end-of-line electrical performance parameter; and
i. retraining the one or more ML Models using the obtained metrology data to improve the performance of the AI system over time.

### Claim 2 (depends on 1)

The method of claim 1, wherein the mask work defines a curvilinear pattern for the semiconductor device, and wherein the AI system is trained on a dataset including curvilinear mask works.

### Claim 3 (depends on 1)

The method of claim 1, wherein the multi-step semiconductor manufacturing process further includes at least one of:
a. wet cleans,
b. surface passivation,
c. plasma ashing,
d. rapid thermal processing (RTP),
e. millisecond thermal processing,
f. laser anneal, or
g. furnace anneals.

### Claim 4 (depends on 1)

The method of claim 1, wherein the metrology data includes at least one of:
a. critical dimension (CD) measurements,
b. overlay measurements,
c. film thickness measurements, or
d. defect inspection data.

### Claim 5 (depends on 1)

The method of claim 1, wherein the end-of-line electrical performance parameter includes at least one of:
a. threshold voltage,
b. saturation current,
c. leakage current, or
d. operating frequency.

### Claim 6 (depends on 1)

The method of claim 1, wherein adjusting the process parameter of the multi-step semiconductor manufacturing process includes modifying at least one of:
a. exposure dose in the photolithography step,
b. etch time in the etching step,
c. implantation dose in the ion implantation step,
d. deposition time in the CVD, PVD, or ALD steps,
e. oxidation time in the thermal oxidation step, or
f. polishing time in the CMP step.

### Claim 7 (depends on 1)

The method of claim 1, further comprising:
a. wafer testing to verify electrical performance of the manufactured semiconductor device;
b. die preparation of the manufactured semiconductor device; and
c. IC packaging of the manufactured semiconductor device.

### Claim 8

A system for optimizing a semiconductor device manufacturing process, comprising:
a. an input interface configured to receive a mask work defining a pattern for a semiconductor device;
b. a memory storing one or more machine learning (ML) models trained on a dataset of mask works and corresponding optimized manufacturing process parameters;
c. one or more processors;
d. a non-transitory computer-readable medium storing instructions that, when executed by the one or more processors, cause the system to:
i. input the received mask work into the one or more ML Models,
ii. generate, using the one or more ML Models, a set of optimized process parameters for a multi-step semiconductor manufacturing process based on the input mask work, and
iii. output the generated set of optimized process parameters;
e. a semiconductor manufacturing apparatus configured to:
i. manufacture the semiconductor device using the generated set of optimized process parameters, the semiconductor manufacturing apparatus including at least one of:
1. a photolithography tool,
2. an etching tool,
3. an ion implantation tool,
4. a chemical vapor deposition (CVD) tool,
5. a physical vapor deposition (PVD) tool,
6. an atomic layer deposition (ALD) tool,
7. a thermal oxidation tool, or
8. a chemical-mechanical polishing (CMP) tool, and
ii. collect metrology data during the manufacturing of the semiconductor device; and
f. wherein the instructions further cause the system to:
i. predict, by inputting the collected metrology data into the one or more ML Models, an end-of-line electrical performance parameter of the semiconductor device,
ii. determine that the predicted end-of-line electrical performance parameter deviates from a target value,
iii. adjust a process parameter of the multi-step semiconductor manufacturing process based on the deviation to optimize the end-of-line electrical performance parameter, and
iv. retrain the one or more ML Models using the collected metrology data.

### Claim 9 (depends on 8)

The system of claim 8, wherein the dataset of mask works and corresponding optimized manufacturing process parameters includes data related to at least one of cleaning, photoresist coating, photoresist baking, exposure, ion implantation, etching, chemical vapor deposition (CVD), physical vapor deposition (PVD), thermal treatments, or chemical-mechanical polishing (CMP).

### Claim 10 (depends on 8)

The system of claim 8, wherein the one or more ML Models are configured to generate optimized process parameters for a curvilinear photomask.

### Claim 11 (depends on 8)

The system of claim 8, wherein the instructions further cause the system to:
a. receive a mask set defining multiple layers of the semiconductor device; and
b. generate, using the one or more ML Models, optimized process parameters for each layer of the mask set.

### Claim 12 (depends on 8)

The system of claim 8, wherein the metrology data includes data related to at least one of critical dimensions, overlay, film thickness, or defects.

### Claim 13 (depends on 8)

The system of claim 8, wherein the end-of-line electrical performance parameter includes at least one of transistor threshold voltage, leakage current, or device speed.

### Claim 14 (depends on 8)

The system of claim 8, wherein adjusting the process parameter includes adjusting at least one of exposure dose, focus, etch time, etch gas composition, deposition temperature, or CMP pressure.

### Claim 15 (depends on 8)

The system of claim 8, wherein the one or more ML Models comprise a neural network trained using a dataset of mask works, corresponding optimized manufacturing process parameters, and end-of-line electrical performance parameters.

### Claim 16 (depends on 8)

The system of claim 8, wherein the instructions further cause the system to:
a. simulate the multi-step semiconductor manufacturing process using the generated set of optimized process parameters; and
b. predict the end-of-line electrical performance parameter based on the simulation.

### Claim 17 (depends on 8)

The system of claim 8, wherein the instructions further cause the system to:
a. identify a root cause of the deviation of the predicted end-of-line electrical performance parameter from the target value; and
b. suggest a corrective action to address the root cause.
