Performance predictors for semiconductor-manufacturing processes
Methods, systems, and computer programs are presented for predicting the performance of semiconductor manufacturing equipment operations. One method includes an operation for obtaining machine-learning (ML) models, each model related to predicting a performance metric for an operation of a semiconductor manufacturing tool. Further, each ML model utilizes features defining inputs for the ML model. The method further includes an operation for receiving a process definition for manufacturing a product with the semiconductor manufacturing tool. One or more ML models are utilized to estimate a performance of the process definition used in the semiconductor manufacturing tool. Additionally, the method includes presenting, on a display, results showing the estimate of the performance of the manufacturing of the product. In some aspects, the use of hybrid models improves the predictive accuracy of the system by augmenting the capabilities of data-driven models with the reinforcement provided by the physics-based models.
1 . A method comprising:
obtaining a plurality of machine-learning (ML) models, each of the plurality of ML models being related to predicting a performance metric for an operation of a semiconductor manufacturing tool, each ML model to utilize a plurality of features defining inputs for the ML model and include at least one semiconductor manufacturing process operation selected from operations including etch/atomic layer etch, electrofill, chemical vapor deposition (CVD), atomic layer deposition (ALD), clean and strip, physical vapor deposition, and chemical/mechanical polishing, the plurality of ML models being based on training data to make data-driven decisions expressed as outputs;
receiving, by one or more processors, a process definition for manufacturing a product with the semiconductor manufacturing tool;
utilizing, by the one or more processors, one or more ML models from the plurality of ML models to estimate a performance of the process definition used in the semiconductor manufacturing tool; and
presenting, on a display, results showing the estimate of the performance of the manufacturing of the product.
2 . The method as recited in claim 1 , wherein creating one ML model from the plurality of ML models comprises:
obtaining training data for the ML model, the training data providing values of the features for the ML model; and
training an ML algorithm to obtain the ML model.
3 . The method as recited in claim 2 , wherein obtaining the training data for the ML model comprises:
performing experiments on the semiconductor manufacturing tool;
measuring values of the features for the experiments; and
using the measured values for the training data.
4 . The method as recited in claim 3 , wherein obtaining the training data for the ML model further comprises:
training a secondary ML model to produce a reduced order model (ROM) for metrology data; and
utilizing an output of the secondary ML model as additional training data.
5 . The method as recited in claim 2 , wherein obtaining the training data for the ML model comprises:
obtaining the training data by performing physics-based simulations for the operation of the semiconductor manufacturing tool.
6 . The method as recited in claim 1 , wherein the plurality of ML models includes chamber, processing matrix, substrate-scale, wafer features, atomistic, and electronic state models.
7 . The method as recited in claim 1 , wherein a chamber ML model is for estimating a geometry of a chamber in the semiconductor manufacturing tool with inputs including design dimensions and outputs including a definition of the geometry of the chamber.
8 . The method as recited in claim 1 , wherein a processing matrix ML model is for analyzing behavior of an environment of a substrate during processing with inputs including one or more of electric field, magnetic B field, current density, chemistry, reaction cross section, reaction pathways, material properties, RF frequency, RF voltage, temperature and RF bias, wherein outputs of a plasma ML model include one or more of charged-species density and fluxes, ambipolar field, electron temperatures, electron energy distribution function (EEDF), ion energy-angle distribution (IEAD), on-wafer fluxes, charge density (surface and volume) source, and loss or generation terms for species.
9 . The method as recited in claim 1 , wherein a substrate-level ML model is for analyzing a performance of a sheath close to the substrate with inputs including one or more of radiofrequency (RF) voltage, electric fields, source terms, reaction collision cross-sections, solution concentrations and reaction pathways, wherein outputs of a sheath ML model include one or more of fluxes on a wafer surface, ion energy and angular distribution, conduction and displacement currents, ion transit times, surface functionalization, and charge density.
10 . The method as recited in claim 1 , wherein a wafer feature ML model is for analyzing a layout of a wafer feature with inputs including one or more of design layout, mask layers, and initial steps, wherein an output of the wafer feature ML model includes a geometric description of the wafer feature.
11 . The method as recited in claim 1 , wherein a wafer chemistry ML model is for analyzing a chemistry of a wafer feature with inputs including one or more of on-wafer fluxes, material properties, reaction pathways, reaction rates, ion angular yields, etch thresholds, sticking coefficients, and accommodation coefficient, wherein outputs of the wafer chemistry ML model include one or more of geometry evolution and front tracking, distribution of species inside the wafer feature, and distribution of ion energies and angles inside wafer features.
12 . A system comprising:
a hardware-based memory comprising instructions; and
one or more computer processors, wherein the instructions, when executed by the one or more computer processors, cause the system to perform operations comprising:
obtaining a plurality of machine-learning (ML) models, each of the plurality of ML models being related to predicting a performance metric for an operation of a semiconductor manufacturing tool, each ML model to utilize a plurality of features defining inputs for the ML model and include at least one semiconductor manufacturing process operation selected from operations including etch/atomic layer etch, electrofill, chemical vapor deposition (CVD), atomic layer deposition (ALD), clean and strip, physical vapor deposition, and chemical/mechanical polishing, the plurality of ML models being based on training data to make data-driven decisions expressed as outputs;
receiving, by one or more processors, a process definition for manufacturing a product with the semiconductor manufacturing tool;
utilizing, by the one or more processors, one or more ML models from the plurality of ML models to estimate a performance of the process definition used in the semiconductor manufacturing tool; and
presenting, on a display, results showing the estimate of the performance of the manufacturing of the product.
13 . The system as recited in claim 12 , wherein creating one ML model from the plurality of machine-learning (ML) models comprises:
obtaining training data for the ML model, the training data providing values of the features for the ML model; and
training an ML algorithm to obtain the ML model.
14 . The system as recited in claim 13 , wherein obtaining the training data for the ML model comprises:
performing experiments on the semiconductor manufacturing tool;
measuring values of the features for the experiments;
using the measured values for the training data; and
obtaining additional training data by performing physics-based simulations for the semiconductor manufacturing tool.
15 . The system as recited in claim 12 , wherein a chamber ML model is for estimating a geometry of a chamber in the semiconductor manufacturing tool with inputs including design dimensions and outputs including a definition of the geometry of the chamber.
16 . The system as recited in claim 12 , wherein a processing matrix ML model is for analyzing behavior of an environment of a substrate during processing with inputs including one or more of electric field, magnetic B field, current density, chemistry, reaction cross section, reaction pathways, material properties, temperature, mass transport, RF frequency, RF voltage, and RF bias, wherein outputs of the processing matrix ML model include one or more of charged-species density and fluxes, ambipolar field, electron temperatures, electron energy distribution function (EEDF), ion energy-angle distribution (IEAD), on-wafer fluxes, charge density (surface and volume) source, and loss or generation terms for species.
17 . A machine-readable storage medium including instructions that, when executed by a machine, cause the machine to perform operations comprising:
obtaining a plurality of machine-learning (ML) models, each of the plurality of ML models being related to predicting a performance metric for an operation of a semiconductor manufacturing tool, each ML model to utilize a plurality of features defining inputs for the ML model and include at least one semiconductor manufacturing process operation selected from operations including etch/atomic layer etch, electrofill, chemical vapor deposition (CVD), atomic layer deposition (ALD), clean and strip, physical vapor deposition, and chemical/mechanical polishing, the plurality of ML models being based on training data to make data-driven decisions expressed as outputs;
receiving, by one or more processors, a process definition for manufacturing a product with the semiconductor manufacturing tool;
utilizing, by the one or more processors, one or more ML models from the plurality of ML models to estimate a performance of the process definition used in the semiconductor manufacturing tool; and
presenting, on a display, results showing the estimate of the performance of the manufacturing of the product.
18 . The machine-readable storage medium as recited in claim 17 , wherein creating one ML model from the plurality of ML models comprises:
obtaining training data for the ML model, the training data providing values of the features for the ML model; and
training an ML algorithm to obtain the ML model.
19 . The machine-readable storage medium as recited in claim 18 , wherein obtaining the training data for the ML model comprises:
performing experiments on the semiconductor manufacturing tool;
measuring values of the features for the experiments;
using the measured values for the training data; and
obtaining additional training data by performing physics-based simulations for the semiconductor manufacturing tool.
20 . The machine-readable storage medium as recited in claim 17 , wherein a chamber ML model is for estimating a geometry of a chamber in the semiconductor manufacturing tool with inputs including design dimensions and outputs including a definition of the geometry of the chamber.