Generating and validating optimized membranes for carbon dioxide separation in binary gas

Patent metadata

Field	Value
Publication number	US-12508550-B2
Application number	US-202117401279-A
Country	US
Kind code	B2
Filing date	Aug 12, 2021
Priority date	Aug 12, 2021
Publication date	Dec 30, 2025
Grant date	Dec 30, 2025

Abstract

A method and system of discovering materials for use in carbon dioxide separation includes extracting references to chemical molecules from online sources. The extracted references are encoded into chemical formulas. Molecular properties are calculated from the encoded chemical formulas. Features are extracted from the chemical formulas. Molecular properties of predicted molecular structures are predicted through a machine learning engine. The predicted molecular properties are based on the calculated molecular properties and extracted features. Target properties for predicted molecular structures are defined. Synthesized molecular structures are generated. The synthesized molecular structures include predicted molecular properties satisfying the defined target properties.

First claim

What is claimed is:

1 . A computer-implemented method of discovering materials for use in carbon dioxide separation, the computer-implemented method comprising:
extracting from sources, references to chemical molecules; encoding the extracted references into one or more chemical formulas; calculating molecular properties from the one or more chemical formulas; extracting features from the one or more chemical formulas; predicting molecular properties of predicted molecular structures related to the carbon dioxide separation, through a machine learning engine, wherein the predicted molecular properties are based on the calculated molecular properties and the extracted features related to the carbon dioxide separation; defining one or more target properties for the predicted molecular structures for the carbon dioxide separation; performing by the machine learning engine, an inverse problem feature search using the defined one or more target properties for the carbon dioxide separation; estimating, by the machine learning engine, feature vector values based on one or more values of the defined one or more target properties; identifying candidate feature vectors satisfying the inverse problem feature search, from the estimated feature vector values; generating proposed synthesized molecular structures for the carbon dioxide separation based on the identified candidate feature vectors, wherein the proposed synthesized molecular structures include predicted molecular properties that satisfy the defined one or more target properties; comparing the proposed synthesized molecular structures and a set of polymer property functionality (PPF) values, wherein the set of PPF values is based on the calculated molecular properties; detecting, based on the comparing, a match between one of the proposed synthesized molecular structures and the set of PPF values for the carbon dioxide separation; and outputting, based on the detected match, the one of the proposed synthesized molecular structures for validation related to the carbon dioxide separation.

2 . The computer-implemented method of claim 1 , wherein the defined one or more target properties include a gas permeability, a glass transition temperature, and a temperature associated with a half-life of a candidate synthesized molecule.

3 . The computer-implemented method of claim 1 , wherein the encoding includes converting the extracted references to the chemical molecules into simplified molecular-input line-entry system (SMILES) string representations.

4 . The computer-implemented method of claim 1 , further comprising:
applying a rules based filtering criteria to the generated proposed synthesized molecular structures; comparing the generated proposed synthesized molecular structures to the rules based filtering criteria; identifying one of the generated proposed synthesized molecular structures as a best match to the rules based filtering criteria; and validating whether the identified generated proposed synthesized molecular structure is usable in the carbon dioxide separation.

5 . The computer-implemented method of claim 1 , further comprising:
selecting a generated proposed synthesized molecular structure of the generated proposed synthesized molecular structures; and performing a molecular dynamics simulation on the selected generated proposed synthesized molecular structure to validate a physical structure of the selected generated proposed synthesized molecular structure.

6 . The computer-implemented method of claim 5 , wherein the validation of the physical structure of the selected generated proposed synthesized molecular structure is based on a gas permeability and a selectivity of the selected generated proposed synthesized molecular structure for the carbon dioxide separation.

7 . A computer program product for discovering materials for use in carbon dioxide separation, the computer program product comprising:
one or more non-transitory computer readable storage media, and program instructions collectively stored on the one or more non-transitory computer readable storage media to perform operations comprising: extracting from sources, references to chemical molecules; encoding the extracted references into one or more chemical formulas; calculating molecular properties from the one or more chemical formulas; extracting features from the one or more chemical formulas; predicting molecular properties of predicted molecular structures related to the carbon dioxide separation, through a machine learning engine, wherein the predicted molecular properties are based on the calculated molecular properties and the extracted features related to the carbon dioxide separation; defining one or more target properties for the predicted molecular structures for the carbon dioxide separation; performing by the machine learning engine, an inverse problem feature search using the defined one or more target properties for the carbon dioxide separation; estimating, by the machine learning engine, feature vector values based on one or more values of the defined one or more target properties; identifying candidate feature vectors satisfying the inverse problem feature search, from the estimated feature vector values; generating proposed synthesized molecular structures for the carbon dioxide separation based on the identified candidate feature vectors, wherein the proposed synthesized molecular structures include predicted molecular properties that satisfy the defined one or more target properties; comparing the proposed synthesized molecular structures and a set of polymer property functionality (PPF) values, wherein the set of PPF values is based on the calculated molecular properties; detecting, based on the comparing, a match between one of the proposed synthesized molecular structures and the set of PPF values for the carbon dioxide separation; and outputting, based on the detected match, the one of the proposed synthesized molecular structures for validation related to the carbon dioxide separation.

8 . The computer program product of claim 7 , wherein the defined one or more target properties include a gas permeability, a glass transition temperature, and a temperature associated with a half-life of a candidate synthesized molecule.

9 . The computer program product of claim 7 , wherein the encoding includes converting the extracted references to the chemical molecules into simplified molecular-input line-entry system (SMILES) string representations.

10 . The computer program product of claim 7 , wherein the operations further comprise:
applying a rules based filtering criteria to the generated proposed synthesized molecular structures; comparing the generated proposed synthesized molecular structures to the rules based filtering criteria; identifying one of the generated proposed synthesized molecular structures as a best match to the rules based filtering criteria; and validating whether the identified generated proposed synthesized molecular structure is usable in the carbon dioxide separation.

11 . The computer program product of claim 7 , wherein the operations further comprise:
selecting a generated proposed synthesized molecular structure of the generated proposed synthesized molecular structures; and performing a molecular dynamics simulation on the selected generated proposed synthesized molecular structure to validate a physical structure of the selected generated proposed synthesized molecular structure.

12 . The computer program product of claim 11 , wherein the validation of the physical structure of the selected generated proposed synthesized molecular

Assignees

• IBM

Inventors

• Giro Ronaldo
• Steiner Mathias B
• Hsu Hsiang Han
• Kishimoto Akihiro
• Takeda Seiji

Classifications

• G06N5/04

  Inference or reasoning models · CPC title

• B01D53/22

  by diffusion (manufacturing semi-permeable membranes B01D67/00; form, structure or properties of semi-permeable membranes B01D69/00; material for semi-permeable membranes B01D71/00) · CPC title

• G06Q30/0201

  Market modelling; Market analysis; Collecting market data · CPC title

• G06N20/00

  Machine learning · CPC title

• Y02C20/40

  of CO2 · CPC title