Quantum-Computed Training Data for Material Solubility Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing solubility prediction methods for materials in solvents are resource-intensive and approximate due to the lack of efficient algorithms for solving the underlying mathematical optimization problem, and they do not consider the quantum nature of nuclei.
Innovation Solution
A method involving a quantum computer to solve the electronic Schrödinger equation and a classical computer to perform path integral molecular dynamics (PIMD) is used to generate a training dataset for a machine learning model, which predicts solubility by calculating free energy changes in solute-solvent systems.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If ab-initio or classical molecular dynamics is used for solubility prediction, then solubility can be calculated, but the calculation is resource-intensive and approximate
Solution Approach 1:
The patent introduces a machine learning model as an intermediary between the quantum mechanical calculations and solubility prediction. The ML model is trained on data from ab-initio or classical MD simulations, allowing it to capture the complex relationships between material properties and solubility. Once trained, the model provides accurate predictions without requiring resource-intensive repeated simulations, thus resolving the contradiction between accuracy and efficiency.
Solution Approach 2:
The patent performs preliminary action by conducting comprehensive ab-initio or classical MD simulations upfront to generate a training dataset. This preliminary data collection phase captures the essential physical chemistry relationships, which are then encoded into the ML model. Subsequent solubility predictions can be made rapidly using the pre-trained model without repeating the expensive simulations.
2Measurement precision
If ab-initio or classical molecular dynamics is used, then solubility prediction is possible, but the quantum nature of nuclei is not considered
Solution Approach 1:
The patent changes the parameters and assumptions of the simulation approach. Instead of using classical MD that treats nuclei classically, the patent employs ab-initio molecular dynamics or path integral molecular dynamics that incorporate quantum mechanical effects. The training data is generated using these quantum-aware methods, allowing the ML model to learn from physically accurate data that accounts for nuclear quantum effects.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach provides a more accurate and efficient method for predicting solubility, enabling the development of materials with specific properties for security features in ID documents and improving the durability and effectiveness of security elements.
Implementation Method 1
solving numerically, on a quantum computer, the electronic Schrödinger equation of a quantum system representing the solute-solvent system
Implementation Method 2
using a thermostat in order to set a specific temperature value T of the solute-solvent system
Implementation Method 3
using a classical computer to apply a path integral molecular dynamics (PIMD) technique to calculate a free energy change between a first state of the solute-solvent system and a second state of the solute-solvent system
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
Disclosed is a method for generating a training dataset to be used in training a machine learning model in order to predict the solubility of a material in at least one solvent, the method comprising: performing the following repeatedly: entering a selection of hypothetical materials (solute and solvent) from a parameter space, the parameter space being defined by a set of material parameters describing the said hypothetical materials; using a thermostat to set a specific temperature value T of the solute-solvent system; using a classical computer to apply a path integral molecular dynamics (PIMD) technique to calculate a free energy change between a first state of the solute-solvent system and a second state of the solute-solvent system, wherein the forces acting on the nuclei of the solute-solvent system which are used by PIMD are determined by solving numerically, on a quantum computer, the electronic Schrödinger equation of a quantum system representing the solute-solvent system; calculating the solubility of the solute in the solvent at the specific temperature value T; adding an entry to the training dataset, the entry indicates the selected solute and the selected solvent and a label, wherein the label indicates the calculated solubility.