System for modeling catalytic materials for efficient and cost-effective hydrogen production

Abstract

An intelligent catalyst discovery and electrolysis optimization system for the production of green hydrogen, consisting of: • a schematic water electrolysis unit configured to simulate the electrolytic splitting of water into hydrogen and oxygen by modeling the hydrogen evolution reaction (HER) at a cathode and the oxygen evolution reaction (OER) at an anode and enabling analysis of catalyst effectiveness in reducing energy input; • a catalyst structure analysis unit configured to visualise and simulate atomic and molecular structures of candidate catalyst materials, including base metals, transition metal-based nanostructures and two-dimensional materials, and to evaluate charge distribution, surface morphology and active site behaviour; • a computational modeling and simulation unit using density functional theory (DFT) for quantum simulations to evaluate the thermodynamic stability, electronic structure, reaction energetics and adsorption energy of potential catalysts; • a machine learning optimization module trained on simulated and experimental data, using supervised and unsupervised learning models to determine optimal catalyst descriptors and predict catalytic activity, stability, and durability; • an experimental validation and test interface comprising electrochemical test environments including three-electrode cells and electrolyzers for performance benchmarking under realistic electrolysis conditions and capable of logging overvoltage, efficiency and degradation behavior; • an energy integration unit configured to interface with renewable energy systems, including photovoltaic systems, to dynamically power the electrolysis process and manage energy consumption through a power management subsystem; • a data processing and control interface to manage cross-system communication between simulation, experimental, and machine learning modules, enabling real-time parameter tuning and visualization of system diagnostics; and • a secure storage and logging unit adapted for storing simulation data, machine learning training sets and experimental results, with features for offline retrieval and integration with cloud platforms for collaborative research.

Description

Field of the invention:

[0001] Given the increasing global demand for clean energy, there is an urgent need to combat climate change, limit the use of fossil fuels, and create a low-carbon economy. Green hydrogen, produced by splitting water with renewable energy, is a clean and flexible energy source. However, the implementation of this concept faces significant challenges, particularly the costs of electrolysis plants and conventionally approved precious metal catalysts for the key hydrogen and oxygen evolution reaction (HER) and oxygen evolution reaction (OER).

[0002] Although platinum-based catalysts are highly efficient, they are expensive and rare, making them impractical for large-scale applications. Researchers are investigating low-cost alternatives such as transition metal alloys and two-dimensional materials, but these often fail to achieve the required efficiency, durability, or stability for use under real-world conditions. To overcome these obstacles, computational modeling, particularly density functional theory (DFT), is being used extensively to commercially explore new catalysts. This process includes steps such as performing preliminary DFT calculations on model systems, calculating the pKa values ​​of acid-base reactions, and typically optimizing their geometric structure.Using DFT, researchers can simulate the interaction of atoms, predict reaction pathways, and identify promising materials without the need for complex experimental trials. However, simulating materials discovery is still slow and resource-intensive.

[0003] This is where artificial intelligence (AI) and machine learning (ML) come into play. AI and ML accelerate the discovery process through rapid, iterative analysis of large data sets. This can predict catalyst properties, support design optimization, and explore large chemical spaces. This enables virtual testing of thousands of potential materials, drastically reducing experimental time and costs.

[0004] Researchers hope to develop more cost-effective and efficient catalysts for green hydrogen production by combining DFT with an AI-assisted optimization process. This approach aims to reduce the cost of electrolysis, increase efficiency, and make green hydrogen a scalable solution for sustainable energy. The project aims to revolutionize catalyst research, overcome economic barriers to smart catalysts, and harness the potential of green hydrogen to lead the world toward a carbon-neutral economy through multidisciplinary computational chemistry, materials, and AI. Background of the invention:

[0005] In recent years, the global demand for clean and sustainable energy solutions has increased due to the need to mitigate climate change, wean countries off fossil fuels, and transition societies to a low-carbon economy. One of the most promising solutions for achieving these goals is green hydrogen, a clean and multifunctional energy carrier that can be produced through water electrolysis, which is based on renewable energy sources such as sunlight or wind power. This process splits water molecules into hydrogen and oxygen. Hydrogen serves as an energy storage medium or fuel and can be used in various sectors such as transportation, industrial processes, and power generation.

[0006] However, a major obstacle to the large-scale adoption of green hydrogen is the high cost of electrolysis plants, particularly the expensive raw materials for the catalysts that drive the reactions in the electrolysis cells. Water splitting involves both the hydrogen evolution reaction (HER) to generate hydrogen and the oxygen evolution reaction (OER) to produce oxygen—the two central electrochemical reactions. These reactions can be catalytically limited and therefore require highly effective catalysts to reduce the activation energy and enable rapid reactions under the demanding conditions of electrolysis.

[0007] Currently, platinum-based catalysts are considered the gold standard for such reactions due to their excellent efficiency and stability. However, platinum is rare and expensive, making it unsuitable for large-scale industrial processes. This is leading to a growing demand for cheaper, abundant substitutes with similar performance characteristics to platinum, but without the precious metal's cost and supply chain challenges.

[0008] Platinum-based catalysts are inefficient and expensive, necessitating an intensive search for alternative materials. Various alternative materials, such as transition metal alloys, base metals, and two-dimensional materials, have been investigated, but they still exhibit poor performance, poor durability, and instability under practical electrolysis operating conditions, especially under extreme conditions such as high current density, high pH, ​​and high temperatures. Therefore, researchers are seeking new catalyst materials that offer the best compromise between efficiency, cost, and stability for large-scale hydrogen production.

[0009] Computational materials science is a powerful tool for the design and discovery of new catalysts to overcome these challenges. Density functional theory (DFT) allows scientists to model the interactions between materials and water molecules at the atomic level, providing valuable insights into the atomic reaction mechanisms that govern HER and OER. Developed in the 1960s, DFT allows researchers to calculate adsorption energies, activation energies, and reaction pathways on many different catalyst surfaces with little or no laboratory effort. Using these results, previously difficult target materials and reaction intermediates can be unraveled through traditional trial-and-error systems. Instead, this computational approach enables the screening and identification of promising candidates.

[0010] However, materials research remains time-consuming and costly due to the large number of candidate materials that must be calculated and tested experimentally. Artificial intelligence (AI) and machine learning (ML) can significantly accelerate the process. AI models can uncover hidden trends and relationships in large datasets of known materials and their properties, which can then be used to develop new catalysts. This is explained in the section "Organizing Data with Machine Learning to Support Experimental Validation." Machine learning algorithms predict material performance in advance, thus avoiding excessive experimentation and saving time and money.

[0011] AI and ML have proven extremely helpful in accelerating the discovery process in many fields. In materials science, they can support the following: Modeling other new catalysts without expensive and slow synthesis. Streamlining catalyst design by identifying the features (e.g., atomic composition, crystal structure, electronic properties) that most strongly correlate with high catalytic performance. Discovering materials that remain unexplored in conventional discovery systems by exploring chemical spaces that would be overwhelming for conventional systems. Accelerating screening processes, allowing literally thousands of potential catalysts to be modeled and optimized in silico before being tested in physical experiments.

[0012] Applying AI to optimize this DFT database of over 2 million 3D transition metal oxides is expected to drastically expand the currently limited search space for catalysts. The project therefore aims to redefine the potential of the catalyst discovery pipeline. Researchers can very quickly narrow down the search for materials for HER and OER using AI, as AI can process large amounts of data and predict material behavior. This could open new avenues for low-cost, high-performance catalysts for green hydrogen production. Summary of the invention:

[0013] These efforts focus on advancing the production of green hydrogen by overcoming a key challenge of water splitting: the need to use expensive and low-performance catalysts. Water electrolysis is the primary approach for producing green hydrogen, an environmentally friendly energy carrier. However, efficient catalysts are needed to enhance the HER and OER, both parts of the water splitting process. Currently, catalysts such as platinum are used due to their high efficiency, but they are rare, expensive, and unsustainable. One such project seeks to find, develop, and optimize new, low-cost catalysts, leveraging high-throughput computing with AI and ML techniques. Main objectives:

[0014] Answer these questions with algorithms for developing low-cost catalysts: Identify and develop suitable alternative catalysts that are abundant, low-cost, and highly efficient for electrolysis processes. 2D transition metal alloys, base metals, and 2D materials are being investigated that can replace the platinum group catalyst and exhibit similar performance in hydrogen production.

[0015] World-class computer models: The project uses state-of-the-art density functional theory (DFT) to model the electronic structure of the materials and their chemical reactivity at atomic resolution. This allows for precise predictions of the performance of these materials in water electrolysis, allowing the selection of the most promising candidates for experimental validation.

[0016] Integration of AI and ML: AI and ML algorithms accelerate the discovery process. These algorithms enable the efficient prediction of the elements with the highest catalytic activity and stability at HER and OER reaction sites. They model the relationships between the many possible compositions and structures of the material and the various catalyzed reactions. This saves considerable time and money compared to experimental testing and forms the basis for a rapid catalyst discovery process.

[0017] Catalyst performance optimization: After selecting potential candidates, the performance of the most promising candidates must be optimized by simulating various conditions (e.g., temperature, pressure, current density) and material configurations. AI and ML algorithms iterate through the catalyst designs over time, while formulating new performance metrics (reaction rate, stability, and efficiency).

[0018] Experimental tests: Based on the calculation results, the best-performing catalysts are subjected to experimental testing in electrolysis cells. The new tests focus on evaluating catalyst performance under realistic conditions, measuring the efficiency, durability, and stability of the catalysts over extended periods of use.

[0019] Hybrid integration with renewable energy sources: The project investigates the synergistic integration of these advanced catalysts with renewable energy systems, such as solar cells, for the hybrid production of green hydrogen. The electrolysis process is powered by solar energy, ensuring that the hydrogen produced is fully sustainable and environmentally friendly. System methodology

[0020] Material selection and design: We begin by selecting a material library including transition metal alloys, base metals, and 2D materials. These materials are selected based on availability, cost, and favorable catalytic properties. Using DFT, various electronic properties of the derived materials are calculated, such as adsorption energies, reaction pathways, surface reactivity, etc.

[0021] Computational screening and modeling: Using DFT calculations and other simulation techniques, the electronic structure of each material is investigated to determine its suitability for HER and OER reactions. We search for platinum-free candidates with high catalytic activity that operate stably under electrolysis conditions.

[0022] AI / ML integration: We use machine learning algorithms to analyze the results of DFT simulations and other data sources. The algorithms learn from the performance and properties of previously tested materials to identify patterns and predict which new materials will perform competitively. The AI ​​models can also support the improvement of material properties by suggesting ideal compositions and structures.

[0023] Catalyst optimization: After identifying statistically promising candidates, simulations of various conditions and configurations (temperature, pressure, current density, etc.) are performed to optimize the catalyst. The goal is to determine the best material properties and operating conditions to achieve optimal catalyst performance.

[0024] Experimental testing: The most promising catalyst candidates are synthesized and tested in electrolysis cells. These catalysts are evaluated for their hydrogen production rate, energy efficiency, durability, and stability for long-term operation in realistic environments. AI and ML models are developed based on the experimental data, which are fed back into the models to optimize material selection and further optimize the process.

[0025] Connection to solar energy: In the final step of the project, the optimized catalysts will be integrated into a hybrid system in which the electrolysis process is powered by renewable energy such as solar power. The goal is a scalable, financially and ecologically efficient system for the production of green hydrogen. Besides the results: • Identification of cost-effective catalysts with good performance in water electrolysis for their use in green hydrogen production. • Reduced dependence on precious metals such as platinum, which increases the scalability and cost-effectiveness of hydrogen production. • New class of fast and effective redox catalysts for the sustainable production of electrolytic hydrogen. • Alternative Volcanoes: A sustainable hydrogen production system integrated with renewable energy sources such as solar power to provide scalable green hydrogen economy. Description of the innovation:

[0026] The work described here enables the artificial intelligence-assisted discovery, development, and validation of complex electrocatalysts for water electrolysis for scalable and sustainable hydrogen production. This intelligent system combines computational chemistry, machine learning algorithms, and experimental validation workflows into a useful methodology, enabling rapid and cost-effective catalyst development. This development significantly reduces the reliance on trial and error by leveraging AI-assisted material prediction and optimization strategies, thus enabling the rapid discovery of high-performance and cost-effective catalyst materials for hydrogen and oxygen evolution reactions (HER & OER).

[0027] According to the schematic representation in Fig.is an intelligent system for catalyst discovery and electrolysis optimization. System 100 includes several interconnected functions that enable data-driven materials discovery, simulation and validation, experimentation, and more. There is a schematic unit 102 for water electrolysis that models the basic electrolytic process in which an electric current is passed through water to produce hydrogen (H 2 ) and oxygen (O 2). Electrochemical reactions take place at two different electrodes: the HER (hydrogen evolution reaction) at the cathode and the OER (oxygen evolution reaction) at the anode, and the unit emphasizes the importance of introducing catalysts to enable the reactions to occur at lower energy inputs. A Catalyst Structure Analysis Unit 104 is designed to map atomic and molecular structures of potential catalyst materials (e.g., base metal alloys, 2D layered materials, transition metal-based nanostructures, etc.) and visualize / simulate their catalytic behavior. In this unit, we contribute to the characterization of these active sites, the charge distribution, and the surface morphology for the interaction zones involved in the catalytic conversion of water molecules.

[0028] Density functional theory (DFT) frameworks for quantum-scale material behavior are integrated into 106, a computational modeling and simulation unit. This unit performs in silico assessments of the thermodynamic stability, reaction energetics, and electronic properties of candidate materials. Finally, along with the simulation results, a machine learning (ML)-based inference engine is applied to cover a variety of electrochemical environments for extrapolating material performance.

[0029] A high-dimensional machine learning optimization module 108 enables parameter optimization based on simulated and experimental data. The module uses supervised and unsupervised learning models to generate significant material descriptors, thus improving prediction accuracy. The catalyst screening model is trained using a feedback loop, leveraging feature extraction, model training, validation, and prediction. This includes the performance of a synthesized catalyst under controlled laboratory conditions via an experimental validation and testing interface 110. This unit includes electrochemical test setups such as three-electrode cells and electrolyzers for measurements of overpotential, durability, and efficiency. It also enables real-time data acquisition, performance logging, and statistical comparisons with benchmark materials.

[0030] An energy integration unit 112 enables the seamless coupling of the electrolysis system with renewable energy sources, particularly photovoltaic systems. This unit guarantees the direct use of clean electricity for hydrogen production, thus improving the environmental efficiency of the process. It consists of an energy management subsystem that tracks energy flow and optimizes energy consumption during the electrolysis process. A data processing and control interface 114 controls the end-to-end communication between the computing subsystems 118, the experimental subsystems 122, and the machine learning subsystems 126. This interface enables the dynamic adjustment of simulation parameters, model inputs, and test protocols, while also providing a graphical overview of catalyst performance metrics and system-wide diagnostics.

[0031] Such storage, even with error-correction protocols, 3,4 would be carried out in a secure storage and logging unit 116, which records all simulation results, training data, and test results for all solid-state storage. It provides offline access to system results and enables the integration of cloud platforms for collaborative research and continuous learning. The combination of these elements creates a fully integrated, intelligent platform for AI-accelerated catalyst discovery and the optimization of hydrogen production systems. This breakthrough means that green hydrogen technologies will now undergo fundamental changes—particularly through significant reductions in production costs, real-time adaptations to renewable energy, and accelerated adoption of clean energy solutions across all sectors.The modular design of the platform allows its extension to other catalytic reactions beyond water electrolysis and demonstrates its broad applicability in sustainable chemistry and energy applications.

Claims

[1] An intelligent catalyst discovery and electrolysis optimization system for the production of green hydrogen, consisting of: • a schematic water electrolysis unit configured to simulate the electrolytic splitting of water into hydrogen and oxygen by modeling the hydrogen evolution reaction (HER) at a cathode and the oxygen evolution reaction (OER) at an anode and enabling analysis of catalyst effectiveness in reducing energy input; • a catalyst structure analysis unit configured to visualise and simulate atomic and molecular structures of candidate catalyst materials, including base metals, transition metal-based nanostructures and two-dimensional materials, and to evaluate charge distribution, surface morphology and active site behaviour; • a computational modeling and simulation unit using density functional theory (DFT) for quantum simulations to evaluate the thermodynamic stability, electronic structure, reaction energetics and adsorption energy of potential catalysts; • a machine learning optimization module trained on simulated and experimental data, using supervised and unsupervised learning models to determine optimal catalyst descriptors and predict catalytic activity, stability, and durability; • an experimental validation and test interface comprising electrochemical test environments including three-electrode cells and electrolyzers for performance benchmarking under realistic electrolysis conditions and capable of logging overvoltage, efficiency and degradation behavior; • an energy integration unit configured to interface with renewable energy systems, including photovoltaic systems, to dynamically power the electrolysis process and manage energy consumption through a power management subsystem; • a data processing and control interface to manage cross-system communication between simulation, experimental, and machine learning modules, enabling real-time parameter tuning and visualization of system diagnostics; and • a secure storage and logging unit adapted for storing simulation data, machine learning training sets and experimental results, with features for offline retrieval and integration with cloud platforms for collaborative research. [2] The system of claim 1, wherein the machine learning optimization module further comprises a reinforcement learning loop that updates model weights based on feedback from experimental test results to iteratively improve catalyst predictions. [3] The system of claim 1, wherein the catalyst structure analysis unit comprises a surface reconstruction module configured to simulate catalyst deterioration and morphological changes under electrolysis operating conditions. [4] The system of claim 1, wherein the computational modeling and simulation unit is configured to calculate adsorption energy profiles and identify transition states for HER and OER intermediates across multiple catalyst surfaces. [5] The system of claim 1, wherein the data processing and control interface provides a graphical user interface (GUI) that visualizes catalyst screening metrics, reaction path simulations, and predictive model accuracy in real time. [6] The system of claim 1, wherein the secure storage and logging unit supports version-controlled data management and cryptographic access protocols to ensure data integrity and security. [7] The system of claim 1, wherein the experimental validation and testing interface is further configured to perform long-term stability tests and accelerated stress tests to validate the durability of the catalyst under variable pH and temperature conditions. [8] The system of claim 1, wherein the energy integration unit comprises real-time energy forecasting algorithms to optimize the electrolysis schedule in response to variable renewable energy supply profiles. [9] The system of claim 1, wherein the machine learning optimization module integrates data from external materials databases to enhance chemical space exploration for catalyst development. [10] The system of claim 1, wherein the catalyst structure analysis unit is configured to generate feature vectors representing crystal lattice distortions, electronic density distributions, and defect states as input for machine learning.

Images