Accelerated Discovery of Seawater HER Catalysts via Artificial Intelligence-Guided Electrodeposition of Transition-Metal Alloys

Hydrogen has the potential to replace fossil fuels in hard-to-decarbonize sectors, thereby supporting global climate goals. Molecular hydrogen can be produced through water electrolysis (or water splitting), and it is considered a clean fuel when generated using renewable electricity via the hydrogen evolution reaction (HER).

Freshwater is scarce globally, but seawater makes up ~97% of Earth’s water. Using seawater instead of purified water removes a major barrier to scaling green hydrogen production. Seawater electrolysis could thus provide essentially unlimited hydrogen production capacity. Yet, seawater is chemically challenging because high chloride concentration causes corrosion and chlorine evolution, while biofouling and impurities reduce efficiency. Hence, developing catalysts that suppress chlorine evolution reaction (CER) while enhancing HER is an active research frontier that advances materials science more broadly.

In this project, binary and ternary alloys of the e.g. Ni–Fe–Cr and Ni–Fe–Mo systems will be prepared by electrodeposition, and their performance toward the HER in seawater will be evaluated. Affordable elements will be considered throughout the project regarding the composition of the electrocatalysts.

To reduce time and resource consumption, artificial intelligence (AI)-driven optimization strategies will be employed to accelerate binary and ternary alloys discovery. Specifically, Bayesian optimization will be used to navigate the high-dimensional compositional and electrochemical parameter space, enabling efficient identification of promising alloy compositions and deposition conditions, with surrogate models providing both performance predictions and uncertainty estimates. Graph neural networks (e.g., Crystal Graph Convolutional Neural Network, Atomistic Line Graph Neural Network) and materials-aware machine-learning models (e.g., Random Forest Regressors, Gradient Boosting Machines such as XGBoost or LightGBM) will be trained to link synthesis parameters and structural characteristics with key electrocatalytic descriptors such as Tafel slope, overpotential, and stability in chloride-rich environments. In parallel, active learning loops will iteratively integrate experimental results into the models, continually refining predictive accuracy and guiding subsequent experiments toward the most informative regions. Explainable AI tools (e.g., SHAP, symbolic regression) will help uncover the fundamental physicochemical factors governing HER selectivity over CER. Furthermore, deep-learning models (e.g., Vision Transformers or U-Net-style architectures) will process electron-microscopy images to automatically extract structural descriptors such as grain size, alloy homogeneity, defect density, porosity, and crystallinity. Integrating these image-derived features into the predictive models will enable quantitative correlations between electrodeposition conditions, nanoscale structure, and HER/CER performance. The most promising catalyst candidates identified computationally will be synthesized and validated at laboratory scale.

Overall, the project aims to advance efficient and cost-competitive seawater electrolysis by combining inexpensive transition-metal alloys, scalable electrodeposition, and state-of-the-art AI methodologies for accelerated catalyst discovery.

Candidates should have:

• Bachelor / Master in Nanoscience & Nanotechnology, Materials Science, Physics or related fields; extra coursework in data science or artificial intelligence is a plus.
• Analytical thinking & problem-solving, ability to interpret experimental data critically.
• Scientific writing and communication skills in English.
• Teamwork and interdisciplinary mindset (materials science, physical chemistry, energy research).
• Knowledge of Physical Chemistry is desirable.
• Knowledge in Python programming language or PyTorch / Tensorflow certification.
• Preliminary exposure to machine/deep learning, statistical modelling or generative AI.

The AI Lab, part of HP’s renowned Technology and Innovation Office, leads strategic research and development in Generative AI to bring state-of-the-art capabilities into HP’s new form factors portfolio. With teams located in Barcelona (Spain), Palo Alto (USA), and Bangalore (India), the Barcelona group focuses on operationalizing the full lifecycle of Generative AI models and developing energy efficient on device solutions that address top user needs while enabling new, transformative user experiences. Our research spans natural language processing, computer vision, and voice-based AI systems, with a strong emphasis on model fine-tuning, quantization (compression / distillation), and evaluation of large language, vision, and audio models.

The Gnm3 group focuses part of its research on the design, synthesis, and characterization of advanced materials with tailored properties for cutting-edge engineering applications. We use electrochemical methods to produce advanced material architectures with high surface-area-to-volume-ratios and compositions with reduced amounts of noble metals. These materials are tested as electrocatalysts for the hydrogen evolution reaction (HER) and, more recently, implemented in proton exchange membrane fuel cells. Sustainability and energy efficiency are guiding principles of our work, shaping both the development of materials and their envisioned applications.