2024 Theses Doctoral

Towards an Atomic-Scale Understanding of Membrane-Coated Electrocatalysts for Hydrogen Production

Qu, Jianzhou

As the demand for climate-neutral energy increases, green hydrogen is becoming an important option as a fuel for a future energy economy. Although green hydrogen can be produced via water electrolysis, this process is expensive compared to fossil hydrogen sources, also because conventional proton exchange membrane (PEM) electrolyzers require highly purified water and rely on platinum-group metals (PGMs) as catalysts for the hydrogen evolution reaction (HER).

To overcome these limitations, one promising direction is to apply a protective coating to the electrocatalyst surface, i.e., developing membrane-coated electrocatalysts (MCECs). The catalytic reaction in MCECs takes place at the buried membrane-catalyst interface that is challenging to characterize with experimental techniques, so that fundamental questions remain unanswered. In this thesis, we employ first-principles calculations and machine-learning methods to investigate the chemical structure at the buried interface of MCECs and the impact of membrane-coating on the HER reaction mechanism.

Using density-functional theory (DFT) calculations, we find that the interaction of SiO2 membranes and Pt surfaces is environment-dependent. By generalizing the concept of Pourbaix diagrams to electrochemical solid-solid interfaces, we establish which bonds are formed between the SiO2 membrane and the Pt(111) surface in aqueous electrolytes with different pH values and subjected to different electrode potentials. We find that the membrane termination changes as a function of the pH and the potential, which affects the adhesion strength and controls the Pt surface area that is accessible for reactant species. The charge transfer between the Pt surface and the SiO₂ membrane is also pH- and potential-dependent and results in changes of the Pt surface d-band states, which are known to correlate with catalytic activity.

Membrane coatings affect the catalytic reaction mechanism. Comparing hydrogen adsorption on bare and coated Pt electrodes, we find that SiO₂ membranes reduce the magnitude of the average hydrogen adsorption energy, decrease the number of binding sites for hydrogen, and alter the adsorption site preferences. On the other hand, the closely contacted interfaces create a confined environment that facilitates the collision of reactive atoms and may decrease activation energies. We find that SiO2 membranes can react with protons to form silanol groups at the interface, which can participate in the HER as proton sources in a Silanol-Heyrovsky mechanism. However, we also find that the Volmer-Tafel mechanism is preferred at the buried interface.

To find a new candidate for PGM-free HER electrocatalysts, we extended our work to other transition metal substrates, i.e., SiO₂/TM (TM = Au, Ag, Cu, Rh). The interface Pourbaix diagrams for these systems reveal that the stable chemical configurations can vary widely with the transition metal species. In particular, Si –OH groups are predicted to be stable for several of the metal species, indicating that the membrane might participate in the HER over these catalyst systems. In addition to the catalytic activity, the transport properties of the silica membrane are also crucial for the performance of a MCEC, but DFT is computationally too demanding for realistic transport simulations. As an alternative, we commenced the construction of machine-learning potential models to enable future transport simulations. We performed molecular dynamics simulations at a wide range of temperatures (500 K to 4000 K) to generate a comprehensive library of crystalline and amorphous SiO2 bulk structures. Using this database, we developed a methodology for the down-selection of data subsets with maximal information content, leveraging a recently developed global structural descriptor (GSDs) method for the representation of atomic structures. The downselection approach is general, can be applied to reduce redundancy in structure data sets, and has applications in active learning.

The research projects conducted as part of this thesis advance the understanding of MCEC for HER and open a new direction for exploring non-PGM MCECs and modeling transport through silica membranes with machine-learning potentials.