# Altered morphology and diffusivity of water confined in MXenes: Machine learning–accelerated computations combined with experiments

## TL;DR

Researchers used machine learning and experiments to study how water behaves when confined between MXene layers, revealing insights for energy storage and transport.

## Contribution

A machine learning-accelerated approach combined with experiments to model water diffusivity in MXene interfaces.

## Key findings

- Water confined in MXene layers shows layer-dependent staging characteristics.
- Water polarization and electrostatic potential are influenced by MXene surface groups.
- A linear combination of exponential model describes water diffusivity based on interfacial properties.

## Abstract

Nanoconfined water exhibits unique properties compared to bulk water due to limited quantities, frustrated hydrogen bonding, and surface interactions, which are fundamental for energy storage and transport applications. We integrate machine learning–accelerated ab initio molecular dynamics with x-ray diffraction (XRD) and inelastic neutron scattering (INS) to systematically analyze the thermodynamic and dynamic behavior of water confined between functionalized (-F, -O, and -OH) two-dimensional (2D) Ti3C2Tx MXene layers. As water intercalates between layers, the interlayer spacing exhibits layer-dependent staging characteristics. The water polarization can be flipped by the count and morphology of intercalated molecules interacting with MXene surface groups, resulting in varying electrostatic potential profiles. On the basis of interfacial electrostatic potential, hydrogen bond lifetime, and molecular orientation, we establish a linear combination of exponential model describing water diffusivity. These computational insights align well with experimental x-ray and neutron measurements, suggesting strategies for tuning water morphology and transport by tailoring MXene surface chemistry and water content for electrochemical energy storage and nanofluidic applications.

Machine learning-accelerated ab initio molecular dynamics address MXene-water interfaces to inform a diffusion model.