Locking of dendrites in potassium metal batteries: A potassiophilic 3D host discovered with machine learning screening

Abstract

The metallic potassium (K) is a promising anode for high-energy K-batteries, offering a high theoretical capacity (687 mAh g(-1)) and low reduction potential (-2.9 V vs. standard hydrogen electrode), but its hostless nature promotes dendrite-driven failure. Here, we report a data-driven discovery and experimental validation of a potassiophilic CuO-covered 3D-Cu current collector (CuO@3D-Cu) that suppresses dendrite formation in potassium metal batteries. A screening workflow combining a crystal graph convolutional neural network and gradient-boosted decision trees to prioritize candidates, validated the top hit with density functional theory adsorption calculations (E-ads = -4.317 eV on CuO), and COMSOL Multiphysics electrochemical-transport modeling quantified mesoscale ion-flux and current-density distributions. Models predict the 3D scaffold together with a potassiophilic surface homogenizes K+ flux and suppresses local current hotspots. Experimentally, CuO@3D-Cu exhibits no measurable nucleation overpotential versus similar to 50 mV for bare 3D-Cu, sustains symmetric cells cycling over 2000 h at a current density of 4 mA cm(-2), and enable potassium-sulfur cells with an initial specific capacity 600 mAh g(-1) and capacity retention of 68.8 % over 100 cycles. Combined modeling and experiments demonstrate that dendrite formation is governed by coupled, nonlinear electrochemical-transport instabilities: concentration-dependent reaction kinetics and local current-density amplification produce threshold behavior in nucleation and growth that is suppressed by the CuO@3D-Cu. The design combines (i) a conductive, porous 3D architecture for uniform current distribution and volumetric accommodation and (ii) a chemisorptive CuO surface to lower nucleation barriers, providing a practical route to stable, scalable K-metal anodes.