Two chemically identical simulated nanoparticles acquire different charges when a polymer settles on one of them. A 23 October preprint reproduces that behaviour with CR-ML, a graph network predicting ionisation probabilities from the changing molecular arrangement.[1]
The physical effect predates this AI method. In their 2021 work, Tine Curk and Erik Luijten showed how allowing surface charges to change could alter the structures that particles assemble into. Surface groups can release or bind ions, and their equilibrium depends on both the surrounding solution and other charged objects nearby. Charge therefore participates in a feedback loop: it affects the forces shaping an assembly, while the assembly changes the electrical environment that determines charge.[2]
This differs from simply redistributing charge across the surface of an object whose total charge stays fixed. Charge regulation allows the amount carried by an individual particle to change through exchange with its surroundings. Chemically identical particles can consequently behave differently when they occupy different local environments.[2]
The conventional LAMMPS procedure tests changes to individual sites. An acid site can switch between neutral and negative; a base site between neutral and positive. The software inserts a compensating ion to preserve the simulation’s electrical neutrality. Other moves exchange pairs of ions with a reservoir representing the solution. Repeating these statistical trials lets the simulation sample different combinations of charge states.[3]
CR-ML replaces hundreds of these trials with one network pass after particle positions evolve. It then samples charges from the predicted probabilities.[1]
In separate polymer-chain tests, CPU charge updates took 7–90 milliseconds versus 430; particle-motion calculations were unchanged.[1]
The surrogate overestimates polymer bridging near the transition between shared attachment and adsorption onto one particle. Mean nanoparticle charges nevertheless track the Monte Carlo reference within 2.5 elementary charges. At the largest tested separation in 1µM salt, the polymer-bearing particle carries about 38 positive charges and the bare one 14, from 256 sites apiece.[1]
Sources & further reading
- Machine Learning Accelerated Simulation of Charge Regulation Effects arXiv · 23 Oct 2026 · Accessed 29 Oct 2026
- Charge Regulation Effects in Nanoparticle Self-Assembly Physical Review Letters · 31 Mar 2021 · Accessed 29 Oct 2026
- fix charge/regulation command LAMMPS documentation · Accessed 29 Oct 2026