[Paper Review] From Accurate Quantum Chemistry to Converged Thermodynamics for Ion Pairing in Solution

Niamh O'Neill, Benjamin X. Shi+7 more|arXiv (Cornell University)|Mar 6, 2026

Machine Learning in Materials Science0 citations

TL;DR

The paper develops a Delta-CCSD(T) machine-learned potential to compute CaCO3 ion pairing in water at CCSD(T) level, achieving quantitative agreement with experiment for free energy, enthalpy, and entropy by combining MP2 baselines, gas-phase corrections, and enhanced sampling.

ABSTRACT

Quantitative prediction of thermodynamic properties in solution is essential for translating atomistic simulations into reliable chemical insight. As an exemplar system, the behaviour of CaCO$_3$ in water has been widely studied to understand its mineralization in seawater, with potential implications for carbon-capture strategies. However, making accurate computational predictions has been a long-standing challenge, requiring both highly accurate electronic structure methods and extensive statistical sampling. Here, we combine advances in machine learning and electronic structure theory to fully resolve the ion pairing free energy of CaCO$_3$ with explicit solvation. We show that achieving quantitative agreement with experiment requires going beyond the standard density functional theory up to the "gold-standard" coupled cluster theory with single, double, and perturbative triple excitations [CCSD(T)]. We generate a set of systematically improvable models, enabling reliable insights into the initial association mechanism of Ca and CO$_3$ ions prior to nucleation while fully quantifying enthalpic and entropic effects. Our results demonstrate that CCSD(T)-level thermodynamic predictions of complex aqueous systems can now be routinely achieved.

Motivation & Objective

• Aim to predict thermodynamics of CaCO3 ion pairing in water with CCSD(T)-level accuracy.
• Develop Delta-learning MLPs to correct MP2 to CCSD(T) in condensed phase.
• Ensure converged potential energy surface and robust sampling for enthalpy and entropy.
• Provide mechanistic insights into ion pairing differences across theory levels.
• Demonstrate that CCSD(T)-level predictions agree with experimental data without empirical fitting.

Proposed method

• Train MP2-based periodic ML potentials as baselines using CP2K implementations for energies and forces.
• Construct Delta-MLP to learn MP2-to-CCSD(T) corrections from gas-phase calcium–carbonate clusters.
• Combine CCSD(T)-MLP with enhanced sampling (OPES) to compute ion pair free energy, enthalpy, and entropy.
• Iteratively refine the periodic and Delta datasets to converge the PMF for ion pairing.
• Use the Symmetrix library to accelerate ML-PES based simulations and enable microsecond-scale cWFT-like sampling.

Experimental results

Research questions

• RQ1 Can CCSD(T)-level accuracy be achieved in explicit solvent for ion pairing using Delta-learning ML potentials?
• RQ2 Do CCSD(T)-level thermodynamics (free energy, enthalpy, entropy) agree with experimental values for CaCO3 ion pairing in water?
• RQ3 How do ion pairing pathways and hydration structures differ across CCSD(T), MP2, RPA, and DFT models?
• RQ4 Is explicit solvent treatment at high level of theory essential to reproduce experimental observations?

Key findings

• CCSD(T)-level ML potentials reproduce experimental ion pair free energy at 300 K for CaCO3 in water.
• Only CCSD(T) consistently matches experiment for free energy, enthalpy, and entropy without fortuitous error cancellation.
• DFT functionals revPBE-D3 and revPBE0-D3 misrepresent CIP/SShIP stability and enthalpy/entropy trends.
• MP2 reproduces CCSD(T) qualitative trends in PMFs, more closely than some DFT functionals.
• CCSD(T)-level PES and solvent treatment reveal distinct ion pairing mechanisms not captured by lower-level theories.
• The approach enables converged thermodynamics for complex aqueous systems at cWFT accuracy and is extendable to other solution-based processes.