Assessing the accuracy of BAMBOO’s viscosity predictions for carbonate solvents

Determine whether the performance of the BAMBOO machine learning force field in predicting viscosities of carbonate solvents—such as vinylene carbonate and fluoroethylene carbonate—is being accurately assessed when benchmarked against currently available experimental viscosity measurements that show substantial inter-study variability.

Background

The paper compares BAMBOO’s viscosity predictions for pure solvents to experimental measurements, noting good agreement for linear carbonates but discrepancies for cyclic carbonates, particularly vinylene carbonate (VC). The authors highlight that literature-reported viscosities for the same solvent can differ markedly—for example, fluoroethylene carbonate (FEC) at 40°C ranges from 2.24 to 4.1 cP—making validation nontrivial.

Given this variability, the authors explicitly state that it remains uncertain whether their model’s viscosity performance on some carbonates is accurately assessed using the available experimental results, raising an open question about evaluation reliability in the presence of inconsistent reference data.