About this Abstract |
| Meeting |
2026 TMS Annual Meeting & Exhibition
|
| Symposium
|
Computational Thermodynamics and Kinetics
|
| Presentation Title |
Can Simulations Reproduce Experiments? |
| Author(s) |
Elizabeth A. Holm |
| On-Site Speaker (Planned) |
Elizabeth A. Holm |
| Abstract Scope |
Computer simulations rarely match experimental results exactly. This discrepancy is usually attributed to shortcomings in the physical model or input data used in the simulation. However, even after improving the model and/or inputs, many experimental results remain stubbornly difficult to replicate computationally. This leads us to ask what are the innate uncertainties present in computational models. We take polycrystalline grain growth as a model system. Using molecular dynamics simulations, we anneal multiple statistically and thermodynamically identical atomic configurations. We find that individual grain growth trajectories vary significantly from run to run. We trace these differences to discrete events that have an extreme and nonlinear dependence on initial conditions. Because we can never know experimental initial conditions with sufficient fidelity, we conclude that simulations can never reproduce experiments exactly. On the other hand, simulations can predict the distribution of possible outcomes for experimental results, providing physical insights even without precise prediction. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Computational Materials Science & Engineering, Modeling and Simulation, Other |