What is Unique to Computer Experiments?

• May look very closely ($\vec{r}_i(t)$, $\vec{v}_i(t)$) at the behavior of any atomic subset of the system. In principle, any function of the atomic positions and velocities, whether time-averaged or instantaneous, is computable.

• May modify potential function $V(\vec{R})$ arbitrarily. Examination of what nature does not do may provide a new understanding for what it does do. For example, we have investigated the contribution of torsional transitions to the anharmonicity of protein dynamics by comparing simulations of MbCO dynamics performed with and without infinitely high barriers that prohibit these transitions. Our conclusion: Dihedral transitions account for nearly all the motional anharmonicity of dried MbCO but for less than half of the motional anharmonicity of hydrated MbCO [18].

• May mutate structures or environments slowly (Free Energy Perturbation Theory) and approximate differences in free-energy differences, $\Delta \Delta G$. By taking differences in $\Delta G$, the states of zero energy are consistently defined and errors due to approximations in the simulation protocol tend to subtract out. See `Classical vs. Quantum Mechanics: The Harmonic Oscillator in One Dimension' above.