Make properly use of timescale separation

[mvassaux]

At the moment, atomic systems are simulated with unrealistically accelerated dynamics, assuming that slower dynamics would not be beneficial, as the system is continuously in equilibrium.

This assumption is too strong...

---

Let's consider the following example:
During a microscale experiment, a homogeneous volume of $1 \mu m^3$ of polymer material is submitted to a tensile strain of $\epsilon=0.1$ at a strain rate of $\dot{\epsilon}=10^{-2}s^{-1}$, the experiment, therefore, lasts $0.1s$.

How can we predict the material's response without simulating its dynamics during $0.1s$?

Our current assumption is that the material is not identically strain-rate dependent at each scale. The material can be assumed to relax faster at the atomic scale than at the continuum scale.

However, isn't this equivalent to assume that slow mechanisms, that leads to continuum viscoelastic effects, such as creep, are not caused by atomic dynamics? This assumption might be correct, as their origin can be captured at the molecule/chain length-scale. In turn, that scale needs to be considered as well.

---

Therefore, the more general question is: how can it be ensured that relevant mechanisms of a certain timescale are not missed by not simulating the dynamics at that given scale long enough?