Shear Viscosity Calculation for Isotropic Molecules

User discussions
1

Title: Difficulty Achieving Converged Shear Viscosity for Bio-Based Asphalt System Using GROMACS

Hello everyone,

I am currently working on molecular dynamics simulations of a bio-based asphalt system using GROMACS, and I would like to ask for advice regarding shear viscosity calculations.

I initially attempted to calculate viscosity using the Green–Kubo approach (gmx energy -vis) from equilibrium MD simulations. However, despite running simulations up to 100 ns (and currently extending to 400 ns), the running viscosity integral does not appear to converge. The shear viscosity continuously increases without reaching a stable plateau.

To address this issue, I also started exploring non-equilibrium MD approaches. At the moment, I am testing:

  • deform

  • cos-acceleration

My current setup is:

  • NVT production simulation

  • Nose–Hoover thermostat

  • timestep = 1 fs

  • bio-based asphalt/asphaltene-like molecular system

  • highly viscous and heterogeneous material

For the deform method, I used:

deform = 0 0 0 1e-4 0 0

based on a published asphaltene viscosity study.

I am now considering switching to:

cos-acceleration = 0.05

since the literature I am following applied a shear rate of (1 \times 10^{-7} , \mathrm{fs}^{-1}) in their NEMD simulations.

I would appreciate advice on several points:

  1. For highly viscous asphalt/asphaltene systems, is Green–Kubo generally expected to be very difficult to converge? or is there any other method to calculate the density of my molecules?

  2. Between deform and cos-acceleration, which method is usually more reliable for viscosity calculations in such systems?

  3. How can I determine whether my chosen shear strength/acceleration is still within the linear regime?

  4. Are there recommended simulation lengths or averaging strategies for these kinds of materials?

  5. Does anyone have experience calculating viscosity for asphalt/asphaltene systems specifically in GROMACS?

Any suggestions or references would be greatly appreciated.

Thank you very much.

2

Green-Kubo converges very slowly for high viscosity liquids.

Shear deformation is preferred over cos-acceleration, as there is a constant shear over the whole system. Use GROMACS version 2026, or at least 2025 for this.

For any non-equilibrium method you need to be in the linear regime. Establishing this requires at least two simulations at different shear rates.

I have no experience with asfalt, but this sounds extremely viscous to me, so might be difficult to converge.

3

Thank you for your reply and explanation.

When you mentioned using at least two simulations at different shear rates to establish the linear regime, do you mean:

  1. running two separate simulations with different shear rates and comparing the results afterward, or

  2. applying two shear rates within the same .mdp setup?

Also, since my asphalt system converges very slowly, would you recommend starting with a relatively high shear rate first and then reducing it gradually to check whether the viscosity remains consistent?

Thank you again for your help.

4

You need to run at least two independent simulations with different shear rates to check if the viscosity matches. If not, then you need to run at a lower shear rate. For very viscous systems there is the risk that you cannot reach a low enough shear rate within reasonable simulation times.