Richard J.A.M. Stevens

Professor at the University of Twente

Wind energy · Turbulence · Environmental flows

Publication 48 · Thermal convection

Comparison of computational codes for direct numerical simulations of turbulent Rayleigh-Bénard convection

G.L. Kooij, M.A. Botchev, E.M.A. Frederix, B.J. Geurts, S. Horn, D. Lohse, E.P. van der Poel, O. Shishkina, R.J.A.M. Stevens, R. Verzicco, Computers & Fluids 166, 1-8 (2018).

Read the article Open-access version

Main finding

Different simulation codes can agree on global heat transport while under-resolution still distorts local temperature structures. The cross-code comparison validates global observables, not earlier local boundary-layer claims or a hardware-independent performance ranking.

Temperature fields from two simulation codes at three resolutions, with a magnified detail
How to read the figure. Temperature at a fixed height from two independent codes, one per row, at increasing resolution, with the rightmost column magnifying the boxed region. The small-scale structure differs visibly between the codes at coarser resolutions even where the global heat transport agrees. Open the full-resolution figure. Figure 5. G.L. Kooij et al. (2018). No separate licence is stated here; consult the original publication and credited source before reuse.

Why this matters

Different codes can agree on heat flow while still drawing visibly wrong small-scale temperature patterns when under-resolved.

Research context

Shishkina et al. (2010) provides the GL/PB-based boundary-layer resolution criterion used to design the meshes. This benchmark supplies much stronger global refinement evidence, but does not isolate or universally validate the earlier lower-bound formula. Prandtl-Blasius temperature and velocity boundary layer profiles in turbulent… (read the finding) and Horizontal structures of velocity and temperature boundary layers in… (read the finding) use a different 2D in-house finite-difference lineage and focus on local conditional profiles. The present comparison tests global Nu across 3D code families; it does not reproduce their dynamic-profile observables.

View in the complete publication list Related thermal convection research