The walls make the difference
A quasi-two-dimensional, or Q2D, model matched the results of a fully three-dimensional calculation in the tested cases when the spanwise walls were electrically insulated. When those walls were conducting, the two models were noticeably different, although both showed strong magnetic damping at large Hartmann numbers.
The study asks whether that reduced Q2D calculation can reproduce the results of a full three-dimensional solution for the same laterally heated magnetoconvection problem. In other words, it tests whether a lower-dimensional numerical description preserves the outputs of the full calculation in the cases examined.
What the test compared
The test set included square-cross-section boxes and horizontally elongated boxes with a length-to-height ratio of 8. The width-to-height ratio ranged from 1 to 10, with comparisons between insulated and conducting spanwise walls.
Both formulations used the same finite-volume schemes. Steady Q2D states used Newton iteration, while steady 3D flows used semi-implicit time integration with a three-time-layer fractional time-step scheme and TPF/TPT direct solvers.
Those results are conditional on a particular setup. The magnetic field must be spanwise, reflection symmetry must hold across that direction, and the spanwise boundaries must be perfectly electrically and thermally insulated. In that regime, Q2D is governed by a combined parameter equal to twice the Hartmann number divided by the box width-to-height ratio, rather than the Hartmann number used by the full three-dimensional formulation.
Insulation held the models close
With width ratio 1 and aspect ratios 1 and 8, the 3D and Q2D results compared very well when the spanwise boundaries were electrically insulated, regardless of the electrical conditions on the other four boundaries. The result was specific to those tested configurations.
In the insulated square-cavity width sweep, averaged characteristics stayed close to Q2D predictions at a fixed value of 200 for the governing parameter. They did not, however, show a clearly asymptotic trend as the width increased.
Conducting walls changed the outcome
Conducting spanwise walls changed the picture. The two models were noticeably different, while their common behavior at large Hartmann numbers was strong flow damping toward the purely thermally conducting state. The study describes agreement under that damping as qualitative rather than quantitative.
In conducting-spanwise width sweeps, the four monitored characteristics approached asymptotic values smaller than the insulated-wall values. At the high fixed value of the governing parameter, 2000, all three Nusselt numbers, the study's heat-transfer figures, became 0.125, the purely conducting value.
Numerical resolution was another fault line. At very large Hartmann numbers, the required grid stretching reached 16; weaker stretching produced an incorrect decay of all four characteristics with increasing width ratio.
Oscillations added a warning
The reduced model also reproduced the onset of oscillatory flow approximately. In the square-cavity case with aspect ratio 1, Prandtl number 0.054 and Hartmann number 100, the flow became oscillatory at width ratios of 2 or more. The Q2D critical Rayleigh number was reported as 1.08 million, close to the fully 3D results for widths at or above 2.
For selected oscillatory configurations, the dimensionless Q2D frequency was about 0.50, compared with 0.521 in the full 3D calculation at width ratio 2 and 0.518 at width ratio 5.
But matching the threshold and frequency did not guarantee identical fully developed oscillatory states. Three-dimensional oscillation amplitudes varied with spanwise position more strongly than the steady fields, and the authors expected the Q2D and 3D results to diverge in regimes beyond the oscillation onset.
What remains untested
The paper's conclusion is deliberately limited. It emphasizes that the result rests on particular examples and that Q2D still requires validation even when the stated conditions hold. An extension for perfectly conducting spanwise boundaries remains future work.
Two important gaps remain. The dependence of the critical Rayleigh number on width ratio was not established, and every computation in the horizontally elongated boxes produced a single convective circulation. Multicellular states were not observed in those simulations.
For numerical users, the boundary condition is the practical dividing line in this study: a spanwise magnetic field, reflection symmetry and perfectly insulated spanwise boundaries accompanied close Q2D-3D agreement in the tested cases. Conducting spanwise walls instead produced noticeable model differences, and high-field calculations required careful control of grid stretching.
Paper data and sources
Original title: Validation of quasi-two-dimensional model of convection in a transverse magnetic field
Authors: Alexander Gelfgat
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-28
DOI: Not available
Original paper · Full text