Iterative Construction of n-Dimensional Navier-Stokes Solutions with Non-Gradient Convection
Abstract
Exact periodic solutions of the incompressible Navier-Stokes equations arise when the convective field is a pure gradient and can be absorbed into the pressure. We construct solutions outside this class, in arbitrary spatial dimension, by an iteration in which each step solves a linear diffusion problem forced by the transverse part of the preceding convective field. The construction is dimension-independent, and for a cyclic family of initial fields the first two iterates are available in closed form. Carried out numerically in three dimensions, the iteration converges: at Re = 10 successive iterates contract with an observed factor 0.43 and the limit is a fixed point of the discretised map to a relative defect of 8.2 x 10^-5. The observed factor exceeds unity by Re = 30, which identifies the practical range of the iteration rather than a proved convergence boundary. We also show what goes wrong when the projection is omitted, as it was in an earlier iteration of the author's: the iterates cease to be divergence-free and an apparent growth appears that belongs to the expansion and not to the flow. For an n-dimensional generalisation of the initial field the fraction of the convective field not absorbed by pressure is exactly 2*sqrt(2)/3 in every dimension n >= 3, and the longitudinal fraction exactly 1/3. The divergence of each returned source is a sharp diagnostic, failing by sixteen orders of magnitude at the step where the projection is omitted; the cell kinetic energy is a weaker one, detecting gross failure only.
Related
- Transitioning to the ultimate regime of convection in three-dimensional direct numerical simulations
- Growth of helicity in salt-finger convection in the two-dimensional three-component limit
- Validation of quasi-two-dimensional model of convection in a transverse magnetic field
- A method for measuring and analyzing four-dimensional flow fields in viscous Rayleigh-Bénard convection
- Gradient-Based Construction of Collisionless Steady-State Guiding-Center Distributions in Tokamaks and Stellarators
BibTeX
@misc{thambynayagam2026iterative,
title = {{Iterative Construction of n-Dimensional Navier-Stokes Solutions with Non-Gradient Convection}},
author = {R. K. Michael Thambynayagam},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2609.22372},
eprint = {2609.22372},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.22372},
}
RIS
TY - UNPB
TI - Iterative Construction of n-Dimensional Navier-Stokes Solutions with Non-Gradient Convection
AU - R. K. Michael Thambynayagam
PY - 2026
DA - 2026/10/26
PB - arXiv
DO - 10.48550/arxiv.2609.22372
UR - https://arxiv.org/abs/2609.22372
AB - Exact periodic solutions of the incompressible Navier-Stokes equations arise when the convective field is a pure gradient and can be absorbed into the pressure. We construct solutions outside this class, in arbitrary spatial dimension, by an iteration in which each step solves a linear diffusion problem forced by the transverse part of the preceding convective field. The construction is dimension-independent, and for a cyclic family of initial fields the first two iterates are available in closed form. Carried out numerically in three dimensions, the iteration converges: at Re = 10 successive iterates contract with an observed factor 0.43 and the limit is a fixed point of the discretised map to a relative defect of 8.2 x 10^-5. The observed factor exceeds unity by Re = 30, which identifies the practical range of the iteration rather than a proved convergence boundary. We also show what goes wrong when the projection is omitted, as it was in an earlier iteration of the author's: the iterates cease to be divergence-free and an apparent growth appears that belongs to the expansion and not to the flow. For an n-dimensional generalisation of the initial field the fraction of the convective field not absorbed by pressure is exactly 2*sqrt(2)/3 in every dimension n >= 3, and the longitudinal fraction exactly 1/3. The divergence of each returned source is a sharp diagnostic, failing by sixteen orders of magnitude at the step where the projection is omitted; the cell kinetic energy is a weaker one, detecting gross failure only.
ER -