pith. sign in

arxiv: 2603.08892 · v2 · submitted 2026-03-09 · ⚛️ physics.flu-dyn · physics.comp-ph

Aliasing and phase shifting in pseudo-spectral simulations of the incompressible Navier-Stokes equations

Clovis Lambert , Jason Reneuve , Pierre Augier This is my paper

Pith reviewed 2026-05-15 13:16 UTC · model grok-4.3

classification ⚛️ physics.flu-dyn physics.comp-ph
keywords phase-shiftingdealiasingpseudo-spectral methodsNavier-Stokes equationsaliasing errorsturbulence simulationdirect numerical simulationcomputational fluid dynamics
0
0 comments X

The pith

Phase-shifting dealiasing replaces the 2/3 truncation rule in pseudo-spectral Navier-Stokes simulations and delivers up to threefold speedups at comparable accuracy.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper establishes that phase-shifting methods cancel aliasing errors that arise from quadratic nonlinearities in discrete Fourier space by evaluating those terms on grids shifted by small fractions of the mesh spacing and combining the results. This cancellation is exact for some time-stepping schemes and approximate for others, allowing the same maximum physical wavenumber to be retained on a coarser grid than the conventional 2/3 dealiasing rule requires. Because the standard truncation rule can consume up to 80 percent of runtime in three-dimensional runs, the smaller grid translates directly into lower memory use and faster wall-clock time. Tests on the transition of Taylor-Green vortices and on forced homogeneous isotropic turbulence at Reynolds number based on the Taylor microscale of 200 show speedups reaching a factor of three relative to a fourth-order Runge-Kutta scheme with 2/3 truncation, while the error in energy spectra and other statistics remains small. The algorithms, including extensions to forced flows, are supplied as open-source code so that the technique can be adopted immediately.

Core claim

Phase-shifting methods cancel aliasing contributions from quadratic nonlinearities in discrete Fourier space by evaluating nonlinear terms on shifted grids. Depending on the time-stepping scheme, the cancellation is exact or approximate. When applied to the incompressible Navier-Stokes equations, these methods allow the same physical resolution to be achieved on a coarser numerical grid than the conventional 2/3 truncation rule, producing speedups of up to a factor of three at comparable accuracy on the transition of Taylor-Green vortices and forced isotropic turbulence at Re_lambda = 200.

What carries the argument

Phase-shifting dealiasing, in which the nonlinear terms are computed on grids shifted by fractions of the mesh spacing and then linearly combined to remove aliasing errors.

If this is right

  • The same maximum wavenumber can be retained while using a grid whose linear size is only two-thirds that demanded by conventional dealiasing.
  • Memory footprint and operation count drop proportionally, producing measured speedups of up to three times relative to RK4 with 2/3 truncation.
  • The cancellation works with both exact and approximate time integrators, including extensions to forced flows.
  • An open-source implementation removes the previous barrier to adoption in existing pseudo-spectral codes.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Higher-Reynolds-number direct numerical simulations become feasible on the same hardware because each grid point now carries a larger fraction of the resolved spectrum.
  • The shift-and-combine procedure may be combined with other cost-reduction techniques such as adaptive mesh refinement or mixed-precision arithmetic.
  • The interaction of phase-shifting with different three-dimensional truncation geometries remains to be mapped systematically beyond the cases tested.

Load-bearing premise

Phase-shifting exactly or approximately cancels the aliasing contributions that arise from quadratic nonlinearities when the nonlinear terms are evaluated on appropriately shifted grids.

What would settle it

A direct comparison of energy spectra or transition times between a phase-shifting run and a standard 2/3-truncation run at the identical maximum wavenumber that shows large, systematic discrepancies.

Figures

Figures reproduced from arXiv: 2603.08892 by Clovis Lambert, Jason Reneuve, Pierre Augier.

Figure 1
Figure 1. Figure 1: FIG. 1. Illustration of aliasing generated by nonlinear terms in one-dimensional spectral space, [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2. Common truncations used without phase shifting. Modes are set to zero in the gray [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3. Numerical and exact solutions of Eq. ( [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4. Temporal evolution of (a) total kinetic energy [PITH_FULL_IMAGE:figures/full_fig_p010_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5. Spectral statistics averaged between [PITH_FULL_IMAGE:figures/full_fig_p010_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: FIG. 6. Cross-section of the velocity field component [PITH_FULL_IMAGE:figures/full_fig_p011_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: FIG. 7. Maximum error (max [PITH_FULL_IMAGE:figures/full_fig_p015_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: FIG. 8. (a) 3D visualisation of regions removed only by double aliases truncation, as in Rogallo [PITH_FULL_IMAGE:figures/full_fig_p017_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: FIG. 9. Same as Fig [PITH_FULL_IMAGE:figures/full_fig_p018_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: FIG. 10. Distribution of computation time across different tasks for a simulation with RK4 at [PITH_FULL_IMAGE:figures/full_fig_p019_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: FIG. 11. Computation time multiplied by the number of cores versus grid size [PITH_FULL_IMAGE:figures/full_fig_p020_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: FIG. 12. Same as Fig [PITH_FULL_IMAGE:figures/full_fig_p020_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: FIG. 13. Same as Fig [PITH_FULL_IMAGE:figures/full_fig_p021_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: FIG. 14. Same as Fig [PITH_FULL_IMAGE:figures/full_fig_p022_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: FIG. 15. Spectral error in % (see Eq. ( [PITH_FULL_IMAGE:figures/full_fig_p022_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: FIG. 16. Distribution of computation time across different tasks (Fourier transforms, operators, [PITH_FULL_IMAGE:figures/full_fig_p023_16.png] view at source ↗
Figure 17
Figure 17. Figure 17: FIG. 17. Speedup (elapsed time of the RK4 simulation at [PITH_FULL_IMAGE:figures/full_fig_p024_17.png] view at source ↗
Figure 18
Figure 18. Figure 18: FIG. 18. Same as Fig [PITH_FULL_IMAGE:figures/full_fig_p024_18.png] view at source ↗
Figure 19
Figure 19. Figure 19: FIG. 19. Same as Fig [PITH_FULL_IMAGE:figures/full_fig_p025_19.png] view at source ↗
Figure 20
Figure 20. Figure 20: FIG. 20. Spectral error in % relative to the reference simulation (RK4 with [PITH_FULL_IMAGE:figures/full_fig_p026_20.png] view at source ↗
Figure 21
Figure 21. Figure 21: FIG. 21. Spectral error in % relative to the reference simulation (RK4 with [PITH_FULL_IMAGE:figures/full_fig_p027_21.png] view at source ↗
Figure 22
Figure 22. Figure 22: FIG. 22. Same as Fig [PITH_FULL_IMAGE:figures/full_fig_p029_22.png] view at source ↗
Figure 23
Figure 23. Figure 23: shows (a) the energy spectra and (b) the energy fluxes and cumulated dissipation, averaged in time from 63 to 66. The results are remarkably similar across all simulations and the reference (black lines). Despite small variations at large scales (see the inset of [PITH_FULL_IMAGE:figures/full_fig_p030_23.png] view at source ↗
Figure 1
Figure 1. Figure 1: presents the scalings of various time schemes with the time step for size N = 32 and Ct = 1. In this example, any scheme that does not contain phase-shifting containing aliasing errors as in the first example of the quadratic model presented in the core of the article. FIG. 1. Maximum error (max|S(k, dt) − Sexact(k, dt)|) for one time step versus time increment dt for all time schemes implemented at N = 32… view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2. Distribution of times spent in different tasks (Fourier transforms, operators, time stepping) [PITH_FULL_IMAGE:figures/full_fig_p039_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3. (a) Compensated energy spectra (3D spectra in dotted lines) versus wavenumber modulus [PITH_FULL_IMAGE:figures/full_fig_p039_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4. Temporal evolution of (a) total kinetic energy [PITH_FULL_IMAGE:figures/full_fig_p040_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5. Same as Fig [PITH_FULL_IMAGE:figures/full_fig_p040_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: FIG. 6. Same as Fig [PITH_FULL_IMAGE:figures/full_fig_p041_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: FIG. 7. Same as Fig [PITH_FULL_IMAGE:figures/full_fig_p041_7.png] view at source ↗
read the original abstract

Pseudo-spectral methods are widely used for direct numerical simulations of turbulence, but the standard 2/3 truncation rule for dealiasing is computationally expensive -- accounting for up to 80% of the total cost in three dimensions. Phase shifting methods provide a more efficient alternative by canceling aliasing errors the combination of nonlinear terms evaluated on shifted grids, allowing the same physical resolution to be achieved on a coarser numerical grid. Despite their use in high-resolution turbulence codes, these methods remain poorly documented in the literature and no open-source implementation exists. This paper presents a comprehensive analysis of phase-shifting dealiasing for pseudo-spectral simulations of the incompressible Navier-Stokes equations. We derive the aliasing mechanism from quadratic nonlinearities in discrete Fourier space and explain how phase-shifting cancels aliasing contributions exactly or approximately depending on the time-stepping scheme. We describe and compare several algorithms -- including the exact and approximate RK2 phase-shifting schemes of Patterson Jr and Orszag (1971) and Rogallo (1981), and an extension to forced flows -- and discuss their interaction with different truncation geometries in three dimensions. All algorithms are implemented in the open-source framework Fluidsim, providing the first publicly available implementation of phase-shifting dealiasing for pseudo spectral Navier-Stokes solvers. We evaluate the methods on two test cases: the transition to turbulence of Taylor-Green vortices and forced homogeneous isotropic turbulence at $Re_\lambda = 200$. Phase-shifting methods achieve speedups of up to a factor of 3 compared to RK4 with 2/3 truncation at the same maximum wavenumber, with small accuracy loss.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

0 major / 2 minor

Summary. The manuscript derives the aliasing mechanism for quadratic nonlinearities in discrete Fourier space for pseudo-spectral incompressible Navier-Stokes simulations. It shows how phase-shifting on shifted grids cancels aliasing contributions exactly for the Patterson-Orszag RK2 scheme and approximately for other schemes such as Rogallo's, extends the approach to forced flows and 3D truncation geometries, implements all variants in the open-source Fluidsim framework, and reports quantitative benchmarks on Taylor-Green vortex transition and forced HIT at Re_λ=200. The central claim is that phase-shifting achieves up to 3× speedup versus RK4+2/3 truncation at fixed k_max with only small accuracy loss.

Significance. If the derivation and numerical results hold, the work is significant because it supplies the first publicly available open-source implementation of phase-shifting dealiasing together with a self-contained analytic treatment and reproducible benchmarks on two standard test cases. The reported factor-of-3 speedup directly addresses the dominant cost of the 2/3 rule (up to 80 % of runtime in 3D) while preserving the same maximum wavenumber, and the explicit treatment of truncation geometries and forced-flow extensions fills a documented gap in the literature.

minor comments (2)
  1. [Abstract and §4] The abstract and introduction state that phase-shifting yields 'small accuracy loss,' but the quantitative error metrics (e.g., kinetic-energy spectra or enstrophy) and their comparison to the 2/3-rule baseline are only summarized; a short table or explicit statement of the observed relative errors at equivalent k_max would strengthen the claim.
  2. [Section on truncation geometries] In the discussion of 3D truncation geometries, the interaction between the phase-shift vectors and the spherical versus cubic cutoff is described but not illustrated; a single schematic diagram showing the shifted wavenumber supports would improve clarity for readers implementing the method.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the detailed summary, positive significance assessment, and recommendation of minor revision. No specific major comments were listed in the report, so we have no points to address point-by-point at this stage. We will incorporate any minor editorial or clarification changes suggested during the revision process.

Circularity Check

0 steps flagged

Derivation self-contained; no circular reductions identified

full rationale

The paper derives aliasing directly from quadratic nonlinearities in discrete Fourier space using standard properties of the discrete Fourier transform, then shows exact or approximate cancellation under phase-shifting for specific time-stepping schemes (Patterson-Orszag RK2, Rogallo). These steps rely on explicit algebraic manipulation of convolution terms rather than any fitted parameter or self-referential definition. Prior schemes are cited from independent 1971/1981 literature with no author overlap, and the central speedup claim (factor ~3 at fixed k_max) is obtained from explicit numerical benchmarks on Taylor-Green transition and Re_λ=200 HIT, not from any prediction that reduces to the input data by construction. The open-source Fluidsim implementation supplies reproducible code, confirming the analysis is externally falsifiable and independent of the present paper's fitted values or self-citations.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claim rests on the standard incompressible Navier-Stokes equations in Fourier space and the algebraic properties of discrete Fourier transforms; no new free parameters, ad-hoc constants, or postulated entities are introduced.

axioms (2)
  • domain assumption Incompressible Navier-Stokes equations expressed in Fourier space
    The paper starts from the standard Fourier-transformed form of the incompressible NS equations.
  • standard math Quadratic nonlinearities produce aliasing via discrete Fourier convolution
    Derivation of the aliasing mechanism follows directly from the convolution theorem applied to the discrete Fourier transform.

pith-pipeline@v0.9.0 · 5597 in / 1382 out tokens · 60011 ms · 2026-05-15T13:16:12.373934+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

30 extracted references · 30 canonical work pages

  1. [1]

    RK2 phase-shift exact

    General description We consider equations that can be written in spectral space as: ∂tS=σS+N(S),(16) whereSrepresents the state of the system (potentially containing several variables),σ(k) is a real or complex coefficient depending on the wavenumber, andN(S) is an operator with at most second-order nonlinearities. The incompressible Navier-Stokes equatio...

    work page

  2. [2]

    RK2 phase-shift approx

    Application to the quadratic 1D model We first use the one-dimensional (1D) model defined by Eq. (11) to verify the behavior of the different algorithms and our implementations. Fig. 7 displays the maximum of|S(k, dt)− Sexact(k, dt)|as a function of the time stepdtfor standard schemes (Euler, RK2, and RK4) and for the two RK2 phase-shifting schemes descri...

    work page

  3. [3]

    Since we compare resolutions, these simulations are run in parallel with a reasonable number of cores for each grid size

    Effect of resolution for RK4 truncated simulations We consider simulations dealiased with 2/3 truncation at four grid sizes:N= 128, 256, 400, 640, and 768 (Table II). Since we compare resolutions, these simulations are run in parallel with a reasonable number of cores for each grid size. 19 N Re scheme Ct kmax˜ηnumber of proc. CPU.h speedup error (%) 768 ...

    work page

  4. [4]

    We already studied in Section III D how aliasing errors affect a simulation at smallk max˜η= 0.75

    Effect ofk max˜ηon aliasing error Having quantified the effect of resolution on fully dealiased simulations, we now examine how resolution affects aliasing errors. We already studied in Section III D how aliasing errors affect a simulation at smallk max˜η= 0.75. Here, we consider analogous simulations but for kmax˜η >3 (see Table III). Fig. 14 shows the e...

    work page

  5. [5]

    RK2 phase-shift random

    Profiles and speedup We profile the different algorithms implemented in Fluidsim on short Taylor-Green simu- lations as described in Section III D, focusing on time-stepping performance. To isolate the effect of the time-stepping algorithm, we always compare in this subsection simulations run with the same number of cores. Fig. 16 compares the distributio...

    work page

  6. [6]

    RK2 phase-shift exact

    Effect of time scheme and Reynolds number We now present three series of four simulations each, listed in Table IV. The three series correspond to different (Re, k max˜η) combinations: (1600,1), (2800,0.6), and (1600,1.6). Each series consists of four simulations: a reference run with the RK4 scheme andC t = 2/3; a second RK4 run withC t = 1; and two runs...

    work page

  7. [7]

    RK2 phase-shift random

    Quality and speedup To generalize the conclusions of the previous section, we explore whether intermediate combinations of truncation and phase shifting can improve the compromise between speedup and accuracy. Because the speedup trends appear to be independent ofk max, we fixk max = 108 (kmax˜η≈1.27) and compare all combinations of time scheme and trunca...

    work page 2020

  8. [10]

    Delache, C

    A. Delache, C. Cambon, and F. Godeferd, Scale by scale anisotropy in freely decaying rotating turbulence, Physics of Fluids26, 025104 (2014)

    work page 2014

  9. [12]

    K. P. Iyer, K. R. Sreenivasan, and P. Yeung, Circulation in high Reynolds number isotropic turbulence is a bifractal, Physical Review X9, 041006 (2019)

    work page 2019

  10. [13]

    Yeung, K

    P. Yeung, K. Ravikumar, S. Nichols, and R. Uma-Vaideswaran, GPU-enabled extreme-scale turbulence simulations: Fourier pseudo-spectral algorithms at the exascale using OpenMP offloading, Computer Physics Communications306, 109364 (2025)

    work page 2025

  11. [15]

    K. J. Burns, G. M. Vasil, J. S. Oishi, D. Lecoanet, and B. P. Brown, Dedalus: A flexible framework for numerical simulations with spectral methods, Physical Review Research2, 023068 (2020)

    work page 2020

  12. [16]

    P. D. Mininni, D. Rosenberg, R. Reddy, and A. Pouquet, A hybrid MPI–OpenMP scheme for scalable parallel pseudospectral computations for fluid turbulence, Parallel Computing37, 316 (2011)

    work page 2011

  13. [17]

    Mortensen and H

    M. Mortensen and H. P. Langtangen, High performance Python for direct numerical simula- tions of turbulent flows, Computer Physics Communications203, 53 (2016)

    work page 2016

  14. [18]

    C. C. Lalescu, B. Bramas, M. Rampp, and M. Wilczek, An efficient particle tracking algo- rithm for large-scale parallel pseudo-spectral simulations of turbulence, Computer Physics Communications278, 108406 (2022)

    work page 2022

  15. [19]

    S. G. Chumakov, A priori study of subgrid-scale flux of a passive scalar in isotropic homoge- neous turbulence, Physical Review E78, 036313 (2008)

    work page 2008

  16. [20]

    A. V. Mohanan, C. Bonamy, M. C. Linares, and P. Augier, FluidSim: Modular, object-oriented Python package for high-performance CFD simulations, Journal of Open Research Software 7, 10.5334/jors.239 (2019)

    work page doi:10.5334/jors.239 2019

  17. [21]

    kinematic fitting

    C. Lambert, J. Reneuve, and P. Augier, Dataset for article ”aliasing and phase shifting in pseudo-spectral simulations of the incompressible navier-stokes equations”, 10.5281/zen- odo.19554006 (2026). 34

    work page doi:10.5281/zen- 2026

  18. [22]

    G. I. Taylor and A. E. Green, Mechanism of the production of small eddies from large ones, Proceedings of the Royal Society of London. Series A, Mathematical and Physical Sciences 158, 499 (1937)

    work page 1937

  19. [24]

    The documentation for this module is available in the Fluidsim documentation at https://fluidsim.readthedocs.io/en/latest/generated/fluidsim.base.time_ stepping.pseudo_spect.html

    work page

  20. [25]

    See Supplemental Material at URL-will-be-inserted-by-publisher

    work page

  21. [26]

    J. R. Debonis, Solutions of the Taylor-Green vortex problem using high-resolution explicit finite difference methods (2013). [20]https://fluidsim.readthedocs.io/en/latest/generated/fluidsim.base.time_ stepping.pseudo_spect.html

    work page 2013

  22. [27]

    Eswaran and S

    V. Eswaran and S. Pope, An examination of forcing in direct numerical simulations of turbu- lence, Computers & Fluids16, 257 (1988). Supplementary material: Aliasing and phase shifting in pseudo-spectral simulations of the incompressible Navier-Stokes equations Clovis Lambert, 1 Jason Reneuve, 1 and Pierre Augier 1,∗ 1Laboratoire des Ecoulements G´ eophys...

    work page 1988

  23. [28]

    Canuto, M

    C. Canuto, M. Y. Hussaini, A. Quarteroni, A. Thomas Jr,et al.,Spectral methods in fluid dynamics(Springer Science & Business Media, 1988)

    work page 1988

  24. [29]

    Canuto, M

    C. Canuto, M. Y. Hussaini, A. Quarteroni, and T. A. Zang,Spectral methods: evolution to complex geometries and applications to fluid dynamics(Springer Science & Business Media, 2007)

    work page 2007

  25. [30]

    Cooley and J

    J. Cooley and J. Tukey, An algorithm for the machine calculation of complex Fourier series, Mathematics of Computation19, 297 (1965)

    work page 1965

  26. [31]

    A. V. Mohanan, C. Bonamy, and P. Augier, FluidFFT: Common API (C++ and Python) for Fast Fourier Transform HPC libraries, Journal of Open Research Software7, 10.5334/jors.238 (2019)

    work page doi:10.5334/jors.238 2019

  27. [32]

    Carleson, On convergence and growth of partial sums of fourier series, Acta Mathematica 116, 135 (1966)

    L. Carleson, On convergence and growth of partial sums of fourier series, Acta Mathematica 116, 135 (1966). 9

    work page 1966

  28. [33]

    O. H. Hald, Convergence of Fourier methods for Navier-Stokes equations, Journal of Compu- tational Physics40, 305 (1981)

    work page 1981

  29. [34]

    Patterson Jr and S

    G. Patterson Jr and S. A. Orszag, Spectral calculations of isotropic turbulence: Efficient removal of aliasing interactions, The Physics of Fluids14, 2538 (1971)

    work page 1971

  30. [35]

    R. S. Rogallo,Numerical experiments in homogeneous turbulence, Vol. 81315 (National Aero- nautics and Space Administration, 1981)

    work page 1981