🔬 physics

A carrier-wave factored one-way Navier--Stokes method for boundary-layer instability modelling

This paper introduces M-OWNS, a computationally efficient spatial marching method that combines carrier-wave factoring with a recursive one-way Navier-Stokes framework to model boundary-layer instability with significantly reduced numerical cost and coarser resolution requirements compared to standard approaches across various incompressible and hypersonic flow conditions.

Original authors: E. J. Badcock, S. Mughal

Published 2026-07-08

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: E. J. Badcock, S. Mughal

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to predict how a ripple in a river grows into a massive wave that eventually crashes the boat. In the world of aerodynamics, this "river" is the layer of air flowing over an airplane wing, and the "ripple" is a tiny disturbance that could turn smooth, laminar flow into chaotic turbulence.

Scientists have been trying to simulate this for decades. The paper you're asking about introduces a new, smarter way to do these simulations, called M-OWNS. Here is how it works, explained through everyday analogies.

The Problem: The "Upstream Noise"

To predict how a wave grows, you usually have to solve complex math equations.

The Old Way (Global Methods): Imagine trying to solve the whole river's behavior at once, from the source to the ocean. It's incredibly accurate, but it requires a supercomputer and takes forever.
The "Marching" Way (Standard OWNS): Instead of solving everything at once, you walk downstream step-by-step, calculating the wave at each spot. This is much faster. However, there's a catch: the math equations naturally include "noise" that travels upstream (against the flow). If you try to march forward while this upstream noise is still in the math, the calculation gets confused and explodes.
The "Filter" Solution: To make the marching work, scientists have to use a "filter" to delete the upstream noise. But to get a clear picture of the wave, you have to take very tiny, slow steps. It's like trying to walk through a forest while wearing thick foggy glasses; you have to take baby steps to avoid tripping.

The Innovation: The "Carrier Wave" Trick

The authors (Badcock and Mughal) combined two existing ideas to create M-OWNS. Think of it as putting on a pair of specialized glasses that change how you see the river.

The Carrier Wave (The Moving Walkway):
Imagine the ripple you are tracking is riding on a moving walkway (the "carrier wave"). In the old method, you had to calculate the ripple's movement relative to the stationary ground. In the new method, you calculate the ripple's movement relative to the moving walkway.
- Why this helps: If the walkway moves at the same speed as the ripple, the ripple looks almost stationary to you. It stops jittering around. This means you don't need to take tiny, baby steps anymore; you can take long, confident strides.
The One-Way Projection (The One-Way Street):
The method also keeps the "filter" that deletes the upstream noise. But because the "moving walkway" trick made the math so much simpler, the filter doesn't have to work as hard.

The Result: Walking Faster Without Tripping

The paper claims that by combining these two tricks, they can simulate the growth of these air ripples much faster and with less computer power than before, without losing accuracy.

The "Fixed" vs. "Iterating" Strategy:
- Iterating: Imagine you adjust your speed on the moving walkway at every single step to perfectly match the ripple. This is the most precise but takes a tiny bit of extra calculation time.
- Fixed: Imagine you set the walkway speed once at the beginning based on a good guess. The paper shows that even this "lazy" approach works incredibly well. It allows them to take steps 2 to 8 times larger than the old methods.

Real-World Tests

The authors tested this new method on three different "rivers":

Subsonic Flow: Air flowing over a flat plate (like a wing) at normal speeds.
3D Crossflow: Air flowing over a curved, swept cylinder (like a wingtip).
Hypersonic Flow: Air moving at Mach 4.5 (five times the speed of sound), which is extremely hot and chaotic.

In all cases, including when they threw "random noise" at the start of the simulation, M-OWNS captured the details of the waves correctly. In fact, the old method (Standard OWNS) often failed or produced nonsense unless they used a massive number of tiny steps. M-OWNS got the same (or better) results with far fewer steps.

The Bottom Line

The paper doesn't claim this will immediately design a new airplane. Instead, it claims to have found a mathematical shortcut.

Think of it like this: If you want to count the grains of sand on a beach, the old way was to count every single grain one by one. The new method (M-OWNS) is like realizing you can group the sand into buckets and count the buckets, because you know exactly how many grains are in each bucket. You get the same total count, but you finish the job in a fraction of the time.

This allows scientists to run complex simulations of how airplanes might transition from smooth flight to turbulence much faster, potentially saving time and money in the design process.

Technical Summary: M-OWNS for Boundary-Layer Instability Modelling

Problem Statement
Predicting laminar–turbulent transition requires tracking disturbance evolution through amplification in developing base flows. While local eigenvalue analysis (LST) is computationally cheap but limited by parallel-flow assumptions, and global linearised harmonic Navier–Stokes (LHNS) solvers are general but prohibitively expensive due to memory and matrix inversion constraints, spatial marching methods offer a middle ground. However, standard spatial marching faces a fundamental difficulty: the linearised Navier–Stokes operator supports upstream-propagating acoustic modes that can destabilise a naive downstream march.

Existing solutions have trade-offs:

Parabolised Stability Equations (PSE): Use carrier-wave factoring to shift the downstream spectrum toward the origin, allowing coarser resolution. However, they rely on regularisation strategies (e.g., neglecting streamwise pressure gradients or implicit damping) to suppress upstream modes, which introduces characteristic errors. Furthermore, PSE suffers from a minimum step-size restriction; refining the step to capture multi-modal or non-modal disturbances often re-captures the unstable upstream modes, destabilising the march.
One-Way Navier–Stokes (OWNS): Removes upstream modes entirely via a recursive projection (OWNS-R), retaining the streamwise pressure gradient and avoiding PSE regularisation errors. However, OWNS does not employ carrier-wave factoring. Consequently, it must resolve eigenvalues at their full wavenumbers from the origin, necessitating fine streamwise resolution. This fine resolution widens the integrator's spectral window, demanding a more complex projection to suppress a broader range of upstream modes, which is limited by numerical error accumulation in the recursion parameters.

Methodology: M-OWNS
The paper presents M-OWNS, a spatial marching method that integrates the carrier-wave factoring of PSE into the recursive one-way Navier–Stokes (OWNS-R) framework.

Carrier-Wave Factoring: The disturbance state $\phi$ is decomposed into a slowly varying envelope $\tilde{\phi}$ and a rapidly varying carrier wave $\exp(i\Theta_0)$ , where $d\Theta_0/dx = \alpha_0$ . This shifts the spatial operator's spectrum by a scalar $\alpha_0$ .
One-Way Projection: The OWNS-R projection operator $P$ is applied to the factored system. The authors prove that the same projection operator that parabolises the unfactored system also parabolises the carrier-wave factored system, removing upstream-propagating modes without requiring modification of the projection parameters.
Carrier Selection ( $\alpha_0$ ):
- Iterated Variant: $\alpha_0$ is updated at each station using the Rayleigh quotient, converging toward the dominant eigenvalue (similar to PSE).
- Fixed-Carrier Variant: $\alpha_0$ is set to a representative value derived from the base flow (e.g., the vorticity/entropy branch endpoint $\alpha_1 = (\omega - \beta W_\infty)/U_\infty$ ). This variant requires no iterations and maintains the same per-step cost as unfactored OWNS.

Key Contributions

Spectral Resolution Advantage: A leading-order analysis demonstrates that for any excited eigenfunction whose eigenvalue lies closer to the carrier wavenumber $\alpha_0$ than to the origin, the factored system resolves the mode at a coarser streamwise step size than the unfactored system. The factoring effectively translates the integrator's "resolution disc" from the origin to the cluster of excited modes.
Decoupling Resolution from Regularisation: By combining factoring with the one-way projection, M-OWNS achieves the resolution benefits of PSE (coarser steps) without the step-size restrictions or regularisation errors (e.g., neglected pressure gradients) that plague PSE.
Robustness in Multi-Modal Regimes: The method is shown to be effective for both single-mode disturbances (where iteration is optimal) and multi-modal or broadband disturbances (where a fixed carrier is preferred to avoid Rayleigh quotient convergence issues associated with non-orthogonal eigenfunctions).

Results and Validation
The method was validated across incompressible and subsonic flat-plate boundary layers, three-dimensional crossflow disturbances over a swept cylinder, and a Mach 4.5 hypersonic boundary layer. Four forcing configurations were tested: eigenfunction inlet forcing, wall suction/blowing, multi-mode freestream forcing, and randomised inlet forcing.

Resolution Efficiency: In all test cases, M-OWNS captured disturbance amplitudes, acoustic radiation fields, and modal synchronisation sequences at coarser streamwise resolutions than unfactored OWNS.
- For the swept cylinder case, unfactored OWNS required $N_x \approx 8000$ steps for convergence, while fixed-carrier M-OWNS converged at $N_x \approx 1000$ (an 8-fold reduction), and iterated M-OWNS converged at $N_x \approx 150$ (a 47-fold reduction).
- For the hypersonic randomised inlet case (the most stringent multi-modal test), unfactored OWNS required $N_x \approx 6000$ , while fixed-carrier M-OWNS converged at $N_x \approx 3000$ (a 2-fold reduction).
Accuracy: M-OWNS results were validated against PSE, unfactored OWNS, and fully elliptic LHNS solvers. In the wall suction/blowing case, M-OWNS results matched the LHNS reference, whereas unfactored OWNS at the same resolution under-predicted amplitudes.
Cost: The total computational cost (total linear solves) was reduced by factors of two to eight across the test cases. The fixed-carrier variant achieved these reductions with identical per-step costs to unfactored OWNS, as it eliminates the iterative overhead.

Significance
The paper claims that M-OWNS offers a distinct numerical resolution and efficiency advantage in modelling disturbance and instability state evolution. By shifting the resolved spectral region to the physically relevant cluster of eigenvalues, the method allows for significantly coarser streamwise steps without sacrificing accuracy or introducing the regularisation errors inherent in PSE. The fixed-carrier variant is highlighted as particularly significant, as it achieves these gains without increasing the per-step computational cost or requiring iterative convergence, making it a robust alternative for both single-mode and complex multi-modal instability problems. The authors note that the method is applicable to general flows where the excited spectrum clusters near a specific wavenumber, providing a practical solution to the resolution constraints of standard OWNS.