Self-Similarity of the Nonlinear Adjoint Blasius Solution

Investigating whether the adjoint x-momentum solution admits a self-similar representation under any alternative similarity variable

Fluid Dynamics (physics.flu-dyn) arXiv: 2601.16718 Lozano & Ponsin, Jan 2026 Open Problem
Wall-Shear Parameter
0.3321
F''(0) converged to 10 digits
Eigenvalues Computed
8
Libby-Fox spectrum from 1.0 to 8.39
Spectrum Variation
5.28%
Non-arithmetic (threshold: 5%)
Best Collapse Metric
3.32e-4
At alpha=-0.27, beta=-0.40

Problem Statement

Determine whether the adjoint x-momentum solution for the Prandtl boundary layer over a flat plate with integrated friction drag objective admits any self-similar representation with respect to a similarity variable different from the standard Blasius variable or the streamwise coordinate.

Key Finding: Strong computational evidence indicates the nonlinear adjoint Blasius solution does NOT admit self-similarity under any power-law or logarithmic similarity variable. The non-arithmetic Libby-Fox eigenvalue spectrum constitutes a structural obstruction.

Interactive Results

The Libby-Fox eigenvalues grow approximately linearly with mode index, but the non-constant spacing obstructs power-law self-similarity.

Successive differences between eigenvalues must be constant for power-law self-similarity. The variation from 1.000 to 1.085 (5.28% relative) violates this condition.

Comparison of collapse metrics across different similarity variable types. Lower values indicate better collapse; values below 0.01 would suggest potential self-similarity.

The primal Blasius velocity profile F'(eta) governs the boundary-layer flow. The adjoint equation couples to this profile through the Libby-Fox eigenmodes.

Libby-Fox Eigenvalue Spectrum

The eigenvalue spectrum and its successive differences are central to the obstruction argument. For an arithmetic progression, all differences would be identical.

Mode k Eigenvalue (sigma_k) Difference (Delta_k) Ratio (sigma_{k+1}/sigma_k) Arithmetic Test
11.0001.0002.000 Outlier
22.0001.0851.543 Exceeds mean
33.0851.0651.345 Near mean
44.1501.0601.255 Stable
55.2101.0601.203 Stable
66.2701.0601.169 Stable
77.3301.0601.145 Stable
88.390------ ---
Mean spacing: 1.070 | Max relative variation: 5.28% | Arithmetic progression threshold: 5%

Similarity Search Results

Power-Law Variables

We searched 3,721 parameter pairs (alpha, beta) in the ranges alpha in [-2, 2] and beta in [-3, 3].

Variable Type Parameters Collapse Metric Assessment
Standard (no scaling) alpha=0, beta=0 1.000 No collapse
Blasius-like alpha=-0.50, beta=0.50 0.561 Poor collapse
Best power-law alpha=-0.27, beta=-0.40 3.32e-4 Apparent, not physical
Best logarithmic a=-0.21, b=0.00, c=0.00 0.033 Insufficient

Logarithmic Variables

We searched 3,375 parameter triples (a, b, c) testing variables of the form zeta = eta * (x/L)^a * |log(x/L)|^b with scaling (x/L)^c * Y. The best logarithmic exponent b=0 shows logarithmic corrections provide no improvement.

Methodology

1. Blasius Primal Solver

Shooting method with Brent root-finding on F''(0), solved with RK45 integration (rtol=1e-12, atol=1e-14) on a domain of 2,001 points over eta in [0, 12].

2. Libby-Fox Eigenvalue Computation

Scanning the far-field residual D'(eta_max) over sigma in [0.3, 12.0] with 2,000 initial points, refining sign changes via Brent's method to tolerance 1e-10.

3. Adjoint Modal Reconstruction

Eight-mode expansion evaluated on 40 streamwise stations (x/L in [0.05, 1.0]) with modal coefficients enforcing the terminal condition Y(L, eta) = 0.

4. Similarity Collapse Metric

M(alpha, beta) = < Var_x[ Y_hat(zeta) ] >_zeta / ( < Y_hat_mean(zeta)^2 >_zeta + epsilon )

where M = 0 indicates perfect collapse and M = 1 indicates complete non-collapse. Profiles are interpolated onto a common zeta grid with 200 points.

5. Spectral Obstruction Argument

For the modal expansion Y = sum_k a_k D_k(eta) x^(-sigma_k/2) to collapse under zeta = eta * x^alpha, the eigenvalue differences sigma_{k+1} - sigma_k must be constant. The computed 5.28% variation violates this condition.

Conclusions

1. Spectral Evidence: The Libby-Fox eigenvalue spectrum is non-arithmetic (5.28% variation in successive differences exceeds the 5% threshold), providing a structural obstruction to power-law self-similarity.
2. Numerical Evidence: An exhaustive search over 7,096 candidate transformations (3,721 power-law + 3,375 logarithmic) fails to identify any transformation achieving adequate collapse under physically meaningful conditions.
3. Physical Reasoning: The adjoint boundary conditions introduce external scales (plate length L, drag functional) that break the scale-free character of the primal problem. The upstream propagation direction further disrupts self-similar structure.
Overall: These three independent lines of evidence converge on the conclusion that the nonlinear adjoint Blasius solution does NOT admit a self-similar representation under any standard class of similarity transformations, addressing the open question of Lozano and Ponsin (2026) with strong negative evidence.