Falkner-Skan Boundary Layer Solution

• The Falkner-Skan solution is a foundational similarity approach that defines the velocity profile in a steady, two-dimensional laminar boundary layer subject to variable pressure gradients.
• It features a third-order nonlinear ODE with boundary conditions representing no slip at the wall and normalized far-field velocity, bridging classical Blasius flow to flows with favorable or adverse pressure gradients.
• Advanced numerical, spectral, and physics-informed neural network methods are employed to ensure precise computation, convergence acceleration, and robust analysis of skin-friction coefficients and flow separation.

The Falkner-Skan solution is the foundational similarity solution describing the velocity profile in a steady, two-dimensional laminar boundary layer over a wedge surface subject to a power-law outer flow. The governing ordinary differential equation encapsulates the transition between classic flat-plate (Blasius) flow and flows with favorable or adverse pressure gradients determined by a parameter β, which is a function of the wedge angle or pressure-gradient exponent. The resulting profile f(η) satisfies a third-order nonlinear boundary value problem on the semi-infinite domain η ∈ &&&0&&&, &&&1&&&, Weyburne, 2016). Zero pressure gradient (Blasius flow) corresponds to β = 0; β > 0 indicates a favorable gradient; β < 0, an adverse gradient leading towards boundary-layer separation.

2. Asymptotic Behavior, Uniqueness, and Branch Structure

Uniqueness and regularity of solutions depend on the parameter β. For $\beta \in (\beta_{\rm crit}, 1)$ with $\beta_{\rm crit} \approx -0.1988$, the solution f(η) is unique and monotonic with $f''(0) > 0$, and the velocity profile approaches unity algebraically: $$f'(\eta) \sim 1 - C(\beta)\,\eta^{-p} + o(\eta^{-p}),\quad p = \frac{2}{2-\beta}$$ (Iyer, 2024, Iyer et al., 2022). For β near the critical value, two solution branches exist (normal and reversed flow), which merge at $\beta_{\rm crit}$ where $f''(0) = 0$ (Fazio, 2012). For β beyond this threshold, separation occurs and classical solutions do not exist.

Table: Skin-friction coefficient $f''(0)$ for selected β

β	$f''(0)$	Type
1.000	1.000000000	stagnation-point
0.500	0.927680040	Homann flow
0.000	0.469599988	Blasius flow
-0.120	0.281760524	reverse flow
-0.1988	0.000000000	separation onset

Values shown are accurate to 12 digits, with comparison benchmarks in (Belden et al., 2019, Ganapol, 2010).

3. Analytical Approximations and Convergence Acceleration

Series representations and convergence acceleration are central for high-precision results. The Maclaurin expansion,

$f(\eta) = \sum_{k=0}^{\infty} a_k\,\eta^k,$

with $a_2 = f''(0)/2$, admits a nonlinear three-term recurrence generating all coefficients (Ganapol, 2010, Belden et al., 2019). However, convergence radius $R(\beta)$ is finite:

• $R = 4.024$ for β = 0 (Blasius)
• Decreases as $\beta \to -0.1988$

To extend the solution beyond series convergence, Wynn-ε acceleration and continuous analytical continuation (CAC) are employed to reach $\eta \gtrsim 100$ with stable precision (Ganapol, 2010). Asymptotic approximants reconcile near-wall series with far-field expansions, providing uniformly accurate closed-form profiles (Belden et al., 2019): $$f_A(\eta) = \eta + B(\beta) - B(\beta)\left[1 + \sum_{n=1}^N A_n\,\eta^n\right]^{-1},$$ where the displacement thickness constant $B(\beta) = \lim_{\eta \to \infty} [f(\eta)-\eta]$.

4. Numerical and Spectral Solution Methods

Numerical resolution on the infinite domain can leverage finite-difference schemes on quasi-uniform grids, allowing the last node to be placed at infinity and exact imposition of far-field boundary conditions. Logarithmic or algebraic stretching maps transform [0,1] → [0,∞), and non-standard finite differences provide second-order accuracy in mesh size with optimal Richardson extrapolation for enhanced precision (Fazio et al., 2012). Pseudospectral discretizations using Hermite functions yield spectrally accurate results due to analytic decay properties and natural boundary enforcement (Parand et al., 2010).

Modern approaches embed the ODE into neural network frameworks. Physics-Informed Neural Networks (PINNs), trained with collocation points and boundary penalties, can achieve sub-percent errors for strong gradients (Eivazi et al., 2021). Deep architectures employing Legendre and Chebyshev polynomial blocks further accelerate evaluation and differentiation, with operational matrices replacing autodifferentiation and yielding 2–5× speedup (Aghaei et al., 2023).

Hybrid metaheuristic-shooting solvers, such as the Jaya–Runge–Kutta method, combine global parameter search robustness with RK4 integration, outperforming classical heuristics in accuracy and stability (Guo et al., 2020).

5. Scaling Laws and Similarity Considerations

Similarity reduction crucially depends on proper scaling. The Prandtl–Plus scaling convention fails for general Falkner–Skan flows except for the sink-flow (β = 1), due to non-constant friction velocity ratios. A new scaling using $(u_0, \delta_0)$ tied to the wall-shear and local viscous length restores similarity for all β: $$\eta = \frac{y}{\delta_0(x)}, \quad U^* = \frac{u(x,y)}{u_0(x)}, \quad \beta = \frac{2m}{m+1}$$ (Weyburne, 2016).

6. Extension: Hydromagnetic, Symmetry, and Analytical Integrability

Hydromagnetic generalizations introduce a Lorentz-force term yielding a modified Falkner–Skan ODE: $f''' + f f'' + \beta(1 - f'^2) - M^2(f' - 1) = 0,$ with M the Hartmann number (Parand et al., 2010, Haas et al., 2010). Spectral decomposition and Newton iteration achieve high-precision agreement with reference solutions.

From the symmetry perspective, the Falkner–Skan ODE admits only trivial classical Lie point symmetries, with nonclassical invariances appearing only for PDE generalizations (Balaji et al., 2020). Closed-form meromorphic solutions exist solely for Λ = –1, corresponding to the Chazy II class; all other β require numerical or semi-analytical treatment (Conte et al., 2021).

7. Regularity, Stability, and Advanced Existence Theory

Higher regularity and nonlinear stability for solutions to the stationary Prandtl system near Falkner–Skan profiles have been rigorously established. Perturbations with data in hybrid weighted Sobolev norms decay with enhanced rates due to favorable pressure gradients, and existence theorems guarantee arbitrarily smooth solutions provided side-constraint algebra is satisfied (Iyer, 2024, Iyer et al., 2022).

• (Ganapol, 2010) "Highly Accurate Solutions of the Blasius and Falkner-Skan Boundary Layer Equations via Convergence Acceleration"
• (Belden et al., 2019) "Asymptotic Approximant for the Falkner-Skan Boundary-Layer equation"
• (Fazio, 2012) "Blasius Problem and Falkner-Skan model: Töpfer's Algorithm and its Extension"
• (Fazio et al., 2012) "Finite difference schemes on quasi-uniform grids for Bvps on infinite intervals"
• (Parand et al., 2010) "An approximate solution of the MHD Falkner-Skan flow by Hermite functions pseudospectral method"
• (Iyer, 2024) "Stability of the Favorable Falkner-Skan Profiles for the Stationary Prandtl Equations"
• (Iyer et al., 2022) "Higher Regularity Theory for a Mixed-Type Parabolic Equation"
• (Aghaei et al., 2023) "Solving Falkner-Skan type equations via Legendre and Chebyshev Neural Blocks"
• (Guo et al., 2020) "Integrated intelligent Jaya Runge-Kutta method for solving Falkner-Skan equations for Various Wedge Angles"
• (Weyburne, 2016) "The Prandtl Plus Scaling Failure and its Remedy"
• (Eivazi et al., 2021) "Physics-informed neural networks for solving Reynolds-averaged Navier–Stokes equations"
• (Conte et al., 2021) "Closed-form meromorphic solutions of some third order boundary layer ordinary differential equations"
• (Balaji et al., 2020) "The Leading Edge Problem in Fluid Mechanics"