Complete classification of the solution set of Weyl's boundary-layer problem $(A_\lambda)$ (the Falkner–Skan equation) for $\lambda > 1$

A 1942 PNAS note by Hermann Weyl proves, by a Schauder-type fixed-point argument, that the boundary-layer problem $(A_\lambda)$ — $w'''+2ww''+2\lambda(k^2-w'^2)=0$ on the half-line $z>0$ with $w(0)=w'(0)=0$ and $w'(\infty)=k$, for arbitrary $k>0$ and $\lambda\ge 0$ — has a solution, and it leaves open whether that solution is unique and how the solution depends on the parameters. After the scaling $w(z)=\sqrt{k/2}\,f(\sqrt{2k}\,z)$, problem $(A_\lambda)$ is exactly the Falkner–Skan equation $f'''+ff''+\beta(1-f'^2)=0$, $f(0)=f'(0)=0$, $f'(\infty)=1$, with $\beta=\lambda$, the canonical similarity reduction of the steady two-dimensional boundary-layer equations for flow past a wedge. The uniqueness component of Weyl's question has since been resolved: the solution is unique for $0\le\lambda\le1$ (with no monotonicity or bound assumption on $f'$), while for $\lambda>1$ uniqueness fails — for every integer $N$ there is a solution with at least $N$ oscillations of $f'$, so the problem has infinitely many solutions. What remains genuinely open — the open core pursued by the subsequent literature (Hastings–Troy; Swinnerton-Dyer–Sparrow) after Weyl's existence result and the later uniqueness results — is the complete rigorous classification of the solution set for $\lambda>1$: whether the known reverse-flow branches exhaust it, at which bifurcation values of $\lambda$ branches are created and destroyed, and whether the conjectured symbolic-dynamics (horseshoe-type) description of the admissible solution set holds in the parameter ranges in which it was made.

Background

A boundary layer is the thin region of retarded fluid next to a solid surface in a high-Reynolds-number flow: viscosity is important inside it, while outside it the flow is essentially inviscid. The Prandtl boundary-layer equations are the standard asymptotic reduction of the Navier–Stokes equations for this region. For steady two-dimensional flow of a slightly viscous incompressible fluid past a wedge, a similarity transformation (measuring distance from the wall in units that grow like the square root of the coordinate along the wall, and introducing a stream function $f$ whose derivative $f'$ gives the tangential velocity profile) reduces the problem to a single third-order nonlinear ordinary differential equation, the Falkner–Skan equation, $f'''+ff''+\beta\,(1-f'^2)=0,\qquad f(0)=f'(0)=0,\quad f'(\infty)=1,$ where primes denote differentiation in the similarity variable $\eta\ge 0$ and $\beta$ is a parameter fixed by the wedge angle ($\beta=0$ is the flat-plate case, $\beta=1$ stagnation-point flow). The physically relevant outputs are the wall shear stress $f''(0)$ and the shape of the velocity profile $f'$. A solution is called normal if $0<f'<1$ along the whole half-line (the velocity increases monotonically from zero at the wall to the free-stream value), and a reverse-flow solution if $f'<0$ somewhere (fluid moving backwards near the wall, i.e. a separated flow region).

In two PNAS notes (1941, 1942) Weyl studied the equivalent formulation $(A_\lambda)$: $w'''+2ww''+2\lambda(k^2-w'^2)=0$ for $z>0$, $w(0)=w'(0)=0$, $w'(\infty)=k$. He first treated $\lambda=0$ (the flat plate) and $\lambda=1/2$ by a rapidly converging process of successive approximations, and then, for arbitrary $\lambda>0$, took recourse to a fixed-point existence argument of the kind introduced in the memoir of Birkhoff and Kellogg (1922) and later put in abstract form by Schauder and Leray — an existence principle asserting that a continuous self-map of a compact convex set has a fixed point, which yields existence but carries no information about uniqueness. He also eliminated the parameter $k$ by differentiating the equation to a fourth-order equation invariant under the scaling $w\mapsto Kw(Kz)$, so that the wall shear in the original variables is a pure multiple $\sqrt{2}\,k^{3/2}f''(0)$ of the normalized one; the parameter $k$ therefore carries no genuinely new content, and the open substance of the question is about $\lambda$ alone. Numerically, the landscape of solutions in $\beta$ had been explored since Hartree (1937) — via the shooting method, which turns the boundary-value problem into an initial-value problem by guessing the unknown initial value $f''(0)$ and adjusting it until the far condition is met — who found a lower critical value $\beta_{\min}\approx-0.1988$ below which no solution exists, and Stewartson (1954), who identified additional solution branches in the negative-$\beta$ range; but rigorous classification results came only much later, and for $\lambda>1$ they remain incomplete (see the problem statement).

For $\lambda>1$ the known solutions comprise the unique normal solution together with an infinite family of reverse-flow solutions carrying an increasing number of oscillations of $f'$. Swinnerton-Dyer and Sparrow (1995, 2002) studied how these branches are created as $\lambda$ passes through positive integer values; as Sparrow (1997) summarizes the program, a sequence of bifurcations creates a large set of periodic and other orbits 'from infinity' at each positive integer value of $\lambda$, and a second sequence destroys these orbits as $\lambda$ increases — a destruction sequence he described as still the subject of active research. The conjectured mechanism is a Smale horseshoe — the standard model of chaos in which stretching and folding of a region of phase space traps an invariant Cantor set indexed by symbol sequences, so that each symbol sequence corresponds to one orbit. A description of orbits by such sequences is called symbolic dynamics. Whether this horseshoe-type picture, and the branch structure built on it, gives the true complete classification of the admissible solutions for $\lambda>1$ is what remains open.

Problem Statement

The source question (Weyl's 1942 PNAS note) asks, for the problem $(A_\lambda)$: $w'''+2ww''+2\lambda(k^2-w'^2)=0$ for $z>0$ with $w(0)=w'(0)=0$, $w'(\infty)=k$, $k>0$, $\lambda\ge 0$ (equivalently, after scaling, the Falkner–Skan boundary-value problem $f'''+ff''+\lambda(1-f'^2)=0$, $f(0)=f'(0)=0$, $f'(\infty)=1$), whether the solution whose existence Weyl proved is unique on the half-line and how the solutions depend on $\lambda$. The branch-classification question stated below is the open core pursued by the subsequent literature (Hastings–Troy; Swinnerton-Dyer–Sparrow) after Weyl's existence result and the later uniqueness results. The uniqueness component is settled: the solution is unique for $0\le\lambda\le1$, and it is non-unique for $\lambda>1$, where for every integer $N$ there is a solution with at least $N$ oscillations (relative minima of $f'$). The surviving open core is the complete rigorous classification of the solution set for $\lambda>1$. Here 'admissible' follows the terminology of Swinnerton-Dyer and Sparrow, as summarized by Sparrow (1997): whole-line solutions of the Falkner–Skan equation $f'''+ff''+\lambda(1-f'^2)=0$ on $\mathbb{R}$ whose $f'$ remains bounded; the class includes periodic orbits (P-orbits), orbits with $f'\to1$ as $\eta\to\pm\infty$ (Q-orbits), and the lines $f'=\pm1$. Half-line boundary-value solutions ($f(0)=f'(0)=0$, $f'(\infty)=1$) typically have unbounded behaviour for $\eta<0$ and are therefore not in general admissible; component (i) below quantifies over the half-line boundary-value solutions, components (ii)–(iii) over the admissible class of the associated flow, and any correspondence relied on between the two classes must be made explicit.

(i) Exhaustiveness: determine, for each given $\lambda>1$, all solutions of the Falkner–Skan boundary-value problem on the half-line: whether the known branches — defined here as the unique normal solution together with the reverse-flow and oscillating solutions carrying finitely many oscillations (relative minima) of $f'$, as constructed by Craven–Peletier and Hastings–Troy — exhaust the solution set, or whether solutions of a genuinely new type exist, to be exhibited and rigorously verified. (ii) Bifurcation structure: determine the critical values of $\lambda$ at which solution branches of the associated three-dimensional flow are created or destroyed, and describe the global bifurcation structure of the admissible solutions for all $\lambda>1$ (the Swinnerton-Dyer–Sparrow picture places the creation of large sets of periodic and other orbits as $\lambda$ passes through positive integer values; the second, destruction sequence of bifurcations is described by Sparrow (1997) as still the subject of active research). (iii) Symbolic dynamics: prove or refute, over the parameter ranges in which they were made, the Hastings–Troy conjecture (1985, stated for $\lambda>2$ and proved rigorously only for $\lambda$ close to 2) that infinitely many periodic admissible solutions are organized by a Smale-horseshoe-type symbolic dynamics, together with the Swinnerton-Dyer–Sparrow picture of strange invariant sets created as $\lambda$ crosses positive integer values — i.e. that specifying a finite symbol sequence determines a unique admissible solution, giving the 'strange invariant set' structure they describe.

A complete classification determines, for each $\lambda>1$, the full solution structure. Components (i)–(iii) are all part of the target, and the classification sought covers all of $\lambda>1$; because branch creation and destruction occur at parameter-dependent bifurcation values, it may be determined parameter-interval by parameter-interval.

References

1. Weyl, H. (1941). Concerning the Differential Equations of Some Boundary Layer Problems. Proceedings of the National Academy of Sciences 27(12), 578-583. doi:10.1073/pnas.27.12.578. https://doi.org/10.1073/pnas.27.12.578
2. Weyl, H. (1942). Concerning the Differential Equations of Some Boundary Layer Problems. II. Proceedings of the National Academy of Sciences 28(3), 100-102. doi:10.1073/pnas.28.3.100. https://doi.org/10.1073/pnas.28.3.100
3. Weyl, H. (1942). On the Differential Equations of the Simplest Boundary-Layer Problems. Annals of Mathematics 43(2), 381-407. doi:10.2307/1968875. https://doi.org/10.2307/1968875
4. Birkhoff, G. D.; Kellogg, O. D. (1922). Invariant points in function space. Transactions of the American Mathematical Society 23(1), 96-115. doi:10.1090/S0002-9947-1922-1501192-9. https://doi.org/10.1090/S0002-9947-1922-1501192-9
5. Hartree, D. R. (1937). On an equation occurring in Falkner and Skan's approximate treatment of the equations of the boundary layer. Mathematical Proceedings of the Cambridge Philosophical Society 33(2), 223-239. doi:10.1017/s0305004100019575. https://doi.org/10.1017/s0305004100019575
6. Stewartson, K. (1954). Further solutions of the Falkner-Skan equation. Mathematical Proceedings of the Cambridge Philosophical Society 50(3), 454-465. doi:10.1017/s030500410002956x. https://doi.org/10.1017/s030500410002956x
7. Coppel, W. A. (1960). On a differential equation of boundary-layer theory. Philosophical Transactions of the Royal Society of London. Series A 253(1023), 101-136. doi:10.1098/rsta.1960.0019. https://doi.org/10.1098/rsta.1960.0019
8. Craven, A. H.; Peletier, L. A. (1972). On the uniqueness of solutions of the Falkner-Skan equation. Mathematika 19(1), 129-133. doi:10.1112/s0025579300005052. https://doi.org/10.1112/s0025579300005052
9. Craven, A. H.; Peletier, L. A. (1972). Reverse flow solutions of the Falkner-Skan equation for λ > 1. Mathematika 19(1), 135-138. doi:10.1112/s0025579300005064. https://doi.org/10.1112/s0025579300005064
10. Hastings, S. P.; Troy, W. (1985). Oscillatory solutions of the Falkner-Skan equation. Proceedings of the Royal Society of London. Series A 397(1813), 415-418. doi:10.1098/rspa.1985.0021. https://doi.org/10.1098/rspa.1985.0021
11. Hastings, S. P.; Troy, W. C. (1988). Oscillating solutions of the Falkner-Skan equation for positive β. Journal of Differential Equations 71(1), 123-144. doi:10.1016/0022-0396(88)90041-1. https://doi.org/10.1016/0022-0396(88)90041-1
12. Swinnerton-Dyer, H. P. F.; Sparrow, C. T. (1995). The Falkner-Skan Equation I. The Creation of Strange Invariant Sets. Journal of Differential Equations 119(2), 336-394. doi:10.1006/jdeq.1995.1094. https://doi.org/10.1006/jdeq.1995.1094
13. Sparrow, C.; Swinnerton-Dyer, H. P. F. (2002). The Falkner-Skan Equation II: Dynamics and the Bifurcations of P- and Q-Orbits. Journal of Differential Equations 183(1), 1-55. doi:10.1006/jdeq.2001.4100. https://doi.org/10.1006/jdeq.2001.4100
14. Wang, J.; Gao, W. (1999). Uniqueness, existence and nonexistence of normal solutions to a problem in boundary layer theory. Tohoku Mathematical Journal (2) 51(4), 539-553. doi:10.2748/tmj/1178224718. https://www.jstage.jst.go.jp/article/tmj1949/51/4/51_4_539/_article
15. Lan, K. Q.; Yang, G. C. (2008). Positive Solutions of the Falkner–Skan Equation Arising in the Boundary Layer Theory. Canadian Mathematical Bulletin 51(3), 386-398. doi:10.4153/CMB-2008-039-7. https://doi.org/10.4153/CMB-2008-039-7
16. Yang, G. C.; Lan, K. Q. (2011). Nonexistence of the reversed flow solutions of the Falkner-Skan equations. Nonlinear Analysis: Theory, Methods & Applications 74(16), 5327-5339. doi:10.1016/j.na.2011.05.017. https://doi.org/10.1016/j.na.2011.05.017
17. Yang, G. C.; Lan, K. Q. (2013). Systems of singular integral equations and applications to existence of reversed flow solutions of Falkner-Skan equations. Communications on Pure and Applied Analysis 12(6), 2465-2495. doi:10.3934/cpaa.2013.12.2465. https://doi.org/10.3934/cpaa.2013.12.2465
18. Padé, O. (2003). On the solution of Falkner-Skan equations. Journal of Mathematical Analysis and Applications 285(1), 264-274. doi:10.1016/S0022-247X(03)00402-5. https://doi.org/10.1016/S0022-247X(03)00402-5
19. Fazio, R. (2019). The iterative transformation method. International Journal of Non-Linear Mechanics 116, 181-194. doi:10.1016/j.ijnonlinmec.2019.06.011. Open-access manuscript: arXiv:2003.07971, https://arxiv.org/abs/2003.07971
20. Sparrow, C. T. (1997). Bifurcations in the Falkner-Skan equation. In: Judd, K.; Mees, A.; Teo, K. L.; Vincent, T. L. (eds.), Control and Chaos. Mathematical Modelling, vol. 8, pp. 91-105. Birkhäuser Boston. doi:10.1007/978-1-4612-2446-4_6. https://doi.org/10.1007/978-1-4612-2446-4_6