Nonlinear Stokes Phenomenon Insights

• Nonlinear Stokes phenomenon is the smooth, rapid switching of exponentially small contributions in nonlinear differential and difference equations as solutions cross specific curves or surfaces.
• Exponential asymptotics coupled with error function smoothing reveal universal profiles that transition formal series beyond all orders.
• Applications in q-Painlevé equations, Hamiltonian systems, and matrix integrals provide analytic connection formulas and geometric insights into Stokes sets.

The nonlinear Stokes phenomenon describes the rapid, but in fact smooth, activation or deactivation of exponentially small contributions across specific curves or hypersurfaces in the domain of solutions to nonlinear differential equations, difference equations, or PDEs. Unlike the linear Stokes phenomenon, which concerns piecewise constants in the asymptotics of linear systems, the nonlinear case involves sectoral normalization jumps of analytic or formal solutions in singular perturbation problems, Painlevé equations, $q$-difference hierarchies, and multidimensional boundary-layer systems. The distinction between formal and true analytic solutions manifests as the presence of exponentially small “beyond all orders” contributions that switch across Stokes curves, surfaces, or rays, with the nature of these transitions governed by the nonlinear interactions of singularities in the Borel or Laplace plane.

1. Foundational Concepts and Definition

The nonlinear Stokes phenomenon generalizes the linear jump in asymptotics associated with “Stokes directions” to settings where the governing equations are nonlinear, or where multiple interacting singularities are present. At an irregular singularity, the analytic classification of solutions includes, beyond local formal invariants, a set of nonlinear automorphisms (“Stokes automorphisms”) associated with overlapping sectoral summations of the same formal normal form. These automorphisms, defined as analytic isotropies between the sectoral realizations, encode the nonlinear Stokes data, and their explicit computation provides the analytic connection formulae relating true solutions in adjacent sectors of the complex domain (Klimes, 2017).

Exponential asymptotics reveals that the late terms in a divergent power series expansion for a nonlinear singularly-perturbed system exhibit factorial-over-power growth, with singulant functions $\chi$ or $\eta$ emerging as solutions to eikonal-type equations. Optimally truncating the formal series leaves an exponentially small remainder, which changes rapidly but smoothly across Stokes curves or surfaces defined by conditions such as $\Im \chi = 0$ and $\Re \chi > 0$ (Howls et al., 7 Oct 2024, Joshi et al., 2018, Johnson-Llambias et al., 2020).

2. Exponential Asymptotics and Stokes Smoothing

The exponential asymptotics program involves extracting and resumming beyond-all-orders terms in a divergent asymptotic sequence. For a broad class of problems—ODEs, difference equations, and PDEs—these exponentially small contributions become non-negligible when their prefactors undergo rapid transition across Stokes loci. In classical work, Dingle and Berry demonstrated that, after optimal truncation of a divergent series and asymptotic expansion of the resulting remainder, the Stokes switching is universally smoothed: the discontinuous jump is replaced with an error function profile, localized near the Stokes set (Johnson-Llambias et al., 2020, Howls et al., 7 Oct 2024).

Recent advances rigorously confirm that higher-order (multi-singulant) Stokes phenomena are also universally smoothed, with the corresponding multipliers expressed as Gaussian-convolution error functions generalizing the simple error function smoothing of the lowest order. At each level, the “ghost” exponentially small contributions emerge only in the vicinity of the relevant Stokes locus and decay rapidly away from it (Howls et al., 7 Oct 2024).

3. Nonlinear Stokes Phenomenon in Specific Systems

a) $q$-Difference Painlevé Equations:

For the $q$-Painlevé I equation, formal expansions in (small) $\epsilon$ yield a quartic leading-order equation, whose roots produce distinct local asymptotic solutions (“Type A” and “Type B”). Exponential asymptotics reveals Stokes curves described by $q$-spirals, across which exponentially small corrections associated with different singulant branches $e^{-\chi_j / \epsilon}$ either become active or are suppressed. The jump in the Stokes multiplier, calculated through local analysis and Berry smoothing, is described via an error function in the relevant scaled variables. Multiple branch points and interacting singularities in this nonlinear $q$-difference context lead to “sectoral” solutions with overlapping regions and “nonlinear” Stokes transitions between them (Joshi et al., 2018).

b) Non-autonomous Hamiltonian Systems and Painlevé Confluence:

In the confluence limit of non-autonomous Hamiltonian systems (e.g., the degeneration of Painlevé VI to Painlevé V), two regular singularities merge into an irregular singular one, producing complex sectoral gluing data for normalized solutions. The nonlinear Stokes phenomenon is captured by Stokes automorphisms, which mediate the analytic connection between 1-summed sectoral normalizations and constitute the “wild monodromy group”—the nonlinear analogue of classical monodromy. The analytic equivalence class of families is determined by both the formal Birkhoff–Siegel invariants and the full set of nonlinear Stokes data (Klimes, 2017).

c) Matrix Integrals and Tau Functions:

For integrals such as the Kontsevich matrix model, which converge in the $n \to \infty$ limit to isomonodromic tau functions of the Painlevé I hierarchy, the nonlinear Stokes phenomenon manifests as exponentially small differences between sectoral tau functions. Each such tau function shares the same divergent formal expansion but is distinct by an exponential connection factor as parameters cross Stokes rays in the coupling parameter space. These global connection formulas, characterized by exponential functions of explicit residue sums, realize the nonlinear Stokes phenomenon at a fully non-perturbative level (Bertola et al., 2016).

4. Geometry and Computation of Stokes Sets

The geometry of Stokes sets in nonlinear problems extends beyond rays to lines, surfaces, or higher-dimensional manifolds depending on the underlying system. In three-dimensional free-surface flow, for example, the Stokes phenomenon is realized by switching on exponentially small waves across three-dimensional “Stokes surfaces.” The relevant singulant function, solving an eikonal PDE with boundary conditions derived from the physical problem, defines manifolds $\Im \chi(x, y, z) = 0$, whose computation may require numerical ray-tracing when analytic expressions are unavailable (Johnson-Llambias et al., 2020).

In difference equations, the $q$-lattice structure induces a $q$-spiral geometry for the Stokes sets, contrasting with the straight rays of differential equations. The multiplicity of singularities and the periodicity of leading-order solutions generate a tessellation of domains with distinct exponentially small sectoral contributions (Joshi et al., 2018).

5. Higher-Order and Multi-Singulant Effects

The higher-order Stokes phenomenon addresses the switching of subdominant array of exponentially small contributions, which originate from nested or interacting singularities in the Borel plane or complex domain. The activation of these sub-subdominant terms is no longer governed by a simple error function multiplier but by higher-level special functions—explicitly, a Gaussian-convolution of the error function in the second-level hyperterminant case. These “ghost-like” smooth profiles explain fine structure in the appearance and disappearance of late terms in multisingulant transseries, and are robustly present in nonlinear equations, difference equations, and nonlinear PDEs (e.g., Painlevé, telegraph equation, nonlinear ODEs) (Howls et al., 7 Oct 2024).

6. Analytic Classification, Monodromy, and Connection Data

The analytic classification of nonlinear systems near irregular singular points is controlled by a complete set of formal invariants (e.g., Birkhoff–Siegel normal forms, Lenard polynomials in Painlevé hierarchies) together with the full set of nonlinear Stokes (sectoral) automorphisms or connection matrices. In isomonodromic tau functions, each true analytic solution is associated to a sector, and the connection formulas (exponentials of contour integrals involving residues of squared Lax connections) encode the jump as one crosses Stokes rays in parameter space. Analytic equivalence is achieved if and only if the totality of these invariants is matched modulo allowed symmetries (Klimes, 2017, Bertola et al., 2016).

7. Examples and Applications

Significant progress has been made in elucidating the nonlinear Stokes phenomenon in a range of contexts:

• Low-Froude flows (Stokes surfaces demarcating wave and wave-free regions) (Johnson-Llambias et al., 2020)
• $q$-Painlevé dynamics (q-spiral Stokes structures and multi-branch sectorality) (Joshi et al., 2018)
• Kontsevich integral and isomonodromic tau functions (sectoral τ-functions differing by explicit exponential connection factors) (Bertola et al., 2016)
• Multisingulant transseries (arising in higher-level hyperasymptotic expansions, with corresponding ghost-hump smoothing) (Howls et al., 7 Oct 2024)

These results confirm that, across a wide spectrum of nonlinear systems, sectoral solutions related by nonlinear Stokes data, rigorous smoothing of jumps in exponentially small terms, and intricate geometric structure of Stokes sets are fundamental to understanding the global and local analytic structure of solutions.

---

Key References:

• "Smoothing of the higher-order Stokes phenomenon" (Howls et al., 7 Oct 2024)
• "Three-dimensional exponential asymptotics and Stokes surfaces for flows past a submerged point source" (Johnson-Llambias et al., 2020)
• "Nonlinear $q$-Stokes phenomena for $q$-Painlevé I" (Joshi et al., 2018)
• "Stokes phenomenon and confluence in non-autonomous Hamiltonian systems" (Klimes, 2017)
• "The Kontsevich matrix integral: convergence to the Painlevé hierarchy and Stokes' phenomenon" (Bertola et al., 2016)