Canards in a bottleneck

1. Problem Statement

The paper investigates the stationary profiles of a nonlinear Fokker-Planck equation describing the unidirectional cross-sectional average flow of pedestrians in a corridor with a single entrance and exit. The specific focus is on corridors containing a bottleneck—a region where the width has a unique, global, non-degenerate minimum in the interior of the domain.

The mathematical model is a 1D area-averaged PDE derived from a 2D convection-diffusion equation:
$$\partial_x (k(x) j(x)) = 0$$
where the flux is $j(x) = -\varepsilon \partial_x \rho + \rho(1-\rho)$, $\rho$ is the pedestrian density, $\varepsilon$ is a small diffusion coefficient, and $k(x)$ represents the variable width of the corridor. The system is subject to nonlinear boundary conditions:

• Inflow (at $x=0$): $j = \alpha(1-\rho)$, where $\alpha$ is the inflow rate.
• Outflow (at $x=1$): $j = \beta\rho$, where $\beta$ is the outflow rate.

The primary goal is to characterize the stationary solutions $\rho(x)$ in the singular limit ($\varepsilon \to 0$) and understand how the bottleneck geometry and boundary rates ($\alpha, \beta$) influence the formation of boundary layers and density transitions.

2. Methodology

The authors employ Geometric Singular Perturbation Theory (GSPT), a dynamical systems approach for singularly perturbed ordinary differential equations.

• Slow-Fast Decomposition: The problem is reformulated as a 3D autonomous system with a small parameter $\varepsilon$.
  • Fast variable: Density $\rho$ (evolves on the scale $\tau = x/\varepsilon$).
  • Slow variables: Flux $j$ and spatial coordinate $x$ (evolve on the scale $x$).
• Limiting Problems:
  • Layer Problem ($\varepsilon=0$ in fast time): Describes rapid transitions (boundary layers) where $\rho$ jumps between branches of the critical manifold.
  • Reduced Problem ($\varepsilon=0$ in slow time): Describes the slow evolution of the system constrained to the critical manifold $C_0 = \{ (j, x, \rho) : j = \rho(1-\rho) \}$.
• Critical Manifold Analysis: The manifold $C_0$ is a folded surface with an attracting branch ($C_0^a$, $\rho > 1/2$), a repelling branch ($C_0^r$, $\rho < 1/2$), and a fold line $F$ at $\rho = 1/2$.
• Canard Phenomenon: The core of the analysis focuses on the behavior at the bottleneck's minimum ($x^*$). The authors identify a folded saddle point on the fold line. They demonstrate the existence of canard solutions—trajectories that follow the repelling branch of the critical manifold for a significant time before jumping. This allows solutions to transition from high density to low density inside the bottleneck, a phenomenon not possible in monotone geometries.
• Shooting Strategy: The authors construct singular solutions by matching the forward evolution of the inflow boundary conditions with the backward evolution of the outflow conditions, specifically analyzing their intersection at the bottleneck location $x^*$.

3. Key Contributions

1. Identification of Canard Solutions in BVPs: The paper establishes that the bottleneck geometry induces canard solutions in a boundary value problem (BVP). These solutions allow the system to pass through the repelling branch of the critical manifold, facilitating a transition from high to low density within the narrowest part of the corridor.
2. Classification of 8 Solution Regimes: By varying the inflow ($\alpha$) and outflow ($\beta$) rates, the authors divide the parameter space into 8 distinct regions ($G_1$ to $G_8$). Each region corresponds to a unique structural profile of the stationary solution.
3. Detailed Bifurcation Diagram: The authors construct a comprehensive bifurcation diagram showing how the solution structure changes as $\alpha$ and $\beta$ vary. This includes transitions between:
   • High-density states with entrance boundary layers.
   • Low-density states with exit boundary layers.
   • Transitional states involving canards and boundary layers at both ends.
4. Analytical and Numerical Validation: The theoretical predictions derived from GSPT are rigorously proven for the singular limit and corroborated by numerical experiments for small but non-zero $\varepsilon$ ($10^{-3}$) using both cosine-shaped and "supergaussian" bottleneck profiles.

4. Key Results

The analysis reveals three main types of stationary profiles, further subdivided into 8 types based on boundary layer configurations:

• Type I (High Density): Solutions remain on the attracting branch $C_0^a$ (high density, $\rho > 1/2$) throughout the domain, with a boundary layer only at the entrance (if $\alpha$ is small) or exit.
• Type II (Low Density): Solutions remain on the repelling branch $C_0^r$ (low density, $\rho < 1/2$), with boundary layers only at the exit (if $\beta$ is small) or entrance.
• Type III (Transitional/Canard): This is the novel contribution. Solutions start at high density, follow the attracting branch, pass through the canard point at the bottleneck minimum ($x^*$), traverse the repelling branch, and exit at low density.
  • Depending on $\alpha$ and $\beta$, these transitional profiles can have boundary layers at the entrance, the exit, or both.
  • Region $G_3, G_4, G_5, G_6$ correspond to these transitional profiles. The size of these regions in the $(\alpha, \beta)$ parameter space is inversely proportional to the bottleneck width; narrower bottlenecks expand the parameter space where canard-mediated transitions occur.

Flux Behavior:

• In low/high density regimes, the flux $J$ is determined by the boundary rates ($J \approx \alpha(1-\alpha)$ or $J \approx \beta(1-\beta)$) and is largely independent of the bottleneck width.
• In the transitional (canard) regime, the flux is limited by the bottleneck capacity: $J \approx \frac{1}{4} \min k(x)$. The flux scales linearly with the minimum width of the bottleneck.

5. Significance

• Pedestrian Dynamics: The results provide a mechanistic explanation for how pedestrian flows behave in constrained spaces (e.g., stadiums, subway stations). It explains how high-density crowds can transition to low-density flows through a bottleneck without forming a jam at the exit, provided the inflow/outflow rates fall within specific ranges.
• Mathematical Physics: The paper extends the application of GSPT and canard theory from initial value problems to nonlinear boundary value problems. It demonstrates that canards are not just mathematical curiosities but structurally stable features in physical models with geometric constraints.
• Modeling Complexity: The study highlights that simple monotonic assumptions about corridor width are insufficient to capture complex flow dynamics. The presence of a local minimum (bottleneck) fundamentally alters the solution space, introducing new regimes of behavior (the "maximum flux" phase) that are critical for safety and capacity planning.

In summary, the paper successfully bridges geometric singular perturbation theory with pedestrian flow modeling, revealing that bottlenecks act as "canard generators" that enable complex density transitions, fundamentally altering the bifurcation structure of the flow compared to monotone corridors.