Propagation phenomena in KPP-bistable periodic patchy environments

Abstract: This paper first investigates the propagation dynamics of solutions to the Cauchy problem for a one-dimensional reaction-diffusion equation in a spatially periodic environment consisting of two distinct patch types. The novelty of this work lies in the systematic analysis of a KPP-bistable heterogeneous framework. In this setting, the respective patch lengths, the linear stability of the zero solution and the positive periodic steady state, and the magnitude of the initial data play crucial roles in the long-time dynamics. We first establish persistence properties of the species, showing that uniform persistence holds when the zero steady state of the associated periodic patch model is unstable, while local persistence is obtained under additional suitable conditions. Using a dynamical systems approach, we further establish spreading properties and demonstrate the existence of pulsating traveling waves in two different cases, depending on whether the trivial solution is unstable or stable. Finally, we present two sets of sufficient conditions characterizing species extinction.

• The paper establishes explicit critical patch-length thresholds determining when the zero state is stable or unstable, revealing how KPP and bistable patch geometry controls persistence.
• The analysis proves uniform persistence and a positive spreading speed equal to the minimal pulsating-wave speed when the zero state is unstable, while also establishing positive-speed fronts in selected bistable regimes.
• The results identify extinction for sufficiently small data and for large populations trapped in unfavorable bistable patches, while leaving general propagation, blocking, homogenization, and diffusion-limit behavior open.

Model and context

The paper by Griette, Hamel, Zhang, and Zhao studies the Cauchy problem for a one-dimensional reaction-diffusion equation in a spatially periodic environment composed of alternating patches of two types: patches governed by KPP (monostable) kinetics and patches governed by bistable kinetics. The equation is

$$u_t - d(x)u_{xx} = f(x,u), \qquad t>0,\ x\in\mathbb{R}\setminus S,$$

with density continuity and flux-jump transmission conditions at the interface sets $S_1$ and $S_2$: $u_x(t,x^-)=\sigma u_x(t,x^+)$ at $S_1$ and $\sigma u_x(t,x^-)=u_x(t,x^+)$ at $S_2$, where $\sigma=(1-\alpha)/\alpha$ encodes the probability $\alpha$ that an individual crossing an interface enters a patch of type 1. Within each patch the coefficients are constant, $d(x)=d_i$ and $S_1$0, with $S_1$1 of KPP type (positive on $S_1$2, sublinear in the KPP sense) and $S_1$3 of bistable type (zeros at $S_1$4, with $S_1$5). The period is $S_1$6, where $S_1$7 is the length of patch type $S_1$8. This interface convention—density continuous, flux discontinuous—follows the framework of Hamel, Lutscher, and Zhang (2608.17474), and differs from the Maciel–Lutscher conditions, which preserve flux while allowing density jumps.

The stated contribution is the first systematic analysis of propagation dynamics in a periodic patch model mixing monostable and bistable nonlinearities. The main analytical difficulty, emphasized by the authors, is that in the purely KPP setting the spreading speed is characterized by the principal eigenvalue of the linearization at zero, whereas in the mixed KPP-bistable setting the sign of $S_1$9 alone does not determine the large-time dynamics; the stability of the positive periodic steady state, the patch lengths, and the magnitude of the initial data all intervene.

Stability thresholds for the trivial state

Linearizing at zero yields a periodic principal eigenvalue problem with the same interface conditions. The sign of $S_2$0 determines linear stability of $S_2$1, and the authors derive explicit critical lengths. For fixed $S_2$2, there is a threshold $S_2$3 given by

$S_2$4

such that $S_2$5 is unstable ($S_2$6) if $S_2$7 and stable if $S_2$8. As $S_2$9, $u_x(t,x^-)=\sigma u_x(t,x^+)$0 converges to a finite limit

$u_x(t,x^-)=\sigma u_x(t,x^+)$1

so that if $u_x(t,x^-)=\sigma u_x(t,x^+)$2, the trivial state is unstable regardless of how large the bistable patches are. Symmetrically, for $u_x(t,x^-)=\sigma u_x(t,x^+)$3 there is a threshold $u_x(t,x^-)=\sigma u_x(t,x^+)$4, increasing in $u_x(t,x^-)=\sigma u_x(t,x^+)$5, such that $u_x(t,x^-)=\sigma u_x(t,x^+)$6 is unstable for $u_x(t,x^-)=\sigma u_x(t,x^+)$7 and stable for $u_x(t,x^-)=\sigma u_x(t,x^+)$8. These monotonicity properties yield a sharp stability transition in the $u_x(t,x^-)=\sigma u_x(t,x^+)$9 parameter plane. The authors note that whether $S_1$0 is stable at $S_1$1 remains unclear in this mixed setting.

Persistence

Two persistence results are established. First, when $S_1$2 is unstable ($S_1$3), the authors prove uniform persistence: for every nonnegative, nonzero, compactly supported initial datum,

$S_1$4

The proof uses the Dirichlet principal eigenvalue on a large interval, which converges to $S_1$5 from above, to construct a small subsolution $S_1$6 dominated by $S_1$7; the monotone solution issued from this subsolution converges locally uniformly to a positive bounded solution of the stationary problem, which lies below $S_1$8 for large times. As a by-product of the same construction, the authors prove that when $S_1$9 the stationary problem admits a minimal positive bounded periodic solution $\sigma u_x(t,x^-)=u_x(t,x^+)$0, below every positive bounded solution; this resolves uniqueness of the positive periodic steady state, which is not automatic in the presence of bistable patches.

Second, a local persistence result holds independently of the stability of zero. If $\sigma u_x(t,x^-)=u_x(t,x^+)$1 (the bistable kinetics are favorable in the integral sense), then for any $\sigma u_x(t,x^-)=u_x(t,x^+)$2 there is $\sigma u_x(t,x^-)=u_x(t,x^+)$3 such that, whenever $\sigma u_x(t,x^-)=u_x(t,x^+)$4 and $\sigma u_x(t,x^-)=u_x(t,x^+)$5 on an interval of size $\sigma u_x(t,x^-)=u_x(t,x^+)$6 inside a bistable patch, the solution satisfies $\sigma u_x(t,x^-)=u_x(t,x^+)$7 on every bounded interval $\sigma u_x(t,x^-)=u_x(t,x^+)$8. The argument is a localization: as $\sigma u_x(t,x^-)=u_x(t,x^+)$9 with $S_2$0 fixed, the solution near the center of a large bistable patch converges locally to the ODE flow $S_2$1 starting at $S_2$2, which tends to $S_2$3; a compactly supported profile $S_2$4 solving $S_2$5 with maximum above $S_2$6 then serves as a subsolution, and comparison propagates the conclusion to all bounded intervals.

Spreading speed and pulsating traveling waves in the monostable regime

When $S_2$7, the authors invoke the abstract theory of Liang–Zhao for monostable evolution systems. The solution maps $S_2$8 on $S_2$9 form a monotone, translation-invariant (for shifts in $\sigma=(1-\alpha)/\alpha$0), compact semiflow with exactly two periodic fixed points, $\sigma=(1-\alpha)/\alpha$1 and $\sigma=(1-\alpha)/\alpha$2. After shifting coordinates so the configuration is symmetric under reflection, the rightward and leftward spreading speeds coincide, giving a single asymptotic speed $\sigma=(1-\alpha)/\alpha$3 such that, for compactly supported $\sigma=(1-\alpha)/\alpha$4 with $\sigma=(1-\alpha)/\alpha$5:

• for $\sigma=(1-\alpha)/\alpha$6, $\sigma=(1-\alpha)/\alpha$7;
• for $\sigma=(1-\alpha)/\alpha$8, $\sigma=(1-\alpha)/\alpha$9, for any nonzero $\alpha$0.

Positivity of $\alpha$1 follows from an induction argument showing $\alpha$2. The authors then prove that pulsating traveling waves $\alpha$3 connecting $\alpha$4 to $\alpha$5 exist if and only if $\alpha$6; hence the asymptotic spreading speed coincides with the minimal wave speed. This parallels the classical KPP-patch theory but requires the minimal-solution machinery because bistable patches preclude a direct variational or purely linear characterization of $\alpha$7; notably, no explicit formula for $\alpha$8 is provided.

Pulsating fronts in the bistable regime

A second propagation result covers the case where $\alpha$9 is stable. Assume $d(x)=d_i$0. Then there exist $d(x)=d_i$1 and $d(x)=d_i$2 such that for all $d(x)=d_i$3 and $d(x)=d_i$4 the problem admits a pulsating front connecting a positive periodic steady state $d(x)=d_i$5 to $d(x)=d_i$6, with strictly positive speed $d(x)=d_i$7.

The proof is the most technical part of the paper and verifies the abstract bistable-semiflow framework of Fang–Zhao. The key steps are:

• Existence and linear stability of $d(x)=d_i$8: a periodic subsolution built from $d(x)=d_i$9 in each bistable patch generates, by monotone iteration, a positive periodic steady state $S_1$00; its principal eigenvalue satisfies $S_1$01, with no restriction on $S_1$02.
• Strong stability from below: via a contradiction argument combining $S_1$03 uniformly as $S_1$04, $S_1$05 (proved by blow-up/compactness in two cases according to $S_1$06) and the unboundedness of solutions of $S_1$07, the authors obtain $S_1$08, so $S_1$09 is a strict subsolution for small $S_1$10.
• Instability of intermediate steady states: the Dancer–Hess connecting-orbit theorem yields at least one steady state $S_1$11 with $S_1$12; a delicate phase-plane analysis (Hamiltonian first integral, Sturm comparison, and exclusion of constant, periodic, and ground-state limits) shows that all such $S_1$13 are unstable when $S_1$14 is small and $S_1$15 large, and that no steady state lies strictly between $S_1$16 and $S_1$17 or between $S_1$18 and $S_1$19.
• Counter-propagation: spreading speeds $S_1$20 and $S_1$21 are shown to be positive on each monostable layer, which is stronger than the Fang–Zhao counter-propagation condition.
• Positive wave speed: a contradiction argument rules out $S_1$22 (a subsolution argument using $S_1$23) and $S_1$24 (a stationary-front profile argument contradicting $S_1$25).

The conclusion is that the system exhibits genuine bistable dynamics with two stable periodic states, $S_1$26 and $S_1$27, separated by a moving front. The authors emphasize that whether solutions propagate in this regime—i.e., whether large initial bumps can cross interfaces and invade neighboring patches—remains an open problem, the obstacle being the lack of translation invariance of the patchy model.

Extinction

Two extinction mechanisms are characterized. First, if $S_1$28 and $S_1$29 is sufficiently small, then $S_1$30; the proof compares $S_1$31 with an exponentially decaying supersolution $S_1$32 built from the principal eigenpair. The admissible threshold $S_1$33 here depends on $S_1$34, hence on $S_1$35.

Second, a stronger extinction result holds for large initial data concentrated in an unfavorable bistable patch: if $S_1$36, $S_1$37, $S_1$38, and $S_1$39 lies inside a bistable patch at distance at least $S_1$40 from the interfaces with $S_1$41, then $S_1$42 uniformly. The proof combines two ingredients. A finite-time spreading estimate shows that $S_1$43 remains below $S_1$44 outside the initial bistable patch up to a fixed time $S_1$45, uniformly in $S_1$46. Inside the patch, the solution is dominated by a pair of symmetric fronts for a modified bistable nonlinearity $S_1$47 with negative integral, which drive the solution below $S_1$48 everywhere by time $S_1$49. A separate lemma then shows that for $S_1$50 and $S_1$51 large, any solution with initial datum bounded by a constant $S_1$52 independent of $S_1$53 goes extinct; this uniformity is essential because the eigenvalue threshold of the first extinction theorem degenerates as $S_1$54 ($S_1$55 when $S_1$56). The uniform lower bound $S_1$57 and the uniform positivity of the eigenfunction on KPP patches (proved by a compactness argument excluding zeros on patches and interfaces) are the technical core. By contrast, the authors note that when $S_1$58 is fixed and $S_1$59 is sufficiently small, whether extinction occurs remains open.

Limitations and open problems

The paper is explicit about several gaps. The characterization of spreading is complete only in the monostable regime ($S_1$60) and for pulsating-front existence (not propagation) in the bistable regime; blocking—propagation failure depending on the relative patch scales—is conjectured to be possible but not rigorously characterized. Whether large localized initial data propagate across interfaces when $S_1$61 is stable is open. The case $S_1$62 fixed with $S_1$63 small is unresolved for extinction. The authors also propose a companion model with hostile patches, $S_1$64 in patch 2, for which the bistable-patch arguments do not apply (no positive equilibrium in patch 2), and ask for the spreading speed and its explicit characterization. Finally, the homogenization limit $S_1$65—whether the hybrid system converges to an effective KPP, bistable, or intermediate equation—and the fast/slow diffusion limits $S_1$66 or $S_1$67 are left open.

Conclusion

The paper provides a complete qualitative picture of the Cauchy problem for a KPP-bistable periodic patch model with flux-jump interface conditions: sharp stability thresholds for the trivial state in terms of explicit critical patch lengths, uniform persistence and a spreading speed equal to the minimal pulsating-wave speed when $S_1$68 is unstable, local persistence and bistable pulsating fronts with positive speed when $S_1$69 is stable but the bistable patches are large and favorable, and two extinction mechanisms covering small data and large unfavorable bistable patches. The central methodological contribution is the adaptation of monostable and bistable semiflow theories to a heterogeneous setting in which the spreading behavior is governed jointly by the linear stability of both steady states, the geometry of the patches, and the initial data, rather than by linearization at the trivial state alone.