Physical Blowup Dynamics in Nonlinear Systems

• Physical blowup dynamics is the study of finite-time singularity formation in nonlinear systems, where mass collapses, field amplitudes diverge, or gradients steepen.
• The topic examines model-dependent phenomena in areas such as chemotaxis, nonlinear Schrödinger equations, and fluid dynamics, emphasizing energy competition and critical thresholds.
• Analytical methods like dynamic rescaling and modulation enable the extraction of asymptotic profiles and invariant manifolds that organize the singular behavior.

Searching arXiv for the cited topic and closely related blowup-dynamics work. Physical blowup dynamics denotes finite-time singularity formation in nonlinear evolution equations and dynamical systems: mass collapsing into smaller and smaller regions, particle collisions/aggregation, infinite field amplitude, shock-type blowup, or a macroscopic synchronous event, depending on the model (Calvez et al., 2013, Miller, 5 Aug 2025, Papadopoulos et al., 21 Aug 2025). Across chemotaxis, dispersive wave equations, geometric flows, turbulence models, fluid models, driven condensates, and mean-field neural dynamics, blowup is typically organized by a small number of recurrent structures: a free energy or Lyapunov functional, a critical profile or static state, a threshold manifold, and a rescaled or compactified description in which the singular regime becomes asymptotically regular (Dejak et al., 2013, Bizoń et al., 2019, Matsue, 2023).

1. Blowup as a model-dependent singularity concept

In continuum aggregation models, blowup means that a density develops a singularity in finite time, typically concentrating mass into a Dirac delta or an arbitrarily narrow peak. In the one-dimensional self-attracting diffusive particle model, the particle-system analogue is collision: an inter-particle spacing $X_{i+1}(t)-X_i(t)\to 0$ in finite time, so that the trajectory hits the boundary of the ordered configuration set $\Pi$ (Calvez et al., 2013). In the reduced two-dimensional Keller–Segel system, blowup is described by $\|\rho(\cdot,t)\|_{L^\infty}\to\infty$ and, in distributional form, $\rho(x,t)\rightharpoonup 8\pi\,\delta(x)$ as $t\to T$ (Dejak et al., 2013).

In dispersive and geometric equations, the singular observable depends on the critical norm. For the focusing nonlinear Schrödinger equation in the $L^2$-supercritical, $H^1$-subcritical regime, blowup means that the maximal existence time $T_{\max}$ is finite and $\|\nabla u(t)\|_{L^2}\to\infty$ as $t\to T_{\max}$ (Gustafson et al., 2022). For the 3D mass-critical half-wave equation, the blowup alternative is expressed by $\Pi$0, and the constructed minimal-mass regime is tracked through $\Pi$1 (Georgiev et al., 2020). For the energy-critical Landau–Lifshitz flow, the singularity is a type-II concentration of a rescaled harmonic map $\Pi$2 (Xu et al., 2020).

Blowup is not restricted to amplitude divergence. In the infinite-dimensional vorticity equation obtained from the $\Pi$3 limit of axisymmetric, swirl-free Euler, the singularity is of Burgers shock type: the characteristic map ceases to be invertible and gradients blow up like $\Pi$4 (Miller, 5 Aug 2025). In the delayed Poissonian mean-field neural model, blowup is a divergence of the mean-field interaction rate, with a finite fraction of neurons spiking simultaneously, thereby marking a macroscopic synchronous event (Papadopoulos et al., 21 Aug 2025). A related but distinct phenomenon is grow-up: global existence with unbounded growth of the $\Pi$5 norm, as in threshold NLS solutions with $\Pi$6 (Gustafson et al., 2022).

2. Energetic mechanisms and competing effects

A recurrent mechanism is competition between a regularizing contribution and a concentrating one. In the self-attracting diffusive particle system, the continuum free energy is

$\Pi$7

with $\Pi$8, $\Pi$9 (Calvez et al., 2013). The entropy $\|\rho(\cdot,t)\|_{L^\infty}\to\infty$0 is convex and encodes diffusion, while the interaction energy favors concentration. In the discrete Lagrangian approximation, this becomes a competition between discrete entropy

$\|\rho(\cdot,t)\|_{L^\infty}\to\infty$1

and discrete interaction energy

$\|\rho(\cdot,t)\|_{L^\infty}\to\infty$2

with blowup or global existence decided by which term dominates (Calvez et al., 2013).

The reduced Keller–Segel system has the gradient-flow energy

$\|\rho(\cdot,t)\|_{L^\infty}\to\infty$3

and blowup reflects chemotactic drift overwhelming diffusion when the critical mass $\|\rho(\cdot,t)\|_{L^\infty}\to\infty$4 is exceeded (Dejak et al., 2013). In physical language, the model describes organisms diffusing while moving up gradients of their own chemoattractant, so aggregation is driven by a self-generated attractive potential (Dejak et al., 2013).

Not all blowup-like dynamics arise from conservative focusing. In coherently driven polariton condensates, a repulsive interaction makes a 2D polariton condensate able to accumulate its energy under above-resonance optical pumping; the process begins when the field density reaches the parametric scattering threshold that is inversely proportional to the polariton lifetime, and although the increase in energy may be arbitrarily slow in its beginning, it is followed by a blowup (Gavrilov, 2014). Here the mechanism is a positive feedback loop between above-resonance pumping, interaction-induced blueshift, and parametric scattering, rather than self-focusing in a conservative Hamiltonian model (Gavrilov, 2014).

This suggests that “blowup” is best understood structurally rather than by sign conventions alone. Diffusion, damping, or defocusing interactions do not by themselves preclude singular or explosive regimes; what matters is how they enter the full nonlinear feedback.

3. Thresholds, critical states, and separatrices

A second recurrent structure is a threshold set separating qualitatively different futures. For supercritical focusing wave equations outside a ball, the threshold for blowup is given by a codimension-one stable manifold of the unique static solution with exactly one unstable direction (Bizoń et al., 2019). The ground static state $\|\rho(\cdot,t)\|_{L^\infty}\to\infty$5 plays the role of the critical solution because the Dirichlet boundary at $\|\rho(\cdot,t)\|_{L^\infty}\to\infty$6 breaks scaling symmetry and excludes self-similar blowup profiles (Bizoń et al., 2019).

In the one-dimensional focusing nonlinear Klein–Gordon equation, even initial data fine-tuned to the threshold are trapped by the static solution $\|\rho(\cdot,t)\|_{L^\infty}\to\infty$7 for intermediate times, and the paper formulates the threshold as a codimension-one center-stable manifold $\|\rho(\cdot,t)\|_{L^\infty}\to\infty$8 separating blowup from global regularity (Bizoń et al., 2010). The trapping mechanism depends strongly on the nonlinearity exponent $\|\rho(\cdot,t)\|_{L^\infty}\to\infty$9: fast convergence to $\rho(x,t)\rightharpoonup 8\pi\,\delta(x)$0 for $\rho(x,t)\rightharpoonup 8\pi\,\delta(x)$1, slow convergence for $\rho(x,t)\rightharpoonup 8\pi\,\delta(x)$2, and very slow, oscillatory trapping for $\rho(x,t)\rightharpoonup 8\pi\,\delta(x)$3 (Bizoń et al., 2010).

At the ground-state threshold for focusing NLS, the sign of the virial functional

$\rho(x,t)\rightharpoonup 8\pi\,\delta(x)$4

distinguishes the focusing and dispersive sides when $\rho(x,t)\rightharpoonup 8\pi\,\delta(x)$5 and $\rho(x,t)\rightharpoonup 8\pi\,\delta(x)$6. If $\rho(x,t)\rightharpoonup 8\pi\,\delta(x)$7, the global alternatives are blow-up in one or both time directions, grow-up in one or both time directions, or the special solution $\rho(x,t)\rightharpoonup 8\pi\,\delta(x)$8 up to symmetries (Gustafson et al., 2022). The finite-variance assumption is not needed in this classification (Gustafson et al., 2022).

Energy-critical Landau–Lifshitz flow exhibits an analogous threshold geometry. There exists a codimension one smooth well-localized set of initial data arbitrarily close to the ground state which generates type-II finite-time blowup solutions (Xu et al., 2020). In the 1-equivariant class, Han and Tan construct another family of blowup solutions near the lowest energy steady state, with

$\rho(x,t)\rightharpoonup 8\pi\,\delta(x)$9

and $t\to T$0 arbitrary small in $t\to T$1 (Han et al., 2022).

Thresholds need not be encoded by a single mass parameter. In the self-attracting diffusive particle system, the criteria are inequalities involving second moment and energy; in the particle picture, global existence is guaranteed when the initial interaction intensity $t\to T$2 is small enough, whereas several incompatible criteria force finite-time collision (Calvez et al., 2013). This suggests that the geometry of the threshold can be moment-based, energy-based, or manifold-based, depending on the model.

4. Asymptotic blowup laws and singular profiles

Blowup dynamics is often classified by asymptotic profiles and rates after renormalization. In the reduced Keller–Segel model, the singular core is

$t\to T$3

and the scaling law is non-self-similar: $t\to T$4 as stated in the paper’s formula (1.5) (Dejak et al., 2013). The paper presents a formal derivation and partial rigorous results, and the formula coincides with the formula of Herrero and Velázquez for specially constructed solutions (Dejak et al., 2013).

In mass-supercritical NLS, the dynamic rescaling method leads to a profile equation

$t\to T$5

and the numerically observed stable blow-up dynamics has square-root rate

$t\to T$6

with blow-up profiles converging to the $t\to T$7 profile (Yang et al., 2018). The same study emphasizes the multi-bump structure of other solutions of the profile equation, but the monotone $t\to T$8 is the candidate stable blow-up profile (Yang et al., 2018).

For fractional dispersive models, the scaling laws are different. In the 3D mass-critical half-wave equation, there exists a class of radial minimal-mass blowup solutions with

$t\to T$9

and the decomposition uses modulation parameters $L^2$0 around a rescaled ground state $L^2$1 (Georgiev et al., 2020). In the one-dimensional inhomogeneous mass-critical half-wave equation, radial blowup solutions with $L^2$2 satisfy

$L^2$3

under mild assumptions on the inhomogeneous factor $L^2$4 (Li, 2022).

Geometric flows furnish bubbling rather than scalar self-similarity. For the 2D Landau–Lifshitz equation with $L^2$5, $L^2$6, finite-time blowup solutions are constructed in the form

$L^2$7

for any $L^2$8 (Han et al., 2022). In the energy-critical 1-equivariant Landau–Lifshitz flow with arbitrary $L^2$9, $H^1$0, the corresponding singularity formation has precise scale

$H^1$1

and the blowup rate is independent of the coefficients (Xu et al., 2020).

Not all singular profiles are spatial bubbles. In the infinite-dimensional vorticity equation,

$H^1$2

is exactly 1D inviscid Burgers in the axial variable, and the vorticity is

$H^1$3

so the singularity is a Burgers shock type blowup rather than concentration of a rescaled ground state (Miller, 5 Aug 2025). In shell models of turbulence, blowups generate coherent structures, instantons, which travel through the inertial range in finite time and are described by universal self-similar statistics (Mailybaev, 2013).

5. Physical realizations across model classes

The term “physical blowup dynamics” is therefore genuinely plural. Different models attach the singular event to different observables, but the events are all treated as physically interpretable concentration or collapse mechanisms.

Model class	Blowup observable	Organizing structure
Self-attracting diffusive particles	Particle collisions/aggregation	Competing convexities in free energy
Keller–Segel chemotaxis	Density concentration into $H^1$4	Critical mass and collapsing profile $H^1$5
Supercritical wave / NLKG / NLS	Field-amplitude or gradient blowup	Static state or ground state with one unstable mode
Half-wave equations	Divergence of $H^1$6	Minimal-mass ground-state modulation
Landau–Lifshitz flow	Type-II bubble concentration	Harmonic-map core and codimension-one set
Infinite-dimensional vorticity	Burgers shock type blowup	Axial compression / characteristics
Mean-field neural network	Divergent firing rate, synchronous spike event	Time-change fixed point and buffering
Shell turbulence	Instantons driving the inertial range	Universal self-similar statistics
Polariton condensates	Explosive transition between steady states	Blueshift compensation and parametric scattering

In chemotaxis, blowup models extreme aggregation of organisms moving up gradients of their own chemoattractant (Dejak et al., 2013). In self-gravitating or self-attracting particle systems, the same competition is read as gravitational collapse versus diffusive spread (Calvez et al., 2013). In nonlinear optics, Bose–Einstein condensates, and plasmas, NLS blowup is interpreted as self-focusing, optical beam collapse, or condensate collapse (Yang et al., 2018, Gustafson et al., 2022).

In ferromagnetic spin models, the singularity is a concentrated topological defect or vortex core in the continuum Landau–Lifshitz description (Han et al., 2022, Xu et al., 2020). In high-dimensional Euler-inspired models, blowup is tied to axial compression and vortex stretching becoming Burgers-like in the infinite-dimensional limit (Miller, 5 Aug 2025). In mean-field neural models, blowup is not spatial concentration but a synchronous avalanche: the mean-field rate of interaction diverges in finite time with a finite fraction of neurons spiking simultaneously (Papadopoulos et al., 21 Aug 2025).

A common misconception is that blowup must always be produced by conservative focusing. The driven polariton case is a direct counterexample: the system is driven-dissipative, the interaction is repulsive, and the explosive stage is an instability-mediated transition between multistable states rather than a conservative collapse (Gavrilov, 2014). Another misconception is that blowup always implies $H^1$7-divergence of the primary state variable; shock formation, particle collision, and grow-up show that the singular observable can instead be a gradient, an interaction rate, a second moment, or an inter-particle spacing (Miller, 5 Aug 2025, Calvez et al., 2013, Gustafson et al., 2022).

6. Geometric formulations, analytical tools, and open directions

The mathematical description of physical blowup dynamics repeatedly reduces singular evolution to regular dynamics in transformed variables. In gradient-flow aggregation models, Lagrangian coordinates and pseudo-inverse formulations turn concentration into collision dynamics of finitely many particles (Calvez et al., 2013). In Keller–Segel, a collapsing frame with

$H^1$8

converts the singularity into a modulation problem around a static profile (Dejak et al., 2013). In wave and dispersive equations, dynamic rescaling isolates unstable, quasinormal, and tail modes near a critical solution (Bizoń et al., 2019, Yang et al., 2018).

More abstractly, blow-up behavior for ODEs can be described by means of dynamics at infinity. Matsue’s framework uses closed embeddings of phase spaces and time-scale transformations so that blow-ups are organized by asymptotic phase of invariant sets on the horizon; normally, or partially hyperbolic invariant manifolds on the horizon possessing asymptotic phase are shown to induce blow-ups (Matsue, 2023). This suggests a unifying language in which finite-time singularity corresponds to convergence toward invariant dynamics on a compactified boundary at infinity.

Two recent developments sharpen the selection problem for singular continuations. In the mean-field neural model, physical blowup dynamics are solutions to a fixed-point problem bearing on the time change associated to the McKean–Vlasov equation, and these solutions are recovered from buffered, rate-conserving regularizations in the limit of vanishing buffering (Papadopoulos et al., 21 Aug 2025). In shell models, the anomaly of inertial-range scaling is related analytically to the process of instanton creation using the large deviation principle, so singular structures are tied directly to statistical laws rather than only pointwise asymptotics (Mailybaev, 2013).

Several open directions recur across the literature. Nonradial or nonsymmetric dynamics remains unresolved in the exterior-wave, nonlinear Klein–Gordon, NLS, and Landau–Lifshitz settings (Bizoń et al., 2019, Bizoń et al., 2010, Gustafson et al., 2022, Xu et al., 2020). The stability and uniqueness of particular blowup manifolds or profiles is often only partial, especially outside symmetry classes. In high-dimensional Euler, the Burgers-like infinite-dimensional model strongly suggests a mechanism for finite-time blowup of smooth solutions, but the finite-dimensional Euler equation appears only as a highly singular perturbation (Miller, 5 Aug 2025). In mean-field neural dynamics, the buffered time-change construction suggests a broader admissibility principle for blowup problems with conserved interaction fluxes, but its extension to other singular PDEs remains open (Papadopoulos et al., 21 Aug 2025).

Taken together, these results show that physical blowup dynamics is not a single asymptotic law but a research program: identify the correct singular observable, pass to variables where the singularity is geometrically organized, determine the critical profile or invariant set that governs the approach, and relate the resulting asymptotics back to the original physical mechanism.