---
title: 'Blow-Up Equations: Analysis & Criteria'
url: https://www.emergentmind.com/topics/blow-up-equations
type: topic
---

# Blow-Up Equations: Analysis & Criteria

Blow-up equations are ordinary differential equations, partial differential equations, stochastic partial differential equations, and integro-differential equations whose maximal solutions can cease to exist in finite time through unbounded growth, loss of regularity, or breakdown of the underlying geometric flow. The literature represented here treats blow-up as a structural phenomenon rather than a single pathology: in some settings every nontrivial positive solution becomes unbounded in finite time, in others only derivatives or moments diverge, and in geometric formulations the decisive event is degeneration of a Jacobian or loss of invertibility of the flow map. Across these formulations, blow-up is typically organized by comparison principles, scaling laws, critical integrals, and invariant quantities [1912.01537].

## 1. Definitions and singularity notions

For autonomous ODEs \(y'=f(y)\), a solution is a blow-up solution when its maximal existence time \(t_{\max}\) is finite; this is distinct from grow-up, where divergence occurs only as \(t\to\infty\) [1606.03039]. In semilinear fractional heat equations on \(\mathbb R^n\), the phrase “blow-up property” is used for the stronger statement that every nontrivial nonnegative solution becomes unbounded in finite time [1912.01537]. For nonlinear Volterra integro-differential equations, the same finite-time/non-finite-time dichotomy is retained, but the growth law is mediated by the full past history through a convolution kernel [1710.07583].

Blow-up does not always mean pointwise divergence of the solution itself. For the periodic \(b\)-family,
\[
u_t-u_{txx}+(b+1)uu_x=b\,u_xu_{xx}+uu_{xxx},
\]
finite-time blow-up occurs if and only if the slope becomes unbounded below,
\[
\lim_{t\to T^*}\Big((2b-1)\inf_{x\in\mathbb T}u_x(t,x)\Big)=-\infty,
\]
so the singularity mechanism is wave breaking: the solution remains bounded while \(u_x\to-\infty\) [1407.4084]. In the Lagrangian formulation of Euler–Arnold equations, blow-up is detected by the vanishing of the radial Jacobian factor \(\gamma_r\); once \(\gamma_r\) reaches zero, the flow ceases to be locally invertible, and this forces \(C^1\)-blow-up of the velocity field through \(\inf_r u_r(t,r)\to-\infty\) [2306.09748].

Several papers use probabilistic notions of blow-up. For semilinear parabolic equations driven by Lévy noise on bounded domains, the relevant event is finite-time blow-up in mean square,
\[
\lim_{t\uparrow T^*}\mathbb E\|u(t)\|_{L^2(O)}^2=\infty
\]
for some finite \(T^*>0\) [2404.06953]. For space-time fractional stochastic reaction-diffusion equations, blow-up means non-existence of a global random field solution because moments such as \(\mathbb E|u_t(x)|^2\) or \(\mathbb E|u_t(x)u_t(y)|\) become infinite in finite time [1803.05890]. Other settings distinguish finer singularity classes: the critical wave equation on curved backgrounds admits finite-energy type II blow-up [1303.1967], singular Liouville equations distinguish simple from non-simple blow-up through the spherical Harnack inequality [2305.07264], and the fourth-order parabolic equation \(u_t=\det(D^2u)-\Delta^2u\) exhibits complete finite-time and complete infinite-time blow-up, meaning that every spatial point diverges, possibly with different signs in different regions [2103.02344].

## 2. Thresholds, critical exponents, and dichotomies

A central theme is that blow-up is often governed by a sharp threshold. For the fractional semilinear heat equation
\[
u_t=\mathcal A u+f(u),\qquad \mathcal A=-(-\Delta)^{\alpha/2},\qquad 0<\alpha\le 2,
\]
with \(f\) locally Lipschitz, non-decreasing, convex, positive for \(u>0\), and satisfying the near-zero scaling condition (S), there is an exact equivalence between three statements: every nontrivial nonnegative PDE solution blows up in finite time, every positive solution of
\[
x'(t)=f(x(t))-\frac{n}{\alpha t}x(t)
\]
blows up in finite time, and the critical near-zero divergence condition
\[
\int_0^1 \frac{u^{1+\alpha/n}}{f(u)}\,du=\infty
\]
holds [1912.01537]. The complementary integrability condition yields positive global solutions. For pure powers \(f(u)=u^p\), this reproduces the fractional Fujita exponent
\[
p_F=1+\frac{\alpha}{n},
\]
with universal blow-up for \(1<p\le p_F\) and small-data global existence for \(p>p_F\) [1912.01537].

Nonlocal diffusion with convolution kernels exhibits an analogous Fujita-type structure. For
\[
u_t=Ju-u+F(u)=Au+F(u),
\]
where \(J\) is radially symmetric, \(\int_{\mathbb R^d}J(x)\,dx=1\), and \(\widehat J(\xi)=1-A|\xi|^\alpha+o(|\xi|^\alpha)\) near \(\xi=0\), the backward-kernel functional
\[
W_T(t)=\int_{\mathbb R^d} k_{T-t}(x)u(x,t)\,dx
\]
satisfies a scalar inequality implying blow-up whenever \(W_T(0)>h^{-1}(T)\), with \(h(w)=\int_w^\infty du/F(u)\) [1807.03569]. For \(F(u)=cu^p\), the same critical exponent appears,
\[
p_F=1+\frac{\alpha}{d},
\]
and the supercritical regime \(p>1+\alpha/d\) admits a partial dichotomy: small critical Morrey norm gives global-in-time smooth solutions with decay \(\|u(\cdot,t)\|_\infty=O(t^{-1/(p-1)})\), while large critical Morrey norm gives finite-time blow-up [1807.03569].

Memory terms can alter the mechanism without destroying sharp thresholds. For the Volterra equation
\[
x'(t)=\int_0^t w(t-s)f(x(s))\,ds,\qquad w(0)>0,
\]
finite-time blow-up occurs if and only if
\[
\int_\eta^\infty \frac{du}{\sqrt{\int_0^u f(s)\,ds}}<\infty
\]
for some \(\eta>0\); if the integral diverges for all \(\eta>0\), solutions are global [1710.07583]. The same paper derives sharp asymptotic laws
\[
\lim_{t\to T^-}\frac{F_B(x(t))}{T-t}=\sqrt{2w(0)},\qquad
\lim_{t\to\infty}\frac{F_U(x(t))}{t}=\sqrt{2w(0)},
\]
showing that the local kernel mass at the origin, rather than the detailed global shape of the kernel, controls both explosive and nonexplosive superlinear growth [1710.07583].

## 3. Local breakdown and continuation criteria

Many blow-up theories are formulated as continuation criteria: a strong solution persists as long as a critical quantity stays integrable or bounded. For the periodic \(b\)-family with \(b\in(1,3]\), the main result is local in space. If there exists \(x_0\in\mathbb T\) such that
\[
u_0'(x_0)<-\beta_b|u_0(x_0)|,
\]
then the corresponding strong solution blows up in finite time, with the explicit upper bound
\[
T^*\le \frac{2}{(b-1)\sqrt{u_0'(x_0)^2-\beta_b^2u_0(x_0)^2}}.
\]
Here \(\beta_b\) is defined through the variational quantity \(J(b,\beta)\), and the proof proceeds along characteristics through the combinations \(-u_x\pm \beta u\) and Riccati-type differential inequalities [1407.4084].

For the Thermal Quasi-Geostrophic system on \(\mathbb R^2\), the continuation mechanism is of Beale–Kato–Majda type. If
\[
\int_0^T \big(\|q(t)\|_\infty+\|\nabla b(t)\|_\infty\big)\,dt<\infty,
\]
then the strong solution extends beyond \(T\). Consequently, if the maximal time is finite, then necessarily
\[
\int_0^T \big(\|q(t)\|_\infty+\|\nabla b(t)\|_\infty\big)\,dt=\infty,
\]
and in particular \(\sup_{t\uparrow T}(\|q(t)\|_\infty+\|\nabla b(t)\|_\infty)=\infty\) [2201.06476]. The key analytic input is a logarithmic estimate for the velocity arising from the modified Helmholtz relation \((\Delta-1)\psi=w\), which plays the role that the Biot–Savart law plays in the classical Euler BKM criterion [2201.06476].

The three-dimensional compressible Navier–Stokes equations admit analogous Serrin-type criteria tailored to compressible flow. In the barotropic case, finite-time breakdown of a strong solution implies
\[
\limsup_{t\to T^*}\Big(\|\rho\|_{L^\beta}+\|\operatorname{div}\mathbf u\|_{L^3}\Big)=\infty
\]
for some \(\beta\in(1,\infty)\) depending only on \(\gamma\); in the heat-conducting case the criterion becomes
\[
\limsup_{t\to T^*}\Big(\|\rho\|_{L^\delta}+\|\theta\|_{L^2}+\|\operatorname{div}\mathbf u\|_{L^3}\Big)=\infty
\]
for some \(\delta>1\) depending on the physical parameters [1705.05132]. A notable feature is that these criteria are proved under only the physical viscosity conditions \(\mu>0\) and \(3\lambda+2\mu>0\), without additional restrictions on the Lamé coefficients [1705.05132].

Geometric Euler–Arnold theory replaces pointwise comparison by comparison in an infinite-dimensional function space. In radial variables, the transported momentum identity
\[
\gamma(t,r)^{n-1}\gamma_r(t,r)^2\,\omega(t,\gamma(t,r))=r^{n-1}\omega_0(r)
\]
leads to a Liouville-type comparison principle for a weighted Jacobian quantity \(q(t,r)\). Under positivity and log-supermodularity hypotheses on the Green kernel and a sign condition \(z_0(r)\le 0\), the comparison theorem forces \(q\) and hence \(\gamma_r\) to vanish in finite time, yielding \(C^1\)-breakdown [2306.09748]. In the radial Hunter–Saxton case this reduces to the exact criterion
\[
u_0'(r)+\frac{n-1}{r}u_0(r)<0
\quad\text{for some }r\ge 0,
\]
which is equivalent to blow-up in any dimension [2306.09748].

## 4. Blow-up profiles, rates, and local structure

Besides criteria, several papers construct explicit or asymptotic blow-up profiles. For the critical focusing wave equation on a curved three-dimensional background,
\[
\ddot u-\Delta_g u=|u|^4u,
\]
there exists, for every \(\nu\in(1/2,1]\), a finite-energy type II blow-up solution of the form
\[
u(t,r)=t^{-1-\nu}W\bigl(t^{-1-\nu}r\bigr)+\varepsilon(r,t).
\]
The parameter \(\nu\) yields a continuum of blow-up rates. The construction extends the slow blow-up theory of Krieger, Schlag, and Tataru to curved backgrounds, but the curvature-induced perturbation \(K\) breaks exact scale invariance and limits the method to \(\nu\le 1\) [1303.1967].

The local structure of blow-up can be highly sensitive to oscillation properties. For singular Liouville equations
\[
\Delta u+|x|^{2\alpha}h(x)e^u=0\qquad\text{in }B_1\subset\mathbb R^2,
\]
Wei–Zhang proved vanishing theorems for non-simple blow-up sequences. Wu showed that these vanishing conclusions are genuinely tied to the non-simple regime. In the non-quantized case \(\alpha\notin\mathbb N\cup\{0\}\), there exist blow-up sequences satisfying spherical Harnack near the origin while both
\[
\big|\nabla \log h_k(0)+\nabla V_k(0)\big|\ge c_1
\quad\text{and}\quad
\big|\Delta\log h_k(0)\big|\ge c_1
\]
remain uniformly nonzero. In the quantized case \(\alpha=N\in\mathbb N\), there exist simple blow-up sequences with
\[
\big|\Delta\log h_k(0)\big|\ge c>0,
\]
showing that the non-simple assumption in Wei–Zhang’s Laplacian vanishing theorem is essential [2305.07264].

The fourth-order parabolic equation
\[
u_t=\det(D^2u)-\Delta^2u
\]
provides explicit examples in which blow-up is complete and sign-sensitive. On the square, the disc, and \(\mathbb R^2\), the constructed solutions exhibit both finite-time and infinite-time blow-up; in several families every spatial point diverges, but different regions may approach \(+\infty\) and \(-\infty\) in finite time [2103.02344]. The same paper refines an earlier criterion by showing that, in the initial-Dirichlet weak-solution framework, \(H^2\)-blow-up implies blow-up in the stronger norm \(W^{1,\infty}\), via a Gagliardo–Nirenberg interpolation argument [2103.02344]. A plausible implication is that a priori criteria based only on \(H^2\)-control can miss the actual geometric sharpness of Hessian-driven singularity formation.

## 5. Stochastic, fractional, and coupled hyperbolic extensions

In stochastic parabolic problems, blow-up theory must absorb Itô corrections and jump compensators. For semilinear stochastic parabolic equations on a bounded smooth domain with Dirichlet boundary conditions,
\[
du=\big[a\Delta u+b|u|^{m-1}u\big]dt+\sigma(u,t)\,dW(t)+n(u,t,z)\,\widetilde N(dt,dz),
\]
finite-time mean-square blow-up is proved for both additive Lévy noise and linear multiplicative Lévy noise [2404.06953]. The proof adapts the classical concavity method by introducing
\[
v(t)=\mathbb E\|u(t)\|_{L^2}^2,\qquad I(t)=\int_0^t v(s)\,ds+K,
\]
and deriving
\[
I''(t)I(t)-(1+\varepsilon)(I'(t))^2>0
\]
for suitable \(\varepsilon>0\). In the multiplicative case the noise strength enters through
\[
K:=5\left(\sigma^2+\int_Z n^2(z)\,\nu(dz)\right),
\]
and the restriction \(0\le K\le aX_1\) ensures that the destabilizing noise can be absorbed by Laplacian dissipation via Poincaré’s inequality [2404.06953].

Space-time fractional stochastic reaction-diffusion equations introduce both temporal memory and anomalous diffusion,
\[
\partial_t^\beta u_t(x)=-\nu(-\Delta)^{\alpha/2}u_t(x)+I_t^{1-\beta}[b(u)+\sigma(u)\dot F(t,x)].
\]
Here blow-up means non-existence of a global random field solution because suitable moments become infinite in finite time [1803.05890]. Under superlinear lower bounds such as \(\sigma(x)\ge |x|^{1+\gamma}\) or \(b(x)\ge |x|^{1+\eta}\), Walsh-isometry and Jensen-type arguments lead to nonlinear renewal inequalities of the form
\[
h(t)\ge C+D\int_0^t h(s)^{1+\gamma}(t-s)^{-\theta}\,ds,
\]
which force finite-time blow-up under the stated kernel and initial-data hypotheses [1803.05890]. The resulting theorems cover white noise, spatially colored noise, Riesz kernel covariances, whole-space and bounded-domain geometries, and deterministic drift-plus-noise models [1803.05890].

Weakly coupled Euler–Poisson–Darboux–Tricomi systems show that time-dependent propagation speeds and derivative nonlinearities preserve a sharp blow-up geometry when the two components share the same speed \(t^m\). For
\[
u_{tt}-t^{2m}\Delta u+\frac{\mu_1}{t}u_t+\frac{\nu_1^2}{t^2}u=|\partial_t v|^p,\qquad
v_{tt}-t^{2m}\Delta v+\frac{\mu_2}{t}v_t+\frac{\nu_2^2}{t^2}v=|\partial_t u|^q,
\]
the blow-up region is
\[
\Omega(\tilde N_m,\mu_1,\mu_2,p,q)\ge 0,\qquad
\tilde N_m=N(m+1)-2m,
\]
and the maximal existence time satisfies polynomial or exponential upper bounds depending on whether the problem is subcritical, critical, or double-critical [2306.14768]. The same-speed assumption is used repeatedly in the support property, in the construction of the adjoint test functions, and in closing the coupled ODE comparison argument. The mass terms \(\nu_i\) do not appear in the blow-up region or the lifespan estimate [2306.14768].

## 6. Compactification, dynamical systems at infinity, and numerical validation

A major line of work interprets blow-up as asymptotic dynamics on a compactified phase space. For asymptotically quasi-homogeneous ODEs, quasi-Poincaré compactification sends infinity to a horizon \(\mathcal E\), and a time desingularization produces a vector field that extends continuously to that horizon [1611.06346]. Hyperbolic equilibria and periodic orbits on \(\mathcal E\) then generate finite-time blow-up along their stable manifolds. If \(x_*\) is a hyperbolic equilibrium at infinity, the corresponding original solution satisfies
\[
p(y(t))\sim c\,(t_{\max}-t)^{-1/k},
\]
and, for nonzero components,
\[
y_i(t)\sim c\,(t_{\max}-t)^{-\alpha_i/k}.
\]
The qualitative dynamics at infinity are proved to be topologically equivalent across quasi-Poincaré, directional, and intermediate compactifications [1611.06346].

Validated numerics make this dynamical-systems picture constructive. For polynomial ODEs, admissible compactifications together with a normalized vector field and a quadratic Lyapunov function around a critical point at infinity allow one to prove that a computed orbit truly blows up and to enclose the blow-up time rigorously [1606.03039]. The method uses interval or affine arithmetic to validate a Lyapunov domain and then estimates
\[
t_{\max}=\int_0^\infty \frac{d\tau}{\kappa(T^{-1}(x(\tau)))^{d-1}}
\]
by explicit upper and lower bounds. Poincaré compactification is used when the normalized field is regular at the boundary; parabolic compactification is used when the Poincaré Jacobian becomes singular near infinity [1606.03039].

For unresolved hydrodynamic PDEs, the numerical problem is different: the issue is not formal validation of a known trajectory but distinguishing genuine singularity formation from discretization artifacts. In a complexified Euler–EPDiff model on \(S^2\), a geometrically consistent Zeitlin-type discretization is used to identify a numerical signature of blow-up: the supremum norm of vorticity should show a resolution-dependent fingerprint as the truncation level \(N\) increases [2210.02328]. The proposed diagnostic is not merely large \(\|\omega\|_\infty\), but persistent and stronger growth of \(\|\omega_N(t)\|_\infty\) under refinement, together with computational stability checks based on the structure-preserving discretization. This suggests a practical criterion for future numerical blow-up studies in equations whose analytical status remains open [2210.02328].

Taken together, these results show that blow-up equations are unified less by a single model than by a repertoire of analytic and geometric mechanisms: Fujita-type threshold integrals, Riccati inequalities along characteristics, Liouville comparison in Banach spaces, stochastic concavity, renewal inequalities, explicit profile construction, and compactification at infinity. The common structure is that finite-time breakdown becomes detectable once the correct critical quantity has been identified—whether it is a solution norm, a derivative, a weighted moment, a Jacobian, or a trajectory on the horizon at infinity.

Source: https://www.emergentmind.com/topics/blow-up-equations