Osgood Condition in Differential Equations

• The Osgood condition is an integral criterion that distinguishes finite-time blow-up from global existence by analyzing the divergence of integrals associated with nonlinearities or moduli.
• It is applied to establish uniqueness of solutions in ODEs, transport equations, and backward parabolic problems, often by imposing divergence conditions on moduli of continuity.
• The criterion’s adaptability is evident in its use across semilinear parabolic equations, Hamilton–Jacobi theory, and even stochastic frameworks, underscoring its broad impact.

Searching arXiv for recent and foundational papers on the Osgood condition to ground the article. The Osgood condition is an integral criterion that appears in several closely related forms across analysis. In its classical autonomous form, for $\dot x=f(x)$, it distinguishes finite-time blow-up from global existence through the convergence or divergence of $\int ds/f(s)$; in its modulus form, for vector fields with continuity modulus $\omega$, it distinguishes uniqueness from non-uniqueness through $\int ds/\omega(s)$. The condition originates in ODE theory, but the modern literature uses it in semilinear parabolic equations, transport and continuity equations, Hamilton–Jacobi theory, stochastic heat equations, backward parabolic uniqueness, cosmological singularity analysis, and BSDEs. The same integral divergence remains decisive in some settings and loses sufficiency in others, especially for PDEs with rough data (Kohli, 2015, Laister et al., 2013).

1. Classical ODE formulations

For the scalar autonomous initial value problem

$\dot x = f(x(t)), \qquad x(t_0)=\xi,$

with $f$ continuous and positive in the relevant range, the standard Osgood blow-up criterion states that the solution blows up in finite time if and only if

$\int_{\xi}^{\infty}\frac{ds}{f(s)}<\infty.$

Equivalently, divergence of the integral implies that it takes infinite time for the solution to reach $+\infty$, so the trajectory exists globally forward in time (Kohli, 2015). In this formulation, the maximal time to reach infinity is exactly the separation-of-variables quantity $\int_{\xi}^{\infty} ds/f(s)$.

A second classical formulation concerns uniqueness at the equilibrium $0$. For

$\int ds/f(s)$0

with $\int ds/f(s)$1 continuous, non-decreasing, $\int ds/f(s)$2, and $\int ds/f(s)$3 for $\int ds/f(s)$4, the trivial solution $\int ds/f(s)$5 is the unique local non-negative solution if and only if

$\int ds/f(s)$6

for some $\int ds/f(s)$7 (Laister et al., 2017). If the integral is finite, non-trivial solutions can leave the equilibrium after an arbitrary delay, so uniqueness fails.

These two formulas involve different ends of the state space: the first probes behavior at infinity and controls finite-time explosion, while the second probes behavior near zero and controls departure from an equilibrium. Later literature often uses “Osgood condition” for both, with the relevant endpoint determined by the application.

2. Moduli of continuity, flows, and transport

In flow theory, the Osgood condition is usually imposed on a modulus of continuity $\int ds/f(s)$8. A typical definition is

$\int ds/f(s)$9

or, on $\omega$0,

$\omega$1

Lipschitz moduli satisfy this condition, while Hölder moduli $\omega$2 with $\omega$3 do not. Log-Lipschitz and iterated log-Lipschitz moduli are standard Osgood examples (Fjordholm et al., 17 Feb 2025, Johansson et al., 1 Apr 2026).

This integral divergence is the sharp threshold for uniqueness of characteristics for ODEs driven by $\omega$4-continuous vector fields. In the continuity-equation setting, failure of the Osgood condition is sharp in a strong sense: for any modulus $\omega$5 that fails Osgood, one can construct a divergence-free $\omega$6 for which the associated ODE admits at least two distinct flow maps on a set of initial conditions of positive Lebesgue measure, and the continuity equation admits two distinct solutions with the same absolutely continuous initial datum (Colombo et al., 17 Jan 2026).

The transport-equation literature uses the same threshold in a different way. For Osgood velocity fields, the usual DiPerna–Lions weak formulation may be unavailable because the divergence need not exist as a function or measure. A recent approach replaces the Lebesgue interpretation of $\omega$7 by a Riemann–Stieltjes or Young integral, using finite $\omega$8-variation of the solution and finite $\omega$9-variation of $\int ds/\omega(s)$0 to define the transport term (Fjordholm et al., 17 Feb 2025). In non-local continuity systems, an effective modulus of the form

$\int ds/\omega(s)$1

appears in the Lagrangian stability estimate, and uniqueness follows once the corresponding Osgood integral diverges (Inversi et al., 2023).

The Osgood condition guarantees well-posedness of the flow but does not enforce dynamical simplicity. For any non-Lipschitz Osgood modulus $\int ds/\omega(s)$2, time-periodic $\int ds/\omega(s)$3-continuous velocity fields generically have time-one maps with infinite topological entropy (Johansson et al., 1 Apr 2026). In this sense, Osgood regularity is strong enough for uniqueness and continuity of flows, but weak enough to permit arbitrarily complex dynamics.

3. Semilinear parabolic equations and the limits of the criterion

For semilinear heat equations, the Osgood condition interacts delicately with the regularity class of the initial data. In the bounded-data setting, comparison with the scalar ODE shows that $\int ds/\omega(s)$4 prevents finite-time blow-up and yields global classical solutions. In the zero-initial-data problem

$\int ds/\omega(s)$5

uniqueness of the trivial bounded non-negative solution is equivalent to uniqueness of the trivial solution of the ODE $\int ds/\omega(s)$6, hence equivalent to the near-zero Osgood condition $\int ds/\omega(s)$7; the corresponding PDE non-uniqueness can be proved without any concavity assumption on $\int ds/\omega(s)$8 (Laister et al., 2017).

The bounded-data intuition fails for rougher data. For

$\int ds/\omega(s)$9

there are locally Lipschitz, nonnegative, non-decreasing nonlinearities satisfying an Osgood-type condition, and initial data $\dot x = f(x(t)), \qquad x(t_0)=\xi,$0, $\dot x = f(x(t)), \qquad x(t_0)=\xi,$1, such that no local integral solution exists; more precisely, every candidate integral solution fails to belong to $\dot x = f(x(t)), \qquad x(t_0)=\xi,$2 for any $\dot x = f(x(t)), \qquad x(t_0)=\xi,$3 (Laister et al., 2013). The bounded-domain Dirichlet analogue is equally negative: there are locally Lipschitz, non-decreasing $\dot x = f(x(t)), \qquad x(t_0)=\xi,$4 with

$\dot x = f(x(t)), \qquad x(t_0)=\xi,$5

such that for every finite $\dot x = f(x(t)), \qquad x(t_0)=\xi,$6 one can find $\dot x = f(x(t)), \qquad x(t_0)=\xi,$7 for which the semilinear Dirichlet problem has no local-in-time solution (Laister et al., 2013). These results show that, for unbounded initial data, the Osgood condition is neither a necessary nor a sufficient criterion for local or global solvability in the parabolic PDE setting.

Variants of the same phenomenon persist in non-Gaussian and fractional-time models. For

$\dot x = f(x(t)), \qquad x(t_0)=\xi,$8

with $\dot x = f(x(t)), \qquad x(t_0)=\xi,$9 and $f$0, Osgood-type nonlinearities constructed from plateau functions $f$1 still permit instantaneous blow-up: the critical threshold for non-existence of local $f$2 solutions is

$f$3

while a distinct small-data global existence theory is available in suitable $f$4-based spaces (Solís et al., 2024).

The stochastic heat equation exhibits both sides of the dichotomy. Under an infinite Osgood-type condition on a dominating function $f$5,

$f$6

together with a related growth bound on the diffusion coefficient, one obtains a unique global mild solution and non-explosion in finite time. Under a finite Osgood condition on a convex, nondecreasing drift $f$7,

$f$8

finite-time blow-up with positive probability occurs for suitable initial data (Chen et al., 2023). Across deterministic and stochastic parabolic problems, the Osgood condition remains a sharp ODE-level indicator but not a universal PDE well-posedness criterion.

4. Osgood growth in Hamilton–Jacobi theory

For the evolutionary Hamilton–Jacobi equation

$f$9

on a closed manifold, Osgood conditions appear as growth restrictions in the state variable $\int_{\xi}^{\infty}\frac{ds}{f(s)}<\infty.$0, rather than as continuity moduli or source-term integrals. The Hamiltonian assumption is that for every compact $\int_{\xi}^{\infty}\frac{ds}{f(s)}<\infty.$1 there exists a continuous $\int_{\xi}^{\infty}\frac{ds}{f(s)}<\infty.$2 such that

$\int_{\xi}^{\infty}\frac{ds}{f(s)}<\infty.$3

and

$\int_{\xi}^{\infty}\frac{ds}{f(s)}<\infty.$4

The associated Lagrangian formulation uses the corresponding condition

$\int_{\xi}^{\infty}\frac{ds}{f(s)}<\infty.$5

This Osgood growth hypothesis is strictly more general than monotonicity or uniform Lipschitz continuity in $\int_{\xi}^{\infty}\frac{ds}{f(s)}<\infty.$6: constant and affine $\int_{\xi}^{\infty}\frac{ds}{f(s)}<\infty.$7 recover the non-increasing and Lipschitz cases, respectively (Wang et al., 2014).

The analytic role of the condition is to guarantee completeness of the scalar comparison equation $\int_{\xi}^{\infty}\frac{ds}{f(s)}<\infty.$8. In the variational construction, truncated Lagrangians yield uniformly Lipschitz approximants; the Osgood completeness condition then provides $\int_{\xi}^{\infty}\frac{ds}{f(s)}<\infty.$9-independent bounds on the value component along calibrated curves, which in turn give compactness of velocities and allow passage to the limit. This produces the fundamental solution $+\infty$0, the implicitly variational principle, and representation formulas for viscosity solutions in terms of minimal characteristics (Wang et al., 2014).

In this setting, the Osgood condition is best understood as a one-sided integral growth control that replaces global Lipschitz or properness assumptions in the $+\infty$1-variable. Its significance is not blow-up prevention of a PDE solution in Lebesgue spaces, but control of the characteristic and variational dynamics required for weak KAM theory and viscosity solution representation.

5. Backward parabolic equations and coefficient regularity

For backward parabolic operators, the Osgood condition governs time-regularity of the principal coefficients. A modulus of continuity $+\infty$2 is Osgood if

$+\infty$3

With coefficients $+\infty$4 that are Lipschitz in $+\infty$5 and Osgood continuous in $+\infty$6, backward uniqueness holds in the natural energy space; if the modulus is non-Osgood, counterexamples to uniqueness are known (Santo et al., 2020).

This threshold remains effective even when Osgood continuity fails at a single time. For backward parabolic operators whose principal coefficients are Osgood continuous on every interval $+\infty$7 with a seminorm allowed to blow up like $+\infty$8 as $+\infty$9, while remaining Hölder continuous up to $\int_{\xi}^{\infty} ds/f(s)$0, uniqueness still holds. The proof uses a Carleman weight built directly from the Osgood modulus and tuned to the factor $\int_{\xi}^{\infty} ds/f(s)$1 (Santo et al., 2020).

The same moduli enter conditional stability theory. Under Osgood continuity of the leading coefficients in time, one has a conditional continuous dependence estimate of the form

$\int_{\xi}^{\infty} ds/f(s)$2

for solutions bounded a priori in $\int_{\xi}^{\infty} ds/f(s)$3, where $\int_{\xi}^{\infty} ds/f(s)$4 is an increasing continuous function with $\int_{\xi}^{\infty} ds/f(s)$5 (Casagrande et al., 2019). Earlier work showed that the stronger Log-Lipschitz-type stability estimate can fail for coefficients that are Osgood but not Log-Lipschitz, while a weaker stability statement still survives in a suitable Osgood-adapted functional space (Casagrande et al., 2018).

Here the Osgood condition is neither a source-term criterion nor a continuity-modulus condition for characteristics. It is the precise temporal regularity threshold at which backward uniqueness and nontrivial conditional stability remain valid.

6. Applications, sharpness, and conceptual scope

The Osgood condition has been exported far beyond classical ODEs. In cosmology, reduction of FLRW evolution equations to scalar autonomous ODEs allows Osgood’s criterion to detect finite-time Type $\int_{\xi}^{\infty} ds/f(s)$6 and Type II singularities. The integral

$\int_{\xi}^{\infty} ds/f(s)$7

then determines whether a finite-time singularity occurs, and special initial conditions can force divergence of the integral and avoid blow-up (Kohli, 2015).

In incompressible fluid dynamics, Osgood moduli govern propagation of geometric singularities. For homeomorphism flows generated by Osgood velocity fields, Hölderian cusps and mild divergences are transported by the flow, with the singular structure quantified through

$\int_{\xi}^{\infty} ds/f(s)$8

For 2D Euler, vorticity profiles of $\int_{\xi}^{\infty} ds/f(s)$9 type travel with the fluid, up to bounded perturbations, and slightly more singular structures are generally not propagated (Drivas et al., 2022).

In stochastic analysis, multidimensional BSDEs admit a stochastic one-sided Osgood condition in the $0$0-variable: $0$1 with $0$2 satisfying the Osgood integral condition and $0$3. Combined with stochastic Lipschitz control in $0$4 and stochastic Gronwall-type and Bihari-type inequalities, this yields existence and uniqueness of $0$5 solutions for multidimensional BSDEs with general terminal time (Lai et al., 15 Sep 2025).

Across these theories, “the Osgood condition” is not a single theorem but a family of integral divergence criteria attached to different objects: nonlinearities $0$6, moduli $0$7, growth envelopes $0$8, and one-sided coercivity functions $0$9. What unifies them is the same analytic mechanism: divergence of an integral that prevents either immediate branching of trajectories or finite-time escape. What distinguishes the applications is where that mechanism acts. In ODEs it is often exact; in transport it is the uniqueness threshold for flows; in backward parabolic equations it is the regularity threshold for uniqueness; in Hamilton–Jacobi theory it is a completeness condition for characteristics; and in semilinear PDEs with rough data it can fail spectacularly as a criterion for local solvability.