---
title: Osgood-Type Growth
url: https://www.emergentmind.com/topics/osgood-type-growth
type: topic
---

# Osgood-Type Growth

Searching arXiv for recent and foundational papers on Osgood-type growth and related ODE/PDE/SPDE flow results.
Osgood-type growth denotes a class of non-Lipschitz control conditions governed by an integral-divergence criterion of Osgood type. In its classical form, for an autonomous ODE \(\dot X=b(X)\), one assumes a modulus of continuity \(\omega\) such that \(|b(x)-b(y)|\le \omega(|x-y|)\) and \(\int_0^\delta \frac{ds}{\omega(s)}=+\infty\); this strictly generalizes Lipschitz continuity and still yields uniqueness through a Gronwall–Osgood argument [1107.2496]. In modern analysis, the notion has broadened in two principal directions. First, it appears as a modulus condition on increments of vector fields or coefficients, often only almost everywhere and possibly weighted by an integrable function, as in the regular Lagrangian flow framework of DiPerna–Lions type [1107.2496]. Second, it appears as a growth condition on nonlinearities, typically through integrals such as \(\int^\infty \frac{ds}{f(s)}\), governing finite-time blow-up versus global existence in ODEs and in several PDE and SPDE settings [1303.7183]. Across these contexts, the Osgood condition acts as a threshold: it is weaker than Lipschitz regularity yet still strong enough to enforce uniqueness or prevent finite-time explosion in some regimes, while failure of the condition can produce sharp non-uniqueness or blow-up phenomena [2601.12096].

## 1. Classical criterion and canonical formulations

The classical Osgood criterion concerns scalar or finite-dimensional ODEs with non-Lipschitz modulus control. If a vector field satisfies
\[
|b(x)-b(y)|\le \omega(|x-y|),\qquad \int_0^\delta \frac{ds}{\omega(s)}=+\infty,
\]
then uniqueness persists although \(\omega\) may grow slower than linearly near \(0\) [1107.2496]. This includes the Lipschitz case \(\omega(s)=Ls\), but also moduli such as \(s\log\frac1s\) and \(s\log\frac1s\log\log\frac1s\) for small \(s\) [1107.2496].

A standard equivalent parametrization uses functions of the form
\[
|b(x)-b(y)|\le C\,|x-y|\,r(|x-y|^2),\qquad \int_0^{c_0}\frac{ds}{s\,r(s)}=+\infty,
\]
which again yields an Osgood modulus after setting \(\omega(s)=s\,r(s^2)\) [1107.2496]. In the same spirit, a mixed Osgood–Sobolev formulation for stochastic equations replaces two-sided modulus control by a one-sided estimate
\[
\langle x-y,\; b(x)-b(y)\rangle \le (g_R(x)+g_R(y))\,\rho(|x-y|^2),
\]
where \(\rho\) is nondecreasing and \(\int_0^1 \frac{ds}{\rho(s)}=+\infty\) [1605.02820]. This one-sided structure is natural in transport, stochastic flow, and BSDE settings.

In PDEs with coefficient regularity measured by a modulus of continuity \(\mu\), the Osgood condition is expressed as
\[
\int_0^1 \frac{1}{\mu(s)}\,ds=+\infty.
\]
This is the critical threshold for several backward parabolic uniqueness and stability results: Lipschitz and Log-Lipschitz moduli satisfy it, Hölder moduli \(s^\tau\) with \(\tau\in(0,1)\) do not [1801.07890], [1902.08573].

A distinct but related formulation appears for scalar nonlinear growth:
\[
\int_1^\infty \frac{ds}{f(s)}=\infty.
\]
For the ODE \(\dot x=f(x)\), this is the criterion preventing finite-time blow-up. It underlies several PDE and SPDE results where bounded data behave ODE-like, while rough data may exhibit sharply different behavior [1303.7183], [2310.02153].

## 2. Almost everywhere Osgood continuity and regular Lagrangian flows

A major modern generalization is the “almost everywhere Osgood continuous” vector field introduced in the direct Lagrangian theory of flows. The structural assumption is that there exist negligible sets \(N_t\), a nonnegative function \(g\in L^1([0,T],L^1_{\mathrm{loc}}(\mathbb{R}^d))\), and a strictly increasing \(\rho\in C(\mathbb{R}_+;\mathbb{R}_+)\) with \(\rho(0)=0\) and
\[
\int_0^+\frac{ds}{\rho(s)}=+\infty,
\]
such that for \(x,y\notin N_t\),
\[
|b_t(x)-b_t(y)|\le (g_t(x)+g_t(y))\,\rho(|x-y|).
\]
This is Assumption (H) in the unified treatment of ODEs under Osgood and Sobolev type conditions [1107.2496].

This formulation interpolates between two previously separate theories. If \(g\) is essentially bounded, one recovers classical Osgood continuity:
\[
|b_t(x)-b_t(y)|\le C\,\rho(|x-y|).
\]
If \(\rho(s)=s\) and \(g_t(x)\) is the local maximal function \(C_d M_R|Db_t|(x)\), one recovers the Crippa–De Lellis almost everywhere Lipschitz estimate for Sobolev vector fields [1107.2496]. Example 2.4 in that paper exhibits a vector field \(b=b_1+b_2\) satisfying (H) but not covered by either pure Sobolev almost everywhere Lipschitz control or pure Osgood continuity alone, showing that the class is genuinely larger [1107.2496].

The relevant flow notion is the regular Lagrangian flow \(X:[0,T]\times\mathbb{R}^d\to\mathbb{R}^d\), defined by the integral equation
\[
X_t(x)=x+\int_0^t b_s(X_s(x))\,ds
\]
for almost every \(x\), together with the compressibility bound
\[
(X_t)_\#\mathcal{L}^d \le L\,\mathcal{L}^d.
\]
Under boundedness of \(b\), Assumption (H), and bounded negative divergence
\[
[\mathrm{div}(b)]^- \in L^1([0,T],L^\infty(\mathbb{R}^d)),
\]
there exists a unique regular Lagrangian flow [1107.2496]. This is precisely the sense in which Osgood-type growth extends DiPerna–Lions theory beyond Sobolev or \(BV\) regularity.

## 3. Uniqueness mechanisms: Osgood, Bihari–LaSalle, and flow stability

The analytic core of Osgood-type growth is the divergence of an inverse-modulus integral. In the Lagrangian flow setting, the auxiliary function
\[
\psi_\delta(\xi)=\int_0^\xi \frac{ds}{\rho(s)+\delta},\qquad \delta>0,
\]
is increasing and concave, and satisfies
\[
\lim_{\delta\downarrow 0}\psi_\delta(\xi)=+\infty \quad \text{for every }\xi>0
\]
because \(\int_0^+ \frac{ds}{\rho(s)}=+\infty\) [1107.2496]. This yields a stability estimate for two flows \(X,X'\):
\[
\int_{B(R)} \psi_\delta\big(\|X(\cdot,x)-X'(\cdot,x)\|_{\infty,[0,T]}\big)\,dx
\le (L+L')\|g\|_{L^1([0,T]\times B(\bar R))}
+ \frac{\tilde L}{\delta}\|b-b'\|_{L^1([0,T]\times B(\bar R))},
\]
with \(\bar R = R + T(\|b\|_{L^\infty}+\|b'\|_{L^\infty})\) [1107.2496]. Setting \(b'=b\) forces equality of the two flows almost everywhere.

For nonlocal continuity systems, the same mechanism appears through Bihari–LaSalle rather than averaged concave transforms. The relevant modulus is not \(\omega_V\) alone but the composed modulus
\[
r\mapsto \omega_V\big(r+\rho\,\omega_n(r)\big),
\]
where \(\rho\) is later chosen in terms of total mass [2301.11822]. The Osgood condition is
\[
\int_0^{\rho}\frac{dr}{\omega_V(r+\rho\,\omega_n(r))}=+\infty\qquad \text{for each }\rho>0,
\]
and it implies uniqueness of Lagrangian weak solutions for the system of non-local continuity equations [2301.11822]. The stability estimate takes the form
\[
Q'(t)\le \omega_V\big(Q(t)+\rho\,\omega_n(Q(t))\big)+M,
\]
and when \(M=0\), the Osgood divergence forces \(Q\equiv 0\) [2301.11822].

In transport by Osgood vector fields, the flow distortion is quantified by
\[
\mathcal{M}(z)=\int_z^m \frac{dr}{\varphi(r)},\qquad R(z)=e^{-\mathcal{M}(z)},
\]
leading to
\[
|\phi^{-1}(x,t)-\phi^{-1}(y,t)|
\le R^{-1}\left(e^{\int_0^t [u(s)]_\varphi ds}R(|x-y|)\right)
=: \mu_{[u]_\varphi,t}(|x-y|).
\]
This is the Osgood analogue of exponential flow distortion in the Lipschitz case [2206.14237]. For \(\varphi(r)=r\log\frac{e}{r}\), the distortion becomes a power law with time-dependent exponent; for iterated-log moduli it becomes substantially weaker but still vanishes at \(0\), which is the decisive Osgood feature [2206.14237].

In BSDE theory, the one-sided Osgood condition is expressed as
\[
\left\langle \frac{y_1-y_2}{|y_1-y_2|}\mathbbm{1}_{\{|y_1-y_2|\neq 0\}},\,
g(t,y_1,z)-g(t,y_2,z)\right\rangle
\le u_t\,\rho(|y_1-y_2|),
\]
with \(\rho\) concave, nondecreasing, and satisfying
\[
\int_{0^+}\frac{du}{\rho(u)}=+\infty
\]
[2509.11927]. Stochastic Bihari inequalities then yield uniqueness of \(L^1\) solutions. The 2017 multidimensional \(L^1\) BSDE theory and its 2025 extension both use this mechanism, with the later work allowing stochastic coefficients and a general terminal time [1701.04152], [2509.11927].

## 4. Growth nonlinearities, blow-up criteria, and the ODE–PDE divide

When Osgood-type growth is placed on a scalar nonlinearity \(f\), the key integral becomes
\[
\int_1^\infty \frac{ds}{f(s)}.
\]
For the scalar ODE \(\dot v=f(v)\), divergence of this integral is necessary and sufficient for global existence of all positive solutions [1303.7183]. This motivates the expectation that Osgood nonlinearities should preclude finite-time blow-up more generally.

That intuition is reliable for bounded initial data in semilinear heat equations. If \(u_0\in L^\infty\), comparison with the scalar ODE implies global existence when
\[
\int_1^\infty \frac{ds}{f(s)}=\infty
\]
[1303.7183]. However, the same papers show that this intuition fails dramatically for rougher data. There exist locally Lipschitz, nondecreasing nonlinearities \(f\) satisfying the Osgood condition and yet, for every \(1\le q<\infty\), one can find \(u_0\in L^q\) such that the corresponding semilinear heat equation has no local integral solution [1303.7183], [1307.6688]. The mechanism is that singular initial data produce large short-time lower bounds on the linear heat flow, and the source term \(f(u)\) then becomes non-integrable in space-time. In this sense, Osgood growth controls temporal blow-up in ODEs but not the interaction of diffusion and spatial concentration for unbounded PDE data [1307.6688].

The same phenomenon survives in a non-Gaussian fractional-time setting. For
\[
\partial_t^\alpha (u-u_0)+\Psi_\beta(-i\nabla)u=f(u),
\]
Osgood-type nonlinearities modeled on piecewise-plateau functions \(f^{[k]}\) still satisfy
\[
\int_1^\infty \frac{ds}{f(s)}=\infty,
\]
yet local solutions can fail to exist in \(L_q\) when the Osgood index \(k\) exceeds the threshold
\[
k> q\left(1+\frac{\beta}{\alpha d}\right)
\]
[2405.13151]. This shows that Osgood-type growth on the source term is not, by itself, an existence criterion once fractional diffusion and singular initial data are involved.

A complementary SPDE result restores the classical Osgood intuition in a different regime. For the stochastic heat equation
\[
\partial_t u = \frac12 \Delta u + b(u) + \sigma(u)\dot W
\]
on \(\mathbb{R}^d\), if there exists an increasing function \(h\) such that \(u\mapsto h(u)/u\) is nondecreasing,
\[
\int_1^\infty \frac{1}{h(u)}\,du = \infty,
\]
\[
|b(u)|\le h(|u|),
\]
and \(\sigma\) satisfies the specific growth bound
\[
|\sigma(u)| \le |u|^{1-\alpha/2}\,h(|u|)^{\alpha/2}
\left[\log\Big(\frac{h(|u|)}{|u|}\Big)\right]^{-1/2},
\]
then there exists a unique global mild solution for initial data \(u_0\in V_p=L^p\cap L^\infty\) with \(p\ge (2+d)/\alpha\) [2310.02153]. The proof uses a discrete Osgood argument based on stopping times and a sequence
\[
a_n \sim \frac{3^{n+1}}{h(3^{n+1})},
\]
whose divergent sum is the discrete counterpart of the infinite Osgood integral [2310.02153]. In that drift-dominated regime, the finite versus infinite Osgood dichotomy again marks the blow-up threshold.

## 5. Regularity, transport, and Hamilton–Jacobi settings

In transport theory with Osgood drifts, the central question is not only uniqueness but also propagation and loss of regularity. Divergence-free vector fields with Osgood modulus
\[
|u(x,t)-u(y,t)|\le C\,\varphi(|x-y|),\qquad \int_0^m \frac{dr}{\varphi(r)}=+\infty
\]
generate unique flows [2206.14237]. Yet Sobolev regularity of passive scalars need not propagate. For every admissible growth function \(\Theta\), there exists a divergence-free vector field with associated Osgood modulus
\[
\varphi_\Theta(r)=r\log\frac{e}{r}\,\Theta\Big(\log\frac{e}{r}\Big)
\]
and initial datum \(\theta_0\in H^\sigma(\mathbb{R}^d)\) such that \(\theta(t)\notin H^s(\mathbb{R}^d)\) for every \(t>0\) and every \(s>0\) [2206.14237]. Thus Osgood continuity is a uniqueness threshold for the flow, not a Sobolev propagation threshold for the transported scalar.

The same paper develops a positive theory for modulus-based regularity. If \(\theta_0\) has modulus \(\mu_0\), then \(\theta(t)\) has modulus \(\mu_0\circ \mu_{[u]_\varphi,t}\), where \(\mu_{[u]_\varphi,t}\) is the explicit distortion modulus induced by the flow [2206.14237]. This suggests that Osgood-type growth should be viewed as compatible with a scale-sensitive, non-Sobolev regularity theory.

In Hamilton–Jacobi theory, Osgood-type growth appears as an upper growth condition in the scalar state variable \(u\). For the equation
\[
\partial_tu + H(x,u,D_xu)=0,\qquad u(x,0)=\phi(x),
\]
the relevant condition is that for every compact \(K\subset T^*M\), there exists a continuous \(f_K:[0,\infty)\to[0,\infty)\) such that
\[
\int_0^\infty \frac{1}{f_K(u)}\,du=+\infty,\qquad H(x,u,p)\le f_K(|u|)\quad \forall (x,p)\in K
\]
[1408.3790]. This is strictly weaker than monotonicity or uniform Lipschitz dependence on \(u\). It guarantees, through comparison with \(\dot u=f_K(u)\), that the \(u\)-component of characteristics does not blow up in finite time, allowing an implicit variational principle and representation of viscosity solutions by minimal characteristics [1408.3790].

Backward parabolic equations provide a different role for Osgood moduli. If the principal coefficients are Osgood continuous in time, meaning
\[
\int_0^1 \frac{1}{\mu(s)}\,ds=+\infty
\]
for the modulus \(\mu\), then uniqueness holds in natural function spaces, and one also has conditional stability in the form
\[
\sup_{t\in[0,T']}\|u(t,\cdot)\|_{L^2} \le \Psi(\|u(0,\cdot)\|_{L^2})
\]
for a modulus \(\Psi\) depending on \(\mu\) [1801.07890], [1902.08573]. The Osgood condition is thus the minimal time regularity threshold between uniqueness and known non-uniqueness phenomena. Stronger moduli such as Log-Lipschitz yield more explicit stability rates, while weaker non-Osgood moduli fail even uniqueness [1902.08573].

## 6. Sharpness, thresholds, and contemporary directions

The sharpness question asks whether the Osgood criterion is merely sufficient or actually optimal. Several supplied works support the latter interpretation.

For continuity equations with divergence-free vector fields, the 2026 sharpness result shows that for every modulus \(\omega\) that fails the Osgood condition, one can construct a divergence-free
\[
v\in C_t C_x^\omega
\]
such that the ODE admits at least two distinct flow maps on a set of initial data of positive Lebesgue measure, in fact full measure inside a supporting cube, and the continuity equation has two distinct solutions starting from the same absolutely continuous datum [2601.12096]. The two key innovations are “parallelization,” meaning simultaneous motion across nested scales, and a fixed-point framework adapted to that parallel structure [2601.12096]. This establishes that Osgood is a sharp threshold for uniqueness of both trajectories and continuity equations in that class.

For BSDEs, the same threshold phenomenon appears in the passage from Lipschitz or monotone generators to one-sided Osgood generators. The existence, uniqueness, and stability theory for multidimensional \(L^1\) solutions survives under Osgood-type growth in \(y\), provided one retains the integral divergence and suitable \(z\)-regularity [1701.04152], [2509.11927]. This indicates that the Bihari–Osgood mechanism is robust in stochastic evolution equations, not merely in deterministic flows.

A different perspective is provided by choiceless infinitesimal proofs of global Osgood theorems. In the SPOT framework, the initial value problem
\[
y'(x)=F(x,y(x)),\qquad y(0)=0
\]
with continuous \(F\) admits a unique maximal solution, obtained by adding a positive infinitesimal perturbation and taking a standard part [2311.01374]. This work does not center the usual integral modulus condition; rather, it reinterprets Osgood’s theorem as a global maximality principle arising from infinitesimal upward perturbation. A plausible implication is that “Osgood-type growth” has a broader conceptual role than modulus inequalities alone, encompassing maximality and continuation structures in nonstandard formulations.

In cosmological ODE reductions, Osgood’s criterion serves as an explicit finite-time singularity test. For FLRW models with perfect fluid, viscous fluid, or Chaplygin-type equations of state, the convergence or divergence of
\[
\int_{\theta_0}^\infty \frac{ds}{F(s)}
\]
for the expansion scalar \(\theta\), or the analogous integral for pressure \(p\), detects Type 0 and Type II singularities and identifies parameter-dependent initial data that avoid them [1507.02241]. This suggests that Osgood-type growth is not confined to regularity theory; it also functions as a sharp blow-up diagnostic in applied dynamical systems.

Taken together, the supplied literature presents Osgood-type growth as a unifying threshold concept. In modulus form, it separates uniqueness from non-uniqueness for ODEs, flows, and transport equations [1107.2496], [2601.12096]. In source-term form, it separates finite-time blow-up from global existence for scalar ODEs, but interacts in subtle ways with spatial singularity, diffusion, and noise in PDE and SPDE settings [1303.7183], [2310.02153]. In variational and stochastic contexts, it enables theories that are strictly weaker than Lipschitz while still quantitatively well posed [1408.3790], [2509.11927]. The common mechanism throughout is the divergence of an inverse-growth integral, which acts as the precise analytic boundary between controllable and uncontrollable growth.

Source: https://www.emergentmind.com/topics/osgood-type-growth