---
title: Parabolic Double Phase Equation
url: https://www.emergentmind.com/topics/parabolic-double-phase-equation
type: topic
---

# Parabolic Double Phase Equation

The parabolic double phase equation is a nonlinear parabolic PDE whose prototype is
\[
u_t-\operatorname{div}\!\left(|\nabla u|^{p-2}\nabla u+a(x,t)|\nabla u|^{q-2}\nabla u\right)=0,
\]
with energy density
\[
|\nabla u|^p+a(x,t)|\nabla u|^q.
\]
It is called “double phase” because the diffusion has two competing power-growth regimes: a \(p\)-phase and a \(q\)-phase activated by the nonnegative coefficient \(a(x,t)\). Where \(a(x,t)=0\), the equation reduces to the \(p\)-parabolic equation; where \(a(x,t)\) is positive, the \(q\)-growth part becomes active. In this sense the medium is heterogeneous in both the size of \(\nabla u\) and the space-time location \(z=(x,t)\) [2011.04373].

## 1. Canonical form and structural regimes

The standard model is
\[
u_t-\operatorname{div}\big(|Du|^{p-2}Du+a(z)|Du|^{q-2}Du\big)=0,
\qquad z=(x,t)\in \Omega_T:=\Omega\times(0,T),
\]
with \(a\ge 0\). In the degenerate setting one typically assumes \(2\le p<q\), while the singular setting treats \(\frac{2n}{n+2}<p\le 2\) with \(p<q\) [2601.01571, 2404.19111]. The coefficient \(a\) may be merely bounded and measurable, or Hölder continuous in the parabolic sense \(C^{\alpha,\alpha/2}\), depending on the theorem under consideration [2011.04373, 2511.13454].

A recurrent structural theme is that the equation is locally similar either to a \(p\)-Laplacian flow or to a \(q\)-Laplacian flow. In the Hölder-regular coefficient theory this is formalized by a phase analysis. In the \(q\)-phase, the \(q\)-term dominates in the intrinsic scaling; in the \(p\)-phase, the \(q\)-term is subordinate and the estimates reduce to a \(p\)-Laplacian type argument [2404.19111]. This local phase classification is not merely heuristic: it determines the intrinsic cylinders, the Sobolev–Poincaré inequalities, and the admissible gap conditions between \(p\) and \(q\).

The same double-phase mechanism appears in variable-growth problems. A representative variable-exponent model is
\[
u_t-\operatorname{div}\left((a(z)|\nabla u|^{p(z)-2}+b(z)|\nabla u|^{q(z)-2})\nabla u\right)=f
\quad \text{in }Q_T,
\]
with Lipschitz-continuous exponents and coefficients, together with the balance conditions
\[
a(z)+b(z)\ge \alpha>0,
\qquad |p(z)-q(z)|<\frac{2}{N+2}
\quad\text{in }\overline{Q}_T
\]
[2109.03597]. In that setting, coercivity is preserved because at each point at least one phase is active.

## 2. Weak and strong solution frameworks

For the prototype constant-exponent problem, weak solutions are the natural solution concept. One formulation requires
\[
u\in C(0,T;L^2(\Omega))\cap L^1(0,T;W^{1,1}(\Omega)),
\qquad \int_{\Omega_T}\big(|Du|^p+a(z)|Du|^q\big)\,dz<\infty,
\]
and the integral identity
\[
\int_{\Omega_T}\big(-u\,\varphi_t+\mathcal A(z,Du)\cdot D\varphi\big)\,dz=0
\]
for all \(\varphi\in C_0^\infty(\Omega_T)\), where
\[
\mathcal A(z,\xi)=|\xi|^{p-2}\xi+a(z)|\xi|^{q-2}\xi
\]
[2511.13454]. In more general inhomogeneous problems one replaces the right-hand side by divergence-form data involving a field \(F\) and the same double-phase structure [2601.01571, 2304.09776].

Several works use stronger ambient spaces, for example
\[
u\in C(0,T;L^2(\Omega))\cap L^q(0,T;W^{1,q}(\Omega)),
\]
especially when proving Hölder continuity or boundedness under \(q\)-growth assumptions [2404.19111, 2604.14544]. A separate technical point is that parabolic Lipschitz truncation can recover the more natural energy space without assuming \(Du\in L^q\) a priori [2304.09776].

Strong solutions arise in the variable-growth theory. For the Dirichlet problem with
\[
u_t-\operatorname{div}\Big(|\nabla u|^{p(z)-2}\nabla u+a(z)|\nabla u|^{q(z)-2}\nabla u\Big)=F(z,u),
\]
one obtains existence and uniqueness of strong solutions with
\[
u_t\in L^2(Q_T),
\qquad |\nabla u|^{s(z)}\in L^\infty(0,T;L^1(\Omega)),
\qquad s(z)=\max\{2,p(z)\},
\]
and global higher integrability
\[
|\nabla u|^{p(z)+\delta}\in L^1(Q_T)\quad\text{for every }0<\delta<r^*
\]
under a small-gap condition \(q(z)\le p(z)+\frac r2\) with \(0<r<r^*\) [2010.08306]. Related strong-solution theories with global higher integrability and second-order spatial regularity have also been proved for variable-growth double phase fluxes [2109.03597, 2607.04492].

## 3. Interior regularity theory

The basic interior regularity problem is local boundedness. For the prototype equation with measurable \(a\), one optimal boundedness result proves that weak solutions are locally bounded when
\[
\frac{2N}{N+2}<p<\infty,\qquad 0\le a(x,t)\le M,\qquad q<p\frac{N+1}{N-1}
\]
[2011.04373]. A later bounded-coefficient result establishes local boundedness under
\[
0\le a\in L^\infty(\Omega_T),\qquad 2\le p<q<p+\frac{p}{n},
\]
using a Caccioppoli inequality, a parabolic embedding theorem, and an iteration method [2604.14544]. These theorems address different structural regimes and should not be conflated into a single universal threshold.

For bounded weak solutions with Hölder-continuous coefficients, local Hölder continuity is now available in the degenerate case. If
\[
2\le p<q\le p+\alpha,
\qquad a\in C^{\alpha,\alpha/2}(\Omega_T;\mathbb R_{\ge 0}),
\]
then bounded weak solutions are locally Hölder continuous, and the proof is based on phase analysis and methods for the \(p\)-Laplace equation [2404.19111]. In that framework, the phase analysis determines whether the equation is locally similar to the \(p\)-Laplace or the \(q\)-Laplace equation.

A further step is gradient higher integrability. For bounded solutions to
\[
u_t-\operatorname{div}\left(|Du|^{p-2}Du+a(x,t)|Du|^{q-2}Du\right)=0
\]
with \(a\in C^{\alpha,\alpha/2}(\Omega_T)\), one obtains local higher integrability of
\[
H(z,|Du|):=|Du|^p+a(z)|Du|^q
\]
under the gap condition
\[
q\le p+\alpha.
\]
More generally, if
\[
u\in C(0,T;L^s(\Omega)),\qquad s\ge 2,
\]
then higher integrability holds under
\[
q\le p+\frac{s\alpha}{n+s},
\]
which interpolates between the \(s=2\) threshold and the bounded-solution threshold [2511.13454]. In the singular regime \(\frac{2n}{n+2}<p\le2\), a parallel interpolation theory gives higher integrability under
\[
q\le p+\frac{\alpha(p(n+2)-2n)}{4}
\]
for bounded solutions, or
\[
q\le p+\frac{\alpha\mu_s}{n+s},
\qquad \mu_s=\frac{(p(n+2)-2n)s}{4},
\]
for \(u\in C(0,T;L^s(\Omega))\) [2601.01571].

Regularity beyond Hölder continuity has also been obtained. For bounded continuous weak solutions with \(1<p\le q\le p+1\), bounded \(a\), local spatial Lipschitz continuity of \(a\), and continuity in time, local spatial Lipschitz regularity and time Hölder regularity follow for the equation with gradient-dependent forcing
\[
\partial_t u-\operatorname{div}\Big(|Du|^{p-2}Du+a(z)|Du|^{q-2}Du\Big)=f(z,Du).
\]
In the degenerate regime \(p\ge2\), the time exponent improves to the sharp \(1/2\) [2508.16391].

## 4. Boundary behavior, Harnack theory, and pointwise continuity estimates

Boundary regularity in the parabolic double-phase setting is phase-dependent. For bounded weak solutions of
\[
\partial_t u-\operatorname{div}A(x,t,\nabla u)=0
\]
with prototype
\[
A(x,t,\xi)\sim |\xi|^{p-2}\xi+a(x,t)|\xi|^{q-2}\xi,\qquad 2<p<q,
\]
a Wiener-type sufficient criterion describes continuity up to the lateral boundary. If \(a(x_0,t_0)=0\), regularity is governed by the \(p\)-capacity; if \(a(x_0,t_0)>0\), it is governed by the \(q\)-capacity [2403.06550]. This phase-sensitive criterion leads, under uniform fatness or density assumptions, to boundary Hölder continuity.

Weak Harnack theory also reflects the double-phase threshold phenomena. For non-negative super-solutions of
\[
u_t-\operatorname{div}A(x,t,Du)=0
\]
with \(2<p<q\), a weak Harnack inequality is proved in the case
\[
q>p\frac{n+2}{n}
\]
under the additional assumption
\[
u\in L^s_{\mathrm{loc}}(\Omega_T)\quad\text{with some }s>p\frac{n+2}{n}
\]
[2305.13053]. The need for extra integrability is a structural feature of the supercritical \(q\)-regime.

A different continuity theory controls the oscillation of bounded weak solutions by nonlinear potentials of the data. For
\[
\partial_t u-\operatorname{div}A(x,t,u,Du)=f,
\]
with
\[
A(x,t,u,\xi)\cdot \xi \ge C_0\bigl(|\xi|^p+a(x,t)|\xi|^q\bigr),
\qquad
|A(x,t,u,\xi)|\le C_1\bigl(|\xi|^{p-1}+a(x,t)|\xi|^{q-1}\bigr)+g(x),
\]
and assumptions
\[
2<p<q\le p+\alpha,\qquad 2<p<n,\qquad f\in \widetilde K_p,\qquad g\in K_p,
\]
one obtains local continuity estimates in terms of elliptic Riesz potentials \(F_p\) and \(G_p\) [2502.01097]. This provides a potential-theoretic continuity criterion adapted to the degenerate double-phase parabolic structure.

## 5. Main variants and generalizations

The parabolic double phase literature now includes several distinct model classes, each with its own analytic thresholds.

| Model class | Representative form | Representative result |
|---|---|---|
| Variable-growth local equations | \(u_t-\operatorname{div}\left((a(z)|\nabla u|^{p(z)-2}+b(z)|\nabla u|^{q(z)-2})\nabla u\right)=f\) | Strong solutions, global higher integrability, second-order regularity [2109.03597] |
| Variable-exponent irregular evolution | \(u_t-\operatorname{div}(\mathcal F(z,\nabla u)\nabla u)=f(z)\) | Global Calderón–Zygmund transfer, higher integrability, second-order space regularity [2607.04492] |
| Nonlocal double phase equations | Fractional \(p\)- and \(q\)-phase integral operator with bounded measurable \(a(x,y)\) | Local boundedness of variational solutions [2112.02345] |
| Mixed local/nonlocal equations | \(\partial_t u-\mathrm{div}(a(x,t)|\nabla u|^{q-2}\nabla u)+\mathcal Lu=0\) | Local boundedness; semicontinuity of super- and subsolutions [2306.14160] |
| Normalized double phase flow | \(u_t=|\nabla u|^{2-p}\operatorname{div}\big(|\nabla u|^{p-2}\nabla u+a(x,t)|\nabla u|^{q-2}\nabla u\big)\) | Asymptotic mean value characterization in the viscosity sense [2211.16003] |
| Parabolic systems of double phase type | \(u_t-\operatorname{div}\mathbf A(z,Du)=0\) | Partial regularity: \(Du\) locally Hölder continuous except on a set of measure zero [2510.03849] |

For systems, a recent partial regularity theorem proves that if
\[
\frac{2n}{n+2}<p\le q,\qquad a\in C^{0,\alpha,\alpha/2}(\Omega_T),\qquad
q<\min\left\{p+\frac{\alpha p}{n+2},\,p+1\right\},
\]
then the spatial gradient of any weak solution is locally Hölder continuous except on a set of measure zero [2510.03849]. This is a genuinely parabolic system result, not merely a scalar extension.

The existence, uniqueness, and energy theory for inhomogeneous Dirichlet problems has also been strengthened by a parabolic double-phase Lipschitz truncation method based on a Whitney-type covering result and a related partition of unity in the intrinsic geometry for the double-phase problem [2304.09776].

## 6. Analytic mechanisms, gap conditions, and terminological distinctions

The dominant analytic tools are Caccioppoli inequalities, intrinsic cylinders, reverse Hölder inequalities, De Giorgi iteration, Whitney-type coverings, and covering arguments of Vitali type [2304.09776, 2511.13454]. In the singular higher-integrability theory, the proof uses intrinsic cylinders adapted to the singular scaling, a stopping-time argument splitting into \(p\)-phase and \((p,q)\)-phase cylinders, Caccioppoli inequalities and Poincaré/Gagliardo–Nirenberg inequalities, a proof that the “bad” third case cannot occur thanks to the gap condition, a reverse Hölder inequality for \(H(z,|Du|)\), and a Vitali covering argument plus Fubini to deduce higher integrability [2601.01571].

Gap conditions are a central organizing principle, but the literature does not support a single universal “optimal” formula valid across all settings. Bounded measurable coefficients yield local boundedness under one class of thresholds [2011.04373, 2604.14544]; Hölder-continuous coefficients yield Hölder continuity and higher integrability under \(q\le p+\alpha\) or its interpolative refinements [2404.19111, 2511.13454]; variable-exponent strong-solution theories impose balance conditions such as \(|p(z)-q(z)|<\frac{2}{N+2}\) [2109.03597]; and the coefficient class \(\mathcal Z^\kappa(\Omega_T)\) leads to the parabolic gap bound
\[
2 \le p \le q \le p + \frac{q\kappa}{q - 2\gamma}
\]
for Hölder continuous weak solutions [2606.10590]. That last condition is explicitly described as purely parabolic in nature and stricter than the optimal gap relation associated with the Lavrentiev phenomenon for the elliptic double phase functional [2606.10590].

A further conceptual distinction concerns terminology. In the standard nonlinear-parabolic literature, “double phase” refers to \(p\)- and \(q\)-growth modulated by a nonnegative coefficient \(a(z)\). A different line of work uses “two-phase” for one-dimensional forward-backward diffusion laws with two forward-diffusion phases separated by an unstable interval. There the model equation is
\[
u_t=(\sigma(u_x))_x+b(x,t)u_x+c(x,t)u+f(x,t)
\]
or, in the simpler entropy framework,
\[
u_t=\phi(u)_{xx},
\]
and the emphasis is on phase transitions, hysteresis, nonuniqueness, and generalized Abel equations rather than on \(p/q\)-growth regularity [1606.08546, 1310.7728]. The shared vocabulary reflects phase coexistence, but the PDE structures are distinct.

Taken together, these developments show that the parabolic double phase equation is not a single theorem or a fixed regularity paradigm. It is a class of nonuniformly elliptic or parabolic evolution equations in which the interaction between the growth gap \(q-p\), the regularity or geometry of the coefficient \(a\), and the intrinsic parabolic scaling determines whether one can prove local boundedness, Hölder or Lipschitz continuity, higher integrability, strong solvability, boundary regularity, or only partial regularity.

Source: https://www.emergentmind.com/topics/parabolic-double-phase-equation