Abstract
For a general class of divergence type quasi-linear singular parabolic equations with generalized Orlicz growth, we prove the intrinsic Harnack inequality for positive solutions. This class of singular equations includes new cases of equations with (p, q) nonlinearity and non-logarithmic growth.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction and main results
In this paper, we are concerned with general divergence type singular parabolic equations with nonstandard growth conditions. Let \(\Omega \) be a domain in \(\mathbb {R}^{n}\), \(T>0\), \(\Omega _{T}:= \Omega \times (0, T)\), we study bounded solutions to the equation
Throughout the paper, we suppose that the functions \(\mathbb {A}:\Omega _{T}\times \mathbb {R}^{n} \rightarrow \mathbb {R}^{n}\) are such that \(\mathbb {A}(\cdot , \cdot , \xi )\) are Lebesgue measurable for all \(\xi \in \mathbb {R}^{n}\), and \(\mathbb {A}(x, t, \cdot )\) are continuous for almost all \((x, t)\in \Omega _{T}\). We also assume that the following structure conditions are satisfied
where \(K_{1}\), \(K_{2}\) are positive constants and the function \(g(x, t, v): \Omega _{T}\times \mathbb {R}_{+}\rightarrow \mathbb {R}_{+}\) satisfies the following conditions:
- (\(\mathrm{g}_{0}\)):
-
for all \((x,t)\in \Omega _{T}\) the function \(g(x, t, \cdot )\) is increasing, continuous and
$$\begin{aligned} \lim \limits _{v\rightarrow +0} g(x, t, v)=0, \quad \lim \limits _{v\rightarrow +\infty } g(x, t, v)=+\infty ; \end{aligned}$$ - (\(\mathrm{g}_{1}\)):
-
\(c^{-1}_{0}\leqslant g(x, t, 1) \leqslant c_{0}\) for all \((x,t)\in \Omega _{T}\) and some \(c_{0}>0\);
- (\(\mathrm{g}_{2}\)):
-
for all \((x,t)\in \Omega _{T}\) and for \(w\geqslant v > 0\) there hold
$$\begin{aligned} \left( \frac{w}{v}\right) ^{p-1} \leqslant \frac{g(x, t, w)}{g(x, t, v)} \leqslant \left( \frac{w}{v}\right) ^{q-1}, \quad \frac{2n}{n+1}< p <\min \{2, q\},~~q\ne 2. \end{aligned}$$
Fix \(R_{0}>0\) such that \(Q_{R_{0},R_{0}}(x_{0}, t_{0})\subset \Omega _{T}\), where
In what follows, we assume that
- (\(\mathrm{g}_{\lambda }\)):
-
there exists positive, continuous and non-decreasing function \(\lambda (r)\) on the interval \((0, R_{0})\), \(0<\lambda (r)\leqslant 1\), \(\lim \limits _{r\rightarrow 0} r^{1-\overline{a}}/\lambda (r)=0\), \(\overline{a} \in (0,1)\), such that for any K, \(K_{0}>0\) and some \(c_{1}(K,K_{0})>0, b\geqslant 0\) there holds
$$\begin{aligned} g(x, t, v/r)\leqslant c_{1}(K,K_{0})\, g(y, \tau , v/r) \end{aligned}$$for all \(b r< v \leqslant K \lambda (r)\) and \((x, t), (y, \tau ) \in Q_{r, rK_{0}} (x_{0}, t_{0}) \subset Q_{R_{0},R_{0}}(x_{0},t_{0})\).
- (\(\mathrm{g}_{\mu }\)):
-
there exists positive, continuous and non-increasing function \(\mu (r)\) on the interval \((0,R_{0})\), \(\mu (r)\geqslant 1\), \(\lim \limits _{r\rightarrow 0}r^{1-\overline{a}} \mu (r)=0\), \(\overline{a}\in (0,1)\) such that for any \(K, K_{0} >0\) and some \(c_{2}(K, K_{0})>0, b\geqslant 0\) there holds
$$\begin{aligned} g(x, t, v/r)\leqslant c_{2}(K,K_{0}) \mu (r)\, g(y, \tau , v/r), \end{aligned}$$for all \(b r< v \leqslant K\) and \((x, t), (y, \tau )\in Q_{r, rK_{0}}(x_{0}, t_{0}) \subset Q_{R_{0},R_{0}}(x_{0},t_{0})\).
As one can easily see, for \(\lambda (r)=\mu (r)=1\), \((\mathrm{g}_{\lambda })\) and \((\mathrm{g}_{\mu })\) reduce to the standard Zhikov’s logarithmic condition. We will also need the following technical inequality
which we assume with some \(c_{3}>0\) and for all \(0<r\leqslant \rho \leqslant R_{0}\).
We will establish that nonnegative bounded weak solutions of Eq. (1) satisfy an intrinsic form of the Harnack inequality in a neighborhood of \((x_{0}, t_{0})\) provided that this p.d.e. is singular, i.e., the function \(\psi (x, t, v):=g(x, t, v)/v\) satisfies the following assumptions:
- (\(\psi \)):
-
there exist \(a_{0}, b_{0} \geqslant 0\), \(0<q_{1}<1\) and \(R_{0}=R_{0}(x_{0}, t_{0})>0\) such that
$$\begin{aligned} \dfrac{\psi (x, t, w)}{\psi (x, t, v)}\leqslant \left( \dfrac{v}{w}\right) ^{1-q_{1}} \end{aligned}$$for all \((x, t) \in Q_{R_{0},R_{0}}(x_{0}, t_{0})\subset \Omega _{T}\) and for \(w\geqslant v > b_{0}R_{0}^{-a_{0}}.\)
Condition \((\psi )\) with \(q_{1}=q-1\) and \(b_{0}=a_{0}=0\) is a consequence of \((\mathrm{g}_{2})\) in the case \(q<2\).
Remark 1
We note that the continuity of solutions was proved in [34] under weaker conditions, namely it was assumed that \(\psi (x_{0}, t_{0}, v)\) is non-increasing (”singular” case) or \(\psi (x_{0}, t_{0}, v)\) is non-decreasing (”degenerate” case) for \(v\geqslant b_{0} R_{0}^{-a_{0}}\). Unfortunately, we were unable to prove Harnack’s inequality under similar conditions. The following examples show that additional conditions on the function g(x, t, v) arise naturally.
Example 1
Consider the function
where \(a(x, t)\geqslant 0\), \(q>p\), \( \mathrm {osc}_{Q_{r, r}(x_{0},t_{0})}a(x,t)\leqslant A r^{q-p} \mu _{1}(r)\), \(0<A<+\infty \),
The function \(g_{1}\) satisfies condition (\(\mathrm{g}_{\lambda }\)) with \(\lambda (r)=[\mu _{1}(r)]^{-1/(q-p)}\) and \(b=0\). Indeed,
if \(0< v \leqslant K \lambda (r)\) and (x, t), \((y, \tau )\in Q_{r,r}(x_{0},t_{0})\). Condition \((\mathrm{g}_{\mu })\) with \(\mu (r)=\mu _{1}(r)\) and \(b=0\) is verified similarly.
Note that condition \((\psi )\) can be violated in the case \(q>2\) and \(a(x_{0}, t_{0})=0\).
Example 2
Consider the function
where \(q>p\), \(\beta >0\), \(a(x, t)\geqslant 0\), \(\mathrm {osc}_{Q_{r, r}(x_{0},t_{0})}a(x,t)\leqslant A r^{q-p} \mu _{2}(r)\), \(0<A<+\infty \),
Conditions \((\mathrm{g}_{\lambda })\) and \((\mathrm{g}_{\mu })\) with \(\lambda (r)=[\mu _{2}(r)]^{-\frac{1}{q-p}} (\ln r^{-1})^{-\frac{\beta }{q-p}}, \mu (r)=\mu _{2}(r)\times \) \(\times (\ln r^{-1})^{\beta }\) and \(b=0\) are checked similarly to Example 1. Condition \((\psi )\) is a consequence of condition \((\mathrm{g}_{1})\) in the case \(q+\beta <2\) . In the case \(q+\beta >2\) and \(a(x_{0},t_{0})=0\) condition \((\psi )\) is violated. In the case \(q>2\) and \(a(x_{0},t_{0})>0\) condition \((\psi )\) is also fails . Let us check condition \((\psi )\) in the case \(q<2<q+\beta \) and \(a(x_{0},t_{0})>0\). For this we note that
which implies \((\psi )\) with \(q_{1}=q/2\), \(a_{0}=0\) and \(b_{0}=e^{\frac{2\beta }{2-q}}-1\).
Example 3
The function
where \(0<L<2-p\), \(a(x, t)\geqslant 0\), \(\text {osc}_{Q_{r, r}(x_{0},t_{0})} a(x, t)\leqslant B r \mu _{3}(r)\), \(0<B<+\infty \),
satisfies condition \((\mathrm{g}_{\lambda })\) with \(\lambda (r)=1/\mu _{3}(r)\) and \(b=0\). Indeed,
if \(0< v \leqslant K \lambda (r)\) and \((x, t), (y, \tau )\in Q_{r, r}(x_{0}, t_{0})\). Condition \((\mathrm{g}_{\mu })\) with \(\mu (r)=\ln \mu _{3}(r)\) and \(b=0\) we obtain similarly. Condition \((\psi )\) holds since \(q=p+L<2\).
We note also that the function \(\tilde{g}_{3}(x, t, v):= v^{p-1}(1+\ln (1+a(x, t)v))\) satisfies condition \((\psi )\) in the case \(p<2\) and \(a(x_{0}, t_{0})>0\). To check this we note that
Choose \(R_{0}\) from the condition \(B R_{0}\mu (R_{0})=\frac{1}{2} a(x_{0},t_{0})\). This choice guarantees that
So, if \(v\geqslant \frac{1}{4B}(e^{\frac{p}{2-p}}-1)R_{0}^{-1}\), from the previous we obtain
which implies condition \((\psi )\) with \(b_{0}=\frac{1}{4B}(e^{\frac{p}{2-p}}-1)\), \(a_{0}=1\) and \(q_{1}=p/2<1\). Unfortunately, condition \((\psi )\) is violated for the function \(\tilde{g}_{3}\) in the case \(a(x_{0},t_{0})=0\) and \(p<2\).
Example 4
Consider the functions
where \(p<2\), \(p(x, t)\leqslant q < 2\), \(a(x, t)\geqslant 0\) and
It is obvious that \(g_{4}(x, t, v)\) satisfies conditions \((\mathrm{g}_{\lambda })\),\((\mathrm{g}_{\mu })\) with \(\lambda (r)=\mu (r)=1\) and \(b=0\).
Similarly, the function \(g_{5}(x, t, v)\) satisfies conditions \((\mathrm{g}_{\lambda })\),\((\mathrm{g}_{\mu })\) with \(\lambda (r)=\mu (r)=1\) and \(b=1\). To check condition \((\psi )\) for the function \(g_{4}\), we note
if \(v \geqslant e^{\frac{2}{2-p}}-1\), which implies condition \((\psi )\) with \(b_{0}=e^{\frac{2}{2-p}}-1\), \(a_{0}=0\) and \(q_{1}=p/2\).
Before describing the main results, a few words concern the history of the problem. The study of regularity of minima of functionals with nonstandard growth has been initiated by Kolodij [25, 26], Zhikov [44,45,46,47, 49], Marcellini [30, 31] and Lieberman [29], and in the last thirty years there has been growing interest and substantial development in the qualitative theory of quasi-linear elliptic and parabolic equations with so-called ”log-conditions” (i.e., if \(\lambda (r)=1\) and \(\mu (r)=1\)). We refer the reader to the papers [1, 3,4,5,6,7,8,9,10, 12,13,14, 18, 19, 21,22,24, 36,37,43, 51] for the basic results, historical surveys and references.
The case when conditions \((\mathrm{g}_{\lambda })\) or \((\mathrm{g}_{\mu })\) hold differ substantially from the logarithmic case. To our knowledge, there are few results in this direction. Zhikov [48] obtained a generalization of the logarithmic condition which guarantees the density of smooth functions in Sobolev space \(W^{1, p(x)}(\Omega )\). Particularly, this result holds if \(p(x) \geqslant p>1\), and for every \(x, y \in \Omega \), \(x \ne y\),
We note that the function \(\mu (r)=(\ln r^{-1})^{L}\), \(0\leqslant L \leqslant p/n\) satisfies the above condition. Later Zhikov and Pastuchova [50] under the same condition proved higher integrability of the gradient of solutions to the p(x)-Laplace equation.
Interior continuity, continuity up to the boundary and Harnack’s inequality to the p(x)-Laplace equation were proved by Alkhutov, Krasheninnikova [1], Alkhutov, Surnachev [2] and Surnachev [35] under the conditions
with some \(\gamma ,~c>1\). Particularly, the function \(\mu (r)=(\ln \ln r^{-1})^{L}\), \(0<L<1/c\) satisfies the above condition.
These results were generalized in [32, 33] for a wide class of elliptic and parabolic equations with non-logarithmic Orlicz growth. Later, the results from [32, 33] were substantially refined in [20, 34]. For interior continuity, instead of condition (3), it was required that
In addition, in [20] Harnack’s inequality was proved for quasilinear elliptic equations under the condition
We note that this condition is worse than condition (4), but at the same time it is much better then condition (3).
In this paper, we establish the intrinsic Harnack inequality for nonnegative solutions to Eq. (1) under the similar conditions as (5). To describe our results, let us introduce the definition of a weak solution to Eq. (1).
Definition 1
We say that u is a bounded weak sub (super) solution to
Eq. (1) if
and for any compact set \(E \subset \Omega \) and every subinterval \([t_{1}, t_{2}]\subset (0, T]\) the integral identity
holds true for any testing functions \(\eta \in W^{1,2}(0, T; L^{2}(E))\cap L^{q}(0, T; W_{0}^{1,q}(E))\), \(\eta \geqslant 0\).
It would be technically convenient to have a formulation of weak solution that involves \(u_{t}\). Let \(\rho (x)\in C_{0}^{\infty }(\mathbb {R}^{n})\), \(\rho (x)\geqslant 0\) in \(\mathbb {R}^{n}\), \(\rho (x)\equiv 0\) for \(|x| > 1\) and \(\int \nolimits _{\mathbb {R}^{n}}\rho (x)\,dx=1\), and set
Fix \(t \in (0, T)\) and let \(h>0\) be so small that \(0<t<t+h<T\). In (1.6) take \(t_{1}=t\), \(t_{2}=t+h\) and replace \(\eta \) by \(\int \nolimits _{\mathbb {R}^{n}}\eta (y, t)\rho _{h}(x-y)\,dy\). Dividing by h, since the testing function does not depend on \(\tau \), we obtain
for all \(t \in (0, T-h)\) and for all nonnegative \(\eta \in W^{1,q}_{0}(E)\).
We refer to the parameters \(M:= \sup \limits _{\Omega _{T}}u\), \(K_{1}\), \(K_{2}\), \(c_{0}\), \(c_{1}=c_{1}(M,M)\), \(c_{2}=c_{2}(M,M)\), \(c_{3}\), n, p, q and \(q_{1}\) as our structural data, and we write \(\gamma \) if it can be quantitatively determined a priori only in terms of the above quantities. The generic constant \(\gamma \) may change from line to line. Note that the constants \(a_{0}\), b and \(b_{0}\) can be equal to zero, therefore, in the proof we keep an explicit track of the dependence of the various constants on \(a_{0}\), b and \(b_{0}\).
Our first main result concerns the Harnack inequality for the case of logarithmic growth (i.e., \(\lambda (r)= \mu (r)=1\)). Fix a point \((x_{0}, t_{0})\in \Omega _{T}\) such that \(u(x_{0}, t_{0})>0\) and consider the cylinders
Theorem 1
Let u be a positive bounded weak solution to Eq. (1). Let the conditions (2), \((\mathrm{g}_{0})\)– \((\mathrm{g}_{2})\), \((\psi )\), \((\mathrm{g}_{\lambda })\) and \((\mathrm{g}_{\mu })\) with \(\lambda (r)=\mu (r)=1\) be fulfilled in the cylinder \(Q_{R_{0}, R_{0}}(x_{0}, t_{0})\subset \Omega _{T}\). Then, there exist positive constants c, \(\bar{c}\), \(\delta _{1}\) depending only upon the data such that for all cylinders \(Q_{8\rho , 8\theta }(x_{0}, t_{0}) \subset Q_{8\rho ,8\bar{c}\rho }(x_{0},t_{0}) \subset Q_{R_{0}, R_{0}}(x_{0}, t_{0})\), \(0<\rho \leqslant R_{0}^{a_{1}}\), \(a_{1}=1 +a_{0}/\overline{a}\), either
or
for all times \(|t-t_{0}| \leqslant \rho ^{2}/\psi (x_{0}, t_{0}, \delta _{1} u (x_{0}, t_{0})\rho ^{-1})\), and the numbers \(a_{0}\), \(\bar{a}\) are defined in conditions \((\mathrm{g}_{\lambda })\), \((\mathrm{g}_{\mu })\) and \((\psi )\).
Our next result is the Harnack inequality for the case of non-logarithmic growth. Having fixed \((x_{0}, t_{0})\in \Omega _{T}\), construct the cylinder \(Q_{\rho , {\bar{\theta }}}(x_{0}, t_{0})\), where
Theorem 2
Let u be a positive bounded weak solution to Eq. (1). Let the conditions (2), \((\mathrm{g}_{0})\)– \((\mathrm{g}_{2})\), \((\mathrm{g}_{\lambda })\), \((\mathrm{g}_{\mu })\) and \((\psi )\) be fulfilled in the cylinder
\(Q_{R_{0}, R_{0}}(x_{0}, t_{0})\subset \Omega _{T}\). Assume also that
If additionally \(\int \nolimits _{0}\lambda _{1}(r)r^{-1}\,dr=+\infty \) and \(\lim \limits _{r\rightarrow 0} r^{1-\bar{a}_{1}}/\lambda _{1}(r)=0\) with some \(\bar{a}_{1}\in (0, 1)\), then there exist positive constants c, \(\bar{c}\), \(\delta _{1}\) depending only upon the data such that for all cylinders
either
or
for all times \(|t-t_{0}| \leqslant \rho ^{2}/\psi \left( x_{0}, t_{0}, \delta _{1} \rho ^{-1}\lambda _{1}(\rho )u (x_{0}, t_{0})\right) \).
Remark 2
We note that in the case \(g_{1}(x, t, v)=v^{p-1}+a(x, t) v^{q-1}\),
inequality (12) translates into
Similar result is also valid for the function \(g_{2}(x,t,v)=v^{p-1}+a(x,t)v^{q-1}(1+\ln (1+v))^{\beta }\) and
In addition, note that in the case \(g_{3}(x, t, v)=v^{p-1}(1+L \ln (1+a(x, t)))\) and
inequality (12) can be rewritten as (13).
We would like to mention the approach taken in this paper. To prove our results, we use Di Benedetto’s approach [15], who developed an innovative intrinsic scaling methods for degenerate and singular parabolic equations. For the p-Laplace evolution equation, the intrinsic Harnack inequality was proved in the famous papers [16, 17]. The main stage in the proof of our results is the so-called \(L^{1}_{\mathrm{loc}}\)– \(L^{\infty }_{\mathrm{loc}}\) Harnack inequality.
Theorem 3
Let u be a nonnegative, bounded weak solution to Eq. (1) and let (2), \((\mathrm{g}_{0})\)– \((\mathrm{g}_{2})\), \((\mathrm{g}_{\mu })\) and \((\psi )\) be fulfilled in the cylinder \(Q_{R_{0}, R_{0}}(x_{0}, t_{0})\subset \Omega _{T}\). Then for all cylinders
provided that
the constant \(\gamma \) depends only on the data, \(\bar{c}\) and a. Here,
and \(G^{-1}_{Q}(\cdot )\), \(\psi ^{-1}_{Q}( \cdot )\) and \(\varphi ^{-1}_{Q}(\cdot )\) are the inverse functions to \(G_{Q}(\cdot )\), \(\psi _{Q} (\cdot )\) and \(\varphi _{Q}(\cdot )\) respectively.
Remark 3
Note that by our choices \(r^{-a_{0}/a} \geqslant R_{0}^{-a_{0}}\), therefore, condition \((\psi )\) is applicable and the left-hand side of inequality (15) makes sense. In addition, the function \(\varphi _{Q_{2r,2(t-s)}^{+}(y,2s-t)}(\cdot )\) is strictly increasing (see Lemma 1 below), so, the right-hand side of inequality (14) also makes sense.
Estimate (14) coincides with the well-known \(L^{1}_{\mathrm{loc}}\)– \(L^{\infty }_{\mathrm{loc}}\) form of Harnack’s inequality (see [15]) in the case of p-Laplace evolution equation (\(p<2\)) :
where \(\kappa =p+n(p-2)>0\).
Main difficulty arising in the proof of our main results is related to the so-called theorem on the expansion of positivity. Roughly speaking, having information on the measure of the ”positivity set” of u over the ball \(B_{r}(\bar{x})\) for some time level \(\bar{t}\):
with some \(r>0\), \(N>0\) and \(\alpha (r)\in (0, 1)\), \(\alpha (r)\rightarrow 0\), as \(r\rightarrow 0\), and using the standard Di Benedetto’s arguments, we inevitably arrive at the estimate
for some time level \(t>\bar{t}\) and with some \(\gamma _{1}\), \(\gamma _{2}, \gamma _{3} >1\). This estimate leads us to condition similar to that of (3) (see, e.g., [34]). To avoid this, we use a workaround that goes back to Landis’s papers [27, 28] and his so called ”growth” lemma. So, in Sect. 3 we use the auxiliary solutions and prove the integral and pointwise estimates of these solutions.
The rest of the paper contains the proof of the above theorem. In Sect. 2, we give a proof of \(L^{1}_{\mathrm{loc}}\)– \(L^{\infty }_{\mathrm{loc}}\) Harnack’s inequality, Theorem 3. Section 3 contains the upper and lower bounds for the auxiliary solutions. Finally, in Sect. 4 we prove our main results, Theorems 1 and 2.
2 \(L^{1}_{\mathrm{loc}}\)– \(L^{\infty }_{\mathrm{loc}}\) Harnack type inequality, Proof of Theorem 3
2.1 An auxiliary lemma
The following inequalities will be used in the sequel, they are simple consequences of the condition \((\mathrm{g}_{2})\).
Lemma 1
The following inequalities hold:
if v, w, \(\varepsilon >0, ~ (x, t)\in \Omega _{T}\);
Here
Note that the third and fourth inequalities in Lemma 1 ensure that the functions \(G(\cdot , v)\) and \(\varphi (\cdot , v)\) are increasing, and therefore, the inverse functions \(G^{-1}(\cdot , v)\) and \(\varphi ^{-1}(\cdot , v)\) are exist.
2.2 An \(L^{1}_{\mathrm{loc}}\) form of Harnack’s inequality
For fixed \((y, s)\in \Omega _{T}\), for \(0< r \leqslant R^{a}\), \(a \geqslant 1+a_{0}\) and for \(0<t-s<\bar{c}r\), we will assume later that inequality (15) holds, i.e.,
Proposition 1
Let u be a nonnegative bounded weak solution to Eq. (1) and let (2), \((\mathrm{g}_{0})\)– \((\mathrm{g}_{2})\), \((\mathrm{g}_{\mu })\) and \((\psi )\) be fulfilled in the cylinder \(Q_{R_{0}, R_{0}}(x_{0}, t_{0})\). Then
To prove Proposition 2.1, we need the following lemma.
Lemma 2
Let the conditions of Proposition 1 be fulfilled. Then, there exists \(\gamma >0\) depending only on the data, such that for all \(\sigma \), \(\delta \in (0,1)\) there holds
where \(J:=\sup \limits _{s<\tau <t} \int \nolimits _{B_{r}(y)} u(x, \tau )\,dx\).
Proof
Assume without loss that \(s=0\) and let \(\alpha \in (0, \frac{1}{q-1})\), \(\tilde{q}=q-1\) if \(q<2\) and \(\tilde{q}=q_{1}\) if \(q>2\), \(\beta \in (1, \min \{2, 1/\tilde{q}\})\). Fix \(\sigma \in (0, 1)\) and let \(\zeta \in C_{0}^{\infty }(B_{r}(y))\), \(0\leqslant \zeta \leqslant 1\), \(\zeta =1\) in \(B_{\sigma r}(y)\), \(|\nabla \zeta | \leqslant \frac{1}{(1-\sigma )r}\), and set \(\varepsilon _{1}:= r t^{-\alpha } \varepsilon ^{\beta -1}\).
Using inequality (16), we obtain
First we estimate \(I_{2}\), by \((\mathrm{g}_{2})\), \((\mathrm{g}_{\mu })\), \((\psi )\) (note that by (17) condition \((\psi )\) is applicable), we obtain
To estimate \(I_{1}\) we test identity (7) by \(\eta = (u_{h}+\varepsilon )^{1-\beta } \tau ^{\alpha } \zeta ^{q}\), integrating over (0, t) and letting \(h\rightarrow 0\), we obtain that
from this by inequality (16) we arrive at
Let us estimate the terms on the right-hand side of (22). By our choice of \(\beta \) and Hölder’s inequality, we have
The function \(\psi (\cdot , v)\) is non-increasing for \(v\geqslant b_{0} R^{-a_{0}}_{0}\), so by (17) we estimate the second term on the right-hand side of (22) as follows:
Collecting estimates (21)–(24), we arrive at the required inequality (19). This completes the proof of Lemma 2. \(\square \)
2.3 Proof of Proposition 1
Assume that \(s=0\), and for \(j=0, 1, 2,\ldots \) set
and let \(\zeta _{j}\in C_{0}^{\infty }(\bar{B_{j}})\), \(0\leqslant \xi _{j}\leqslant 1\), \(\zeta _{j}=1\) in \(B_{j}\), \(|\nabla \zeta _{j}| \leqslant 2^{j+2}/r\). Test (1.7) by \(\eta =\zeta _{j}\), integrating over \((t_{1}, t_{2})\subset (0, t)\) for any two time levels \(t_{1}\) and \(t_{2}\), and letting \(h\rightarrow 0\), we obtain that
Fix \(t_{1}\) by the condition
and set \(J_{j}:= \sup \limits _{0< \tau <t} \int _{B_{2r}(y)} u\,dx\). By Lemma 2, choosing \(\delta \) from the condition \(\delta =\delta _{0}\gamma ^{-1} 2^{-j}\), \(\delta _{0}\in (0, 1)\), from the previous we obtain
Iterating this inequality and choosing \(\delta _{0}\) sufficiently small we arrive at (18), which completes the proof of Proposition 1.
2.4 \(L^{1}_{\mathrm{loc}}-L^{\infty }_{\mathrm{loc}}\) estimate of solution
Theorem 3 is a simple consequence of Proposition 1 and the following lemma.
Lemma 3
Let u be a nonnegative bounded weak solution to Eq. (1) and let (2), \((\mathrm{g}_{0})\)– \((\mathrm{g}_{2})\), \((\mathrm{g}_{\mu })\) and \((\psi )\) be fulfilled in the cylinder \(Q_{R_{0}, R_{0}}(x_{0}, t_{0}) \subset \Omega _{T}\). Assume also that (17) holds, then there exists \(\gamma >0\) depending only upon the data, such that
Proof
Assume that \(s=0\) and for fixed \(\sigma \in (0, 1)\) and \(j=0, 1, 2, \ldots \) set
and let \(M_{0}:= \sup \limits _{Q_{0}}u\), \(M_{\sigma }:= \sup \limits _{Q_{\infty }}u\).
Next, let \(\zeta =\zeta _{1}(x)\zeta _{2}(t)\), where \(\zeta _{1}\in C^{\infty }_{0}(B_{j})\), \(\zeta _{2}\in C^{\infty }(-t, t)\),
Define also the sequence \(k_{j}:= k- 2^{-j}k\), where \(k>0\) to be chosen later. Testing (7) by \(\eta =(u_{h}-k_{j+1})_{+} \zeta ^{q}\), integrating over \((t_{j}, t)\) and letting \(h\rightarrow 0\), by inequality (16), we obtain
If
then by \((\mathrm{g}_{\mu })\) and \((\psi )\) we obtain
Therefore, the previous inequalities yield
Let us estimate the second term on the left-hand side of (27). By (16), for any \(b>0\), we have
here we used the notation \(g^{-}_{Q_{0}}(v):= \inf \limits _{(x, t)\in Q_{0}} g(x, t, v)\), \(v>0\). By (27) from the previous, we obtain
Choosing b from the condition \(\dfrac{M_{0}t^{-1}}{k g^{-}_{Q_{0}}(b)} = \dfrac{b}{k^{2}}\), i.e., \(g^{-}_{Q_{0}}(b)b=\dfrac{M_{0}k}{t}\leqslant \dfrac{M_{0}^{2}}{t}\), which by Lemma 2.1 implies
Using Hölder’s inequality, Sobolev embedding theorem, (27) and (28), we arrive at
It follows that \(Y_{j}\rightarrow 0\) as \(j\rightarrow \infty \) provided k is chosen to satisfy
By this choice, we have
Since the function \(\varphi _{Q_{0}}(v)\) is increasing
inequality (29) can be rewritten as
Set
then from the last inequality we have
Using inequality (16), we obtain for any \(\delta \in (0, 1)\)
By standard interpolation arguments, taking into account (26), from the previous we arrive at (25), which completes the proof of Lemma 3. \(\square \)
3 Integral and pointwise estimates of auxiliary solutions
Fix \((x_{0}, t_{0})\in \Omega _{T}\) such that \(Q_{R_{0},R_{0}}(x_{0}, t_{0})\subset \Omega _{T}\) and assume that conditions \((\mathrm{g}_{0})\), \((\mathrm{g}_{2})\), \((\mathrm{g}_{\lambda })\) and \((\psi )\) be fulfilled in \(Q_{R_{0},R_{0}}(x_{0}, t_{0})\). Let \(r<\rho \leqslant R^{a}_{0}\) with some \(a\geqslant 1+a_{0}\), where \(a_{0}\) is the constant from condition \((\psi \)). Fix \((\bar{x},\bar{t})\) such that \(Q_{r,r}(\bar{x},\bar{t}) \subset Q_{R_{0},R_{0}}(x_{0},t_{0})\), \(|\bar{t} -t_{0}| \leqslant \bar{c}\rho \) with some \(\bar{c} >0\), and \(E\subset B_{r}(\bar{x})\), \(|E|>0\), \(0< N\leqslant \lambda (r)\).
We consider the function
as the solution of the following problem
where \(\tau _{1}(\xi ):= \frac{\rho ^{2}}{\psi (\bar{x}, t_{0},\xi N|E|\rho ^{-n-1})},\) and
here \(\xi \in (0,1)\) depends only on the data will be chosen later.
In addition, the integral identity
holds for all \(t\in (\bar{t}, \bar{t}+8\tau _{1}(\xi )-h)\) and for all \(\eta \in W^{1, q}_{0}(B_{8\rho }(\bar{x}))\). Here, \(v_{h}\) is defined similarly to (7). The existence of the solution \(\upsilon \) follows from the general theory of monotone operators. Testing (33) by \(\eta =(v_{h})_{-}\) and \(\eta =(v_{h}-N)_{+}\), integrating over \((\bar{t}, t)\), \(t\in (\bar{t}, \bar{t}+8\tau _{1}(\xi ))\) and letting \(h\rightarrow 0\) we obtain that \(0\leqslant v \leqslant N \leqslant \lambda (r)\).
Lemma 4
Let v be a solution of (30)–(32), then for any \(\xi \in (0,1)\) either
or
and
with constant \(\gamma \) depending only on the known data and \(\bar{c}\).
Proof
For fixed \(\sigma \in (0, 1)\), \(\rho \leqslant s \leqslant s(1+\sigma )\leqslant 2\rho \), and \(j=0, 1, 2, \ldots \) set
and let \(M_{0}:= \sup \limits _{\mathcal {D}_{0}}\,v\), \(M_{\sigma }:= \sup \limits _{\mathcal {D}_{\infty }}\,v\), and consider the function \(\zeta (t) \in C^{\infty }(\mathbb {R}^{1})\), \(0\leqslant \zeta (t)\leqslant 1\),
Testing (33) by \(\eta =(v_{h}-k_{j})_{+} \zeta ^{q}(t)\), integrating over \((\bar{t}, t)\), \(t\in (\bar{t}, \bar{t}+8\tau _{1}(\xi ))\) and letting \(h\rightarrow 0\) and repeating the same arguments as in Sect. 2.4, similarly to (29), using condition \((\mathrm{g}_{\lambda })\) and the fact that \(v \leqslant N \leqslant \lambda (r)\) in \(Q_{1}\), we arrive at
provided that
and
Since
by condition \((\mathrm{g}_{\lambda })\) the last inequality holds if
To estimate the integral on the right-hand side of (37), we test (33) by \(\eta =\min (v_{h}, M_{0})\), integrating over \((\bar{t}, t)\), \(t\in (\bar{t}, \bar{t}+8\tau _{1}(\xi )-h)\) and letting \(h\rightarrow 0\), we obtain for \(v_{M_{0}}:= \min (v, M_{0})\)
From this and (37) since \(v(x, t)=v_{M_{0}}\) for \((x, t) \in \mathcal {D}_{0}\), we obtain
Repeating the iterative arguments similar to that of Sect. 2.4, we arrive at
taking into account (38), (39), we obtain inequality (35).
To prove (36) we set
and let \(M_{0}:= \sup \limits _{\mathcal {D}_{0}}\,v\), \(M_{\sigma }:= \sup \limits _{\mathcal {D}_{\infty }}\,v\), and consider the function \(\zeta (x) \in C^{\infty }(\mathbb {R}^{n})\), \(0\leqslant \zeta (x)\leqslant 1\),
Testing (33) by \(\eta =(v_{h}-k_{j})_{+} \zeta ^{q}(x)\), integrating over \((\bar{t}, t)\), \(t\in (\bar{t}, \bar{t}+8\tau _{1}(\xi ))\) and letting \(h\rightarrow 0\) and repeating the same arguments as in Sect. 2.4, similarly to (29), using condition \((\mathrm{g}_{\lambda })\) and the fact that \(v \leqslant N \leqslant \lambda (r)\) in \(Q_{1}\), we arrive at
which by (16) and condition \((g_{\lambda })\) yields
From this by the Sobolev embedding theorem, we arrive at
here we used the notation \(g^{-}_{\mathcal {D}_{0}}(v)=\min \limits _{\mathcal {D}_{0}}g(x,\tau ,v)\). By inequality (16) , Poincare inequality and (40) from the previous we obtain for any \(\varepsilon \in (0,1)\)
If inequality (36) is violated then the last inequality implies
Repeating the iterative arguments similar to that of Sect. 2, we arrive at the required (36),which completes the proof of the lemma. \(\square \)
The following proposition is the main result of this Section.
Proposition 2
There exist numbers \(\varepsilon _{1}\), \(\alpha _{1}\in (0, 1)\) depending only on the data, \(\bar{c}\) and \(\xi \) such that either (34) holds, or
for all time levels \(t\in (\bar{t}+ \dfrac{1}{8}\tau _{1}(\xi ),\, \bar{t}+2\tau _{1}(\xi ))\).
Proof
Let \(\zeta _{1}\in C^{\infty }_{0}(B_{3\rho }(\bar{x}))\), \(0\leqslant \zeta _{1}\leqslant 1\), \(\zeta _{1}=1\) in \(B_{2\rho }(\bar{x})\), \(|\nabla \zeta _{1}|\leqslant 1/\rho \). Testing (33) by \(\eta =v_{h}-N\zeta ^{q}_{1}\), integrating over \((\bar{t}+ \dfrac{1}{8}\,\tau _{1}(\xi ), t)\), \(t\in (\bar{t}+\dfrac{1}{8}\,\tau _{1}(\xi ),\bar{t}+2\tau _{1}(\xi ))\) and letting \(h\rightarrow 0\) we obtain
Let us estimate the terms on the right-hand side of (42). Further we will assume that (34) is violated, i.e.,
Set \(\tilde{Q}_{1}=B_{4\rho }(\bar{x})\times (\bar{t}+\dfrac{1}{8}\tau _{1}(\xi ),\bar{t}+2\tau _{1}(\xi ))\), by Proposition 1 we obtain
which by Lemma 4 yields for all \(t\in (\bar{t}+\dfrac{1}{8}\tau _{1}(\xi ),\,\bar{t}+2\tau _{1}(\xi ))\)
Let \(\zeta _{2}\in C^{\infty }(\mathbb {R}^{n})\), \(0\leqslant \zeta _{2}\leqslant 1\), \(\zeta _{2}=1\) in \(B_{3\rho }(\bar{x})\setminus B_{2\rho }(\bar{x})\), \(\zeta _{2}=0\) in \(B_{3\rho /2}(\bar{x})\cup (\mathbb {R}^{n}\setminus B_{4\rho }(\bar{x}))\) and \(|\nabla \zeta _{2}| \leqslant \gamma \rho ^{-1}\). Using inequality (16) with \(\varepsilon = N|E| \rho ^{-n-1}\),we obtain
By (43) conditions \((g_{\lambda })\) and \((\psi )\) are applicable, so by our choice of N we have
To estimate \(I_{3}\) we test (33) by \(\eta =v_{h}\zeta ^{q}_{2}\), integrating over \((\bar{t}, \bar{t}+\tau _{1}(\xi ))\) and letting \(h\rightarrow 0\), we arrive at
From this, by condition \((\mathrm{g}_{\lambda })\) and Lemma 4 we obtain
Collecting estimates (42)–(46), we obtain for all \(t\in (\bar{t}+\dfrac{1}{8}\tau _{1}(\xi ),\, \bar{t}+2\tau _{1}(\xi ))\)
choosing \(\varepsilon _{1}\) and \(\xi \) so small that \( \left( \varepsilon _{1}+\xi +\xi ^{1-q_{1}}\right) \,\gamma = \frac{1}{4}\), we arrive at
for all \(t\in (\bar{t}+\dfrac{1}{8}\tau _{1}(\xi ),\, \bar{t}+2\tau _{1}(\xi ))\),which completes the proof of Proposition 2. \(\square \)
4 Harnack’s inequality, proof of Theorems 1, 2
4.1 Expansion of positivity
The following lemma can be found in [34]. In the case of singular p-Laplace evolution equation, this result was proved by Chen and Di Benedetto [11] (see also [15, 17]); in the case when g is independent of (x, t) this lemma was proved in [23, 24].
Lemma 5
Let u be a nonnegative, bounded weak solution to Eq. (1) and let (2), \((\mathrm{g}_{0})\)–\((\mathrm{g}_{2})\), \((\mathrm{g}_{\lambda })\) be fulfilled in the cylinder \(Q_{R_{0}, R_{0}}(x_{0}, t_{0})\subset \Omega _{T}\). Assume also that for some \(0<r\leqslant R_{0}^{a}\), \(a\geqslant 1+a_{0}\), \(0<N\leqslant \lambda (r)\) and some \(\alpha , \delta _{0} \in (0, 1)\), \(\bar{c} >0\) there holds
for all \(t\in (s-\theta , s)\), \(\theta =r^{2}/\psi (y, s, \delta _{0} N/r)\).
Then, there exists \(\sigma _{0}\in (0, 1)\) depending only upon the data and \(\alpha , \delta _{0}\) such that either
or
Here, b, \(a_{0}\) and \(b_{0}\) are the numbers from conditions \((\mathrm{g}_{\lambda })\) and \((\psi )\).
4.2 Proof of Theorems 1, 2
Having fixed \((x_{0}, t_{0})\in \Omega _{T}\) such that \(Q_{R_{0}, R_{0}}(x_{0}, t_{0})\subset \Omega _{T}\), we assume that conditions \((\mathrm{g}_{2})\), \((\mathrm{g}_{\lambda })\), \((\mathrm{g}_{\mu })\) and \((\psi )\) hold in the cylinder \(Q_{R_{0}, R_{0}}(x_{0}, t_{0})\). We will prove Theorems 1 and 2 simultaneously, making the necessary remarks during the proof. Assume without loss that \((x_{0}, t_{0})=(0, 0)\) and set \(u_{0}:=u(0, 0)\). Further we will assume that inequalities (8) and (11) are violated, i.e.,
taking into account that in the logarithmic case (i.e., under the conditions of Theorem 1) \(\lambda (\rho )=\mu (\rho )=1\).
For \(\tau \in (0, 1)\) set \( M_{\tau }:=\sup \limits _{B_{\tau \rho }(0)}u(\cdot , 0) \) and
where \(m>1\) positive number to be fixed later. As \(\tau \rightarrow 1\), \(N_{\tau }\rightarrow +\infty \), whereas \(M_{\tau }<+\infty \). So, the equation \(M_{\tau }=N_{\tau }\) has roots, let \(\tau _{0}\) be the largest root of the above equation \(M_{\tau _{0}}=N_{\tau _{0}}\) and \(M_{\tau }\leqslant N_{\tau }\) for \(\tau >\tau _{0}\).
The conditions of Theorems 1 and 2 imply the continuity of the solution u, therefore \(M_{\tau _{0}}\) is achieved at some \(\bar{x}\in B_{\tau _{0}\rho }(0)\). Choose \(\tau _{1}\) from the condition \((1-\tau _{1})^{-m}=4(1-\tau _{0})^{-m}\), i.e., \(\tau _{1}=1-4^{-\frac{1}{m}}(1-\tau _{0})\), and set \( 2r=(\tau _{1}-\tau _{0})\rho =(1-4^{-\frac{1}{m}})(1-\tau _{0})\rho . \) For these choices \(B_{2r}(\bar{x})\subset B_{\tau _{1}\rho }(0)\), \(M_{\tau _{1}}\leqslant N_{\tau _{1}}\), and
provided that \(m\geqslant c_{3}(1+\beta )\). Here, \(c_{3}\) is the constant from the technical condition of Sect. 1.
Construct the cylinders \(Q^{-}_{2r, \theta }(\bar{x}, 0)\) and \(Q^{-}_{2r,\theta _{0}}(\bar{x},0)\), where
We have an inclusion
where \(\bar{c}= c_{0}\max (M^{2-p},M^{2-q}).\) Indeed, by (50) we have
provided that \(C \geqslant 2^{a_{0}/a}\). So, using condition \((\mathrm{g}_{2})\), we obtain
Lemma 6
The following inequality holds:
where \(m_{1}=n/\varkappa (p)\), \(\varkappa (p)=p+n(p-2)>0\).
Proof
We use Theorem 3 over the pair of cylinder \(Q^{-}_{r,\theta _{0}/2}(\bar{x}, 0)\) and \(Q^{-}_{2r,\theta _{0}}(\bar{x}, 0)\) two times, first for the choices \(s=0\), \(t=\theta _{0}/2\) and second for the choices \(s=-\theta _{0}/2\), \(t=0\). By condition \((\mathrm{g}_{\mu })\)
which by \((\mathrm{g}_{2})\) and (51) implies
provided that \(C> c_{2}^{\frac{1}{2-p}}\). Therefore, Theorem 3 yields
To estimate the first term on the right-hand side of the previous inequality, note
here we used condition \((\mathrm{g}_{\mu })\) and Lemma 1.
By Lemma 1 and inequality (53), we have
From which the required inequality (52) follows, which completes the proof of the lemma. \(\square \)
Lemma 7
There exist numbers \(\delta \), \(\varepsilon \), \(\alpha \in (0, 1)\) depending only on the data such that
for all
where \(m_{1}=n/\varkappa (p)\), \(m_{2}=(2-p)n/\varkappa (p)\).
Proof
We use Theorem 3 over the pair of cylinder \(Q^{-}_{r/2,{\bar{\theta }}_{0}/2}(\bar{x}, 0)\) and \(Q^{-}_{r,{\bar{\theta }}_{0}}(\bar{x}, 0)\) for the choices \(s=0\) and \(t={\bar{\theta }}_{0}\). By (51) condition (15) holds, therefore for all \(|t-t_{0}|\leqslant {\bar{\theta }}_{0}\)
Let us estimate the terms on the right-hand side of the previous inequality. Similarly to (53), (54), we obtain
and by this and Lemma 1
which implies
\(I_{1}\leqslant \gamma \delta ^{\frac{\varkappa (p)}{p(n+1)}}N_{\tau _{0}}\).
Choose \(\delta \) from the condition \(\gamma (\delta +\delta ^{\frac{\varkappa (p)}{p(n+1)}})\leqslant 1/4\), from the previous and Lemma 6 we obtain for all \(|t|\leqslant {\bar{\theta }}_{0}\)
First term on the right-hand side of (56), we estimate similarly to \(I_{1}\)
Choosing \(\varepsilon \) from the condition \(\varepsilon ^{\frac{\varkappa (p)}{p(n+1)}}\gamma (\delta )=1/4\), from (56) we have
which similarly to \(I_{1}\) we estimate as follows
Collecting the last two inequalities, we arrive at
for all \(|t|\leqslant {\bar{\theta }}_{0}\), which proves Lemma 7. \(\square \)
4.3 Proof of Theorem 1
First note that under the conditions of Theorem 1, inequality (55) can be rewritten as
for all \( |t|\leqslant {\bar{\theta }}_{0}:=\dfrac{r^{2}}{\psi _{Q^{-}_{2r,\theta }(\bar{x},0)}( \delta N_{\tau _{0}}/r)}, \) and with some fixed \(\varepsilon \), \(\delta \), \(\alpha \in (0, 1)\) depending only on the data, \(N_{\tau _{0}}=(1-\tau _{0})^{-m}u_{0}\).
Apply Lemma 5 with \(\rho =r\) and \(N=\varepsilon N_{\tau _{0}},\) we obtain that
for all \(x\in B_{2r}(\bar{x})\) and for all times \(-{\bar{\theta }}_{0}+\dfrac{1}{8}\,{\bar{\theta }}_{0, \varepsilon }^{(0)}\leqslant t\leqslant \bar{\theta _{0}}\), \({\bar{\theta }}_{0, \varepsilon }^{(0)}:=r^{2}/\psi (\bar{x}, 0,\)
\(\delta \varepsilon N_{\tau _{0}}r^{-1})\).
After j iterations
for all \(x\in B_{2^{j+1}r}(\bar{x})\) and for all times
By our choices and condition \((\psi )\)
provided \(C\geqslant \left( \dfrac{2}{\sigma _{0}}\right) ^{j} \dfrac{(1-\tau _{0})^{m}}{\delta \varepsilon }\, \dfrac{r}{\rho }\). By our assumptions
Therefore, inequality (58) holds for all times
Choose j from the condition \(\rho \leqslant 2^{j}r\leqslant 2\rho \) and m from the condition \(\sigma _{0}2^{m}=1\), and then choose \(\varepsilon \) smaller if necessary, we arrive at the required (9) for all times \(|t|\leqslant \theta '/2\), provided
By conditions \((\psi )\) and \((\mathrm{g}_{\mu })\) with \(\mu (r)=1\)
From the definition of r and choosing \(m\geqslant \frac{1+q_{1}}{1-q_{1}},\) which we may assume taking \(\sigma _{0}\) smaller if necessary, we obtain
which completes the proof of Theorem 1.
4.4 Proof of Theorem 2
We note that under the conditions of Theorem 2, inequality (55) implies
for all \(|t|\leqslant {\bar{\theta }}^{'}_{0}:=\dfrac{r^{2}}{\psi \left( \bar{x},0,\delta \dfrac{N_{1}}{r}\right) }\), where
As was already mentioned, direct application of inequality (59) leads us to condition (3). We will use auxiliary solutions, defined in Sect. 3.
Set \(E(N,t):=\{B_{r}(\bar{x}) : u(\cdot ,t)\geqslant N \}\) and for \(t\in [-{\bar{\theta }}^{'}_{0},\,-{\bar{\theta }}^{'}_{1}], \quad {\bar{\theta }}^{'}_{1} =\dfrac{r^{2}}{\psi \left( \bar{x},0,\delta \alpha (r) \dfrac{N_{1}}{r}\right) }\)
consider the equation
By our choices \(f(-{\bar{\theta }}^{'}_{0}) \leqslant 0\) and \(f(-{\bar{\theta }}^{'}_{1}) \geqslant 0\). By the continuity of u and by the continuity of the Lebesgue measure, f(t) is continuous, so the previous equation has roots. Let \(\bar{t}\) be the largest root of the above equation, i.e.,
Taking \(\delta \) smaller if necessary, we construct the auxiliary solution v with \(\rho =r\) and \(\xi =\delta \) in the cylinder \(Q_{1}\negthickspace =\negthickspace B_{8r}(\bar{x})\times (\bar{t}, \bar{t} + 8{\bar{\tau }}_{1}), {\bar{\tau }}_{1}=\dfrac{r^{2}}{\psi \left( \bar{x},0,\delta N_{1} \dfrac{|E( N_{1},\bar{t})|}{r^{n+1}}\right) }.\)
If
then by Proposition 2 with some \(\varepsilon _{1},\alpha _{1} \in (0,1)\) depending only on the data
for all \(t\in (\bar{t}+\dfrac{1}{8}{\bar{\tau }}_{1},\bar{t}+2{\bar{\tau }}_{1}),\,\,\) which by Lemma 5 implies
for all \((x,t) \in B_{2r}(\bar{x})\times (\bar{t} +\dfrac{1}{4} {\bar{\tau }}_{1},\,\bar{t} + 2{\bar{\tau }}_{1}),\) and the constant \(\sigma _{0} \in (0,1)\) depends only on the data. By our choice of \(\bar{t}\) and by (55)
moreover, by (55)
We note also that since \(u \geqslant v\) on the parabolic boundary of the cylinder \(B_{8r}(\bar{x})\times (\bar{t}, \bar{t} +8{\bar{\tau }}_{1})\), inequality (60) and the monotonicity condition (10) yield
for all \(t\in (-\dfrac{3}{4}\theta ^{(1)}_{0}, \quad \theta ^{(1)}_{0}), \quad \theta ^{(1)}_{0}=\dfrac{r^{2}}{\psi (\bar{x},0,\delta \alpha (r)\frac{N_{1}}{r})}.\)
Now we can use Lemma 5, which implies that if
then
for all \(t\in (-\dfrac{3}{4}\theta ^{(1)}_{0} + \dfrac{1}{8}\theta ^{(2)}_{0}, \quad \theta ^{(1)}_{0}),\quad \theta ^{(2)}_{0} =\dfrac{(2r)^{2}}{\psi (\bar{x},0,\sigma _{0} \delta \varepsilon _{1} \alpha (r)\frac{N_{1}}{2r})}.\)
After j iterations
and for all times
provided that
By our choices and condition \((\psi )\)
provided that inequality (64) holds.
Therefore, inequality (63) holds for all times
Choose j from the condition \(\rho \leqslant 2^{j}r\leqslant 2\rho \), and m from the condition \(\sigma _{0}2^{m}=1\), then we obtain
for all \(x\in B_{2\rho }(\bar{x})\) and for all times \( |t|\leqslant \frac{1}{2}\theta ^{(1)}_{0}\) provided that (64) holds.
Let us estimate the term on the right-hand side of the previous inequality. By our choices, we have
Choosing
we arrive at the required (12) for all times \(|t|\leqslant \frac{1}{2} \theta ^{(1)}_{0}\), provided that (64) holds.
Inequality (64) holds, if
and moreover, by conditions \((\psi )\) and \((\mathrm{g}_{\lambda })\)
From the definition of r and choosing \(m\geqslant \frac{1+q_{1}}{1-q_{1}}\), we obtain
which completes the proof of Theorem 2.
Data Availability
All data generated or analyzed during this study are included in this published article.
References
Yu. A. Alkhutov, O. V. Krasheninnikova, On the continuity of solutions of elliptic equations with a variable order of nonlinearity, (Russian) Tr. Mat. Inst. Steklova 261 (2008), Differ. Uravn. i Din. Sist., 7–15; translation in Proc. Steklov Inst. Math. 261 (2008), no. 1–10.
Yu. A. Alkhutov, M. D. Surnachev, Behavior at a boundary point of solutions of the Dirichlet problem for the \(p(x)\)-Laplacian, (Russian) Algebra i Analiz 31 (2019), no. 2, 88–117; translation in St. Petersburg Math. J. 31 (2020), no. 2, 251–271.
Yu. A. Alkhutov, V. V. Zhikov, Hölder continuity of solutions of parabolic equations with variable nonlinearity exponent, Translation of Tr. Semin. im. I. G. Petrovskogo No. 28 (2011), Part I, 8–74; J. Math. Sci. (N.Y.) 179 (2011), no. 3, 347–389.
S. Antontsev, V. Zhikov, Higher integrability for parabolic equations of \(p(x,t)\)-Laplacian type, Adv. Differential Equations 10 (2005), no. 9, 1053–1080.
P. Baroni, V. Bögelein, Calderón-Zygmund estimates for parabolic \(p(x, t)\)-Laplacian systems, Rev. Mat. Iberoam. 30 (2014), no. 4, 1355–1386.
P. Baroni, M. Colombo, G. Mingione, Harnack inequalities for double phase functionals, Nonlinear Anal. 121 (2015), 206–222.
P. Baroni, M. Colombo, G. Mingione, Non-autonomous functionals, borderline cases and related function classes, St. Petersburg Math. J. 27 (2016), 347–379.
P. Baroni, M. Colombo, G. Mingione, Regularity for general functionals with double phase, Calc. Var. Partial Differential Equations 57 (2018), Paper No. 62, 48 pp.
V. Bögelein, F. Duzaar, Hölder estimates for parabolic \(p(x, t)\)-Laplacian systems, Math. Ann. 354 (2012), no. 3, 907–938.
K. O. Buryachenko, I. I. Skrypnik, Local Continuity and Harnack Inequality for Double-Phase Parabolic Equations, Potential Anal. (2020), https://doi.org/10.1007/s11118-020-09879-9
Y. Z. Chen, E. Di Benedetto, Hölder estimates of solutions of singular parabolic equations with measurable coefficients, Arch. Rational Mech. Anal. 118 (1992), 257–271.
M. Colombo, G. Mingione, Bounded minimisers of double phase variational integrals, Arch. Rational Mech. Anal. 218 (2015), no. 1, 219–273.
M. Colombo, G. Mingione, Regularity for double phase variational problems, Arch. Rational Mech. Anal. 215 (2015), no. 2, 443–496.
M. Colombo, G. Mingione, Calderon-Zygmund estimates and non-uniformly elliptic operators, J. Funct. Anal. 270 (2016), 1416–1478.
E. Di Benedetto, Degenerate Parabolic Equations, Springer-Verlag, New York, 1993.
E. Di Benedetto, U. Gianazza, V. Vespri, Harnack estimates for quasi-linear degenerate parabolic differential equations, Acta Math. 200 (2008), 181–209.
E. Di Benedetto, U. Gianazza, V. Vespri, Forward, backward and elliptic Harnack inequalities for non-negative solutions to certain singular parabolic differential equations, Ann. Scuola Norm. Sup. Pisa Cl. Sci. 5 (2010), no. 9, 385–422.
L. Diening, P. Harjulehto, P. Hästö, M. Růžička, Lebesgue and Sobolev Spaces with Variable Exponents, in: Lecture Notes in Mathematics, 2017, Springer, Heidelberg, 2011, x+509 pp.
Ding Mengyao, Zhang Chao, Zhou Shulin, Global boundedness and Hölder regularity of solutions to general \(p(x,t)\)-Laplace parabolic equations, Math. Methods Appl. Sci. 43 (2020), no. 9, 5809–5831.
O. V. Hadzhy, I. I. Skrypnik, M. V. Voitovych, Interior continuity, continuity up to the boundary and Harnack’s inequality for double-phase elliptic equations with non-logarithmic growth, arXiv:201210960v1 [math.AP].
P. Harjulehto, P. Hästö, Orlicz Spaces and Generalized Orlicz Spaces, in: Lecture Notes in Mathematics, vol. 2236, Springer, Cham, 2019, p. X+169 https://doi.org/10.1007/978-3-030-15100-3
P. Harjulehto, P. Hästö, Út V. Lê, M. Nuortio, Overview of differential equations with non-standard growth, Nonlinear Anal. 72 (2010), no. 12, 4551–4574.
S. Hwang, G. M. Lieberman, Hölder continuity of bounded weak solutions to generalized parabolic \(p\)-Laplacian equations I: degenerate case, Electron. J. Differential Equations, 2015 (2015), no. 287, 1–32.
S. Hwang, G. M. Lieberman, Hölder continuity of bounded weak solutions to generalized parabolic \(p\)-Laplacian equations II: singular case, Electron. J. Differential Equations, 2015 (2015), no. 288, 1–24.
I. M. Kolodij, On boundedness of generalized solutions of elliptic differential equations, Vestnik Moskov. Gos. Univ. 1970 (1970), no. 5, 44–52.
I. M. Kolodij, On boundedness of generalized solutions of parabolic differential equations, Vestnik Moskov. Gos. Univ. 1971 (1971), no. 5, 25–31.
E. M. Landis, Some questions in the qualitative theory of second-order elliptic equations (case of several independent variables), Uspehi Mat. Nauk, 109 (1963), 18, no. 1, 3–62 (in Russian).
E. M. Landis, Second Order Equations of Elliptic and Parabolic Type, in: Translations of Mathematical Monographs, vol. 171, American Math. Soc., Providence, RI, 1998.
G. M. Lieberman, The natural generalization of the natural conditions of Ladyzhenskaya and Ural’tseva for elliptic equations, Comm. Partial Differential Equations 16 (1991), no. 2-3, 311–361.
P. Marcellini, Regularity of minimizers of integrals of the calculus of variations with non standard growth conditions, Arch. Rational Mech. Anal. 105 (1989), no. 3, 267–284.
P. Marcellini, Regularity and existence of solutions of elliptic equations with \(p,q\)-growth conditions, J. Differential Equations 90 (1991), no. 1, 1–30.
M. A. Shan, I. I. Skrypnik, M. V. Voitovych, Harnack’s inequality for quasilinear elliptic equations with generalized Orlicz growth, Electron. J. Differential Equations 2021 (2021), no. 27, 1–16.
I. I. Skrypnik, M. V. Voitovych, \(\cal{B}_{1}\) classes of De Giorgi-Ladyzhenskaya-Ural’tseva and their applications to elliptic and parabolic equations with generalized Orlicz growth conditions, Nonlinear Anal. 202 (2021) 1121-1135.
I. I. Skrypnik, M. V. Voitovych, On the continuity of solutions of quasilinear parabolic equations with generalized Orlicz growth under non-logarithmic conditions, Ann.Mat.Pura Appl., https://doi.org/10.1007/510231-021-01161-y.
M. D. Surnachev, On Harnack’s inequality for \(p(x)\)-Laplacian (Russian), Keldysh Institute Preprints 10.20948/prepr-2018-69, 69 (2018), 1–32.
M. D. Surnachev, On the weak Harnack inequality for the parabolic \(p(x)\)- Laplacian, Asymptotic Analysis, https://doi.org/10.3233/ASY-211746.
M. V. Voitovych, Pointwise estimates of solutions to 2m-order quasilinear elliptic equations with m-(p,q) growth via Wolff potentials, Nonlinear Anal. 181 (2019), 147–179.
Y. Wang, Intrinsic Harnack inequalities for parabolic equations with variable exponents, Nonlinear Anal. 83 (2013) 12–30.
P. Winkert, R. Zacher, Global a priori bounds for weak solutions to quasilinear parabolic equations with nonstandard growth, Nonlinear Anal. 145 (2016), 1–23.
M. Xu, Y. Chen, Hölder continuity of weak solutions for parabolic equations with nonstandard growth conditions, Acta Math. Sin. (Engl. Ser.) 22(3) (2006) 793–806.
F. Yao, Hölder regularity of the gradient for the non-homogeneous parabolic \(p(x, t)\)-Laplacian equations, Math. Methods Appl. Sci., 37 (2014), no. 12, 1863–1872.
F. Yao, Hölder regularity for the general parabolic \(p(x, t)\)-Laplacian equations, NoDEA Nonlinear Differential Equations Appl. 22 (2015), no. 1, 105–119.
C. Zhang, S. Zhou, X. Xue, Global gradient estimates for the parabolic \(p(x, t)\)-Laplacian equation, Nonlinear Anal. 105 (2014) 86–101.
V. V. Zhikov, Questions of convergence, duality and averaging for functionals of the calculus of variations, (Russian) Izv. Akad. Nauk SSSR Ser. Mat. 47 (1983), no. 5, 961–998.
V. V. Zhikov, Averaging of functionals of the calculus of variations and elasticity theory, (Russian) Izv. Akad. Nauk SSSR Ser. Mat. 50 (1986), no. 4, 675–710, 877.
V. V. Zhikov, On Lavrentiev’s phenomenon, Russian J. Math. Phys. 3 (1995), no. 2, 249–269.
V. V. Zhikov, On some variational problems, Russian J. Math. Phys. 5 (1997), no. 1, 105–116 (1998).
V. V. Zhikov, On the density of smooth functions in Sobolev-Orlicz spaces, (Russian) Zap. Nauchn. Sem. S.-Peterburg. Otdel. Mat. Inst. Steklov. (POMI) 310 (2004), Kraev. Zadachi Mat. Fiz. i Smezh. Vopr. Teor. Funkts. 35 [34], 67–81, 226; translation in J. Math. Sci. (N.Y.) 132 (2006), no. 3, 285–294.
V. V. Zhikov, S. M. Kozlov, O. A. Oleinik, Homogenization of differential operators and integral functionals, Springer-Verlag, Berlin, 1994.
V. V. Zhikov, S. E. Pastukhova, On the improved integrability of the gradient of solutions of elliptic equations with a variable nonlinearity exponent, (Russian) Mat. Sb. 199 (2008), no. 12, 19–52; translation in Sb. Math. 199 (2008), no. 11–12, 1751–1782.
V. V. Zhikov, S. E. Pastukhova, On the property of higher integrability for parabolic systems of variable order of nonlinearity, (Russian) Mat. Zametki 87 (2010), no. 2, 179–200; translation in Math. Notes 87 (2010), no. 1–2, 169–188.
Acknowledgements
This work was supported by Ministry of Education and Science of Ukraine (Project Number 0121U109525) and by the Volkswagen Foundation project ”From Modeling and Analysis to Approximation”.
Author information
Authors and Affiliations
Corresponding author
Additional information
In memory of DiBenedetto.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
About this article
Cite this article
Skrypnik, I.I. Harnack’s inequality for singular parabolic equations with generalized Orlicz growth under the non-logarithmic Zhikov’s condition. J. Evol. Equ. 22, 45 (2022). https://doi.org/10.1007/s00028-022-00794-7
Accepted:
Published:
DOI: https://doi.org/10.1007/s00028-022-00794-7