Large time behavior of solutions for the porous medium equation with a nonlinear gradient source

Nan Li; Pan Zheng; Chunlai Mu; Iftikhar Ahmed

doi:10.1186/1687-2770-2014-98

Research
Open Access

Large time behavior of solutions for the porous medium equation with a nonlinear gradient source

Nan Li¹,
Pan Zheng^{2, 3}Email author,
Chunlai Mu² and
Iftikhar Ahmed²

Boundary Value Problems20142014:98

https://doi.org/10.1186/1687-2770-2014-98

Received: 23 December 2013
Accepted: 3 April 2014
Published: 6 May 2014

Abstract

This paper deals with the large time behavior of non-negative solutions for the porous medium equation with a nonlinear gradient source $u_{t} = Δ u^{m} + {| \nabla u^{l} |}^{q}$ , $(x, t) \in Ω \times (0, \infty)$ , where $l \geq m > 1$ and $1 \leq q < 2$ . When $l q = m$ , we prove that the global solution converges to the separate variable solution $t^{- \frac{1}{m - 1}} f (x)$ . While $m < l q \leq m + 1$ , we note that the global solution converges to the separate variable solution $t^{- \frac{1}{m - 1}} f_{0} (x)$ . Moreover, when $l q > m + 1$ , we show that the global solution also converges to the separate variable solution $t^{- \frac{1}{m - 1}} f_{0} (x)$ for the small initial data $u_{0} (x)$ , and we find that the solution $u (x, t)$ blows up in finite time for the large initial data $u_{0} (x)$ .

MSC:35K55, 35K65, 35B40.

Keywords

large time behavior
separate variable solution
porous medium equation
gradient source
blow-up

1 Introduction

In this paper, we investigate the large time behavior of non-negative solutions for the following initial-boundary value problem:

{\begin{cases} u_{t} = Δ u^{m} + {| \nabla u^{l} |}^{q}, & (x, t) \in Ω \times (0, \infty), \\ u (x, t) = 0, & (x, t) \in \partial Ω \times (0, \infty), \\ u (x, 0) = u_{0} (x), & x \in Ω, \end{cases}

(1.1)

where

l \geq m > 1

,

1 \leq q < 2

, Ω is a bounded domain of

R^{N}

(

N \geq 2

) with smooth boundary ∂ Ω, and the initial function is

u_{0} (x) \in C_{0} (\bar{Ω}) = {z \in C (\bar{Ω}) : z = 0 on \partial Ω}, u_{0} (x) \geq 0, u_{0} (x) ≢ 0 .

(1.2)

Equation (1.1) arises in the study of the growth of surfaces and it has been considered as a mathematical model for a variety of physical problems (see [1, 2]). For instance, in ballistic deposition processes, the evolution of the profile of a growing interface is described by the diffusive Hamilton-Jacobi type equation (1.1) with $m = l = 1$ (see [3]).

One of the particular feature of problem (1.1) is that the equation is a slow diffusion equation with nonlinear source term depending on the gradient of a power of the solution. In general, there is no classical solution. Therefore, it turns out that a suitable framework for the well-posedness of the initial-boundary value problem (1.1) is the theory of viscosity solutions (see [4–6]), so we first define the notion of solutions.

Definition 1.1 A non-negative function

u (x, t) \in C (\bar{Ω} \times (0, \infty))

is called a solution of (1.1), if

u (x, t)

is a viscosity solution to (1.1) in

Ω \times (0, \infty)

and satisfies

u (x, t) = 0, (x, t) \in \partial Ω \times (0, \infty) and u (x, 0) = u_{0} (x) \in C_{0} (\bar{Ω}), x \in \bar{Ω} .

(1.3)

Under some assumptions, the global (local) existence in time, uniqueness and regularity of solutions to reaction-diffusion equations with gradient terms have been extensively investigated by many authors (see [7–12] and the references therein). In particular, in [1], Andreucci proved the existence of solutions for the following degenerate parabolic equation with initial data measures:

{\begin{cases} u_{t} = Δ u^{m} + {| \nabla u^{l} |}^{q}, & (x, t) \in R^{N} \times (0, \infty), \\ u (x, 0) = μ, & x \in R^{N}, \end{cases}

(1.4)

where $m \geq 1$ , $q l \geq m$ , $0 < q < 2$ , and $N \geq 1$ .

The main purpose of this paper is to further study the large time behavior of non-negative solutions

u (x, t)

to (1.4) with homogeneous Dirichlet boundary conditions. In recent years, many authors have investigated the asymptotic behavior of solutions to the viscous Hamilton-Jacobi equations (see [3, 4, 7, 8, 10, 13–21] and the references therein). For example, for the special case

m = l = 1

, Gilding [22] studied the large time behavior of solutions to the following Cauchy problem:

{\begin{cases} u_{t} = Δ u + {| \nabla u |}^{q}, & (x, t) \in R^{N} \times (0, \infty), \\ u (x, 0) = u_{0} (x), & x \in R^{N}, \end{cases}

(1.5)

and he gave the temporal decay estimates.

In [23], Stinner investigated the asymptotic behavior of solutions for the following one space dimensional viscous Hamilton-Jacobi equation:

{\begin{cases} u_{t} = {({| u_{x} |}^{p - 2} u_{x})}_{x} + {| u_{x} |}^{q}, & (x, t) \in (- R, R) \times (0, \infty), \\ u (\pm R, t) = 0, & t \in (0, \infty), \\ u (x, 0) = u_{0} (x), & x \in (- R, R), \end{cases}

(1.6)

where $R > 0$ , $p > 2$ , and $1 < q < p - 1$ , and he proved that these solutions converge to the steady states by Lyapunov functional.

In higher dimensional case, Barles et al. [24] studied the large time behavior of solutions for the following initial-boundary value problem:

{\begin{cases} u_{t} = div ({| \nabla u |}^{p - 2} \nabla u) + {| \nabla u |}^{q}, & (x, t) \in B (0, 1) \times (0, \infty), \\ u (x, t) = 0, & (x, t) \in \partial B (0, 1) \times (0, \infty), \\ u (x, 0) = u_{0} (x), & x \in B (0, 1), \end{cases}

(1.7)

where $p \geq 2$ , $0 < q < p - 1$ , and they showed that the non-negative radially symmetric solutions converge to the stationary solution.

Recently, in [25], Laurencot et al. extended the case $0 < q < p - 1$ of the problem (1.7) to the case $q \geq p - 1$ , and derived that these solutions converge to two different separate variables solutions according to the cases $q = p - 1$ and $q > p - 1$ in the general bounded domain $Ω \subset R^{N}$ , respectively.

Motivated by the above work, by using the modified comparison argument, the self-similar transformation method, and the half-relaxed limits technique used in [6, 25, 26], we investigate the asymptotic behavior of non-negative solutions to (1.1). Our main results in this paper are stated as follows.

Theorem 1.1 Let

l \geq m > 1

,

1 \leq q < 2

, and

l q = m

. Assume that

u_{0} (x) \in W^{1, \infty} (Ω)

satisfies (1.2). Then there exists a unique solution

u (x, t)

to (1.1) in the sense of Definition 1.1 such that

lim_{t \to \infty} {∥ t^{\frac{1}{m - 1}} u (t) - f ∥}_{\infty} = 0,

(1.8)

where

f (x) \in C_{0} (\bar{Ω})

is the unique positive solution to

- Δ f^{m} - {| \nabla f^{l} |}^{q} - \frac{f}{m - 1} = 0 in Ω, f > 0 in Ω, f = 0 on \partial Ω .

(1.9)

Moreover, we have

| \nabla u (t) | \in L^{\infty} (\bar{Ω})

for all

t \geq 0

and

ℓ [u_{0}] : = sup_{t \geq 0} {{∥ \nabla u (t) ∥}_{\infty}} < \infty .

(1.10)

Theorem 1.2 Let

l \geq m > 1

,

1 \leq q < 2

, and

m < l q \leq m + 1

. Assume that

u_{0} (x) \in W^{1, \infty} (Ω)

satisfies (1.2). Then there exists a unique solution

u (x, t)

to (1.1) in the sense of Definition 1.1 such that

lim_{t \to \infty} {∥ t^{\frac{1}{m - 1}} u (t) - f_{0} ∥}_{\infty} = 0,

(1.11)

where

f_{0} (x) \in C_{0} (\bar{Ω})

is the unique positive solution to

- Δ f_{0}^{m} - \frac{f_{0}}{m - 1} = 0 in Ω, f_{0} > 0 in Ω, f_{0} = 0 on \partial Ω .

(1.12)

Theorem 1.3 Let

l \geq m > 1

,

1 \leq q < 2

, and

l q > m + 1

. Assume that

u_{0} (x) \in W^{1, \infty} (Ω)

satisfies (1.2) and suppose further that there exists

G_{0} \in W^{1, \infty} (Ω)

satisfying (1.2) such that

u_{0} (x) \leq \frac{G_{0} (x)}{ℓ [G_{0}]}, x \in \bar{Ω},

(1.13)

where

ℓ [G_{0}]

is defined in (1.10). Then there exists a unique solution

u (x, t)

to (1.1) in the sense of Definition 1.1 such that

lim_{t \to \infty} {∥ t^{\frac{1}{m - 1}} u (t) - f_{0} ∥}_{\infty} = 0,

(1.14)

where $f_{0} (x) \in C_{0} (\bar{Ω})$ is defined in (1.12).

Theorem 1.4 Let

l \geq m > 1

,

1 \leq q < 2

, and

l q > m + 1

. Assume that there exists a positive constant

K_{0}

depending only on m, l, q, s, and ε such that

\int_{Ω} u_{0}^{s} (x) ϕ_{ε} (x) d x \geq K_{0}

, where

s \in (0, \frac{1}{2})

,

ε \in (0, 1)

, and

ϕ_{ε} (x) = A_{ε} e^{- ε | x |}

with

A_{ε} = \frac{1}{\int_{Ω} e^{- ε | x |} d x}

. Then the solution

u (x, t)

of the problem (1.1) blows up in finite time in the sense of weak solution. Moreover, the upper bound of blow-up time is given as follows:

T \leq \frac{2}{q l - 1} {(\frac{q l}{q l + s - 1})}^{\frac{q (m - 1)}{q l - m}} {(m ε^{2})}^{\frac{1 - q l}{q l - m}} .

(1.15)

Remark 1.1 Compared to the results in [25], we extend the results of p-Laplacian equation to the porous medium equation (1.1) with a nonlinear gradient source.

Remark 1.2 In Theorem 1.4, we only give an upper estimate of the blow-up time. But the lower estimate of the blow-up time is an open problem.

This paper is organized as follows. In Section 2, we establish the comparison lemmas to prove the uniqueness of the positive solution to (1.9) and the identification of the half-relaxed limits. In Section 3, using the comparison principle, we construct the global solutions to obtain the upper bound and Hölder estimate of solutions to (1.1), and we prove Theorems 1.1 and 1.2 by the half-relaxed limits method. Moreover, we give the large time behavior of solutions to (1.1) with the small initial data $u_{0} (x)$ for $l q > m + 1$ , and we prove Theorem 1.3 in Section 4. Finally, we obtain the blow-up case, and we prove Theorem 1.4 in Section 5.

2 Comparison lemmas

In this section, we establish the following comparison lemma between positive supersolutions and non-negative subsolutions to the elliptic equation in (1.9): which is an important tool for the uniqueness of the positive solution to (1.9) and the identification of the half-relaxed limits later.

Lemma 2.1 Let

l \geq m > 1

,

1 \leq q < 2

, and

l q = m

. Assume that

w \in U S C (\bar{Ω})

and

W \in L S C (\bar{Ω})

are respectively a bounded upper semicontinuous (usc) viscosity subsolution and a bounded lower semicontinuous (lsc) viscosity supersolution to

- Δ ζ^{m} - {| \nabla ζ^{l} |}^{q} - \frac{ζ}{m - 1} = 0 in Ω,

(2.1)

such that

w (x) = W (x) = 0 for x \in \partial Ω,

(2.2)

and

W (x) > 0 for x \in Ω .

(2.3)

Then

w (x) \leq W (x) for x \in \bar{Ω} .

(2.4)

Proof The proof is based on the idea as in [25, 27], but with different auxiliary functions.

For $i \geq N_{0}$ large enough, it is easy to see that $Ω_{i} = {x \in Ω : d (x, \partial Ω) > \frac{1}{i}}$ is a non-empty open subset of $R^{N}$ .

Since

{\bar{Ω}}_{i}

is compact and W is lower semicontinuous, the function W has a minimum in

{\bar{Ω}}_{i}

. By the positivity (2.3) of W in

{\bar{Ω}}_{i}

, we have

μ_{i} = min_{x \in {\bar{Ω}}_{i}} {W (x)} > 0 .

(2.5)

Similarly, the compactness of

\bar{Ω} ∖ Ω_{i}

, the upper semicontinuity and boundedness of w imply that w has at least one point of maximum

x_{i}

in

\bar{Ω} ∖ Ω_{i}

and we set

η_{i} = w (x_{i}) = max_{x \in \bar{Ω} ∖ Ω_{i}} {w (x)} .

(2.6)

It follows from $\partial Ω \subset \bar{Ω} ∖ Ω_{i}$ and w vanishes on ∂ Ω by (2.2) that $η_{i} \geq 0$ .

Next, we claim that

lim_{i \to \infty} η_{i} = 0 .

(2.7)

Indeed, owing to the compactness of

\bar{Ω}

and the definition of

Ω_{i}

, there exist

y \in \partial Ω

and a subsequence of

{x_{i}}_{i \geq N_{0}}

(not relabeled) such that

x_{i} \to y

as

i \to \infty

. Since

w (y) = 0

, we deduce from the upper semicontinuity of w that

lim_{ε \to 0} sup {w (x) : x \in B (y, ε) \cap \bar{Ω}} \leq 0 .

Given

ε > 0

small enough, there exists

i_{ε}

large enough such that

x_{i} \in B (y, ε) \cap \bar{Ω} for i \geq i_{ε} .

Therefore, we have

0 \leq η_{i} = w (x_{i}) \leq sup {w (x) : x \in B (y, ε) \cap \bar{Ω}}, i \geq i_{ε} .

Thus

0 \leq lim_{i \to \infty} sup η_{i} \leq sup {w (x) : x \in B (y, ε) \cap \bar{Ω}} .

Letting $ε \to 0$ , we conclude that zero is a cluster point of ${η_{i}}_{i \geq N_{0}}$ as $i \to \infty$ . The claim (2.7) follows from the monotonicity of ${η_{i}}_{i \geq N_{0}}$ .

Now, fix

s \in (0, \infty)

. For

δ > 0

and

i \geq N_{0}

, we define

z_{i} (t, x) = {(t + s)}^{- \frac{1}{m - 1}} w (x) - s^{- \frac{1}{m - 1}} η_{i}, (t, x) \in [0, \infty) \times \bar{Ω}

and

Z_{δ} (t, x) = {(t + δ)}^{- \frac{1}{m - 1}} W (x), (t, x) \in [0, \infty) \times \bar{Ω} .

It follows from the assumptions on w and W that

z_{i}

and

Z_{δ}

are, respectively, a bounded usc viscosity subsolution and a bounded lsc viscosity supersolution to

\partial_{t} ζ - Δ ζ^{m} - {| \nabla ζ^{l} |}^{q} = 0 in (0, \infty) \times Ω,

and satisfy

Z_{δ} (t, x) = 0 \geq - s^{- \frac{1}{m - 1}} η_{i} = z_{i} (t, x), (t, x) \in (0, \infty) \times \partial Ω .

Moreover, if

0 < δ < {[\frac{μ_{i}}{1 + {∥ w ∥}_{\infty}}]}^{m - 1} s,

(2.8)

then it follows from (2.5) and (2.8) that, for

x \in Ω_{i}

,

Z_{δ} (0, x) = δ^{- \frac{1}{m - 1}} W (x) \geq δ^{- \frac{1}{m - 1}} μ_{i} \geq s^{- \frac{1}{m - 1}} {∥ w ∥}_{\infty} \geq z_{i} (0, x) .

For

x \in \bar{Ω} ∖ Ω_{i}

, we deduce from (2.6) that

Z_{δ} (0, x) \geq 0 \geq s^{- \frac{1}{m - 1}} (w (x) - η_{i}) = z_{i} (0, x) .

By the comparison principle [6], we have

z_{i} (t, x) \leq Z_{δ} (t, x), (t, x) \in [0, \infty) \times \bar{Ω},

(2.9)

for $i \geq N_{0}$ and $δ > 0$ satisfying (2.8).

According to (2.8), the parameter δ can be taken to be arbitrarily small in (2.9). Therefore, we deduce that

{(t + s)}^{- \frac{1}{m - 1}} w (x) - s^{- \frac{1}{m - 1}} η_{i} \leq t^{- \frac{1}{m - 1}} W (x), (t, x) \in (0, \infty) \times \bar{Ω},

for $i \geq N_{0}$ .

Passing to the limit as

i \to \infty

, it follows from (2.7) that

{(t + s)}^{- \frac{1}{m - 1}} w (x) \leq t^{- \frac{1}{m - 1}} W (x), (t, x) \in (0, \infty) \times \bar{Ω} .

Finally, let

s \to 0

and take

t = 1

; then we obtain

w (x) \leq W (x), (t, x) \in (0, \infty) \times \bar{Ω} .

The proof of Lemma 2.1 is complete. □

A straightforward consequence of Lemma 2.1 is the uniqueness of the solution to (1.9).

Corollary 2.1 There is at most one positive viscosity solution to (1.9).

By the similar argument, we have the following result to (1.12).

Lemma 2.2 Let

w \in U S C (\bar{Ω})

and

W \in L S C (\bar{Ω})

be respectively a bounded upper semicontinuous (usc) viscosity subsolution and a bounded lower semicontinuous (lsc) viscosity supersolution to

- Δ ζ^{m} - \frac{ζ}{m - 1} = 0 in Ω,

(2.10)

satisfying (2.2) and (2.3). Then

w (x) \leq W (x) for x \in \bar{Ω} .

(2.11)

Proof The proof is similar as in Lemma 2.1, so we omit it here. □

3 Proofs of Theorems 1.1 and 1.2

In this section, we obtain the well-posedness and large time behavior of solutions to (1.1) for $l q \in [m, m + 1]$ , and we prove Theorems 1.1-1.2. To do this, we first obtain the well-posedness to (1.1) by the following proposition.

Proposition 3.1 Assume that $l \geq m > 1$ , $1 \leq q < 2$ , $l q \in [m, m + 1]$ , and $u_{0} (x) \in C_{0} (\bar{Ω})$ satisfies (1.2). Then there exists a unique solution $u (x, t)$ to (1.1) in the sense of Definition 1.1.

Proof The idea of the proof is same as in [25, 28], so we omit here. □

Next, in order to prove the large time behavior of the solution to (1.1), we shall need several steps: Step 1, we will find that the temporal decay rate of ${∥ u (t) ∥}_{\infty}$ is indeed $t^{- \frac{1}{m - 1}}$ . Step 2, we prove the boundary estimates for the large time which guarantee that no loss of boundary condition occurs throughout the time evolution. Step 3, the half-relaxed limits technique is applied to show the expected convergence after introducing self-similar variables. The approach is developed by Laurençot and Stinner in [25, 27]. To do this, we need the following lemmas.

Lemma 3.1 (Upper bound)

Assume that

l \geq m > 1

,

1 \leq q < 2

,

l q = m

, and the initial data

u_{0} (x) \in W^{1, \infty} (Ω)

satisfies (1.2). Then there exists

C_{1} > 0

depending only on m, l, q, Ω, and

{∥ u_{0} ∥}_{\infty}

such that

u (t, x) \leq C_{1} {(1 + t)}^{- \frac{1}{m - 1}}, (t, x) \in (0, \infty) \times \bar{Ω} .

(3.1)

Proof Assume that

x_{0} \notin \bar{Ω}

and

R_{0} > 0

such that

Ω \subset B (x_{0}, R_{0})

. For

(t, x) \in (0, \infty) \times \bar{Ω}

, we define the function

S_{1} (t, x) = A_{1} {(1 + t)}^{- \frac{1}{m - 1}} φ (r), φ (r) = {[\frac{m}{m + 1} (e^{\frac{R (m + 1)}{m}} - e^{\frac{r (m + 1)}{m}})]}^{\frac{1}{m}} with r = | x - x_{0} |,

where

R > R_{0}

and

A_{1} > 0

satisfies the following condition:

A_{1} \geq max {{(\frac{m}{m - 1} e^{\frac{R (m + 1)}{m^{2}}})}^{\frac{1}{m - 1}}, \frac{{∥ u_{0} ∥}_{\infty}}{φ (R_{0})}} .

(3.2)

Since

x_{0} \notin \bar{Ω}

, the function

S_{1} (t, x)

is

C^{\infty}

-smooth in

[0, \infty) \times \bar{Ω}

. Moreover, according the condition

l \geq m > 1

,

1 \leq q < 2

, and

l q = m

, we have

l = m > 1

and

q = 1

. Therefore, for

(t, x) \in (0, \infty) \times Ω

and

r = | x - x_{0} | < R_{0} < R

, it follows from (3.2) that

\begin{aligned} {(1 + t)}^{\frac{m}{m - 1}} {\partial_{t} S_{1} - △ S_{1}^{m} - {| \nabla S_{1}^{l} |}^{q}} (t, x) \\ = - \frac{A_{1}}{m - 1} φ (r) - A_{1}^{m} [\frac{N - 1}{r} {(φ^{m})}_{r} + {(φ^{m})}_{r r}] - A_{1}^{m} {| {(φ^{l})}_{r} |}^{q} \\ = - \frac{A_{1}}{m - 1} φ (r) + A_{1}^{m} [\frac{N - 1}{r} e^{\frac{r (m + 1)}{m}} + \frac{m + 1}{m} e^{\frac{r (m + 1)}{m}}] - A_{1}^{m} e^{\frac{r (m + 1)}{m}} \\ \geq A_{1} [A_{1}^{m - 1} (\frac{N - 1}{r} + \frac{1}{m}) e^{\frac{r (m + 1)}{m}} - \frac{1}{m - 1} e^{\frac{R (m + 1)}{m^{2}}}] \\ \geq A_{1} (\frac{1}{m} A_{1}^{m - 1} - \frac{1}{m - 1} e^{\frac{R (m + 1)}{m^{2}}}) \\ \geq 0 . \end{aligned}

(3.3)

Hence, the condition (3.2) guarantees that

S_{1} (t, x)

is a supersolution to (1.1) in

(0, \infty) \times Ω

. In addition, since

r = | x - x_{0} | < R_{0} < R

for

x \in Ω

, for

(t, x) \in (0, \infty) \times \partial Ω

, we deduce from (3.2) that

u (t, x) = 0 \leq A_{1} {(1 + t)}^{- \frac{1}{m - 1}} φ (R_{0}) \leq S_{1} (t, x),

and

u_{0} (x) \leq {∥ u_{0} (x) ∥}_{\infty} \leq A_{1} φ (R_{0}) \leq S_{1} (0, x), x \in \bar{Ω} .

By the comparison principle, we have

u (t, x) \leq S_{1} (t, x), (t, x) \in [0, \infty) \times \bar{Ω} .

The proof of Lemma 3.1 is complete. □

Lemma 3.2 (Upper bound)

Assume that

l \geq m > 1

,

1 \leq q < 2

,

l q \in (m, m + 1]

, and the initial data

u_{0} (x) \in W^{1, \infty} (Ω)

satisfies (1.2). Then there exists

C_{2} > 0

depending only on m, l, q, Ω, and

{∥ u_{0} ∥}_{\infty}

such that

u (t, x) \leq C_{2} {(1 + t)}^{- \frac{1}{m - 1}}, (t, x) \in (0, \infty) \times \bar{Ω} .

(3.4)

Proof Assume that

x_{0} \notin \bar{Ω}

and

R_{0} > 0

such that

Ω \subset B (x_{0}, R_{0})

. For

(t, x) \in (0, \infty) \times \bar{Ω}

, we define the function

S_{2} (t, x) = A_{2} {(1 + δ t)}^{- \frac{1}{m - 1}} σ (r), σ (r) = {[\frac{m}{m + 1} (R^{\frac{m + 1}{m}} - r^{\frac{m + 1}{m}})]}^{\frac{1}{m}} with r = | x - x_{0} |,

where the positive constants

A_{2}

, R, and δ satisfy the following condition:

\begin{aligned} A_{2} = {[\frac{m^{q - 1} (1 + m (N - 1))}{2 l^{q} R^{\frac{(m + 1) (l - m) q + m (q + m - 1)}{m^{2}}}}]}^{\frac{1}{l q - m}}, R = {[R_{0}^{\frac{m + 1}{m}} + \frac{(m + 1) {∥ u_{0} ∥}_{\infty}^{m}}{m A_{2}^{m}}]}^{\frac{m}{m + 1}} > R_{0} \\ and δ = \frac{(m - 1) (1 + m (N - 1)) A_{2}^{m - 1}}{2 m R^{\frac{1 + m^{2}}{m^{2}}}} . \end{aligned}

(3.5)

Since

x_{0} \notin \bar{Ω}

, the function

S_{2} (t, x)

is

C^{\infty}

-smooth in

[0, \infty) \times \bar{Ω}

. Moreover, for

(t, x) \in (0, \infty) \times Ω

and

r = | x - x_{0} | < R_{0} < R

, it follows from (3.5) and

l q > m

that

\begin{aligned} {(1 + δ t)}^{\frac{m}{m - 1}} {\partial_{t} S_{2} - △ S_{2}^{m} - {| \nabla S_{2}^{l} |}^{q}} (t, x) \\ = - \frac{A_{2} δ}{m - 1} σ (r) - A_{2}^{m} [\frac{N - 1}{r} {(σ^{m})}_{r} + {(σ^{m})}_{r r}] - A_{2}^{l q} {(1 + δ t)}^{- \frac{l q - m}{m - 1}} {| {(σ^{l})}_{r} |}^{q} \\ = - \frac{A_{2} δ}{m - 1} σ (r) + A_{2}^{m} [\frac{1}{m} + N - 1] r^{- \frac{m - 1}{m}} - A_{2}^{l q} {(1 + δ t)}^{- \frac{l q - m}{m - 1}} {| {({(σ^{m})}^{\frac{l}{m}})}_{r} |}^{q} \\ = - \frac{A_{2} δ}{m - 1} σ (r) + A_{2}^{m} [\frac{1}{m} + N - 1] r^{- \frac{m - 1}{m}} - A_{2}^{l q} {(\frac{l}{m})}^{q} R^{\frac{(m + 1) (l - m) q + m q}{m^{2}}} \\ \geq A_{2}^{m} [(N - 1 + \frac{1}{m}) R^{- \frac{m - 1}{m}} - A_{2}^{l q - m} {(\frac{l}{m})}^{q} R^{\frac{(m + 1) (l - m) q + m q}{m^{2}}} - \frac{δ R^{\frac{m + 1}{m^{2}}}}{(m - 1) A_{2}^{m - 1}}] \\ \geq A_{2}^{m} [(\frac{1}{2} (N - 1 + \frac{1}{m}) R^{- \frac{m - 1}{m}} - A_{2}^{l q - m} {(\frac{l}{m})}^{q} R^{\frac{(m + 1) (l - m) q + m q}{m^{2}}}) \\ + (\frac{1}{2} (N - 1 + \frac{1}{m}) R^{- \frac{m - 1}{m}} - \frac{δ R^{\frac{m + 1}{m^{2}}}}{(m - 1) A_{2}^{m - 1}})] \\ \geq 0 . \end{aligned}

(3.6)

Therefore, the condition (3.5) guarantees that

S_{2} (t, x)

is a supersolution to (1.1) in

(0, \infty) \times Ω

. In addition, since

r = | x - x_{0} | < R_{0} < R

for

x \in Ω

, for

(t, x) \in (0, \infty) \times \partial Ω

, we deduce from (3.5) that

u (t, x) = 0 \leq A_{2} {(1 + t)}^{- \frac{1}{m - 1}} σ (R_{0}) \leq S_{2} (t, x),

and

u_{0} (x) \leq {∥ u_{0} (x) ∥}_{\infty} \leq A_{2} σ (R_{0}) \leq S_{2} (0, x), x \in \bar{Ω} .

By the comparison principle, we have

u (t, x) \leq S_{2} (t, x), (t, x) \in [0, \infty) \times \bar{Ω} .

The proof of Lemma 3.2 is complete. □

Lemma 3.3 (Hölder estimate)

Assume that

l \geq m > 1

,

1 \leq q < 2

,

l q \in [m, m + 1]

, and the initial data

u_{0} (x) \in W^{1, \infty} (Ω)

satisfies (1.2). Then there exists

L_{1} > 0

depending only on m, l, q, Ω, and

{∥ u_{0} ∥}_{W^{1, \infty} (Ω)}

such that

\begin{aligned} | u (t, x) | = | u (t, x) - u (t, x_{0}) | \leq \frac{L_{1}}{{(1 + t)}^{\frac{1}{m - 1}}} {| x - x_{0} |}^{\frac{1}{m}}, \\ (t, x, x_{0}) \in [0, \infty) \times \bar{Ω} \times \partial Ω . \end{aligned}

(3.7)

Proof Since the boundary ∂ Ω of Ω is smooth, there exists

R_{Ω} > 0

such that for each

x_{0} \in \partial Ω

, there exists

y_{0} \in R^{N}

satisfying

| x_{0} - y_{0} | = R_{Ω}

and

B (y_{0}, R_{Ω}) \cap Ω = \emptyset

. It follows from the initial data condition

u_{0} (x) \in W^{1, \infty} (Ω)

that

u_{0} (x)

is Lipschitz continuous, i.e., there exists

L_{0} > 0

such that

| u_{0} (x) - u_{0} (y) | \leq L_{0} | x - y |, (x, y) \in \bar{Ω} \times \bar{Ω} .

(3.8)

Next, we define the open subset

U_{κ, x_{0}}

of

R^{N}

by

U_{κ, x_{0}} = {x \in Ω : R_{Ω} < | x - y_{0} | < R_{Ω} + κ},

where

0 < κ < 1

satisfies

Ω_{κ} = {x \in Ω : d (x, \partial Ω) > κ} \neq \emptyset

, and denote the function

S_{κ, x_{0}} (t, x) = A_{3} {(1 + t)}^{- \frac{1}{m - 1}} ψ (r), ψ (r) = {(1 - e^{- \frac{r}{κ}})}^{\frac{1}{m}} with r = | x - y_{0} | - R_{Ω},

for $(t, x) \in (0, \infty) \times {\bar{U}}_{κ, x_{0}}$ and $A_{3} > 0$ .

Moreover, we assume that

A_{3} \geq max {e L_{0}, \frac{e^{\frac{1}{m}} C}{{(e - 1)}^{\frac{1}{m}}}, {(\frac{3 e}{m - 1})}^{\frac{1}{m - 1}}}

(3.9)

and

0 < κ < min {1, \frac{R_{Ω}}{3 (N - 1)}, {(\frac{m^{q}}{3 l^{q} A^{l q - m}})}^{\frac{1}{2 - q}}},

(3.10)

where $C = max {C_{1}, C_{2}} > 0$ , $C_{1}$ , and $C_{2}$ are defined in Lemmas 3.1 and 3.2, respectively.

Since

y_{0} \notin {\bar{U}}_{κ, x_{0}}

, the function

S_{κ, x_{0}} (t, x)

is

C^{\infty}

-smooth in

[0, \infty) \times {\bar{U}}_{κ, x_{0}}

. For

(t, x) \in (0, \infty) \times U_{κ, x_{0}}

, we have

r = | x - y_{0} | - R_{Ω} \in (0, κ)

. By a direct computation, we infer from (3.9)-(3.10) that

\begin{aligned} \frac{{(1 + t)}^{\frac{m}{m - 1}}}{A_{3}^{m}} {\partial_{t} S_{κ, x_{0}} - Δ S_{κ, x_{0}}^{m} - {| \nabla S_{κ, x_{0}}^{l} |}^{q}} (t, x) \\ = - \frac{1}{(m - 1) A_{3}^{m - 1}} ψ (r) - [\frac{N - 1}{r + R_{Ω}} {(ψ^{m})}_{r} + {(ψ^{m})}_{r r}] - A_{3}^{l q - m} {(1 + t)}^{- \frac{l q - m}{m - 1}} {| {(ψ^{l})}_{r} |}^{q} \\ = - \frac{1}{(m - 1) A_{3}^{m - 1}} {(1 - e^{- \frac{r}{κ}})}^{\frac{1}{m}} + \frac{1}{κ^{2}} e^{- \frac{r}{κ}} - \frac{N - 1}{(r + R_{Ω}) κ} e^{- \frac{r}{κ}} \\ - A_{3}^{l q - m} {(1 + t)}^{- \frac{l q - m}{m - 1}} {| {({(ψ^{m})}^{\frac{l}{m}})}_{r} |}^{q} \\ \geq - \frac{1}{(m - 1) A_{3}^{m - 1}} + \frac{1}{κ^{2}} e^{- \frac{r}{κ}} - \frac{N - 1}{κ R_{Ω}} e^{- \frac{r}{κ}} - A_{3}^{l q - m} {| {({(ψ^{m})}^{\frac{l}{m}})}_{r} |}^{q} \\ \geq \frac{1}{κ^{2}} e^{- \frac{r}{κ}} [1 - \frac{N - 1}{R_{Ω}} κ - A_{3}^{l q - m} {(\frac{l}{m})}^{q} κ^{2 - q} - \frac{e}{(m - 1) A_{3}^{m - 1}}] \\ \geq \frac{1}{κ^{2}} e^{- \frac{r}{κ}} [(\frac{1}{3} - \frac{N - 1}{R_{Ω}} κ) + (\frac{1}{3} - A_{3}^{l q - m} {(\frac{l}{m})}^{q} κ^{2 - q}) + (\frac{1}{3} - \frac{e}{(m - 1) A_{3}^{m - 1}})] \\ \geq 0 . \end{aligned}

Therefore,

S_{κ, x_{0}} (t, x)

is a supersolution to (1.1) in

(0, \infty) \times U_{κ, x_{0}}

. In addition, it follows from (3.8)-(3.10) and the mean value theorem that

\begin{aligned} S_{κ, x_{0}} (0, x) & = A_{3} {(1 - e^{- \frac{| x - y_{0} | - R_{Ω}}{κ}})}^{\frac{1}{m}} \geq A_{3} (1 - e^{- \frac{| x - y_{0} | - R_{Ω}}{κ}}) \\ \geq A_{3} \frac{| x - y_{0} | - R_{Ω}}{e κ} = \frac{A_{3}}{e κ} dist (x, \partial B (y_{0}, R_{Ω})) \\ \geq \frac{A_{3}}{e κ} dist (x, \partial Ω) \geq L_{0} dist (x, \partial Ω) \geq u_{0} (x), x \in {\bar{U}}_{δ, x_{0}} . \end{aligned}

Moreover, for

(t, x) \in (0, \infty) \times \partial U_{κ, x_{0}}

, we have either

x \in \partial Ω

or

r = | x - y_{0} | - R_{Ω} = δ

. If

x \in \partial Ω

, then we have

S_{κ, x_{0}} (t, x) = A_{3} {(1 + t)}^{- \frac{1}{m - 1}} {(1 - e^{- \frac{r}{κ}})}^{\frac{1}{m}} \geq 0 = u (x, t), (t, x) \in (0, \infty) \times \partial Ω .

If

r = | x - y_{0} | - R_{Ω} = δ

, by Lemmas 3.1-3.2 and (3.9), we have

S_{κ, x_{0}} (t, x) = A_{3} {(1 + t)}^{- \frac{1}{m - 1}} {(1 - e^{- 1})}^{\frac{1}{m}} \geq C {(1 + t)}^{- \frac{1}{m - 1}} \geq u (t, x) .

By the comparison principle [6] we have

u (t, x) \leq S_{κ, x_{0}} (t, x), (t, x) \in [0, \infty) \times {\bar{U}}_{κ, x_{0}} .

Consequently,

0 \leq u (t, x) - u (t, x_{0}) = u (t, x) \leq A_{3} {(1 + t)}^{- \frac{1}{m - 1}} {(1 - e^{- \frac{| x - y_{0} | - R_{Ω}}{κ}})}^{\frac{1}{m}},

for $(t, x) \in [0, \infty) \times {\bar{U}}_{κ, x_{0}}$ .

Since

| x_{0} - y_{0} | - R_{Ω} = 0

, we have

0 \leq | u (t, x) - u (t, x_{0}) | \leq \frac{A_{3}}{κ^{\frac{1}{m}} {(1 + t)}^{\frac{1}{m - 1}}} {| x - x_{0} |}^{\frac{1}{m}}, x \in {\bar{U}}_{κ, x_{0}} .

(3.11)

Finally, we consider

x \in Ω

and

x_{0} \in \partial Ω

. If

| x - x_{0} | \geq \frac{κ}{2}

, it follows from Lemmas 3.1-3.2 that

| u (t, x) - u (t, x_{0}) | = u (t, x) \leq \frac{2^{\frac{1}{m}} C}{κ^{\frac{1}{m}} {(1 + t)}^{\frac{1}{m - 1}}} {| x - x_{0} |}^{\frac{1}{m}},

where $C = max {C_{1}, C_{2}} > 0$ , and $C_{1}$ and $C_{2}$ are defined in Lemmas 3.1 and 3.2, respectively.

If

| x - x_{0} | < \frac{κ}{2}

, let

y_{0} \in R^{N}

satisfy

| x_{0} - y_{0} | = R_{Ω}

and

B (y_{0}, R_{Ω}) \cap Ω = \emptyset

. Since

x \in Ω

, we have

| x - y_{0} | > R_{Ω} and | x - y_{0} | \leq | x - x_{0} | + | x_{0} - y_{0} | < R_{Ω} + κ,

which implies

x \in U_{κ, x_{0}}

. Therefore, we deduce from (3.11) that

| u (t, x) - u (t, x_{0}) | = u (t, x) \leq \frac{A_{3}}{κ^{\frac{1}{m}} {(1 + t)}^{\frac{1}{m - 1}}} {| x - x_{0} |}^{\frac{1}{m}} .

Choosing

L_{1} = max {\frac{2^{\frac{1}{m}} C}{κ^{\frac{1}{m}}}, \frac{A_{3}}{κ^{\frac{1}{m}}}}

, we have

| u (t, x) | = | u (t, x) - u (t, x_{0}) | \leq \frac{L_{1}}{{(1 + t)}^{\frac{1}{m - 1}}} {| x - x_{0} |}^{\frac{1}{m}}, (t, x, x_{0}) \in [0, \infty) \times \bar{Ω} \times \partial Ω .

The proof of Lemma 3.3 is complete. □

We next proceed as in [29] to deduce the Hölder continuity of $u (x, t)$ from Lemma 3.3. Therefore, we obtain the following corollary.

Corollary 3.1 Assume that

l \geq m > 1

,

1 \leq q < 2

,

l q \in [m, m + 1]

, and the initial data

u_{0} (x) \in W^{1, \infty} (Ω)

satisfies (1.2). Then there exists

L_{2} > 0

depending only on m, l, q, Ω, and

{∥ u_{0} ∥}_{W^{1, \infty} (Ω)}

such that

| u (t, x) - u (t, y) | \leq \frac{L_{2}}{{(1 + t)}^{\frac{1}{m - 1}}} {| x - y |}^{\frac{1}{m}}, (t, x, y) \in [0, \infty) \times \bar{Ω} \times \bar{Ω} .

(3.12)

Proof The proof is similar to the argument in [25, 29], so we omit here. □

Proofs of Theorems 1.1 and 1.2 The proofs are based on the ideas in [25], but we give the details of the argument for the reader’s convenience. Let

U (x, t)

be the solution to the porous medium equation with homogeneous Dirichlet boundary conditions

{\begin{cases} \partial_{t} U - Δ U^{m} = 0, & (t, x) \in (0, \infty) \times Ω, \\ U (t, x) = 0, & (t, x) \in (0, \infty) \times \partial Ω, \\ U (0, x) = u_{0} (x), & x \in Ω . \end{cases}

(3.13)

According to the nonnegativity of

{| \nabla u^{l} |}^{q}

, it follows from the comparison principle [6] that

0 \leq U (t, x) \leq u (t, x), (t, x) \in (0, \infty) \times \bar{Ω} .

(3.14)

We introduce the scaling variable

s = ln t

for

t > 0

and denote the new unknown function v and V by

u (t, x) = t^{- \frac{1}{m - 1}} v (ln t, x), (t, x) \in (0, \infty) \times \bar{Ω}

and

U (t, x) = t^{- \frac{1}{m - 1}} V (ln t, x), (t, x) \in (0, \infty) \times \bar{Ω} .

Then v is a viscosity solution to the following problem:

{\begin{cases} \partial_{s} v - Δ v^{m} - e^{- \frac{(l q - m) s}{m - 1}} {| \nabla v^{l} |}^{q} - \frac{v}{m - 1} = 0, & (s, x) \in (0, \infty) \times Ω, \\ v (s, x) = 0, & (s, x) \in (0, \infty) \times \partial Ω, \\ v (0, x) = u (1, x), & x \in Ω . \end{cases}

(3.15)

In addition, owing to Lemmas 3.1-3.3 and (3.14), we have

V (s, x) \leq v (s, x) \leq C, (s, x) \in [0, \infty) \times \bar{Ω},

(3.16)

and

| v (s, x) - v (s, y) | \leq L_{1} {| x - y |}^{\frac{1}{m}}, (s, x, y) \in [0, \infty) \times \bar{Ω} \times \partial Ω .

(3.17)

Next, for

ε \in (0, 1)

, we define

w_{ε} (s, x) = v (\frac{s}{ε}, x), (s, x) \in [0, \infty) \times \bar{Ω},

and the half-relaxed limits

w_{*} (x) = lim_{(σ, y, ε) \to (s, x, 0)} inf w_{ε} (σ, y) and w^{*} (x) = lim_{(σ, y, ε) \to (s, x, 0)} sup w_{ε} (σ, y),

for $(s, x) \in (0, \infty) \times \bar{Ω}$ .

By (3.16), it is easy to see that

w_{*} (x)

and

w^{*} (x)

are well-defined and do not depend on

s > 0

. Moreover, it readily follows from (3.15) and (3.17) that

w_{*} (x) = w^{*} (x) = 0, x \in \partial Ω .

(3.18)

By a direct computation,

w_{ε} (s, x)

is a solution to the following initial-boundary problem:

{\begin{cases} ε \partial_{s} w_{ε} - Δ w_{ε}^{m} - e^{- \frac{(l q - m) s}{(m - 1) ε}} {| \nabla w_{ε}^{l} |}^{q} - \frac{w_{ε}}{m - 1} = 0, & (s, x) \in (0, \infty) \times Ω, \\ w_{ε} (s, x) = 0, & (s, x) \in (0, \infty) \times \partial Ω, \\ w_{ε} (0, x) = u (1, x), & x \in Ω . \end{cases}

(3.19)

Next, we shall give proofs of Theorems 1.1 and 1.2. To do this, we distinguish the two cases $l q = m$ and $l q \in (m, m + 1]$ .

Case 1.

l q = m

. We use the stability of semicontinuous viscosity solutions [6] to deduce from (3.19) that

w_{*} (x) is a supersolution to (2.1) in Ω

(3.20)

and

w^{*} (x) is a subsolution to (2.1) in Ω .

(3.21)

In addition, by [30],

V (s) \to f_{0}

in

L^{\infty} (Ω)

as

s \to \infty

. Moreover, it follows from (3.16) and the definition of

w_{*} (x)

and

w^{*} (x)

that

f_{0} (x) \leq w_{*} (x) \leq w^{*} (x) \leq C, x \in \bar{Ω} .

(3.22)

Since

f_{0} > 0

in Ω by [30], we deduce from (3.22) that

w_{*} (x)

and

w^{*} (x)

are positive and bounded in Ω and vanish on ∂ Ω by (3.18). Owing to (3.20) and (3.21), we infer from Lemma 2.1 that

w^{*} (x) \leq w_{*} (x), x \in \bar{Ω} .

By (3.22), we have

w^{*} (x) = w_{*} (x), x \in \bar{Ω} .

Setting $f (x) = w_{*} (x) = w^{*} (x)$ , we deduce from (3.18), (3.20), (3.21), and (3.22) that $f (x) \in C_{0} (\bar{Ω})$ is a positive viscosity solution to (2.1) and solves (1.9). Therefore, the existence of a positive solution to (1.9) is proved. Moreover, by Corollary 2.1, we obtain the uniqueness of solution to (1.9).

Furthermore, it follows from the equality

w_{*} (x) = w^{*} (x)

that

{∥ w_{ε} (1) - f ∥}_{\infty} \to 0 as ε \to 0,

i.e.,

lim_{s \to \infty} {∥ v (s) - f ∥}_{\infty} = 0 .

(3.23)

Therefore, we infer from the scaling transformation that

lim_{t \to \infty} {∥ t^{\frac{1}{m - 1}} u (t) - f ∥}_{\infty} = 0 .

(3.24)

Finally, Corollary 3.1 gives the last statement of Theorem 1.1. The proof of Theorem 1.1 is complete.

Case 2.

l q \in (m, m + 1]

. We use once more the stability of semicontinuous viscosity solutions [6] to deduce from (3.19) that

w_{*} (x) is a supersolution to (2.10) in Ω

(3.25)

and

w^{*} (x) is a subsolution to (2.10) in Ω .

(3.26)

In addition, by [30],

V (s) \to f_{0}

in

L^{\infty} (Ω)

as

s \to \infty

. Moreover, it follows from (3.16) and the definition of

w_{*} (x)

and

w^{*} (x)

that

f_{0} (x) \leq w_{*} (x) \leq w^{*} (x) \leq C, x \in \bar{Ω} .

(3.27)

Since

f_{0} > 0

in Ω by [30] and is a solution to (2.10), we can apply Lemma 2.2 to conclude that

w_{*} (x) \leq f_{0}

in

\bar{Ω}

. Owing to (3.27), we have

w_{*} (x) = w^{*} (x) = f_{0} (x), x \in \bar{Ω} .

Therefore, we deduce from (3.19) that

{∥ w_{ε} (1) - f_{0} ∥}_{\infty} \to 0 as ε \to 0,

i.e.,

lim_{s \to \infty} {∥ v (s) - f_{0} ∥}_{\infty} = 0 .

(3.28)

Thus, we infer from the scaling transformation that

lim_{t \to \infty} {∥ t^{\frac{1}{m - 1}} u (t) - f_{0} ∥}_{\infty} = 0 .

(3.29)

The proof of Theorem 1.2 is complete. □

4 Proof of Theorem 1.3

In this section, we shall consider the well-posedness and the large time behavior of solutions to (1.1) with the small initial data $u_{0} (x)$ for $l q > m + 1$ by the method used in [25]. To do this, we need the following lemma.

Lemma 4.1 Let

l \geq m > 1

and

1 \leq q < 2

. Assume that G is the corresponding solution to (1.1) with the initial data

G_{0} \in W^{1, \infty} (Ω)

satisfying (1.2) for the case

l q = m

, and denote

F (t, x) = \frac{1}{ℓ [G_{0}]} G (\frac{t}{ℓ {[G_{0}]}^{m - 1}}, x), (t, x) \in [0, \infty) \times \bar{Ω},

(4.1)

where the parameter $ℓ [G_{0}]$ is defined in (1.10). Then $F (t, x)$ is a solution to (1.1) with the initial data $G_{0} / ℓ [G_{0}]$ and $l q = m$ such that $| \nabla F^{l} | \leq 1$ for $(t, x) \in [0, \infty) \times \bar{Ω}$ . Moreover, $F (t, x)$ is a supersolution to (1.1) for $l q > m$ .

Proof According to the definition of $ℓ [G_{0}]$ in (1.10), it is easy to see that $| \nabla F^{l} | \leq 1$ for $(t, x) \in [0, \infty) \times \bar{Ω}$ .

Next, let

φ (t, x) \in C^{2} ((0, \infty) \times Ω)

and

F - φ

has a local minimum at

(t_{0}, x_{0}) \in (0, \infty) \times Ω

. Since ℱ is smooth with respect to the time variable and Hölder continuous with respect to the space variable, we obtain

| \nabla φ^{l} (t_{0}, x_{0}) | \leq 1 .

(4.2)

Moreover, introducing

ψ (t, x) : = ℓ [G_{0}] φ (ℓ {[G_{0}]}^{m - 1} t, x)

for

(t, x) \in [0, \infty) \times \bar{Ω}

, the function

G - ψ

has a local minimum at

(t_{0} ℓ {[G_{0}]}^{1 - m}, x_{0})

such that

\partial_{t} ψ (t_{0} ℓ {[G_{0}]}^{1 - m}, x_{0}) - Δ ψ^{m} (t_{0} ℓ {[G_{0}]}^{1 - m}, x_{0}) - {| \nabla ψ^{l} (t_{0} ℓ {[G_{0}]}^{1 - m}, x_{0}) |}^{\frac{m}{l}} \geq 0,

i.e.,

\partial_{t} φ (t_{0}, x_{0}) - Δ φ^{m} (t_{0}, x_{0}) - {| \nabla φ^{l} (t_{0}, x_{0}) |}^{\frac{m}{l}} \geq 0 .

(4.3)

Therefore, $F (t, x)$ is a supersolution to (1.1) with $l q = m$ . In a similar way, it can be shown that $F (t, x)$ is also a subsolution. Hence, $F (t, x)$ is a solution to (1.1) with $l q = m$ .

Furthermore, we deduce from (4.2), (4.3), and

l q > m

that

\begin{aligned} \partial_{t} φ (t_{0}, x_{0}) - Δ φ^{m} (t_{0}, x_{0}) - {| \nabla φ^{l} (t_{0}, x_{0}) |}^{q} \\ \geq {| \nabla φ^{l} (t_{0}, x_{0}) |}^{\frac{m}{l}} (1 - {| \nabla φ^{l} (t_{0}, x_{0}) |}^{q - \frac{m}{l}}) \geq 0 . \end{aligned}

(4.4)

The proof of Lemma 4.1 is complete. □

Proposition 4.1 Let

l \geq m > 1

,

1 \leq q < 2

, and

l q > m + 1

. Assume that the initial data

u_{0} (x) \in W^{1, \infty} (Ω)

satisfies (1.2), moreover, there exists

G_{0} \in W^{1, \infty} (Ω)

satisfying (1.2) such that

u_{0} (x) \leq \frac{G_{0} (x)}{ℓ [G_{0}]}, x \in \bar{Ω},

(4.5)

where the parameter

ℓ [G_{0}]

is defined in (1.10). Then there exists a unique solution u to (1.1) in the sense of Definition 1.1 and it satisfies

u (t, x) \leq F (t, x), (t, x) \in [0, \infty) \times \bar{Ω},

(4.6)

where $F (t, x)$ is defined in (4.1).

Proof On the one hand, the solution U to the porous medium equation (3.13) is clearly a subsolution to (1.1) in $(0, \infty) \times Ω$ .

On the other hand, it follows from Lemma 4.1 that the function $F (t, x)$ is a supersolution to (1.1) in $(0, \infty) \times Ω$ . Therefore, $F (t, x)$ is a supersolution to (3.13).

Since

U = F = 0

on

(0, \infty) \times \partial Ω

and

U (0, x) = u_{0} (x) \leq F (0, x)

for

x \in \bar{Ω}

by (4.5), we infer from the comparison principle [6] that

U (t, x) \leq F (t, x), (t, x) \in [0, \infty) \times \bar{Ω} .

This property and the simultaneous vanishing of U and ℱ on $(0, \infty) \times \partial Ω$ allow us to use the classical Perron method to establish the existence of a solution $u (x, t)$ to (1.1) in the sense of Definition 1.1 which satisfies (4.6). The uniqueness next follows from the comparison principle [6]. The proof of Proposition 4.1 is complete. □

Proof of Theorem 1.3 We notice that Lemma 3.2 is also valid in that case. It readily follows from Lemma 3.3 and Proposition 4.1 that

\begin{aligned} 0 \leq u (t, x) = u (t, x) - u (t, x_{0}) \leq F (t, x) \leq \frac{L_{1}}{{(1 + t)}^{\frac{1}{m - 1}}} {| x - x_{0} |}^{\frac{1}{m}}, \\ (t, x, x_{0}) \in [0, \infty) \times \bar{Ω} \times \partial Ω . \end{aligned}

The convergence proof is similar to that performed in the proof of Theorem 1.2 for $l q \in (m, m + 1]$ . The proof of Theorem 1.3 is complete. □

5 Proof of Theorem 1.4

In this section, when $l q > m + 1$ , we shall prove that the solution $u (x, t)$ of (1.1) blows up in finite time for the large initial data $u_{0} (x)$ in the sense of weak solution by the method used in [26].

In order to obtain a blow-up condition corresponding to (1.1), we have to modify the function

e^{- ε {| x |}^{2}}

used in [31–33], and introduce a test function

ϕ_{ε} (x)

as follows:

ϕ_{ε} (x) = A_{ε} e^{- ε | x |} with A_{ε} = \frac{1}{\int_{Ω} e^{- ε | x |} d x} .

Proof of Theorem 1.4 Suppose that

u (x, t)

is the solution of the problem (1.1) and T is the blow-up time of the solution. For

s \in (0, \frac{1}{2})

, we denote

W (t) = \frac{1}{s} \int_{Ω} u^{s} ϕ_{ε} (x) d x, t \in (0, T) .

(5.1)

By a direct calculation, we have

\begin{aligned} W^{'} (t) = & \int_{Ω} u^{s - 1} ϕ_{ε} u_{t} d x \\ = & \int_{Ω} u^{s - 1} ϕ_{ε} Δ u^{m} d x + \int_{Ω} u^{s - 1} ϕ_{ε} {| \nabla u^{l} |}^{q} d x \\ \geq & - (s - 1) \int_{Ω} u^{s - 2} ϕ_{ε} \nabla u^{m} \cdot \nabla u d x + ε \int_{Ω} u^{s - 1} ϕ_{ε} \nabla u^{m} \cdot \frac{x}{| x |} d x \\ + \int_{Ω} u^{s - 1} ϕ_{ε} {| \nabla u^{l} |}^{q} d x \\ \geq & m (1 - s) \int_{Ω} u^{m + s - 3} ϕ_{ε} {| \nabla u |}^{2} d x - m ε \int_{Ω} u^{m + s - 2} ϕ_{ε} | \nabla u | d x \\ + \int_{Ω} u^{s - 1} ϕ_{ε} {| \nabla u^{l} |}^{q} d x . \end{aligned}

(5.2)

By Young’s inequality, we obtain

ε \int_{Ω} u^{m + s - 2} ϕ_{ε} | \nabla u | d x \leq \frac{1}{2} \int_{Ω} u^{m + s - 3} ϕ_{ε} {| \nabla u |}^{2} d x + \frac{ε^{2}}{2} \int_{Ω} u^{m + s - 1} ϕ_{ε} d x .

(5.3)

Since

0 < s < \frac{1}{2}

, it follows from (5.2), (5.3), and Poincaré’s inequality that

\begin{aligned} W^{'} (t) & \geq \int_{Ω} u^{s - 1} ϕ_{ε} {| \nabla u^{l} |}^{q} d x - \frac{m ε^{2}}{2} \int_{Ω} u^{m + s - 1} ϕ_{ε} d x \\ = {(\frac{q l}{q l + s - 1})}^{q} \int_{Ω} ϕ_{ε} {| \nabla u^{\frac{q l + s - 1}{q}} |}^{q} d x - \frac{m ε^{2}}{2} \int_{Ω} u^{m + s - 1} ϕ_{ε} d x \\ \geq {(\frac{q l}{q l + s - 1})}^{q} \int_{Ω} ϕ_{ε} u^{q l + s - 1} d x - \frac{m ε^{2}}{2} \int_{Ω} u^{m + s - 1} ϕ_{ε} d x . \end{aligned}

(5.4)

According to

q l > m + 1

,

0 < s < \frac{1}{2}

,

\int_{Ω} ϕ_{ε} (x) d x = 1

and Hölder’s inequality, we have

\int_{Ω} u^{m + s - 1} ϕ_{ε} d x = \int_{Ω} u^{m + s - 1} ϕ_{ε}^{\frac{m + s - 1}{q l + s - 1}} ϕ_{ε}^{\frac{q l - m}{q l + s - 1}} d x \leq {(\int_{Ω} u^{q l + s - 1} ϕ_{ε} d x)}^{\frac{m + s - 1}{q l + s - 1}} .

(5.5)

Thus, by (5.4) and (5.5), we obtain

\frac{d W}{d t} \geq {(\int_{Ω} u^{q l + s - 1} ϕ_{ε} d x)}^{\frac{m + s - 1}{q l + s - 1}} [{(\frac{q l}{q l + s - 1})}^{q} {(\int_{Ω} u^{q l + s - 1} ϕ_{ε} d x)}^{\frac{q l - m}{q l + s - 1}} - \frac{m ε^{2}}{2}] .

(5.6)

Owing to

q l > m + 1

,

0 < s < \frac{1}{2}

,

\int_{Ω} ϕ_{ε} (x) d x = 1

, and Jensen’s inequality, we have

{(\int_{Ω} u^{q l + s - 1} ϕ_{ε} d x)}^{\frac{q l - m}{q l + s - 1}} \geq {(\int_{Ω} u^{s} ϕ_{ε} d x)}^{\frac{q l - m}{s}} .

(5.7)

Therefore, it follows from (5.6) and (5.7) that

\frac{d W}{d t} \geq \frac{1}{2} {(\frac{q l}{q l + s - 1})}^{q} {(\int_{Ω} u^{s} ϕ_{ε} d x)}^{\frac{q l + s - 1}{s}} = \frac{1}{2} {(\frac{q l}{q l + s - 1})}^{q} s^{\frac{q l + s - 1}{s}} W^{\frac{q l + s - 1}{s}} (t)

(5.8)

as long as

W (t) \geq \frac{1}{s} {(\frac{q l + s - 1}{q l})}^{\frac{q s}{q l - m}} {(m ε^{2})}^{\frac{s}{q l - m}} for all t \in [0, T) .

(5.9)

Taking

K_{0} = {(\frac{q l + s - 1}{q l})}^{\frac{q s}{q l - m}} {(m ε^{2})}^{\frac{s}{q l - m}} .

Since the initial data

u_{0} (x)

satisfies

\int_{Ω} u_{0}^{s} (x) ϕ_{ε} (x) d x \geq K_{0},

we have

W (0) \geq \frac{1}{s} K_{0} = \frac{1}{s} {(\frac{q l + s - 1}{q l})}^{\frac{q s}{q l - m}} {(m ε^{2})}^{\frac{s}{q l - m}} .

(5.10)

Therefore, $w (t)$ increases and remains above $\frac{1}{s} K_{0}$ for all $t \in [0, T]$ .

By (5.8), integrating over

(0, t)

yields

W (t) \geq {(W^{\frac{1 - q l}{s}} (0) - K_{1} t)}^{- \frac{s}{q l - 1}} with K_{1} = \frac{q l - 1}{2} {(\frac{q l}{q l + s - 1})}^{q} s^{\frac{q l - 1}{s}} .

(5.11)

Hence, it follows from (5.10) and (5.11) that the solution $u (x, t)$ of (1.1) blows up in finite time, $T = \frac{1}{K_{1}} W^{\frac{1 - q l}{s}} (0)$ .

Moreover, by (5.10), we obtain the upper estimate on the blow-up time T of the solution

u (x, t)

as follows:

T \leq \frac{2}{q l - 1} {(\frac{q l}{q l + s - 1})}^{\frac{q (m - 1)}{q l - m}} {(m ε^{2})}^{\frac{1 - q l}{q l - m}} .

(5.12)

The proof of Theorem 1.4 is complete. □

Declarations

Acknowledgements

The authors would like to express sincere gratitude to the referees for their valuable suggestions and comments on the original manuscript. The first author is supported by Scientific Research Fund of Sichuan Provincial Science and Technology Department (2014JY0098); the second author is supported in part by the Fundamental Research Funds for the Central Universities, Project No. CDJXS 12 10 00 14; the third author is supported in part by NSF of China (11371384).

Authors’ Affiliations

(1)

Department of Applied Mathematics, Southwestern University of Finance and Economics, Chengdu, 610074, P.R. China

(2)

College of Mathematics and Statistics, Chongqing University, Chongqing, 401331, P.R. China

(3)

College of Mathematics and Physics, Chongqing University of Posts and Telecommunications, Chongqing, 400065, P.R. China

References

Andreucci D: Degenerate parabolic equations with initial data measures. Trans. Am. Math. Soc. 1997, 349: 3911-3923. 10.1090/S0002-9947-97-01530-4MathSciNetView ArticleGoogle Scholar
Gilding BH, Guedda M, Kersner R: The Cauchy problem for $u_{t} = Δ u + {| \nabla u |}^{q}$ . J. Math. Anal. Appl. 2003, 284: 733-755. 10.1016/S0022-247X(03)00395-0MathSciNetView ArticleGoogle Scholar
Li YX, Souplet P: Single-point gradient blow-up on the boundary for diffusive Hamilton-Jacobi equations in planar domains. Commun. Math. Phys. 2010, 293: 499-517. 10.1007/s00220-009-0936-8MathSciNetView ArticleGoogle Scholar
Bardi M, Capuzzo-Dolcetta I: Optimal Control and Viscosity Solutions of Hamilton-Jacobi-Bellman Equations. Birkhäuser, Boston; 1997.View ArticleGoogle Scholar
Barles G Mathématiques et Applications 17. In Solutions de Viscosité des Equations d’Hamilton-Jacobi. Springer, Berlin; 1994.Google Scholar
Crandall MG, Ishii H, Lions PL: User’s guide to viscosity solutions of second order partial differential equations. Bull. Am. Math. Soc. (N.S.) 1992, 27: 1-67. 10.1090/S0273-0979-1992-00266-5MathSciNetView ArticleGoogle Scholar
Arrieta JM, Bernal AR, Souplet P: Boundedness of global solutions for nonlinear parabolic equations involving gradient blow-up phenomena. Ann. Sc. Norm. Super. Pisa, Cl. Sci. 2004, 3: 1-15.MathSciNetGoogle Scholar
Benachour SD, Dăbuleanu-Hapca S, Laurençot P: Decay estimates for a viscous Hamilton-Jacobi equation with homogeneous Dirichlet boundary conditions. Asymptot. Anal. 2007, 51: 209-229.MathSciNetGoogle Scholar
Ding JT, Guo BZ: Global existence and blow-up solutions for quasilinear reaction-diffusion equations with a gradient term. Appl. Math. Lett. 2011, 24: 936-942. 10.1016/j.aml.2010.12.052MathSciNetView ArticleGoogle Scholar
Souplet P, Zhang QS: Global solutions of inhomogeneous Hamilton-Jacobi equations. J. Anal. Math. 2006, 99: 355-396. 10.1007/BF02789452MathSciNetView ArticleGoogle Scholar
Xia L, Yao ZA: Existence, uniqueness and asymptotic behavior of solutions for a singular parabolic equation. J. Math. Anal. Appl. 2009, 358: 182-188. 10.1016/j.jmaa.2009.04.039MathSciNetView ArticleGoogle Scholar
Zhou WS, Lei PD: A one-dimensional nonlinear heat equation with a singular term. J. Math. Anal. Appl. 2010, 368: 711-726. 10.1016/j.jmaa.2010.03.066MathSciNetView ArticleGoogle Scholar
Barles G, Souganidis PE: On the large time behaviour of solutions of Hamilton-Jacobi equations. SIAM J. Math. Anal. 2000, 31: 925-939. 10.1137/S0036141099350869MathSciNetView ArticleGoogle Scholar
Benachour S, Dabuleanu S: Large time behavior for a viscous Hamilton-Jacobi equation with Neumann boundary condition. J. Differ. Equ. 2005, 216: 223-258. 10.1016/j.jde.2005.02.017MathSciNetView ArticleGoogle Scholar
Laurençot P: Convergence to steady states for a one-dimensional viscous Hamilton-Jacobi equation with Dirichlet boundary conditions. Pac. J. Math. 2007, 230: 347-364. 10.2140/pjm.2007.230.347View ArticleGoogle Scholar
Namah G, Roquejoffre JM: Remarks on the long time behaviour of the solutions of Hamilton-Jacobi equations. Commun. Partial Differ. Equ. 1999, 24: 883-893. 10.1080/03605309908821451MathSciNetView ArticleGoogle Scholar
Qi YW, Wang MX: The self-similar profiles of generalized KPZ equation. Pac. J. Math. 2001, 201: 223-240. 10.2140/pjm.2001.201.223MathSciNetView ArticleGoogle Scholar
Quittner P, Souplet P: Superlinear Parabolic Problems. Blow-up, Global Existence and Steady States. Birkhäuser, Basel; 2007.Google Scholar
Roquejoffre JM: Convergence to steady states or periodic solutions in a class of Hamilton-Jacobi equations. J. Math. Pures Appl. 2001, 80: 85-104. 10.1016/S0021-7824(00)01183-1MathSciNetView ArticleGoogle Scholar
Shi PH, Wang MX: Global solution of the fast diffusion equation with gradient absorption terms. J. Math. Anal. Appl. 2007, 326: 602-621. 10.1016/j.jmaa.2006.02.094MathSciNetView ArticleGoogle Scholar
Tchamba TT: Large time behavior of solutions of viscous Hamilton-Jacobi equations with superquadratic Hamiltonian. Asymptot. Anal. 2010, 66: 161-186.MathSciNetGoogle Scholar
Gilding BH: The Cauchy problem for $u_{t} = Δ u + {| \nabla u |}^{q}$ : large-time behaviour. J. Math. Pures Appl. 2005, 84: 753-785. 10.1016/j.matpur.2004.11.003MathSciNetView ArticleGoogle Scholar
Stinner C: Convergence to steady states in a viscous Hamilton-Jacobi equation with degenerate diffusion. J. Differ. Equ. 2010, 248: 209-228. 10.1016/j.jde.2009.09.019MathSciNetView ArticleGoogle Scholar
Barles G, Laurençot P, Stinner C: Convergence to steady states for radially symmetric solutions to a quasilinear degenerate diffusive Hamilton-Jacobi equation. Asymptot. Anal. 2010, 67: 229-250.MathSciNetGoogle Scholar
Laurençot P, Stinner C: Convergence to separate variable solutions for a degenerate parabolic equation with gradient source. J. Dyn. Differ. Equ. 2012, 24: 29-49. 10.1007/s10884-011-9238-xView ArticleGoogle Scholar
Mu CL, Li YH, Wang Y: Life span and a new critical exponent for a quasilinear degenerate parabolic equation with slow decay initial values. Nonlinear Anal., Real World Appl. 2010, 11: 198-206. 10.1016/j.nonrwa.2008.10.048MathSciNetView ArticleGoogle Scholar
Laurençot P, Stinner C: Refined asymptotics for the infinite heat equation with homogeneous Dirichlet boundary conditions. Commun. Partial Differ. Equ. 2011, 36: 532-546.View ArticleGoogle Scholar
Da Lio F: Comparison results for quasilinear equations in annular domains and applications. Commun. Partial Differ. Equ. 2002, 27: 283-323. 10.1081/PDE-120002788MathSciNetView ArticleGoogle Scholar
Kawohl B, Kutev N: Comparison principle and Lipschitz regularity for viscosity solutions of some classes of nonlinear partial differential equations. Funkc. Ekvacioj 2000, 43: 241-253.MathSciNetGoogle Scholar
Manfredi JJ, Vespri V: Large time behavior of solutions to a class of doubly nonlinear parabolic equations. Electron. J. Differ. Equ. 1994, 2: 1-17.MathSciNetGoogle Scholar
Guo JS, Guo YY: On a fast diffusion equation with source. Tohoku Math. J. 2001, 53: 571-579. 10.2748/tmj/1113247801MathSciNetView ArticleGoogle Scholar
Li YH, Mu CL: Life span and a new critical exponent for a degenerate parabolic equation. J. Differ. Equ. 2004, 207: 392-406. 10.1016/j.jde.2004.08.024MathSciNetView ArticleGoogle Scholar
Mukai K, Mochizuki K, Huang Q: Large time behavior and life span for a quasilinear parabolic equation with slowly decaying initial values. Nonlinear Anal. TMA 2000, 39: 33-45. 10.1016/S0362-546X(98)00161-8MathSciNetView ArticleGoogle Scholar

Copyright

This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited.