We discuss the conditions under which blow-up occurs for the solutions of discrete p-Laplacian parabolic equations on networks S with boundary ∂S as follows: ut(x,t)=Δp,ωu(x,t)+λ|u(x,t)|q-1u(x,t), (x,t)∈S×(0,+∞); u(x,t)=0, (x,t)∈∂S×(0,+∞); u(x,0)=u0≥0, x∈S¯, where p>1, q>0, λ>0, and the initial data u0 is nontrivial on S. The main theorem states that the solution u to the above equation satisfies the following: (i) if 0<p-1<q and q>1, then the solution blows up in a finite time, provided u¯0>ω0/λ1/q-p+1, where ω0:=maxx∈S∑y∈S¯ω(x,y) and u¯0:=maxx∈S u0(x); (ii) if 0<q≤1, then the nonnegative solution is global; (iii) if 1<q<p-1, then the solution is global. In order to prove the main theorem, we first derive the comparison principles for the solution of the equation above, which play an important role throughout this paper. Moreover, when the solution blows up, we give an estimate for the blow-up time and also provide the blow-up rate. Finally, we give some numerical illustrations which exploit the main results.
1. Introduction
In this paper, we discuss the blow-up property and global existence of solutions to the following discrete p-Laplacian parabolic equation:
(1)utx,t=Δp,wux,t+λux,tq-1ux,t,x,t∈S×0,+∞,ux,t=0,x,t∈∂S×0,+∞,ux,0=u0≥0,x∈S¯,
where q>0, p>1 and λ>0.
The continuous case of this equation has been studied by many authors, assuming some conditions q, p, and λ, in order to get a blow-up solution or global solution (see [1–5]). For example, they consider the case 1<p<2, q>0, and λ>0 in [2], the case q>p-1>1 in [5], the case p=q>2 in [1, 3], the case 1<p<2, q>0, and λ>0 in [4], respectively, and so on.
On the other hand, the long time behavior (extinction and positivity) of solutions to evolution p-Laplace equation with absorption on networks is studied in the paper [6, 7].
The goal of this paper is to give a condition on p, q, and λ for the solution to (1) to be blow-up or global. In fact, we prove the following as one of the main theorems.
Theorem 1.
Let u be a solution of (1). Then one has the following.
If 0<p-1<q and q>1, then the solution blows up in a finite time, provided u¯0>ω0/λ1/(q-p+1), where ω0:=maxx∈S∑y∈S¯ω(x,y) and u¯0:=maxx∈Su0(x).
If 0<q≤1, then the nonnegative solution is global.
If 1<q<p-1, then the solution is global.
In order to prove the above theorem, we give comparison principles for the solutions of (1) in Section 2. Moreover, when the solutions to (1) blow up, we derive the blow-up rate as follows:
(2)λq-1T-t-1/q-1≤maxx∈Sux,t≤q-1T-t2q-p/q-1λq-1T-t-ω0λq-p+1/q-1q-1T-t2q-p/q-1×q-1T-t2q-p/q-1-1/q-1,
where ω0:=maxx∈S∑y∈S¯ωx,y, and as a consequence
(3)limt→T-T-t1/q-1maxx∈S¯ux,t=1λq-11/q-1.
We organized this paper as follows. In Section 2, we discuss the preliminary concepts on networks and the discrete version of comparison principles on networks. In Section 3, we are devoted to find out blow-up conditions of the solution and the blow-up rate with the blow-up time. Finally, in Section 4, we give some numerical illustrations to exploit the main results.
2. Preliminaries and Discrete Comparison Principles
In this section, we start with some definitions of graph theoretic notions frequently used throughout this paper (see [8–10], for more details).
For a graph G=G(V,E), we mean finite sets V of vertices (or nodes) with a set E of two-element subsets of V (whose elements are called edges). The set of vertices and edges of a graph G are sometimes denoted by V(G) and E(G), or simply V and E, respectively. Conventionally, we denote by x∈V or x∈G the facts that x is a vertex in G.
A graph G is said to be simple if it has neither multiple edges nor loops, and G is said to be connected if, for every pair of vertices x and y, there exists a sequence (called a path) of vertices x=x0,x1,…,xn-1, xn=y, such that xj-1 and xj are connected by an edge (called adjacent) for j=1,…,n.
A graph S=S(V′,E′) is said to be a subgraph of G(V,E), if V′⊂V and E′⊂E.
A weight on a graph G is a function ω:V×V→[0,+∞) satisfying
ω(x,x)=0, x∈V,
ω(x,y)=ω(y,x) if x~y,
ω(x,y)=0 if and only if x≁y.
Here x~y means that two vertices x and y are connected (adjacent) by an edge in E. A graph associated with a weight is said to be a weight graph or a network.
For a subgraph S of a graph G(V,E), the (vertex) boundary ∂S of S is the set of all vertices z∈V∖S but is adjacent to some vertex in S; that is,
(4)∂S:=z∈V∖S∣z~yforsomey∈S.
By S¯, we denote a graph, whose vertices and edges are in both S and ∂S.
Throughout this paper, all subgraphs S and S¯ in our concern are assumed to be simple and connected.
For a function u:S¯→R, the discrete p-Laplacian Δp,ω on S is defined by
(5)Δp,ωu(x):=∑y∈S¯uy-uxp-2uy-uxω(x,y)
for x∈S.
The rest of this section is devoted to prove the comparison principle for the discrete p-Laplacian parabolic equation:
(6)utx,t=Δp,ωux,t+λux,tq-1ux,t,x,t∈S×0,+∞,ux,t=0,x,t∈∂S×0,+∞,ux,0=u0≥0,x∈S¯,
where λ>0, q>0, p>1, and the initial data u0 is nontrivial on S, in order to study the blow-up occurrence and global existence which we begin in the next section.
Now, we state the comparison principles and some related corollaries.
Theorem 2.
Let T>0 (T may be +∞), λ>0, q≥1, and p>1. Suppose that real-valued functions u(x,·),v(x,·)∈C[0,T) are differentiable in (0,T) for each x∈S¯ and satisfy
(7)utx,t-Δp,ωux,t-λux,tq-1u(x,t)≥vtx,t-Δp,ωvx,t-λvx,tq-1v(x,t),iiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiii(x,t)∈S×0,T,ux,t≥vx,t,(x,t)∈∂S×0,T,ux,0≥vx,0,x∈S¯.
Then u(x,t)≥v(x,t) for all (x,t)∈S¯×[0,T).
Proof.
Let T′>0 be arbitrarily given with T′<T. Then, by the mean value theorem, for each x∈S and 0≤t≤T′,
(8)ux,tq-1ux,t-vx,tq-1vx,t=qξx,tq-1ux,t-vx,t
for some ξ(x,t) lying between u(x,t) and v(x,t). Then it follows from (7) that we have
(9)utx,t-Δp,ωux,t-λqξx,tq-1ux,t≥vtx,t-Δp,ωvx,t-λqξx,tq-1vx,t
for all (x,t)∈S×(0,T′]. Let u~,v~:S¯×[0,T′]→R be the functions defined by
(10)u~x,t:=e-2λqLtux,t,v~x,t:=e-2λqLtvx,t,
where L:=maxr≤Mrq-1 and M:=maxx∈S¯,t∈[0,T′]ux,t,vx,t.
Then inequality (9) can be written as
(11)u~tx,t-v~tx,t-e2qλLp-2tΔp,ωu~x,t-Δp,ωv~x,t+λq2L-ξx,tq-1u~x,t-v~x,t≥0
for all (x,t)∈S×0,T′. Since S¯×[0,T′] is compact, there exists (x0,t0)∈S¯×[0,T′] such that
(12)(u~-v~)(x0,t0)=minx∈S¯min0≤t≤T′(u~-v~)x,t.
Then we have only to show that (u~-v~)(x0,t0)≥0. Suppose that (u~-v~)(x0,t0)<0, on the contrary. Since (u~-v~)(x,t)≥0 on both ∂S×[0,T′] and S¯×0, we have (x0,t0)∈S×0,T′. Then we have
(13)u~tx0,t0≤v~tx0,t0,Δp,ωu~x0,t0≥Δp,ωv~x0,t0.
Since ξq-1x,t≤maxr≤Mrq-1=L, we have
(14)2L-ξx0,t0q-1u~-v~x0,t0≤Lu~-v~x0,t0<0.
Combining (13) and (14), we obtain
(15)u~tx0,t0-v~tx0,t0-e2qλLp-2t0[Δp,ωu~x0,t0-Δp,ωv~x0,t0]+λq2L-ξx0,t0q-1[u~x0,t0-v~x0,t0]<0
which contradicts (11). Therefore, u~x,t-v~x,t≥0 for all (x,t)∈S×(0,T′] so that we get ux,t≥vx,t for all (x,t)∈S¯×0,T, since T′<T is arbitrarily given.
When p≥2, we obtain a strict comparison principle as follows.
Corollary 3 (strict comparison principle).
Let T>0 (T may be +∞), λ>0, q≥1, and p≥2. Suppose that real-valued functions u(x,·),v(x,·)∈C0,T are differentiable in (0,T) for each x∈S¯ and satisfy
(16)utx,t-Δp,ωux,t-λux,tq-1u(x,t)≥vtx,t-Δp,ωvx,t-λvx,tq-1v(x,t),iiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiii(x,t)∈S×0,T,ux,t≥vx,t,(x,t)∈∂S×0,T,ux,0≥vx,0,x∈S¯.
If u0(x*)>v0(x*) for some x*∈S, then ux,t>vx,t for all (x,t)∈S×(0,T).
Proof.
First, note that u≥v on S¯×0,T by Theorem 2. Let T′>0 be arbitrarily given with T′<T and let τ:S¯×[0,T′]→R be a function defined by
(17)τx,t:=ux,t-vx,t,x,t∈S¯×0,T′.
Then τ(x,t)≥0 for all (x,t)∈S¯×[0,T′]. Since τ(x*,0)>0 and ux*,tq-1u(x*,t)≥vx*,tq-1v(x*,t) for all 0<t≤T′, we obtain from inequality (16) that
(18)τtx*,t-Δp,ωux*,t-Δp,ωvx*,t≥0,
for all 0<t≤T′. Then, by the mean value theorem, for each y∈S¯ and t with 0≤t≤T′, it follows that
(19)uy,t-ux*,tp-2uy,t-ux*,t-vy,t-vx*,tp-2vy,t-vx*,t=p-1ηx*,y,tp-2τy,t-τx*,t,
and ηx*,y,t<2M, where M:=max0<t≤T′ux*,t,vx*,t.
Then inequality (18) gives
(20)τtx*,t≥∑y∈S¯p-1ηx*,y,tp-2τy,t-τx*,tωx*,y≥-dp-12Mp-2τx*,t,
where d=∑y∈S¯ωx*,y. This implies that
(21)τx*,t≥τx*,0e-dp-12Mp-2t>0,t∈(0,T′].
Now, suppose that there exists (x0,t0)∈S×0,T′ such that
(22)τx0,t0=minx∈S,0<t≤T′τx,t=0.
Then
(23)τtx0,t0≤0,Δp,ωux0,t0≥Δp,ωvx0,t0.
Hence, inequality (18) gives
(24)0≤τtx0,t0-Δp,ωux0,t0-Δp,ωvx0,t0≤0.
Therefore,
(25)Δp,ωux0,t0=Δp,ωvx0,t0;
that is,
(26)∑y∈S¯uy,t0-ux0,t0p-2uy,t0-ux0,t0iiiiiiiiiii-vy,t0-vx0,t0p-2iiiiiiiiii×uy,t0-ux0,t0p-2vy,t0-vx0,t0ωx0,y=0,
which implies that τy,t0=0 for all y∈S¯ with y~x0. Now, for any x∈S¯, there exists a path:
(27)x0~x1~⋯~xn-1~xn=x,
since S¯ is connected. By applying the same argument as above inductively, we see that τ(x,t0)=0 for every x∈S¯. This gives a contradiction to (21).
For the case 0<q<1, it is well known that (6) may not have unique solution, in general, and the comparison principle in usual form as in Theorem 2 may not hold. Instead, with a strict condition on the parabolic boundary, we obtain a similar comparison principle as follows.
Theorem 4.
Let T>0 (T may be +∞), λ>0, q>0, and p>1. Suppose that real-valued functions ux,·,vx,·∈C0,T are differentiable in 0,T for each x∈S¯ and satisfy
(28)utx,t-Δp,ωux,t-λux,tq-1u(x,t)≥vtx,t-Δp,ωvx,t-λvx,tq-1v(x,t),iiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiii(x,t)∈S×0,T,ux,t>vx,t,(x,t)∈∂S×0,T,ux,0>vx,0,x∈S¯.
Then ux,t≥vx,t for all (x,t)∈S×0,T.
Proof.
Let T′>0 and δ>0 be arbitrarily given with T′<T and 0<δ<min(x,t)∈Γ[u(x,t)-v(x,t)], respectively, where Γ:={(x,t)∈S¯×[0,T′]∣t=0orx∈∂S} (called a parabolic boundary).
Now, let a function τ:S¯×(0,T′]→R be a function defined by
(29)τx,t:=ux,t-vx,t-δ,x,t∈S¯×0,T′.
Then τ(x,t)>0 on Γ. Now, we suppose that minx∈S,0<t≤T′τ(x,t)<0. Then there exists (x0,t0)∈S×0,T′ such that
τ(x0,t0)=0,
τy,t0≥τx0,t0=0, y∈S,
τ(x,t)>0, (x,t)∈S×(0,t0).
Then
(30)τt(x0,t0)≤0
and
(31)Δp,ωux0,t0≥Δp,ωvx0,t0,
since
(32)uy,t0-ux0,t0≥vy,t0-vx0,t0.
Hence, (28) gives
(33)0≥τtx0,t0≥λux0,t0q-1ux0,t0-vx0,t0q-1vx0,t0=λvx0,t0+δq-1vx0,t0+δIIIIIII-vx0,t0q-1vx0,t0>0,
which leads to a contradiction. Hence, τ(x,t)≥0 for all (x,t)∈S×0,T′ so that we have u(x,t)≥v(x,t) for all (x,t)∈S×(0,T), since δ and T′ are arbitrary.
Using the same method as in Corollary 3, we obtain a strict comparison principle as follows.
Corollary 5 (strict comparison principle).
Let T>0 (T may be +∞), λ>0, q>0, and p≥2. Suppose that real-valued functions ux,·,vx,·∈C0,T are differentiable in 0,T for each x∈S¯ and satisfy
(34)utx,t-Δp,ωux,t-λux,tq-1u(x,t)≥vtx,t-Δp,ωvx,t-λvx,tq-1v(x,t),IIIIIIIIIIIIIIIIIIIIIIIIIIIIIII(x,t)∈S×0,T,ux,t>vx,t,x,t∈∂S×0,T,ux,0>vx,0,x∈S¯.
If u0(x*)>v0(x*) for some x*∈S, then ux,t>vx,t for all (x,t)∈S×0,T.
3. Blow-Up and Blow-Up Estimates
In this section, we discuss the blow-up phenomena of the solutions to discrete reaction-diffusion equation defined on networks, which is a main part of this paper.
We first introduce the concept of the blow-up as follows.
Definition 6 (blow-up).
One says that a solution u to an equation defined on a network S¯ blows up in finite time T, if there exists x∈S such that |u(x,t)|→+∞ as t↗T-.
According to the comparison principle in the previous section, we are guaranteed to get a solution to
(35)utx,t=Δp,ωux,t+λuqx,t,x,t∈S×0,∞,ux,t=0,x,t∈∂S×0,∞,ux,0=u0≥0,x∈S¯,
when p>1, q>0, λ>0, and the initial data u0 is nontrivial on S.
We now state the main theorem of this paper as follows.
Theorem 7.
Let u be a solution of (35). Then one has the following.
If 0<p-1<q and q>1, then the solution blows up in a finite time, provided u¯0>ω0/λ1/(q-p+1), where ω0:=maxx∈S∑y∈S¯ω(x,y) and u¯0:=maxx∈Su0(x).
If 0<q≤1, then the nonnegative solution is global.
If 1<q<p-1, then the solution is global.
Proof.
First, we prove (i). We note that u(x,t)≥0, for all x,t∈S¯×0,+∞, by Theorem 2. Assume that 0<p-1<q, q>1, and u¯0>ω0/λ1/(q-p+1), where u¯0:=maxx∈Su0(x). For each t>0, let xt∈S be a node such that uxt,t:=maxx∈Sux,t. In fact, we note that maxx∈Sux,t is differentiable, for almost all t>0. Then (35) can be written as follows:
(36)utxs,s=Δp,ωuxs,s+λuqxs,s=∑y∈S¯uy,s-ux,sp-2×uy,s-uxs,sωxs,y+λuqxs,s≥-ω0up-1xs,s+λuqxs,s
for almost all s>0. We need to show that maxx∈Sux,t>u¯0, for all t>0. Since u(x,t)≥0 on S×(0,∞) and
(37)lims→0+utx,t=lims→0+∑y∈S¯uy,s-uxs,sp-2iiiiiiiiiiiiiiiiiii×uy,s-uxs,sωxs,yiiiiiiiiiiiiiiiiiii∑y∈S¯+λuqxs,s≥lims→0+-ω0up-1xs,s+λuqxs,s=-ω0u¯0p-1+λu¯0q>0,u(xs,s) is increasing in some interval (0,s1). Suppose that there exists s>0 somewhere at which u(xs,s)≤u¯0. Then now take the interval (0,s1) to be maximal on which u(xs,s)>u¯0, s∈(0,s1), and u(xs,s)=u¯0. Then there exists s*∈(0,s1) such that ut(xs*,s*)<0 but
(38)0>utxs*,s*=∑y∈S¯uy,s*-uxs*,s*p-2×uy,s*-uxs*,s*ωxs*,s*+λuqxs*,s*≥-ω0up-1xs*,s*+λuqxs*,s*>0,
which leads to a contradiction. Thus it follows that u(xs,s)>u¯0, s∈(0,+∞).
Let F:[u¯0,+∞)→(0,Fu¯0] be a function defined by
(39)Fy:=∫y+∞ds-ω0sp-1+λsq<+∞,y≥u¯0.
We note that -ω0sp-1+λsq>0, for s≥u¯0, since u¯0>ω0/λ1/(q-p+1).
Then F is a decreasing continuous function from u¯0,+∞ onto 0,Fu¯0 with its inverse function G. Integrating (36) from 0 to t, we have
(40)t≤∫0tutxs,s-ω0up-1xs,s+λuqxs,sdu=∫u¯0uxt,tds-ω0sp-1+λsq.
This can be written as
(41)Fuxt,t≤Fu¯0-t
and, equivalently,
(42)uxt,t≥GFu¯0-t,
which implies that uxt,t blows up, as t→Fu¯0.
Secondly, we prove (ii). Consider the following ODE problem:
(43)ddtzt=λzqt,t>0,z0=u¯0+1.
Then, we have
(44)zt=1-qλt+z1-q01/1-q,q≠1,zt=z0eλt,q=1,
for every t≥0.
Take vx,t:=zt, for all x∈S¯ and t≥0. Then it is easy to see that v(x,t)>u(x,t), (x,t)∈∂S×(0,+∞), v(x,0)=z(0)>u¯0, x∈S¯, and
(45)vtx,t-Δp,ωvx,t-λvqx,t=ddtzt-λzqt=0.
Thus, 0≤u(x,t)≤v(x,t)=z(t) for every (x,t)∈S¯×(0,+∞) by Theorem 4. This implies that u must be global.
Finally, we prove (iii). Consider the following eigenvalue problem:
(46)-Δp,ωϕx=λ1ϕxp-2ϕx,x∈S,ϕx=0,x∈∂S.
Note that it is well known that λ1>0 and ϕx>0, for all x∈S (see [11, 12]).
Now, take v(x,t):=kϕ(x), x∈S¯, t≥0. Choosing k>0 so large that kϕ(x)>u¯0 and kϕ(x)>λ/λ11/(p-1-q), then we see that v(x,0)=kϕ(x)≥u0(x)=u(x,0), x∈S, and
(47)vtx,t-Δp,ωvx,t-λvqx,t=λ1kϕxp-1-λkϕxq≥0.
Therefore, 0≤u(x,t)≤v(x,t)=kϕ(x) for every (x,t)∈S¯×(0,+∞) by Theorems 2 and 4, which is required.
Remark 8.
(i) When the solution blows up in the above, the blow-up time T can be estimated as
(48)u¯01-qλ(q-1)≤T≤F(u¯0):=∫u¯0+∞ds-ω0sp-1+λsq.
In fact, the first inequality is derived as follows. By the definition of maximum function u(xt,t), (35) gives
(49)utxs,s≤λuqxs,s,
for almost all s>0. Then integrating both sides, we have
(50)t≥∫0tutxs,sλuqxs,sds=∫u¯0uxt,tdsλsq,
so that we obtain T≥∫u¯0+∞ds/λsq=u¯01-q/λ(q-1), by taking the limit as t→T-.
(ii) In the above, if u¯0:=maxx∈Su0(x) is not sufficiently large, then the solution may be global. This can be seen in the numerical examples in Section 4.
(iii) In the above, the case where 1<p-1=q was not discussed. As a matter of fact, the solution to (35) in this case may blow up or not, depending on the magnitude of the parameter λ. Each case is illustrated in Section 4. A full argument will be discussed in a forthcoming paper.
We now derive the lower bound, the upper bound, and the blow-up rate for the maximum function of blow-up solutions.
Theorem 9.
Let u be a solution of (35), which blows up at a finite time T, q>p-1>0, and q>1. Then one has the following.
The lower bound is
(51)maxx∈Su(x,t)≥λ(q-1)(T-t)-1/(q-1),0<t<T.
The upper bound is
(52)maxx∈Sux,t≤λ(q-1)(T-t)-α(T-t)(2q-p)/(q-1)-1/(q-1),IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII0<t<T,
where α:=ω0λ(q-p+1)/(q-1)(q-1)(2q-p)/(q-1) and ω0=maxx∈S∑y∈S¯ωx,y.
The blow-up rate is
(53)limt→T-T-t1/q-1maxx∈S¯ux,t=1λq-11/q-1,0<t<T.
Proof.
First, we prove (i). As in the previous theorem, let xt∈S be a node such that uxt,t:=maxx∈Sux,t, for each t>0. Then it follows from (35) that
(54)utxs,s≤λuqxs,s,
for almost all s>0. Then integrating from t to T, we get
(55)λT-t≥∫tTutxs,suqxs,sds=∫uxt,t+∞dssq=1q-1u1-q(xt,t).
Hence, we obtain
(56)u(xt,t)≥λ(q-1)(T-t)-1/(q-1),0<t<T.
Next, we prove (ii). Since the solution u is positive, we get
(57)utxs,s≥-∑y∈S¯up-1xs,sωxs,y+λuqxs,s≥-ω0up-1xs,s+λuqxs,s=uqxs,sλ-ω0up-1-qxs,s,
for almost all s>0 and ω0=maxx∈S∑y∈S¯ωx,y. Then, it follows from (i) (lower bound) that we have
(58)ut(xs,s)≥uqxs,sλ-ω0λ(q-1)(T-t)(q-p+1)/(q-1).
Integrating from t to T, we get
(59)uxt,t≤λ(q-1)(T-t)-αT-t(2q-p)/(q-1)-1/(q-1),iiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiii0<t<T,
where α:=ω0λ(q-p+1)/(q-1)q-1(2q-p)/(q-1).
Finally, (iii) can be easily obtained by (i) and (ii).
4. Examples and Numerical Illustrations
In this section, we show numerical illustrations to exploit our results in the previous section.
Now, consider a graph S={x1,…,x29} with the boundary ∂S={x30,x31} and the weight
(60)ωxi,xj=0.05,j=i+1,0.1,j=i+2,0.05,i=1,j=30,i=29,j=31,0,otherwise,
where i=1,…,28 (see Figure 1). Then, we note that ω0:=maxx∈S∑y∈S¯ω(x,y)=0.3.
For the graph S¯ (see Figure 1), consider q=3, p=2.5, λ=0.5, and the initial data u0 given by Table 1.
Initial data of u.
Node i
u0(xi)
1
0.7
2
0.9
3
1.1
4
1.3
5
1.5
6
1.3
7
1.1
8
0.9
9
0.7
10
0
11
0.7
12
0.9
13
1.1
14
1.3
15
1.5
16
1.3
17
1.1
18
0.9
19
0.7
20
0
21
0.7
22
0.9
23
1.1
24
1.3
25
1.5
26
1.3
27
1.1
28
0.9
29
0.7
30
0
31
0
Then 1<p-1=1.5<q=3 and maxx∈Su0(x)=1.5>ω0/λ1/q-p+1≒0.711. Then Figure 2 shows that the solution to (35) blows up and the computed blow-up time T is estimated as T≒0.4617817 and
(61)0.444≒∫1.5+∞ds0.5s3≤T≤∫1.5+∞ds-0.3s1.5+0.5s3≒0.553.
Behavior of each node for q=3 and p=2.5.
On the other hand, consider a small initial data u0 given by Table 2.
Initial data of u.
Node i
u0(xi)
1
0.002
2
0.004
3
0.006
4
0.008
5
0.01
6
0.008
7
0.006
8
0.004
9
0.002
10
0
11
0.002
12
0.004
13
0.006
14
0.008
15
0.01
16
0.008
17
0.006
18
0.004
19
0.002
20
0
21
0.002
22
0.004
23
0.006
24
0.008
25
0.01
26
0.008
27
0.006
28
0.004
29
0.002
30
0
31
0
Then maxx∈Su0(x)=0.01≯ω0/λ1/(q-p+1)≒0.711 and Figure 3 shows that the solution to (35) is global.
For the graph S¯ (see Figure 1), consider q=3, p=1.5, λ=0.1, and the initial data u0 given by Table 3.
Initial data of u.
Node i
u0(xi)
1
1
2
1.5
3
2
4
2.5
5
3
6
2.5
7
2
8
1.5
9
1
10
0
11
1
12
1.5
13
2
14
2.5
15
3
16
2.5
17
2
18
1.5
19
1
20
0
21
1
22
1.5
23
2
24
2.5
25
3
26
2.5
27
2
28
1.5
29
1
30
0
31
0
Then 0<p-1=0.5<q=3 and maxx∈Su0(x)=3>ω0/λ1/(q-p+1)≒1.55. Then Figure 4 shows that the solution to (35) blows up and the computed blow-up time T is estimated as T≒0.5864884 and
(62)0.555≒∫3+∞ds0.1s3≤T≤∫3+∞ds-0.3sp-1+0.1sq≒0.610.
For the graph S¯ (see Figure 1), consider q=1.5, p=3, λ=0.1, and the initial data u0 given by Table 3 in Example 2. Then 1<q=1.5<p-1=2 and Figure 5 shows that the solution to (35) is global.
For the graph S¯ (see Figure 1), consider q=0.5, p=3, λ=0.1, and the initial data u0 given by Table 3 in Example 2. Then 0<q=0.5≤1 and Figure 6 shows that the solution to (35) is global.
For the graph S¯ (see Figure 1), consider q=2, p=3, λ=2, and the initial data u0 given by Table 3 in Example 2. Then 1<q=p-1=2 and Figure 7 shows that the solution to (35) blows up.
On the contrary, when λ=0.00001, the solution to (35) is global, as seen in Figure 8.
Behavior of each node for q=2 and p=3.
Behavior of each node for q=2 and p=3.
5. Conclusion
We discuss the conditions under which blow-up occurs for the solutions of discrete p-Laplacian parabolic equations on networks S with boundary ∂S:
(63)ut(x,t)=Δp,ωu(x,t)+λux,tq-1u(x,t),iiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiix,t∈S×0,+∞,ux,t=0,(x,t)∈∂S×(0,+∞),u(x,0)=u0≥0,x∈S¯,
where p>1, q>0, λ>0, and the initial data u0 is nontrivial on S.
The main theorem states that the solution u to the above equation satisfies the following:
if 0<p-1<q and q>1, then the solution blows up in a finite time, provided u¯0>ω0/λ1/(q-p+1), where ω0:=maxx∈S∑y∈S¯ω(x,y) and u¯0:=maxx∈Su0(x);
if 0<q≤1, then the nonnegative solution is global;
if 1<q<p-1, then the solution is global.
In addition, we give an estimate for the blow-up time and the blow-up rate for the blow-up solution. Finally, we give some numerical illustrations which exploit the main results.Conflict of Interests
The authors declare that there is no conflict of interests regarding the publication of this paper.
Acknowledgments
This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korean Government (MOE) (no. 2012R1A1A2004689) and Sogang University Research Grant of 2014 (no. 201410044).
FujiiA.OhtaM.Asymptotic behavior of blowup solutions of a parabolic equation with the p-LaplacianLiuW.A note on blow-up for fast diffusive p-Laplacian with sourcesLiY.XieC.Blow-up for p-Laplacian parabolic equationsYinJ.JinC.Critical extinction and blow-up exponents for fast diffusive p-Laplacian with sourcesZhaoJ.LiangZ.Blow-up rate of solutions for p-Laplacian equationChungY.-S.LeeY.-S.ChungS.-Y.Extinction and positivity of the solutions of the heat equations with absorption on networksLeeY.-S.ChungS.-Y.Extinction and positivity of solutions of the p-Laplacian evolution equation on networksChungF. R. K.ChungS.-Y.BerensteinC. A.ω-harmonic functions and inverse conductivity problems on networksCvetkovicD. M.DoobM.SachsH.ParkJ.-H.ChungS.-Y.Positive solutions for discrete boundary value problems involving the p-Laplacian with potential termsParkJ.-H.ChungS.-Y.The Dirichlet boundary value problems for p-Schrödinger operators on finite networks