PROCEEDINGS OF THE

AMERICAN MATHEMATICAL SOCIETY

Volume 145, Number 12, December 2017, Pages 5235–5245
http://dx.doi.org/10.1090/proc/13647
Article electronically published on June 16, 2017

EXISTENCE AND UNIQUENESS OF SINGULAR SOLUTIONS

OF p-LAPLACIAN WITH ABSORPTION FOR DIRICHLET
BOUNDARY CONDITION

NGUYEN ANH DAO AND JESUS ILDEFONSO DÍAZ

(Communicated by Catherine Sulem)

Abstract. In this paper, we consider the existence and uniqueness of singular

solutions of degenerate parabolic equations with absorption for zero homoge-
neous Dirichlet boundary condition. Moreover, we also get some estimates of
the short time behavior of singular solutions.

1. Introduction
Given a smooth bounded Ω ⊂ RN , we study the existence and uniqueness of
singular solutions of the following equation:

∂t u − Δp u + uq = 0, in Ω × (0, ∞),
(1)
u = 0, on ∂Ω × (0, ∞),
where Δp u = div(|∇u|p−2 ∇u), with p > 2, q > 1. The singular solutions of (1) refer
to the fundamental solutions, the very singular solutions, and the large solutions.
Throughout this paper we assume, without loss of generality, that 0 ∈ Ω, and such
a fundamental solution and a VSS have the singularity at (0, 0).
By a fundamental solution of (1), we mean a continuous nonnegative function
u(x, t) satisfying (1) in the sense of distribution. Moreover,
(2) u(x, 0) = 0, ∀x = 0,
and there is a finite c > 0 such that
(3) lim u(x, t) = cδ0 ,
t→0
where δ0 is a Dirac mass concentrated at 0. Since a VSS is more singular than any
fundamental solution, then it also satisfies (2), and

(4) lim u(x, t) = +∞;
t→0 Ω
see e.g. [4], [17], [15], [14], [16].
It is well known that both kinds of solutions play a crucial role in studying the
long time behavior of more general solutions; see [15], [14], and references therein.

Received by the editors August 26, 2016 and, in revised form, January 2, 2017.
2010 Mathematics Subject Classification. Primary 35K65, 35K15.
Key words and phrases. Degenerate parabolic equations, large solution, very singular solution,
Dirac measure.
This work was partially supported by ITN FIRST of the Seventh Framework Program of the
European Community (grant agreement number 238702).

2017
c American Mathematical Society

5235
5236 NGUYEN ANH DAO AND JESUS ILDEFONSO DÍAZ

In spite of a large amount of papers dealing with VSS, as far as we know, most
of them consider the case Ω = RN . For instance, the existence of a self-similar VSS
for the Cauchy problem associated to (1),
(5) ∂t u − Δp u + uq = 0, in RN × (0, ∞),
was proved by Peletier and Wang, [17], provided that p − 1 < q < p − 1 + Np .
The uniqueness of VSS required a different type of argument and it was proved
by Kamin and Vazquez, [15] (see also Diaz and Saa [10] for a general quasilinear
parabolic equation). In fact, these papers utilized the self-similarity, or scaling
argument in RN × (0, ∞) in order to look for a solution, which is of the form:
−1 p(q − 1)
(6) W (x, t) = t q−1 f (η), η = |x|t−1/β , β= .
q+1−p
In order that W is a VSS, f must satisfy the ODE:
⎧  p−2   N −1  p−2  
⎨ (|f | f ) + η |f | f + β ηf + q−1 f − f = 0, in (0, ∞),
1 1
⎪ q

(7) f (η) ≥ 0, on [0, ∞),

⎪
⎩ f  (0) = 0, lim η p/(q+1−p) f (η) = 0.
η→∞

Moreover, it was shown that


> 0, for 0 ≤ η < η0 ,
f (η)
= 0, for η ≥ η0 ,
(see also the treatment for 1 < p < 2 in [5], [6]).
Concerning the uniqueness of VSS of (1), as far as we know, it has not been
proved yet. Thus, our main result is as follows:
Theorem 1. Let p > 2, and p − 1 < q < p − 1 + Np . Then, there exists a unique
VSS, u(x, t) of (1). Furthermore, we have the short time behavior of u as t → 0 at
the singular point x = 0:
1
(8) lim t q−1 u(0, t) = f (0).
t→0

Remark 2. The result (8) implies that the short time behavior of VSS for any
bounded domain is similar to the one in RN (compare to W (x, t) in (6)).
Clearly, the argument of the proof of the uniqueness of VSS in RN × (0, ∞)
based on the self-similarity of solutions is not applicable to such a bounded domain
Ω in RN . To prove Theorem 1, we show that there exist a minimal VSS and a
maximal VSS. And both solutions are equal. Note that the existence of a minimal
VSS is well known. This one is the convergence of the nondecreasing sequence of
fundamental solutions and we use the large solutions to construct a maximal VSS.
Thus, it is convenient for us to introduce the large solution in what follows.
In [7], M. Crandall, P. Lions, and P. Souganidis considered nonnegative solutions
of the equation:

∂t u − Δu + |∇u|q = 0, in Ω × (0, ∞),
(9)
u = 0, on ∂Ω × (0, ∞),
with unbounded initial data of the form

+∞, in D,
(10) u(0) =
0, in Ω\D,
 EXISTENCE AND UNIQUENESS OF SINGULAR SOLUTIONS 5237

where D is an open subset of Ω. The initial data is comprehended as follows:

u(x, t) → +∞, for any x ∈ D, and u(x, t) → 0, for any x ∈ Ω\D.
This problem is motivated by studying the theory of large deviations of Markov
diffusion processes. The authors showed that there is a unique solution to problem
(9)–(10) when q > 1. Such a solution with initial data (10) is called a large solution.
Many other references involve the study of such types of unbounded initial data
(see, e.g., [1], [2], [3], and references therein). Roughly speaking, a large solution
is more singular than any VSS. Inspired by their works, and also for our purposes
later, to prove the uniqueness of VSS, we shall prove the existence and uniqueness
of large solutions of problem (1). Moreover, the applications of the study of large
solutions applies even to some unexpected contexts such as Control Theory (see
[9]).

Theorem 3. Let p > 2, and q > p − 1. Then, there is a unique large solution of
(1).

The paper is organized as follows: We will give some definitions, and preliminary
results in the next section. Section 3 is devoted to proving the existence and
uniqueness of large solutions. Finally, we prove the existence and uniqueness of
VSS in Section 4.

2. Some definitions and preliminary results

Notation. We denote by B(x, r) the open ball with center at x and radius r > 0.
We also denote by C 0,β (Ω), the β-Hölder continuous space, for β ∈ (0, 1].

Let us define a weak solution of problem (1).

Definition 4. u is called a weak solution of (1) if u ∈ Lploc (0, ∞; W01,p (Ω))∩L∞

loc (Ω×
(0, ∞)) satisfies (1) in the sense of distribution.

Next, we recall some results for the existence, uniqueness, and regularity of weak
solutions of the following problem:
⎧
⎨ ∂t v − Δp v = f, in Ω × (0, ∞),
(11) v = 0, ∂Ω × (0, ∞),
⎩
v(x, 0) = v0 , Ω.

Theorem 5. Let p > 2, f ∈ L∞ (Ω × (0, ∞)), and v0 ∈ C 0,1 (Ω). Then, there exists
a weak bounded solution u of (1). Moreover, there is a positive constant β ∈ (0, 1)
such that
 p−2
β
1
(12) |u(x, t) − u(y, s)| ≤ CuL∞ (Ω×(0,T )) |x − y| + uL∞ (Ω×(0,T )) |t − s|
p p ,

for (x, t), (y, s) ∈ Ω × [0, T ], and C > 0 is a constant not depending on u.

Proof. We skip the proof, and refer to Theorem 1.2, [11] (see also [12], [13] for more
regularity results). 

Next, we consider the large solutions of (1).

5238 NGUYEN ANH DAO AND JESUS ILDEFONSO DÍAZ

3. The existence and uniqueness of large solutions

We prove Theorem 3.
Proof. (i) Existence
For any n ≥ 1, let us put

 1
Dn = {x ∈ D : dist x, Ω\D > )}
n
and construct a nondecreasing sequence of functions ψn ∈ C 0,1 (Ω) such that

n, if x ∈ Dn ,
ψn =
0, if x ∈ Ω\D.
Now, we consider the following equation:
⎧
⎨ ∂t un − Δp un + uqn = 0, in Ω × (0, ∞),
(13) un = 0, ∂Ω × (0, ∞),
⎩
un (x, 0) = ψn (x), Ω.
Thanks to Theorem 5, there exists a unique bounded solution un ∈ C(Ω × [0, ∞))
of (13).
Clearly, z(t) = (q − 1)− q−1 t1−q is a solution of the ODE:
1

 
z (t) + z q (t) = 0, t > 0,
z(0) = +∞,
By the comparison principle, we get
(14) un (x, t) ≤ z(t), ∀(x, t) ∈ Ω × (0, ∞).
We observe that {un }n≥1 is nondecreasing. Thus, there is a function u such that
un ↑ u. The classical argument and regularity result imply that u is a weak contin-
uous solution of (1).
Then, it remains to show that u(0) fulfills condition (10). Indeed, for any x ∈ D,
there is a natural number Nx ∈ N such that x ∈ Dn , ∀n ≥ Nx . Therefore, the
monotonicity of {un }n≥1 yields
lim inf u(x, t) ≥ lim inf un (x, t) = n.
t→0 t→0
The last inequality holds for any n ≥ Nx , thereby proving u(x, 0) = ∞, for x ∈ D.
Next, we show that u(t) converges to 0 in Ω\D as t → 0. For any y ∈ Ω\D, let

−Δp α(x) = 1, B(y, r),
(15)
α = 0, ∂B(y, r),
where r > 0 is small enough such that B(y, r) ⊂ Ω\D.
Put 1
w(x, t) = λeCt e α(x) ,
for any λ > 0, and C > 0 is chosen later such that
(16) ∂t w − Δp w + wq ≥ 0.
If this is done, since w = ∞ on the boundary ∂B(y, r) and un (x, 0) = 0 in B(y, r),
then the comparison principle deduces
(17) un (x, t) ≤ w(x, t), ∀(x, t) ∈ B(y, r) × (0, ∞),
hence
u(x, t) ≤ w(x, t), ∀(x, t) ∈ B(y, r) × (0, ∞).
 EXISTENCE AND UNIQUENESS OF SINGULAR SOLUTIONS 5239

This implies that

1
u(x, 0) ≤ λe α(x) ,
and the conclusion follows as λ → 0.
Now, we demonstrate (16). Indeed, computation yields

Δp α
1 2 
wt = Cw, −Δp w = −wp−1 − 2(p−1) + (p − 1)|∇α|p 2p + 2p−1 ,
α α α
so

1 1 2
wt −Δp w +w = Cw +w
q p−2
w q−(p−1)
− 2(p−1) − (p − 1)|∇α| ( 2p + 2p−1 ) .
p
α α α
We observe that |∇α| is bounded, while w(x, t) → ∞ faster than α−2p as x →
∂B(y, r). Thus, there exists a real positive number δ > 0 such that

1 1 2
w q−(p−1)
− 2(p−1) − (p − 1)|∇α| ( 2p + 2p−1 ) > 0, on {r−δ < |x−y| ≤ r}.
p
α α α
It is important to note that we can choose δ being independent of C. It remains to
take C = C(λ) > 0 large enough such that

1 1 2
Cw + wp−2 wq−(p−1) − 2(p−1) − (p − 1)|∇α|p ( 2p + 2p−1 ) > 0,
α α α
on {|x − y| ≤ r − δ}.
In brief, we get (16), likewise the existence of large solution follows.
(ii) Uniqueness
To prove the uniqueness, we use the scaling argument as in [7]. For any η > 0,
we set
q−(p−1)
uη (x, t) = ηu(η p−1 x, η q−1 t).
Clearly, if u is a large solution of (1) with respect to (Ω, D), then uη is a large
q−(p−1) q−(p−1)
solution of (1) with respect to (η p−1 Ω, η p−1 D).
By the routine argument, we obtain for any large solution v of (1) that
(18) uη (x, t) ≥ v(x, t) ≥ uη (x, t), ∀(x, t) ∈ Ω × (0, ∞),

for any η > 1 > η > 0.
By the continuity of u, we can pass to the limit as η → 1+ , η  → 1− in (18) to
get
u = v, in Ω × (0, ∞).
This concludes the proof of Theorem 3. 

4. The existence and uniqueness of a very singular solution

and its short time behavior
4.1. The existence of a minimal VSS. To construct a minimal VSS, our ar-
gument needs the uniqueness of fundamental solutions of (1). Then, we have the
following lemma.
Lemma 6. Let p > 2, and 0 < q < p − 1 + p
N. Then, there exists a unique
fundamental solution of (1).
5240 NGUYEN ANH DAO AND JESUS ILDEFONSO DÍAZ

Proof. The existence of fundamental solutions is well known. We then focus on the
uniqueness. It suffices to prove for the fundamental solution with initial data δ0 .
Let E be a unique solution of the equation (see [15]):

∂t u − Δp u + uq = 0, in RN × (0, ∞),
(19)
u(0) = δ0 .
We first claim that if v is a fundamental solution of (1) with initial Dirac mass δ0 ,
then
(20) v(x, t) ≤ E(x, t), ∀(x, t) ∈ Ω × (0, ∞).
Indeed, for any τ > 0, let

v(x, τ ), in Ω,
Eτ (0) =
0, outside Ω.
We consider the following problem:

∂t u − Δp u + uq = 0, in RN × (0, ∞),
(21)
u(0) = Eτ (0).
Thanks to the classical result, this equation possesses a unique solution Eτ (x, t)
(see e.g. [11]). Moreover, the strong comparison principle deduces:
v(x, t + τ ) ≤ Eτ (x, t), ∀(x, t) ∈ Ω × (0, ∞).
Clearly, Eτ (x, t) is bounded by the barrier function z(t) for all τ > 0. Then, there
exists a nonnegative function E such that Eτ (x, t) → E(x, t) as τ → 0, uniformly
on any compact set in RN × (0, ∞). Moreover, E is a weak solution of equation
(19), and
v(x, t) ≤ E(x, t), ∀(x, t) ∈ Ω × (0, ∞).
If we can prove E(0) = δ0 , it follows then from the uniqueness of the fundamental
solution of (19) that E = E, hence we get
claim (20). 
Indeed, using the test function ϕ ∈ Cc∞ RN \{0} for (21), we have
(22)
  t 


Eτ (x, t)ϕdx + |∇Eτ |p−2 ∇Eτ · ∇ϕ + Eτq ϕ dxds = Eτ (x, 0)ϕdx
RN RN RN
0

= v(x, τ )ϕdx.
Ω
Passing to the limit as τ → 0 yields:
  t

q 
E(x, t)ϕdx + |∇E|p−2 ∇E · ∇ϕ + E τ ϕ dxds = 0.
RN 0 RN
After that, letting t → 0 deduces:



(23) lim E(x, t)ϕdx = 0, ∀ϕ ∈ Cc∞ RN \{0} .
t→0 RN

N
Next, using the test function ψ ∈ C∞
c R instead of ϕ in (22), and passing τ → 0
yields
  t 

q 
E(x, t)ψdx+ |∇E|p−2 ∇E ·∇ψ+E τ ψ dxds = lim v(x, τ )ψdx = ψ(0),
RN 0 RN τ →0 Ω
 EXISTENCE AND UNIQUENESS OF SINGULAR SOLUTIONS 5241

which implies

(24) lim E(x, t)ψdx = ψ(0).
t→0 RN

By (24) and (23), we get E(0) = δ0 , so claim (20) follows.

Now, let v1 and v2 be two fundamental solutions of (1). Note that the well-known
L1 -contraction principle does not hold for such a bounded domain in general. Thus,
we use the L1 -contraction principle for the following truncation:
(25) S1 (v1 − v2 )(t)L1 (Ω) ≤ S1 (v1 − v2 )(s)L1 (Ω) , for 0 < s < t < ∞,
where  r 
s, if |s| ≤ k,
Sk (r) = Tk (s)ds, and Tk (s) =
0 k sign(s), if |s| > k,
for any k > 0. Note that inequality (25) can be obtained by using the test function
T1 (v1 − v2 ) to the difference between the equations satisfied by v1 and v2 .
Next, we note that 0 ≤ S1 (r) ≤ |r|, for any r ∈ R. It then follows from (25) and
the triangle inequality that
(26)
S1 (v1 −v2 )(t)L1 (Ω) ≤ (v1 −v2 )(s)L1 (Ω) ≤ (v1 − E)(s)L1 (Ω) +(E −v2 )(s)L1 (Ω)
 



= E(x, s) − v1 (x, s) dx + E(x, s) − v2 (x, s) dx.
Ω Ω
Passing s → 0 in (26) yields
S1 (v1 − v2 )(t)L1 (Ω) = 0, for 0 < t < ∞,
which implies v1 (x, t) = v2 (x, t), in Ω × (0, ∞). This concludes the proof of Lemma
6. 
Now, let vcΩ be the unique solution of (1) with initial Dirac cδ0 , for c > 0. Clearly,
{vcΩ }c>0 is the nondecreasing sequence, and it is bounded by z(t). Thus, there is a
Ω
function, say vmin , such that vcΩ converges to vmin
Ω
as c → ∞. We will show that
Ω
vmin is the minimal VSS. It is equivalent to proving that for any VSS u(x, t) of (1),
it holds that
(27) u(x, t) ≥ vcΩ (x, t), ∀(x, t) ∈ Ω × (0, ∞),
for any c > 0.
(c)
To prove (27), we wish to construct a sequence {un }n≥1 such that
⎧ (c)
⎪
⎪ u (x) ≤ u(x, n1 ), ∀x ∈ Ω,
⎪ n
⎪
⎪
⎨
(c)
(28) un L1 (Ω) = c,
⎪
⎪
⎪
⎪
⎪
⎩ lim u(c) = cδ0 .
n
n→∞

Indeed,
 since limt→0 Ω u(x, t)dx = ∞, there is a natural number n0 ∈ N, such that
n )dx > c, for any n ≥ n0 . Moreover, u(., n ) is a continuous function in Ω.
1 1
Ω
u(x,
Then, there is a positive number h = h(n, c) such that

1 
Th u(., ) L1 (Ω) = c.
n
(c)

Thus, un = Th u(., n1 ) is a desired function satisfying (28).
5242 NGUYEN ANH DAO AND JESUS ILDEFONSO DÍAZ

Consider the following equation:

⎧
⎨ ∂t v − Δp v + v = 0, in Ω × (0, ∞),
q

(29) v = 0, on ∂Ω × (0, ∞),

⎩ (c)
v(0) = un , in Ω.
By the classical result, there exists a unique solution of (29), say vn,c (x, t). Thanks
to the strong comparison principle, we get
1
vn,c (x, t) ≤ u(x, t + ), ∀(x, t) ∈ Ω × (0, ∞).
n
It is not difficult to observe that vn,c (x, t) converges to vcΩ (x, t) as n → ∞, the
unique fundamental solution in Lemma 6. Thus, conclusion (27) follows. In other
Ω
words, vmin is a minimal VSS.

Remark 7. By the construction, the sequence {vmin

BR
}R>0 is nondecreasing, and it
converges to V as R → ∞, a self-similar VSS of equation (1) in RN × (0, ∞).
4.2. The existence of a maximal VSS. We have the following result.
Theorem 8. Let p > 2, and p − 1 < q < p − 1 + Np . Then, there is a maximal VSS
of (1).
Proof. Let uε be a unique large solution of problem (1) with initial data

+∞, in B(0, ε),
u(0) =
0, in Ω\B(0, ε).
It is clear that {uε }ε>0 is a nondecreasing sequence. Then, there is a function
vmax such that uε ↓ vmax as ε → 0, which is a weak solution of (1). Moreover, vmax
is continuous on any compact of Ω × [0, ∞)\{(0, 0)} by Theorem 5.
Now, we show that vmax is a maximal VSS.
Indeed, for any x ∈ Ω\{0}, there is a real number εx > 0 such that uε (x, t) → 0
as t → 0, for any ε ∈ (0, εx ). It follows then from the monotonicity of {uε }ε>0 that
vmax (x, 0) = 0. Or vmax fulfills (2).
It remains to prove that for any VSS v of (1), it holds that
(30) v ≤ uε , in Ω × (0, ∞), for any ε > 0.
On the one hand, proceeding as the proof of (17) yields, for any τ > 0,
1
v(x, τ ) ≤ λeCτ e α(x) , ∀x ∈ Ω, |x| ≥ ε/2,
where α(x) is the solution of (15) in B(x, ε/4). Thus,
(31) v(x, τ ) ≤ mε λeCτ , ∀x ∈ Ω, |x| ≥ ε/2,
1
with mε = supy∈B(x,ε/4) {e }.
α(y)

On the other hand, since uε (x, t) → +∞ uniformly on any compact of B(0, ε)

as t → 0, there is a time sτ = s(τ ) > 0 such that
(32) v(x, τ ) ≤ uε (x, s), for any x ∈ B(0, ε/2), ∀s ∈ (0, sτ ).
A combination of (31) and (32) deduces:
v(x, τ ) ≤ mε λeCτ + uε (x, s), for any x ∈ Ω, ∀s ∈ (0, sτ ).
 EXISTENCE AND UNIQUENESS OF SINGULAR SOLUTIONS 5243


Clearly, mε λeCτ + uε (., . + s) is the super solution of (1). Therefore, the strong
comparison principle yields
(33) v(x, t + τ ) ≤ mε λeCτ + uε (x, t + s), ∀(x, t) ∈ Ω × (0, ∞).
Letting s → 0 in (33) gives us
v(x, t + τ ) ≤ mε λeCτ + uε (x, t), ∀(x, t) ∈ Ω × (0, ∞).
The last inequality holds for any τ > 0, so we obtain, after τ → 0,
v(x, t) ≤ mε λ + uε (x, t), ∀(x, t) ∈ Ω × (0, ∞).
Finally, passing λ → 0 yields conclusion (30), or we get Theorem 8. 
Remark 9. We denote by vmaxΩ
, the maximal VSS of equation (1) in Ω × (0, ∞). By
the construction, the sequence {vmax
BR
}R>0 is nondecreasing. Note that this sequence
is also bounded by z(t). Thus,
(34) BR
vmax (x, t) ↑ W (x, t),
for any (x, t) ∈ RN × (0, ∞), as R → ∞. It is not difficult to verify that W is a
self-similar VSS of the Cauchy problem associated with (1) in RN × (0, ∞).
Now, we complete the proof of Theorem 1.
Since W and V are two self-similar solutions of the Cauchy equation (1), they
must satisfy (7). It follows from the uniqueness result of (7) (see [15]) that
(35) W = V, in RN × (0, ∞).
Next, we claim that for any c > 0,
(36) vcBR (x, t) ≤ vcΩ (x, t) + mε λeCt , in BR × (0, ∞),
for any R > 0 large enough such that Ω ⊂⊂ B(0, R), and mε is in (31). Recall
that vcΩ (resp. vcBR ) is the unique solution of equation (1) with initial data cδ0 in
Ω × (0, ∞) (resp. B(0, R) × (0, ∞)).
Indeed, we first observe that (vcΩ + mε λeCt ) is a super-solution of equation (1)
in Ω × (0, ∞). Moreover, using the same analysis as (31) yields
(37) vcBR (x, t) ≤ mε λeCt , for x ∈ B(0, R), |x| ≥ ε/2, and t > 0.
Note that this inequality can be done by using a smoothing effect to the initial data
and the uniqueness result in Lemma 6.
In particular, we get
vcBR (x, t) ≤ mε λeCt , ∀x ∈ ∂Ω.
Thus, the comparison principle (after a smoothing effect to initial data and the
uniqueness result) deduces
(38) vcBR ≤ vcΩ + mε λeCt , in Ω × (0, ∞).
Thus, claim (36) follows from (37) and (38).
Next, passing to the limit as c → ∞ in (36) yields
(39) BR
vmin ≤ vmin
Ω
+ mε λeCt , in BR × (0, ∞).
Letting R → ∞ in (39) deduces:
(40) W = V ≤ vmin
Ω
+ mε λeCt , in RN × (0, ∞).
5244 NGUYEN ANH DAO AND JESUS ILDEFONSO DÍAZ

By combining (34) and (40), we obtain

(41) Ω
vmin ≤ vmax
Ω
≤ W ≤ vmin
Ω
+ mε λeCt , in Ω × (0, ∞).
Thanks to the L1 -contraction in (25), we have, for any t > s > 0,
(42)

Ω 
Ω 
S1 vmax
Ω
− vmin
Ω
(t)L1 (Ω) ≤ S1 vmax − vmin
Ω
(s)L1 (Ω) ≤  vmax − vmin
Ω
(s)L1 (Ω)
It follows from (41) and (42) that
 

Ω 
S1 vmax (t) − vmin (t) dx ≤
Ω
mε λeCs dx = |Ω|mε λeCs .
Ω Ω

Passing s → 0 in the last inequality yields



Ω 
S1 vmax (t) − vmin
Ω
(t) dx ≤ |Ω|mε λ.
Ω

Then, the uniqueness result follows when λ → 0.

To end this part, we prove the short time behavior for the unique VSS u of
equation (1).
Thanks to (41) and the uniqueness, we have
−1
u(0, t) ≤ W (0, t) = t q−1 f (0) ≤ u(0, t) + mε λeCt , ∀t > 0;
or
1 1 1
(43) t q−1 u(0, t) ≤ f (0) ≤ t q−1 u(0, t) + λmε t q−1 eCt .
The result follows by passing t → 0 in (43).
Consequently, we have
Corollary 10. Let uL be the unique large solution of equation (1). Then, uL (x, t)
−1
has the rate t q−1 as t → 0 for any x ∈ D.
Proof. It is sufficient to show that the result holds for x = 0 ∈ D. Let u be the
unique VSS of equation (1). Then, we have
−1 −1
u(0, t) ≤ uL (0, t) ≤ (q − 1) q−1 t q−1 .
This inequality and (43) lead to the conclusion. 

Remark 11. Our argument in this paper can be applied to obtain the same results
for the porous medium equation

∂t u − Δ(um ) + uq = 0, in Ω × (0, ∞),
u = 0, on ∂Ω × (0, ∞);
see our forthcoming paper [8].

Acknowledgement
The authors want to thank Professor S. Kamin for her valuable comments. The
first author would like to express his gratitude to the Instituto de Matemática
Interdisciplinar, Universidad Complutense de Madrid, Spain for their hospitality
and support during his visit.
 EXISTENCE AND UNIQUENESS OF SINGULAR SOLUTIONS 5245

References
[1] C. Bandle, G. Dı́az, and J. I. Dı́az, Solutions d’équations de réaction-diffusion non linéaires
explosant au bord parabolique (French, with English and French summaries), C. R. Acad. Sci.
Paris Sér. I Math. 318 (1994), no. 5, 455–460. MR1267826
[2] M. F. Bidaut-Veron and N. A. Dao, Isolated initial singularities for the viscous Hamilton-
Jacobi equation, Adv. Differential Equations 17 (2012), no. 9-10, 903–934. MR2985679
[3] M.-F. Bidaut-Véron and N. A. Dao, Initial trace of solutions of Hamilton-Jacobi parabolic
equation with absorption, Adv. Nonlinear Stud. 15 (2015), no. 4, 889–921, DOI 10.1515/ans-
2015-0408. MR3405821
[4] H. Brezis, L. A. Peletier, and D. Terman, A very singular solution of the heat equation with
absorption, Arch. Rational Mech. Anal. 95 (1986), no. 3, 185–209, DOI 10.1007/BF00251357.
MR853963
[5] X. Chen, Y. Qi, and M. Wang, Self-similar singular solutions of a p-Laplacian evolution equa-
tion with absorption, J. Differential Equations 190 (2003), no. 1, 1–15, DOI 10.1016/S0022-
0396(02)00039-6. MR1970953
[6] X. Chen, Y. Qi, and M. Wang, Singular solutions of parabolic p-Laplacian with absorption,
Trans. Amer. Math. Soc. 359 (2007), no. 11, 5653–5668, DOI 10.1090/S0002-9947-07-04336-
X. MR2327046
[7] M. G. Crandall, P.-L. Lions, and P. E. Souganidis, Maximal solutions and universal bounds
for some partial differential equations of evolution, Arch. Rational Mech. Anal. 105 (1989),
no. 2, 163–190, DOI 10.1007/BF00250835. MR968459
[8] N. A. Dao, Uniqueness of very singular solution of nonlinear degenerate parabolic equations
with absorption for Dirichlet boundary condition, Elec. Jour. Diff. Equa., 2016 (2016), no.
299, 1–8.
[9] J. I. Diaz, Obstruction and some approximate controllability results for the Burgers equation
and related problems, In Control of partial differential equations and applications (E. Casas
ed.), Lecture Notes in Pure and Applied Mathematics, 174 (1995), Marcel Dekker, Inc., New
York, 63–76.
[10] J. I. Diaz and J. E. Saá, Uniqueness of very singular self-similar solution of a quasilin-
ear degenerate parabolic equation with absorption, Publ. Mat. 36 (1992), no. 1, 19–38, DOI
10.5565/PUBLMAT 36192 02. MR1179599
[11] E. DiBenedetto, Degenerate parabolic equations, Universitext, Springer-Verlag, New York,
1993. MR1230384
[12] E. DiBenedetto and A. Friedman, Regularity of solutions of nonlinear degenerate parabolic
systems, J. Reine Angew. Math. 349 (1984), 83–128. MR743967
[13] E. DiBenedetto and A. Friedman, Hölder estimates for nonlinear degenerate parabolic sys-
tems, J. Reine Angew. Math. 357 (1985), 1–22, DOI 10.1515/crll.1985.357.1. MR783531
[14] S. Kamin and J. L. Vázquez, Fundamental solutions and asymptotic behaviour for
the p-Laplacian equation, Rev. Mat. Iberoamericana 4 (1988), no. 2, 339–354, DOI
10.4171/RMI/77. MR1028745
[15] S. Kamin and J. L. Vázquez, Singular solutions of some nonlinear parabolic equations, J.
Anal. Math. 59 (1992), 51–74, DOI 10.1007/BF02790217. MR1226951
[16] S. Kamin, L. A. Peletier, and J. L. Vázquez, Classification of singular solutions of a nonlinear
heat equation, Duke Math. J. 58 (1989), no. 3, 601–615, DOI 10.1215/S0012-7094-89-05828-6.
MR1016437
[17] L. A. Peletier and J. Y. Wang, A very singular solution of a quasilinear degenerate diffu-
sion equation with absorption, Trans. Amer. Math. Soc. 307 (1988), no. 2, 813–826, DOI
10.2307/2001200. MR940229

Applied Analysis Research Group, Faculty of Mathematics and Statistics, Ton Duc
Thang University, Ho Chi Minh City, Vietnam
E-mail address: daonguyenanh@tdt.edu.vn

Instituto de Matemática Interdisciplinar, Universidad Complutense de Madrid, 28040

Madrid Spain
E-mail address: ildefonso.diaz@mat.ucm.es