Global Existence And Blow-Up Of Generalized Self-Similar

Solutions To Nonlinear Degenerate Di¤usion Equation Not In

Divergence Form

1 Introduction
Many problems and models in physics, chemistry, biology and economics are modeled by partial di¤erential
equations. In this work, we shall give an example of a class of renowned equations, which allow to describe the
di¤usion phenomena; that is, a parabolic PDE and known as nonlinear di¤usion equation not in divergence
form and is written as:
@u @2u
= um 2 ; m 2 (0; 1) [ (1; 1) ; (1)
@t @x
where u = u (x; t) is a nonnegative scalar function of space variables x 2 R and time t > 0:
The nonlinear di¤usion equation not in divergence form (1), is often studied by researchers (see for
example [1, 6, 15–17]), who gave some results of existence and uniqueness of the global solutions in time and
solutions which blow-up in a …nite time, this is applied for a certain class of the function u which satis…ed
some su¢ cient conditions.
In general, for some PDEs which has the characterization of symmetries, (see for example [11, 12, 13]), we
can determine their exact solutions with certain (…nite or in…nite) transformations. Here, a PDE becomes
an ordinary di¤erential equation, in this case the solutions are called self-similar solutions ([3, 7, 9, 16]),
which often play a central role in the study of a PDE, since it is equivalent to these solutions to solve locally
or globally.
C. Wang et al. [16], tackled in detail the existence and the uniqueness of a shrinking self-similar solution
to the equation (1) for m 1: The proposed solution was:

1 2
u (x; t) = ! (t + 1) jxj ; t > 0 and x 2 Rn ;
(t + 1)

where ! is a positive function that satis…es some properties and

1+
0; = ;
m

Mathematics Subject Classi…cations: 34A05, 35B06, 35B44, 35C06.

367
368

are constants chosen so that the solutions exist.

The equation (1) is a representation of a large class of the nonlinear parabolic equations. Obviously, for
m = 0 we recognize the well-known heat equation, which in fact has quite di¤erent properties from the two
nonlinear ranges, m > 1 (usually called not in divergence form) and m 2 (0; 1) : For the case m 2 (0; 1) ; if
we consider the implicit change of variables:
1 1 1
v = pp 1 u p ; with p = ; m 2 (0; 1) ;
1 m
the equation (1) can be written in divergence form of the porous medium equation (see [2, 5, 10, 14])

@v @2
= (v p ) ; p > 1: (2)
@t @x2
The porous medium equation (2), is an equation, which admits the properties of similarity. There are
several known fundamental families of self-similar solutions, maybe the most important one is formed by the
Barenblatt solutions B( ) ; discovered independently by Barenblatt in [2] and by Zeldovich and Kompaneets
in [18], which are written under the following form:
8 1
< 1
p 1 2 2
2 p 1 1

BC (x; t) = t p+1
2p(p+1) C x t p+1 ; for jxj < Ct p+1 ; C > 0; (3)
: 0; otherwise.

Our objective in this work is to study the existence and uniqueness of positive solutions of the degenerate
parabolic PDE (1), under the generalized self-similar form which is:

x
u (x; t) = c (t) f ; x 2 R; t > 0:
a (t)

The functions a (t) and c (t) depend on time t and the basic pro…le f > 0; are not known in advance and are
to be identi…ed.

2 De…nitions and Preliminary Results

To discuss the generalized self-similar solutions:
x
u (x; t) = c (t) f ( ) ; with = ; x 2 R; t > 0: (4)
a (t)

We should …rst deduce the equation satis…ed by the function f ( ) in (4) used for the de…nition of self-similar
solutions.

2.1 Basic Idea to Compute the Self-Similar Solutions

The functional-di¤erential equation resulting from the substitution of expression (4) in the original PDE (1),
should be reduced to the standard bilinear functional equation (see [13]); in this case we obtain the following
equation:
a_ (t) cm+1 (t) m
c_ (t) f ( ) c (t) 2 xf 0 ( ) = f ( ) f 00 ( ) ; with a; c 2 R+ : (5)
a (t) a2 (t)
By expressing x from (4) in terms of ; substituting into (5) and divide by c (t) ; we get the functional
equation in two variables t and ; as follows:

c_ (t) a_ (t) 0 cm (t) m

f( ) f ( ) f ( ) f 00 ( ) = 0:
c (t) a (t) a2 (t)
 369

This can be rewritten as the standard bilinear functional equation:

'1 1 + '2 2 + '3 3 = 0;

with a direct calculation shows that a = a (t) ; c = c (t) and f = f ( ) ; we have:

c_ a_ cm
'1 = ; '2 = ; '3 =
c a a2
and
1 = f; 2 = f0 ; 3 = f m f 00 :
Substituting these expressions into the solution of the three-term functional equation [13], yields the deter-
mining system of ordinary di¤erential equations (see also [3, 4])
8 c_ m
>
> c = k1 ca2 ;
<
m
a_
= k2 ca2 ; (6)
>
>
a
:
f m f 00 = f + f0 :

Where ; ; k1 and k2 are arbitrary constants.

The system of ordinary di¤erential equations (6) depends on many unknown parameters. The functions
a (t) ; c (t) and f ( ) are explicitly determined.

2.2 Statement of the Problem

In this part, we …rst attempt to …nd the equivalent approximate to the following problem of the nonlinear
di¤usion equation not in divergence form:
8
> @u m @2u
>
> @t = u @x2 ; x 0; m 2 (0; 1) [ (1; 1) ;
>
>
>
> u (0; t) = c (t) U; U 0; t > 0;
>
>
< m
@u( a(t);t) c (t) 1 m
@x = (1 m)a(t) lim [u (x; t)] ; > 0; (7)
>
> x! a(t)
>
>
>
>
>
> R
a(t) m 1
>
:
u(x;t)
ds 2 a (t) ; 8m > 1; 8x 2 [0; a (t)) :
u(s;t)
x

Under the generalized self-similar form which is:

x
u (x; t) = c (t) f ( ) ; with = and a; c 2 R+ : (8)
a (t)

According to the preceding part (the system (6)), we consider this problem:
8 m
> f ( ) f 00 ( ) = f ( ) + f 0 ( ) ; 0; m 2 (0; 1) [ (1; 1) ;
>
>
>
>
>
> f (0) = U; U 0;
>
<
f 0 ( ) = 1 m lim f 1 m ( ) ; > 0; (9)
>
> !
>
>
>
> R
>
>
: f1 m ( ) d 2f
1 m
( ); 8m > 1; 8 2 [0; ) ;

in which ; are arbitrary real constants.

370

3 Existence and Uniqueness of Basic Pro…le

In this section we study the existence and uniqueness of a class of positive solutions under the generalized
self-similar form (8) for the problem of the nonlinear di¤usion equation not in divergence form (7).
We …nd also exact solutions under certain conditions, which depend on the similarity coe¢ cients and
:
As in [7] it is necessary to consider the weak solutions of the problem (9).

De…nition 1 (Weak solution) A function f will be called a weak solution with compact support of the
problem (9) if and only if

1) f is a continuous function, bounded and nonnegative on [0; 1) :

2) f has a continuous derivative in a left neighborhood of = :

3) f satis…es the identity:

Z1 Z1
0 1 m 0
f ( )+ f ( ) ' ( )d + + f1 m
( ) ' ( ) d = 0;
m 1 m 1
0 0

for all ' 2 C01 (0; 1) :

We discuss the existence of basic pro…le f of the weak solutions with compact support, then as we shall
see later, f is positive in a right neighborhood of = 0; more speci…cally for some " > 0; there exists a
number " < < 1 such that:
f > 0; on (0; )
and
f 0; on [ ; 1) :
Assume that f is a weak solution of the problem (9) with compact support and let be an arbitrary positive
number, we shall be mainly concerned with proving the existence and uniqueness of a positive solution of
problem (9) on an interval (0; ) which satis…es the boundary conditions:

f (0) = U; f ( ) = 0: (10)

Then we shall show that for suitable and there exists a unique positive solution of the equation:

f m ( ) f 00 ( ) = f ( ) + f0 ( ) ; (11)

in the left neighborhood of = and that this solution can be continued back to = 0: We then ask whether
can be chosen so that condition (10) is satis…ed.
Before dealing with the question of existence, we give the necessary conditions of the parameters and
for the existence of a nontrivial weak solution with compact support of the problem (9).

Lemma 1 The function f is a nontrivial weak solution with compact support of the boundary value problem
(9)–(10), if and only if

(i) is a negative coe¢ cient for m 2 (0; 1) :

(ii) is a positive coe¢ cient for m > 1:

(iii) = 0 and takes strictly positive value.

 371

Proof. Suppose f is a nontrivial weak solution of problem (9) with compact support. Then for some " > 0;
we have:
> 0 in ( "; )
f ; > 0:
= 0 in [ ; 1)
It results that f is a weak solution of (9) which satis…es (10) in the left neighborhood of = : The integration
of (11) starting from to ; where " < < gives:

Z
(1 m) f 0 ( ) = f1 m
( ) + [ (m 1) + ] f1 m
( )d : (12)

The continuity of f and f 0 left of ; ensures the existence of 2( "; ) such that f 0 ( ) < 0:
In fact, if we apply the mean value theorem on the interval ( "; ) ; we have:

f( ) f( ") f( ")
9 2( "; ) such that f 0 ( ) = = < 0:
( ") "

This implies that the member in the left of (12) at the point = is negative for m 2 (0; 1) and positive
for m > 1: Hence, and (m 1) + cannot both be greater than zero for m 2 (0; 1) ; (resp. less than zero
for m > 1) and = 0 implies that > 0. Hence the estimate (iii).
In what follows, because the function f is positive and tends to zero if ! ; the mean value theorem
0
enables us to prove that there exists " 0 < ; such that f ( ) < 0; 8 2 ( 0 ; ) : We divide the proof
into two cases:
(i) For m 2 (0; 1) ; let us consider > 0; for all 2 ( 0 ; ) ; this implies that (m 1) + < 0: It results
from (12) that:
(1 m) f 0 ( ) f m 1 ( ) > [ (m 1) + ] ( ): (13)
In fact, we will use:
8 2[ 0; ) ; 8 2 ( ; ) ; we have f ( ) < f ( ) ;
which implies for any m 2 (0; 1) ; that:

f1 m
( ) < f1 m
( ) and f 1 m
( ) fm 1
( ) < 1;

also
Z
f1 m
( ) fm 1
( )d < :

If ! in (13), the left part is negative and the right part tends towards zero, contradiction, which implies
that < 0 for m 2 (0; 1) :
(ii) For m > 1; we consider < 0; for all 2 ( 0 ; ) ; this implies that (m 1) + > 0; consequently
> 0: It results from (12) that:

Z
0 1 m
(1 m) f ( ) f ( )< (m 1) f1 m
( )d : (14)

If ! in (14), the left part is positive and the right part tends towards zero, contradiction.
Thus, we have already proved that the coe¢ cients

> 0 for m > 1; or < 0 for m 2 (0; 1) ; or = 0 and > 0;

are the only cases for which the nontrivial weak solution with compact support of the boundary value problem
(9)–(10) exists.
372

New explicit solutions for = 0 and m 2 (0; 1) [ (1; 2)

Let be an arbitrary positive number. It is clear from the proof of Lemma 1 that a necessary condition
for the existence of a positive solution of problem (9)–(10) in a left neighborhood of = is either > 0
for m > 1; < 0 for m 2 (0; 1) or = 0 and > 0:
The objective of this part is to show that this condition is also su¢ cient.
We begin by assuming that = 0 and > 0; we can then solve problem (9)–(10) uniquely. After an
elementary computation that the function:
1
m2 2
m

f( ; )= ( ) ; m 2 (0; 1) [ (1; 2) ; 0 < ; (15)

2 (2 m)

is the unique solution of the problem (9)–(10).

In fact, starting from (11), we put = 0 and > 0; we get

f m ( ) f 00 ( ) = f ( ) ;

also
f 00 ( ) f 0 ( ) = f 1 m
( ) f0 ( ) : (16)
0
For m 2 (0; 1) [ (1; 2) ; we have from (9) that f ( ) = 0: Then, the integration of (16) starting from to
gives:
1 0 2
[f ( )] = f2 m ( ) :
2 2 m
Also, after a simple integration from to ; we have the function f ( ; ) which is presented in (15).
Because f (0; ) is a continuous, monotonically increasing function of ; such that

f (0; 0) = 0 and f (0; 1) = 1;

the equation f (0; ) = U is uniquely solvable for every U > 0:

Let (U ) be its solution, then f = f ( ; (U )) is the unique solution of the problem (9)–(10), with
compact support de…ne for all 0 < and satis…es:

f (0; ) = U and f ( ; ) = 0;

where r
m) m 2 (2
(U ) =
U :
m2
In addition, f is a continuous and monotonous decreasing function.
Next, we discuss the case 6= 0 and we give some elementary lemmas for the two following cases <0
for m 2 (0; 1) and > 0 for m > 1:

Lemma 2 Suppose that 0 1 < and f is a positive solution of the boundary value problem (9)–(10) on
[ 1 ; ) : Then

1. f 0 ( ) < 0 on [ 1; ) ; if and only if > 1 m:

2. If < 1 m and f 0 ( 0 ) = 0; for some 0 2[ 1; ) : Then f has a maximum at 0 such that:

[ (m 1) + ]
0 ; for the case m 2 (0; 1)
(m 1)
and
[ (m 1) + ]
0 ; for the case m 2 (1; 1) :
(m 1)
 373

3. Suppose that f is a positive solution of (9) on [0; ). Then:

f 0 (0) < 0; for > :

1 m

4. If
= ;
1 m
we have:
f 0 (0) = 0 and f 0 ( ) < 0; 8 2 (0; ) ; i.e. f ( ) f (0) = U; 8 2 [0; ] : (17)
In this case, we …nd a new explicit solution of the problem (9)–(10) as follows:
1
m 2 2
m
f( ; )= ; m 2 (0; 1) [ (1; 1) ; j j < : (18)
2 (m 1)

Proof. For 2[ 1; ) ; the integral equation (12) gives:

Z
(1 m) f 0 ( ) = f1 m
( ) + [ (m 1) + ] f1 m
( )d : (19)

1. Because < 0 for m 2 (0; 1) ; the right member in (19) is always negative i.e. f 0 ( ) < 0 if and only if
(m 1) + takes a negative value. In the same way, we have > 0 for m 2 (1; 1) ; then the left member
in (19) is always positive i.e. f 0 ( ) < 0 if and only if (m 1) + takes a positive value, therefore

f 0 ( ) < 0 on [ 1; ) when > :

1 m
2. If < 1 m then < 0 for any m 2 (0; 1) [ (1; 1) : According to (11);

f 00 ( 0 ) < 0 when f 0 ( 0 ) = 0:

So f has a maximum at = 0 and strictly decreasing on ( 0 ; ) ; i.e. f 0 ( ) < 0 on ( 0 ; ) : We put = 0

in (19), we get:
Z
1 m
0= 0f ( 0 ) + [ (m 1) + ] f1 m
( )d :
0

For m 2 (0; 1) ; we have < 0 and (m 1) + > 0; then:

1 m 1 m
0f ( 0 ) + [ (m 1) + ] ( 0) f ( 0) 0:

So
[ (m 1) + ]
0 :
(m 1)
In the same way, for m 2 (1; 1) ; we have > 0 and (m 1) + < 0; then we …nd:
[ (m 1) + ]
0 :
(m 1)
3. With = 0; (19) becomes:
Z
(1 m) f 0 (0) = [ (m 1) + ] f1 m
( )d : (20)
0
374

The sign of f 0 (0) results immediately from (20).

4. If
(m 1) + = 0 for any m 2 (0; 1) [ (1; 1) ;
we have from (19):
(1 m) f 0 ( ) = f1 m
( ): (21)
Which implies easily (17).
The integration of (21) starting from to gives:
1 m 2 2
f ( )= ;
m 2 (1 m)
which implies the function f ( ; ) presents in (18) by:
1
m 2 2
m
f( ; )= ; m 2 (0; 1) [ (1; 1) ; j j < :
2 (m 1)
Let (U ) be its solution.
Because f (0; (U )) is a continuous, monotonically increasing function of (U ) ; such that:

f (0; 0) = 0 and f (0; 1) = 1:

Then the equation f (0; (U )) = U is uniquely solvable for every U 0 and f = f ( ; (U )) is the unique
solution of problem (9)–(10), with compact support de…nite for all j j < and it satis…es:

f (0; ) = U and f ( ; ) = 0;

where s
2 (m 1) m
(U ) = U :
m
In addition, f is a continuous and monotonous decreasing function.
We now turn to the question of existence.

Lemma 3 For any m 2 (0; 1) [ (1; 1) and any > 0; the problem (9) with the boundary conditions (10),
has a unique positive solution in the left neighborhood of = :

The proof of this lemma concerns the analysis of the of Banach’s …xed point theorem, which is:

Theorem 4 (Banach’s …xed point [8]) Let X be a non-empty closed subset of a Banach space E; then
any contraction mapping M of X into itself has a unique …xed point.

Now, we prove the Lemma 3.

Proof. Assume that f is a positive solution in the left neighborhood of = : By Lemma 2, f 0 ( ) < 0 for
2( "; ) for some " > 0:
1. For m 2 (0; 1) ; we have < 0: In this case, the writing of (19) is given as follows:
Z
f0 ( )
f0 ( ) = f1 m
( ) + [ (1 m) ] d : (22)
fm ( )

With G (f ) = we have
dG 1
= : (23)
df Rf G(')
Gf 1 m [ (1 m) ] 'm d'
0
 375

By integrating the di¤er-integral equation (23) from 0 to f; we obtain

Zf
d'
G (f ) = : (24)
R' G(g)
0 G'1 m [ (1 m) ] gm dg
0
1
We put H (f ) = 1 G (f ) ; then the equation (24) becomes
Zf
1 d'
H (f ) = : (25)
2 R' H(g)
0 H'1 m 1 m '1 m [ (1 m) ] gm dg
0

Using the principle of contraction of Banach Theorem, we prove now that the equation (25) admits a unique
positive solution to the right of f = 0:
Let X be the set of bounded functions H (f ) on [0; h] ; h > 0; satisfying:
m
0 H (f ) = :
j j (1 m) + j (1 m) j
Let k k be the norm sup de…ned on X; then X is a complete metric space. On X we de…ne the operator:
Zf
1 d'
M (H) (f ) = : (26)
2 R' H(g)
0 H'1 m
1 m'
1 m [ (1 m) ] gm dg
0

First, we prove that M is a mapping from X to X.

Let H 2 X: It is clear that
Z'
1 m 1 m H (g)
H (') ' ' [ (1 m) ] dg
1 m gm
0
Z'
1 m 1 m m
' j j H (') ' j (1 m) j kHk g dg
1 m
0

'1 m
: (27)
1 m
Therefore, starting from (26), we have:
Zf
1 d' (m 1) hm
M (H) (f ) 2 2 : (28)
1 m m
0 1 m'

Thus, M (H) is well de…ned on X and M (H) : [0; h] ! R is nonnegative and continuous. The right side of
(28) shows that we can …nd h0 > 0; such that h h0 and
kM (H)k ; for all H 2 X:
So M is a mapping from X to X for some h h0 :
In the next step, we show that M is a contracting mapping on X. Let H1 ; H2 2 X and h h0 ; be a
positive real number, we put:
Z'
1 m 1 m H (g)
K (H) = H (') ' ' [ (1 m) ] dg:
1 m gm
0
376

Then

kK (H2 ) K (H1 )k j j kH2 (') H1 (')k '1 m

j (1 m) j
+ kH2 (') H1 (')k '1 m
1 m
j (1 m) j
j j+ kH2 (') H1 (')k '1 m
:
1 m

So
Zf Zf
1 d' 1 d'
kM (H1 ) M (H2 )k = 2 2
K (H1 ) K (H2 )
0 0
Zf
1 K (H2 ) K (H1 )
= 2 d'
K (H1 ) K (H2 )
0

2 Zf
(1 m) kK (H2 ) K (H1 )k
2 2 d'
'2(1 m)
0
(1 m) [j j (1 m) + j (1 m) j]
kH2 (') H1 (')k :
mh m 2 2

Hence, there exists a h1 2 (0; h0 ] such that if h h1 ; M is a contraction on X: After the principle of
contraction of Banach [8], M has a unique …xed point in X; consequently, there exists a unique positive
solution for the problem (9) in the interval ( "; ) for some " > 0:
2. After using some techniques, we use the same steps for the case m 2 (1; 1) ; to prove the existence of
a single positive solution for the problem (9).
We now continue f backwards as a function of : By the standard theory, this can be done uniquely so
long as f remains positive and bounded. There are now three possibilities:

(A) f ( ) ! 1 as ! 1 for some 1 2 [0; ) ;

(B) f ( ) can be continued back to = 0;

(C) f ( ) ! 0 as ! 2 for some 2 2 [0; ) :

We begin by ruling out possibility (A).

Lemma 5 Let 2 [0; ) be a positive real number. If f is a positive solution of the problem (9) on ( ; ) ;
then f is bounded on ( ; ) and
1
2 m
m
sup f ( ) maxfj (m 1) + 2 j ; j jg ; for any m 2 (0; 1) [ (1; 1) :
< < 2 jm 1j

Proof. Assume that f is a positive solution of (9) on ( ; ). We prove this lemma for the following two
cases: (i) (m 1) + < 0; (ii) (m 1) + 0:
1. For the case m 2 (0; 1) ; < 0:
(i) If (m 1) + < 0, in this case f 0 ( ) < 0; 8 2 ( ; ) by Lemma 2, i.e f ( ) < f ( ) ; 8 2 ( ; )
and from (19),

(1 m) f 0 ( ) f1 m
( ) + [ (m 1) + ] f 1 m
( )( ); 8 2 ( ; ):
 377

So, for any < < ; we have:

(1 m) f m 1
( ) f0 ( ) + (m 1) ( )+ ( )
+ (m 1) ( ): (29)
The integration of (29) from to gives:
1 m m 1
f ( ) + (m 1) ( ) ( )
m 2
m
fm ( ) [ (1 m) ( ) 2 ]( ):
2 (1 m)
So
2
m
sup f m ( ) [ (m 1) + 2 ] : (30)
< < 2 (1 m)
(ii) If (m 1) + 0, by equation (19), we have for any < < ;
(1 m) f m 1
( ) f0 ( ) : (31)
The integration of (31) from to gives:
1 m m 2 2
f ( ) :
m 2
This implies that:
2
m
sup f m ( ) : (32)
< < 2 (1 m)
2. The case m 2 (1; 1) ; > 0.
(i) If (m 1) + < 0, by equation (19), we have for any < < ;
(1 m) f m 1
( ) f0 ( ) : (33)
After the integration of (33) from to ; we have:
2
m
sup f m ( ) : (34)
< < 2 (m 1)
(ii) If (m 1) + 0, 8 2 ( ; ) ; we have from (19):
Z
0 1 m
(1 m) f ( ) = f ( ) + [ (m 1) + ] f1 m
( )d ;

f1 m
( )+ [ (m 1) + ] f 1 m
( ):
2
Then
(1 m) f m 1
( ) f0 ( ) + [ (m 1) + ] : (35)
2
After the integration (35) from to ; we have:
2
m
sup f m ( ) [ (m 1) + 2 ] : (36)
< < 2 (m 1)
Note that the bounds of (30), (32), (34) and (36) are independent of and consequently, f ( ) can not be
unlimited at ! :
The following lemmas distinguish between the possibilities (B) and (C).
First, we give an elementary lemma for the case m 2 (0; 1) ; < 0:
378

Lemma 6 Assume that f is a positive solution of problem (9)–(10) in the left neighborhood of = : Then

(i) If (m 1) + 2 < 0; then f ( ) > 0 on [0; ) :

(ii) If (m 1) + 2 = 0; then f ( ) > 0 on (0; ) and f (0) = 0:
(iii) If (m 1) + 2 > 0; there exists an 2 (0; ) such that f ( ) > 0 on ( ; ) and f ( ) = 0:

Proof. For the case m 2 (0; 1) ; < 0; assume that f is a positive solution of the problem (9)–(10) on [0; ) :
The integration of (19) from to gives:

Z Z
d (m 1) + 2
f( )= + d : (37)
m 1 fm 1 ( ) m 1 fm 1 ( )

Hence the lemma.

Now, we give an elementary lemma for the case m 2 (1; 1) ; > 0:

Lemma 7 Assume that f is a positive solution of the problem (9)–(10) in the left neighborhood of = :
Then

(i) If (m 1) + 0; then f ( ) > 0 on [0; ) :

(ii) If (m 1) + 0 and (m 1) + 2 > 0; then f ( ) > 0 on [0; ) :
(iii) If (m 1) + 2 0; then there exists an 2 (0; ) such that f ( ) > 0 on ( ; ) and f ( ) = 0:

Proof. For the case m 2 (1; 1) ; > 0; assume that f is a positive solution of the problem (9)–(10) on
[0; ) :
(i) If (m 1) + 0; then (m 1) + 2 > 0; also we have from (19) that:

(1 m) f 0 ( ) f1 m
( ) (38)

and the integration of (38) from to gives:

m 2
fm ( ) 2
: (39)
2 (m 1)

It is easy to see from (39) that f ( ) is always positive on [0; ) :

Now we prove the estimate (ii) and (iii). If (m 1) + 0; we have from (19) that:

(1 m) f 0 ( ) f1 m
( )+ [ (m 1) + ] f 1 m
( ):
2
Which implies after an integration from to ; that
m
fm ( ) ( + [ (m 1) + 2 ]) ( ):
2 (m 1)

We observe in this case the validity of estimate (ii) and (iii). We can …x as follows:

[ (m 1) + 2 ]
; for the case < (m 1) + 2 0:

Hence the lemma.

New explicit solutions for (m 1) + 2 = 0 and m 2 (0; 1) [ (1; 2)
Let be an arbitrary positive number.
 379

Starting from (37) for m 2 (0; 1) ; if we put (m 1) + 2 = 0; we have:

Z
d
f( )= : (40)
m 1 fm 1 ( )

R
The change of variable g ( ) = f1 m
( ) d ; gives g ( ) = 0 and g 0 ( ) = f1 m
( ) < 0; 8 2 (0; ) : Then
the equation (40) becomes:
1
f ( ) = [ g 0 ( )] 1 m
= g( ): (41)
m 1
The last equation can be resolved for any m 2 (0; 1) [ (1; 2) :
In fact, from the proof of Lemma 1 that a necessary condition for the existence of a positive solution of
problem (9)–(10) in a left neighborhood of = ; is either > 0 for m > 1; < 0 for m 2 (0; 1) or = 0
and > 0:
Then the value m 1 is always positive. In this case the equation (41) is equivalent to

1 m
g0 ( ) gm 1
( )= 1 m
: (42)
m 1

The integration of (42) starting from to gives:

1 m
1 m 1 2 m 2 m
g ( )= :
m 2 m m 1

From (41), we have after an elementary computation that:

1
m m 2 m 2 m
m
f( ; )= ; 0 < ; (43)
(m 1) (2 m)

is a continuous function, for any choice of > 0; with compact support and satis…es:

f (0; ) = f ( ; ) = 0:

Theorem 8 The following statements (i) and (ii) hold.

(i) Assume that U = 0: Then for every > 0 there exists a solution f ( ; ) of the boundary value problem
(9)–(10) which is positive in (0; ) if and only if (m 1) + 2 = 0 and < 0 for m 2 (0; 1) ; or if
(m 1) + 2 = 0 and > 0 for m > 1:

(ii) Assume that U > 0: Then the boundary value problem (9)–(10) admits a unique solution and there
exists a unique (U ) > 0 such that f ( ; (U )) is positive on (0; ) if and only if (m 1) + 2 < 0
and 0 for m 2 (0; 1) ; or if (m 1) + 2 > 0 and 0 for m > 1:

Proof. By Lemma 1 a necessary condition for the existence of such a solution is that 0 for m 2 (0; 1)
or 0 for m > 1:
For = 0; > 0; we already gave the solution (15) of (19) for m 2 (0; 1) [ (1; 2) which is:
1
m2 2
m

f( ; )= ( ) ; 0 < :
2 (2 m)
380

This solution is continuous and satis…es for every U > 0; that f (0; ) = U and f ( ; ) = 0; where
r
2 (2 m) m
(U ) = U :
m2
In the following, for the case 6= 0; we put < 0 for m 2 (0; 1) and > 0 for m > 1:
We already proved in Lemma 3, the local existence of positive solution in the left neighborhood of =
of problem (9)–(10). This local solution is unique and can be continued back to = 0 as a positive solution
with f (0) > 0 if and only if < 0 and (m 1)+2 < 0 for m 2 (0; 1) ; also if > 0 and (m 1)+2 > 0
for m > 1 (by Lemmas 6 and 7).
Now, the boundary condition (10) at = 0 is satis…ed if we can …nd a (U ) such that

f (0; ) = U: (44)

If only one such a exists, the solution is unique. We distinguish two cases:
(i) U = 0: By Lemmas 6 and 7, the equation (44) can only be satis…ed if (m 1) + 2 = 0: Moreover
(44) is then satis…ed for any > 0: Then there exists a nontrivial weak solution f with compact support of
the problem (9)–(10) with the property:

f ( ) > 0 on (0; ) ;
f ( ) = 0 on f0g [ [ ; 1) :

(ii) U > 0: It follows from Lemmas 6 and 7 that now a necessary condition for (44) to have a solution is
that < 0 and (m 1) + 2 < 0 for m 2 (0; 1) or > 0 and (m 1) + 2 > 0 for m > 1:
With this intention, let us suppose that f ( ; ) is a solution of the problem (9)–(10) on (0; ) ; then for
2
any > 0 the function mf ( ; ) is a solution of (9)–(10) on (0; ) : If = 1 ; then:
2
m f (0; 1) = U: (45)

Because f (0; 1) > 0 in the cases (m 1) + 2 < 0 for m 2 (0; 1) or (m 1) + 2 > 0 for m > 1: We
obtain a unique solution = (U ) of (45).
Thus, the function f ( ; (U )) is the unique solution of (9)–(10), with the property:

f ( ) > 0 on [0; ) ;
f ( ) = 0 on [ ; 1) :

Hence the theorem.

We denote by (z)+ the positive part of z; which is z if z > 0 and else is zero.
Now, we give the principal theorem of this work.

Theorem 9 (Global existence and blow-up of self-similar solutions) Let a (t) ; c (t) be positive real
functions of t; which satisfy:
a (0) = c (0) = 1; a_ (0) = ; c_ (0) = : (46)
Then, for f 2 C ([0; ] ; R+ ) ; the problem (7) admits an exact solution in the form:
x
u (x; t) = c (t) f ( ) ; with = ; x 2 R; t > 0; (47)
a (t)

if the basic pro…le f is a solution of the problem (9) on [0; ] and satis…es in each point:

f m ( ) f 00 ( ) = f ( ) + f0 ( ) :

Thereupon, we separate the following cases:

 381

(1) If m + 2 < 0; the functions a (t) ; c (t) are given by:

8
< a (t) = (1 ( m + 2 ) t) m+2
;
8t > 0: (48)
: c (t) = (1 ( m + 2 ) t) m+2 ;

(2) If m + 2 = 0; the functions a (t) ; c (t) in this case are given by:
(
a (t) = e t ;
8t > 0: (49)
c (t) = e t ;

In each case (1) and (2). The problem (7) admits a global solution in time under the generalized
self-similar form, this solution de…ned for all t > 0. Moreover, if < 0; we have

lim u (x; t) = 0 for all x 2 R:

t!+1

(3) If m + 2 > 0; the functions a (t) ; c (t) are given by:

8
< a (t) = (1 ( m + 2 ) t) m+2 ;
+
0 < t < T: (50)
:
c (t) = (1 ( m + 2 ) t)+ m+2 ;
1
The moment T = m+2 represents the maximal existence value of the functions a (t) ; c (t) : Moreover;
if > 0; the problem (7) admits a solution under the generalized self-similar form, which blows up in
a …nite time. The solution is de…ned for all t 2 (0; T ) ; where T represents the blow-up time of the
solution such that:
1
for all x 2 R; lim u (x; t) = +1; with T = > 0:
t!T m+2

Proof. We have already proved in Theorem 8 the existence of solution f of (9), which is with compact
support [0; ] if and only if

1. (m 1) + 2 = 0 and [ < 0 for m 2 (0; 1) or > 0 for m > 1] ;

2. (m 1) + 2 < 0 and 0 for m 2 (0; 1) ;
3. (m 1) + 2 > 0 and 0 for m > 1:

Now, to determine the functions a (t) ; c (t) ; just solve the system (6) which is
( c_ cm
c = k1 a2 ;
m
(51)
a_
a = k2 ca2 :

The conditions (46), imply k1 = and k2 = ; then the system (51), can be resolved as follows
( c_ cm
c = a2 ; a_ c_
) = ;
a_ cm a c
a = a2 :

then we deduce after an integration from 0 to t that:

a (t) = c (t) : (52)

382

If we replace (52) in (51), we obtain

m+2
1
c dc = dt: (53)

If m + 2 6= 0; we obtain easily the solution of (53) as follows

m+2
c (t) = (1 ( m + 2 ) t)+

in the same way we …nd:

a (t) = (1 ( m + 2 ) t)+m+2 :

We deduce that the functions a (t) ; c (t) are globally de…ned if m + 2 < 0; and a (t) ; c (t) are maximal
functions if m + 2 > 0; and well de…ned if and only if

1
0<t<T = :
m+2

If m + 2 = 0; the functions a (t) ; c (t) are de…ned globally, we obtain in this case:
( t
a (t) = e ;
8t > 0:
c (t) = e t ;

We notice from this theorem that we have two time behaviors of functions a (t) ; c (t) ; their behaviors depend
on parameters of similarity ; :
In (1) and (2) i.e m + 2 0; the functions a (t) ; c (t) are de…ned globally in time. Now, the pro…le f
is a bounded positive function for all x 2 R: If < 0 we have:

lim c (t) = 0:
t!+1

In each case (1) and (2). Thus:

x
lim u (x; t) = lim c (t) f = 0:
t!+1 t!+1 a (t)

(3) In the case m + 2 > 0; we have a (t) ; c (t) given in (50), is well de…ned if and only if:

1
0<t<T = :
m+2

We recall that the solution blows up in …nite time if there exists a time T < +1; which we call it the blow-up
time, such that the solution is well de…ned for all 0 < t < T; while

sup ju (x; t)j ! +1; when t ! T :

x2R

If > 0; the value T represents the blow-up time of the solution, thus lim c (t) = +1 and
t!T

x
lim u (x; t) = lim c (t) f = +1:
t!T t!T a (t)
 383

4 Examples
Example 1 (Blow-up solution) For = 0 and > 0; we have from Theorem 9 that m + 2 = m > 0
and
1 1
a (t) = 1; c (t) = (1 mt)+ m ; where 0 < t < T = :
m
In this case, the self-similar form is a solution of separation of variables u (x; t) = c (t) f (x) ; where f is
given in (15), by:
1
m2 2
m

f (x) = ( x) ; m 2 (0; 1) [ (1; 2) ; 0 x< :

2 (2 m)

Then, the solution of (1) in the generalized self-similar form (47) for = 0; > 0 is given by:
8 1
< m2
( x)
2 m
; 0 x < ; m 2 (0; 1) [ (1; 2) :
2(2 m)(1 mt)
u (x; t) =
:
0; otherwise.
1
This solution blows up when t ! m: In this case
r
2 (2 m) m
(U ) = U ; for every U > 0:
m2
Example 2 (Global and blow-up solutions) Now, we present the new explicit solutions on the general-
ized self-similar form of the equation (1). Because the pro…le f is bounded with compact support, then it is
an integrable function on R; where
Z
f ( ) d = M; for some M > 0:
R

Therefore, for m 2 (0; 1) [ (1; 1) and Z

u (s; t) ds = cm (t) :
R

We can …nd explicitly a new exact solution of (1). Indeed

Z Z Z
s
u (s; t) ds = c (t) f ds = a (t) c (t) f ( ) d = cm (t) ;
R R a (t) R

this implies that

cm 1 (t)
= M: (54)
a (t)
According to the formulas a (t) and c (t) in (48) and (49), the equality (54) implies that:

= (1 m) :

Because < 0 for m 2 (0; 1) and > 0 for m > 1; then < 0:
For (m 1) + = 0; we already gave the solution (18) of (9) for m 2 (0; 1) [ (1; 1) which is:
1
m 2 2
m
f( ; )= ; 0 < :
2 (m 1)

Now, we determine the functions a (t) and c (t) : We have from Theorem 9

m + 2 = m + 2 (1 m) = (2 m) :
384

If
(2 m) 6= 0; i:e: m 6= 2;
the functions a (t) and c (t) are given by
( m 1
a (t) = (1 (2 m) t)+m 2
;
1 m 6= 2; 0 < t < T:
m 2
c (t) = (1 (2 m) t)+ ;

Where T represents the maximal existence value

T = (21 m) ; if m > 2;
T = +1; if m < 2:

We obtain the exact solution of (1), for m 2 (0; 1) [ (1; 2) [ (2; 1) as follows
8 1
>
< c (t) m 2 2
2(1 m)
m 2
m

2(m 1) x (1 (2 m) t)+ ; jxj < r1 ;

u (x; t) = (55)
>
:
0; otherwise,

where m 1
r1 = (1 (2 m) t)+m 2
:
This solution is de…ned globally in time if m 2 (0; 1) [ (1; 2) ; and blows up when
1
t! ; for m 2 (2; 1) :
(2 m)
In this case s
2 (m 1) m
(U ) = U ; for every U > 0:
m
If m = 2; the functions a (t) and c (t) are written as (49), in this case we obtain = and
( q
2
e t x2 e2 t ; jxj < e t ;
u (x; t) =
0; otherwise.

This solution is de…ned globally in time. In this case

U
(U ) = p ; for every U > 0:

For the functions a (t) and c (t) that satisfy the parameters of the classical self-similar form, i.e.

c (t) = t ; a (t) = t :

The self-similar solution (8) of the equation (1) is written as:

u (x; t) = t f ( ) ; with = xt ;

where and are exponents which satisfy the similarity condition ([11], [12], [13])

m + 2 + 1 = 0; (56)

and the function f is the self similar pro…le which to be determined by the solution of the following di¤ erential
equation:
f m ( ) f 00 ( ) = f ( ) + f ( ) :
 385

If = (1 m) ; the similarity condition (56) is equivalent to:

m+2 +1= (2 m) + 1 = 0;

this implies that:

1 1 m
= and = :
m 2 m 2
We get for m 2 (0; 1) that ; < 0: So the solution (55) for c (t) = t ; a (t) = t ; is written as:
1
1 m 2 2(1 m) m m 1
U (x; t) = t m 2 x2 t m 2 ; for jxj < t m 2 ; m 2 (0; 1) :
2 (2 m)
Also 8 1
1 m
>
< 1
1
m 1 m
m(1 m) 2 2
2(1 m) m
1 m
m 1

1 m t m 2
2(2 m) x t m 2 ; jxj < t m 2 ;
U (x; t) = (57)
>
:
0; otherwise.
1
If we put for m 2 (0; 1) the parameter p = > 1; then the solution (57) of the equation (1) in the classical
1 m
self-similar form is written as: p
U (x; t) = p p 1 B p (x; t) ; p > 1:
Where B( ) is the Barenblatt solutions of the porous medium equation (2), which are given in (3) under the
following form:
8 1
< t p+11
p 1 2
x 2
t
2
p+1
p 1 1
; for jxj < t p+1 ; p > 1;
2p(p+1)
B (x; t) =
:
0; otherwise.

Example 3 (Global solution) We present the second new explicit solutions on the generalized self-similar
form of the equation (1). For 0 < m 6= 1 and
Z
cm (t)
u (s; t) ds = ;
R a (t)
we can …nd explicitly a new exact solution of (1). In fact
Z Z Z
s cm (t)
u (s; t) ds = c (t) f ds = a (t) c (t) f ( ) d = ;
R R a (t) R a (t)
this implies that
cm 1 (t)
= M: (58)
a2 (t)
According to the formulas a (t) and c (t) in (48) and (49), the equality (58) implies that:
(1 m)
= ; then < 0:
2
For (m 1) + 2 = 0; we gave the solution (43) of (9) for m 2 (0; 1) [ (1; 2) which is:
1
m m 2 m 2 m
m
f( ; )= ; 0 < :
(m 1) (2 m)

Now, we determine the functions a (t) and c (t) : We have from Theorem 9

m+2 = m+ (1 m) = < 0:
386

The functions a (t) and c (t) are given by:

( 1 m
a (t) = (1 t) 2 ; 8t > 0:
c (t) = 1 1 t ;

We obtain the exact solution of (1), for m 2 (0; 1) [ (1; 2) as follows:

8 1
< 1 mxm 2 m
x2 m c
(1 m)(2
2
m)
(t)
m
; x 2 r2 ;
1 t (m 1)(2 m)am (t)
u (x; t) =
:
0; otherwise,

where 1 m
r2 = 0; (1 t) 2
:

This solution is de…ned globally in time for any m 2 (0; 1) [ (1; 2) : In this case > 0:

5 Conclusion
In this paper we have discussed the existence and uniqueness of positive solutions for a class of nonlinear
di¤usion equation not in divergence form, under the generalized self-similar form. Also we have found new
solutions; the behavior of these solutions depends on some parameters (that satisfy some conditions), which
make their existence global or local, and we generalized the families of self-similar solutions of the porous
medium equation, which is formed by the Barenblatt solutions.
Acknowledgments. The authors are deeply grateful to the reviewers and editors for their insightful
comments that have helped to improve the quality of this research work that is supported by the General
Direction of Scienti…c Research and Technological Development (DGRSTD)- Algeria.

References
[1] W. Arendt and R. Chill, Global existence for quasilinear di¤usion equations in isotropic nondivergence
form, Vol. IX(2010), 523–539.
[2] G. I. Barenblatt, On some unsteady motions of a liquid or a gas in a porous medium, Akad. Nauk SSSR.
Prikl. Mat. Meh., 16(1952). 67–78.
[3] N. Benhamidouche, Exact solutions to some nonlinear PDEs, traveling pro…les method, Electron. J.
Qual. Theory Di¤er. Equ., 2008, 7 pp.
[4] N. Benhamidouche and Y. Arioua, New method for constructing exact solutions to non-linear PDEs,
Int. J. Nonlinear Sci., 7(2009), 395–398.
[5] L. Ca¤arelli and J. L. Vàzquez, Viscosity solutions for the porous medium equation, Proc. Sympos.
Pure Math., 65(1999), 13–26.
[6] R. Ferreira, A. D. Pablo and J. D. Rossi, Blow-up for a degenerate di¤usion problem not in divergence
form, Indiana Univ. Math. J., 55(2006), 955–974.
[7] B. E. Gilding and L. A. Peletier, On a class of similarity solution of the porous medium equation I, J.
Math. Anal. Appl., 55(1976), 351–364.
[8] A. Granas and J. Dugundji, Fixed Point Theory, Springer-Verlag, New York 2003.
[9] J. Hulshof and J. L. Vàzquez, Self-similar solutions of the second kind for the modi…ed porous medium
equation, European J. Appl. Math., 5(1994), 391–403.