Вы находитесь на странице: 1из 14

Meccanica 35: 119132, 2000.

c 2000 Kluwer Academic Publishers. Printed in the Netherlands.


Shock Propagation in a One-Dimensional Flow Through
Deformable Porous Media
E. COMPARINI
1
and M. UGHI
2
1
Universit` a di Firenze, Dipartimento di Matematica U. Dini, Viale Morgagni 67/A; 50134 Firenze, Italy
2
Universit` a di Trieste, Dipartimento di Scienze Matematiche, Piazzale Europa 1; 34127 Trieste, Italy
(Received: 3 November 1999; accepted in revised form: 5 June 2000)
Abstract. We consider a one-dimensional incompressible ow through a porous medium undergoing deforma-
tions such that the porosity and the hydraulic conductivity can be considered to be functions of the ux intensity.
The medium is initially dry and we neglect capillarity, so that a sharp wetting front proceeds into the medium. We
consider the open problem of the continuation of the solution possibly in the case of onset of singularities, which
can be interpreted as a local collapse of the medium. In particular we analyze the case in which the boundary
pressure has a piecewise constant derivative.
Sommario. Consideriamo il usso unidimensionale di un uido incomprimibile in un mezzo poroso deformabile,
in cui la porosit` a e la conduttivit` a idraulica dipendono dallintensit` a del usso. Trascurando fenomeni di capillarit` a,
una frontiera regolare penetra nella zona asciutta (inizialmente occupante lintero mezzo) dividendola dalla zona
bagnata. Studiamo il problema, nora aperto, della continuazione della soluzione nel caso di eventuali singolarit` a
interpretabili sicamente come collassi locali del mezzo. In particolare in questo lavoro viene analizzato il caso
in cui la pressione sul bordo sso ha derivata costante a tratti.
Key words: Conservation laws, Free boundary, Filtration, Incompressible ow.
1. Introduction
In a recent paper [4] the authors have considered a generalization of the classical Green-Ampt
model for the penetration of a wetting front in a dry porous medium [1, 5]. They assume that
the physical parameters k (hydraulic conductivity) and (porosity) depend on the volumetric
velocity q. The model takes thus into account the possibility of ow-induced deformations on
microscopic scale. In summary the model gives rise to the following free boundary problem:
p
x
=
q
k(q)
, 0 < x < s(t ), t > 0, (1.1)
q
x
+

(q)
q
t
= 0, 0 < x < s(t ), t > 0, (1.2)
p(0, t ) = p
0
(t ) > 0, t > 0, (1.3)
p(s(t ), t ) = 0, t > 0, (1.4)
s(t ) =
q(s(t ), t ))
(q(s(t ), t )
, t > 0, s(0) = 0, (1.5)
where p is the pressure, p
0
is a given bounded function, s is the thickness of the wet region.
120 E. Comparini and M. Ughi
In the Green-Ampt model and k are constants while here we assume that and k are
given functions of q such that (in agreement with physical intuition, see [4])
C
3
,

< 0,

0,
0
> 0 q > 0, (1.6)
k C
2
, k

0, k

0, k k
0
> 0 q > 0. (1.7)
From a mathematical point of view the problem (1.1)(1.5) is a free boundary problem for
a non-homogeneous reducible quasilinear hyperbolic system, which is strictly hyperbolic in
the assumption

< 0 (see [6, 7]). References about free boundary problems for hyperbolic
equations can be found in [11].
Let also mention that p and q are the Riemann invariants of the problem. The domain
{0 < x < s(t ), t > 0} is typical in the sense of Li Ta-Tsien (see [6]) because the characteristic
starting from the origin does not enter in the domain ( s > 0,

> 0), but the boundary


conditions (1.3) and (1.4) are not standard.
However, in the above-mentioned paper [4], local in time existence and uniqueness are
given, provided p
0
(0) = 0, but it is hinted that there is the possibility of the development of
singularities of the solution, which can be interpreted as a local collapse of the medium. In
such a case it is an open question whether and how the solution can be continued after the
onset of the above-mentioned singularities.
The aim of the present paper is to give a rst answer to this question in the particular case
in which the boundary pressure p
0
(t ) has a piecewise constant derivative. In this assumption
we will give an almost explicit solution to the problem (1.1)(1.5), containing the relevant
features of the phenomenon.
It turns out that there is a striking difference between the cases of p
0
(t ) increasing or
decreasing. In the rst case (see Section 2) we will have a solution with innitely many shocks
starting from the free boundary and travelling towards the surface x = 0, while in the second
case (see Section 3) the solution will be continuous for 0 x s(t ), t > 0. In both cases the
solution approaches the one of the Green-Ampt model with the same p
0
(t ) for t tending to
.
This type of qualitative behaviour of the solution is not unusual for this kind of hyperbolic
problems, see e.g. [7, 8].
The global existence for a general p
0
(t ) concave is given in Section 4. In a forthcoming
paper [2] the general case of p
0
(t ) convex will be dealt with the problem describing this case
that presents many difculties related to the possible onset of the mentioned singularities of
the solution (see Remark 4.1).
2. Explicit Example with Increasing Injection Rate
Through this and the following section we will assume that p
0
(t ) is given by
p
0
(0) = 0, p
0
(t ) =
_
p
0

, 0 < t < T
0
,
p
0
+
, t > T
0
,
(2.1)
where T
0
, p
0

, p
0
+
are given positive constants.
It turns out that we can construct explicit solutions whose behaviour is very different in the
two cases
(i) p
0

< p
0
+
(increasing injection rate),
(ii) p
0

> p
0
+
(decreasing injection rate).
(2.2)
Shock Propagation in a One-Dimensional Flow 121
In the rst case we will have a solution with shocks (precisely a piecewise constant solution),
in the second one the solution will be continuous for any t .
In this section we consider the case (i), i.e. p
0

< p
0
+
. Let us dene
q
0
= F
1
( p
0

), Q = F
1
( p
0
+
), (2.3)
where F(z) is an increasing function given by
F(z) =
z
2
k(z)(z)
. (2.4)
Let us remark that the solution of the Green-Ampt model (i.e. assuming and k to be constant)
corresponding to boundary data (2.1) is
q =
_

_
q
0
, 0 < t < T
0
,
F
1
_
_
_
_
_
_
p
+
0
_
1 +
p

0
p
+
0
T
0
t T
0
_
2
1 +
p

0
p
+
0
_
t
2
(t T
0
)
2
1
_
_
_
_
_
_
_
, T
0
< t.
(2.5)
In order to describe the solution in the case of varying with q we need the following two
lemmas, whose proof will be given after the statement of the main result in this section.
LEMMA 2.1. The following increasing sequence {q
i
} is uniquely dened
q
0
given by (2.3),
q
i+1
: p
0
+
= A
i
(q
i+1
) = F(q
i+1
) +f
i
(q
i+1
), i = 0, 1, . . .
(2.6)
where
f
i
(z) =
q
i
z
(q
i
) (z)
_
q
i
k(q
i
)

z
k(z)
_
. (2.7)
Moreover
lim
i
q
i
= Q.
Giving the sequence {q
i
} we can dene the following sequences:

i
=
q
i1
q
i
(q
i1
) (q
i
)
< 0, i = 1, 2, . . . (2.8)
_

_
s(T
0
) =
q
0
(q
0
)
T
0
,
s(T
i+1
) = s(T
0
)

i
k=1
_
1
q
k

k
_
= s(T
i
) +
q
i+1
(q
i+1
)
(T
i+1
T
i
),
(2.9)
_
_
_
T
0
given in (2.1),
T
i+1
= T
i

s(T
i
)

i+1
, i = 0, 1, . . .
(2.10)
We will prove the following:
122 E. Comparini and M. Ughi
Figure 1. q(x, t ).
LEMMA 2.2. The sequence {T
i
} is increasing and
lim
i
T
i
= +, lim
i
T
i+1
T
i
= +.
We are now able to give the explicit solution of the problem in terms of the sequences just
dened, in the following Theorem:
THEOREM 2.1. In the assumptions (2.1) and (2.2)(i), and given the sequences of Lemma 2.1
and (2.8)(2.10) we have
(i) there exists a piecewise constant solution for any t > 0, given by
q = q
0
, s =
q
0
(q
0
)
t, 0 < x < s(t ), 0 < t < T
0
,
q(x, t ) =
_
q
i
, 0 < x <
i+1
(t ),
q
i+1
,
i+1
(t ) < x < s(t ),
T
i
< t < T
i+1
, i = 0, 1, . . .
s(t ) = s(T
i
) +
q
i+1
(q
i+1
)
(t T
i
),
i+1
(t ) = s(T
i
) + (t T
i
)

i+1
.
(ii) For t + we have
q
i+1
q
i
0 as t ,
so that (see Lemma 2.1) q(x, t ) Q for t , uniformly on compact subsets.
(iii) For t +
s(t )
Q
(Q)
,
q
0
(q
0
)
t < s(t ) <
Q
(Q)
t.
The meaning of this solution is that at every T
i
the material drastically changes at the free
boundary s(T
i
) and this change propagates with nite speed

i+1
towards the surface x = 0;
when the shock line reaches the surface at t = T
i+1
there is another instantaneous drastic
change on the free boundary at x = s(T
i+1
) and the process goes on repeatedly.
We have an innite sequence of shocks (i.e. of drastic changes in the material), each
travelling from the free boundary to the surface with nite speed.
Shock Propagation in a One-Dimensional Flow 123
Figure 2. Domain.
We remark that both the solution of the Green-Ampt model (2.5) (which is continuous at
T
0
also in the case of p
0
increasing) and the one of the preceeding example tend to the same
value Q for t .
The solution q(x, t ) in this case is represented in Figure 1; whereas in Figure 2 we have
the domain, with the shock lines
i
. Concerning the gures in the paper we refer to [9].
Let us begin with the proof of Lemma 2.1.
Proof of Lemma 2.1. By induction: assume that q
0
< q
i
< Q, we will prove that there
exists a unique q
i+1
solution of (2.6) and that
q
0
< q
i
< q
i+1
< Q.
First of all let us remark that f
i
(z) dened in (2.7) is increasing in z for z > q
i
.
Since F(z) is increasing in z and lim
zq
i
+ f
i
(z) = 0, we have that A
i
(z) is increasing in z
and lim
zq
i
+ A
i
(z) = F(q
i
), A
i
(z) > F(z), for z > q
i
. But by assumption p

0
= F(q
0
) <
F(q
i
) < F(Q) = p
+
0
. Therefore, there exists a unique q
i+1
> q
i
such that (2.6) holds and
q
i+1
< Q. So the sequence {q
i
} is well dened and increasing; since it is bounded above by
Q it has to tend to a nite limit q. But by the regularity assumptions made on and k we have
lim
i
f
i
(q
i+1
) = 0. (2.11)
In fact
f
i
(q
i+1
) =
1

(
i
)
d
dz
_
z
k
_

z=
i
(q
i+1
q
i
)
where q
i
<
i
< q
i+1
, q
i
<
i
< q
i+1
.
Since
lim
i

(
i
)
d
dz
_
z
k
_

z=
i
= C <
(2.11) holds. Now from (2.6) we get
p
+
0
= F(Q) = F( q) + lim
i
f
i
(q
i+1
) = F( q)
and hence q = Q, thus ending the proof of Lemma 2.1.
124 E. Comparini and M. Ughi
Proof of Lemma 2.2. Let us rst prove that the sequence {

i
} dened in (2.8) is decreasing
and
lim
i

i
=
1

(Q)
. (2.12)
We can write

i
=
1

(
i
)
, q
i1
<
i
< q
i
, i = 1, 2, . . . (2.13)

i+1
=
1

(
i
)
,
i
< q
i
<
i
< q
i+1
.
Since is convex

(
i
)

(
i
) < 0,

i
is decreasing; the limit in (2.12) follows immediately
from Lemma 2.1 (q
i
Q) and (2.13). Since

i+1
< 0 we have that T
i
and s(T
i
) are
increasing sequences; moreover, it follows from (2.10) that
T
i+1
T
i

s(T
0
)

i+1
> s(T
0
)

(Q) > 0. (2.14)


Therefore T
i
+ as i .
From (2.9) in a similar way, taking into account that z/(z) is an increasing function, we
get
s(T
i+1
) s(T
i
) >
q
0
(q
0
)
(T
i+1
T
i
)
q
0
(q
0
)
s(T
0
)

(Q) > 0. (2.15)


Therefore also s(T
i
) + as i .
Again from (2.10) we then get that
T
i+1
T
i
s(T
i
)

(Q)
and hence T
i+1
T
i
+, thus concluding the proof of Lemma 2.2.
Let us remark that from (2.9) it also follows that
q
0
(q
0
)
T
i
s(T
i
)
Q
(Q)
T
i
, i. (2.16)
Let us now consider the main theorem
Proof of Theorem 2.1. We have
q(x, t ) = q
0
, s(t ) =
q
0
(q
0
)
t, 0 x s(t ), 0 < t < T
0
.
In t = T
0
, p(t ) increases and so q(s(t ), t ) has to increase abruptly. Hence there is a shock line

1
(t ) starting in x = s(T
0
), t = T
0
and we look for a solution of the form
q =
_
q
0
, 0 < x <
1
(t )
q
1
,
1
(t ) < x < s(t )
T
0
< t < T
0
+, (2.17)
Shock Propagation in a One-Dimensional Flow 125
where (from the usual shock conditions [3]) we have

1
=
q
0
q
1
(q
0
) (q
1
)
,
1
= s(T
0
) +

1
(t T
0
),
s(t ) = s(T
0
) +
q
1
(q
1
)
(t T
0
), (2.18)
and q
1
has to be such that
p
0
(t ) =
_
s
0
q
k(q)
dx =
1
_
q
0
k(q
0
)

q
1
k(q
1
)
_
+s
q
1
k(q
1
)
,
that is (since s = q
1
/(q
1
) and

1
is given in (2.18))
p
+
0
= A
0
(q
1
) = F(q
1
) +f
0
(q
1
),
that is (2.6) with i = 0 (see also (2.4) and (2.7) for the denition of F and f
i
). So by Lemma
2.1 we have that there exists a unique q
1
satisfying (2.6) and q
0
< q
1
< Q.
This solution can be continued only up to the time T
1
at which
1
reaches x = 0, that is
T
1
= T
0

s(T
0
)

1
.
In fact, if we assume that for t > T
1
, q q
1
, then to satisfy the condition
p
0
(t ) =
_
s
0
q
1
k(q
1
)
dx
we should have p
0
+
= F(q
1
), i.e. q
1
= Q while q
1
< Q.
This means (since q(x, T
1

) q
1
) that, at time T
1
, q(s(T
1
), T
1
) has to change abruptly
and a new shock line x =
2
(t ) starts at x = s(T
1
), t = T
1
. So we look for a solution as is
given in the Theorem 2.1 with i = 1 and q
2
has to satisfy condition (2.6) of Lemma 2.1 with
i = 1. Repeating the argument at each step we get point (i) and then by means of the Lemmas
2.1 and 2.2 points (ii) and (iii) (recalling that s = q/).
3. Explicit Example with Decreasing Injection Rate
In this section we consider the case p
+
0
< p

0
. We dene q
0
, Q according to (2.3) and (2.4):
now we have Q < q
0
.
Also in this case we need rst the following.
LEMMA 3.1. The following decreasing sequence {q
i
} is uniquely dened
q
0
= F
1
( p

0
),
q
i+1
: p
+
0
= B
i
(q
i+1
) = F(q
i+1
) +G(q
i+1
) G(q
i
), i = 0, 1, . . .
(3.1)
where F is given by (2.4) and G(z) is such that
G

(z) =
1

(z)
d
dz
_
z
k
_
> 0, G(0) = 0. (3.2)
Moreover lim
i
q
i
= Q.
126 E. Comparini and M. Ughi
Proof. Assuming that Q < q
i
< q
0
, we prove that there is a unique solution q
i+1
of (3.1)
and Q < q
i+1
< q
i
. Fix q
i
and consider the function B
i
(z). For z < q
i
, B
i
is an increasing
function (since F

+G

> 0) and B
i
(q
i
) = F(q
i
) so that B
i
(z) < F(z) for z < q
i
.
Therefore, since by assumption p
+
0
< F(q
i
) < p

0
, there exists a unique q
i+1
such that
(3.1) holds and q
i+1
< q
i
. Moreover, since
F(Q) = p
+
0
= B
i
(q
i+1
) < F(q
i+1
),
we have q
i+1
> Q. So the sequence q
i
is decreasing and bounded below by Q and con-
sequently it converges to a limit q.
Since G is continuous we have from (3.1) that
p
+
0
= F( q) +G( q) G( q) = F( q)
so that q = Q.
Now let us begin to describe the qualitative behaviour of the solution. The discontinuity of
p
0
(t ) in T
0
causes now (since q(s(t ), t ) has to be decreasing) the start of a rarefaction wave in
s(T
0
), T
0
. Then for an interval starting at T
0
and to be dened later, the solution is continuous
and the relation between q(0, t ) = q
0
(t ) and q(s(t ), t ) = q
s
(t ) is given as in [4] by (see (2.4)
and (3.2))
p
+
0
= F(q
s
) +G(q
s
) G(q
0
) = H(q
s
) G(q
0
). (3.3)
Since for T
0
< t < T
0
+ , q
0
(t ) = q
0
then q
s
(t ) has to be a constant satisfying (3.1) with
i = 0. By Lemma 3.1 then q
s
(t ) is given by q
1
. To see how long this solution can be prolonged
let us dene the two characteristic lines
1

,
1
+
starting in s(T
0
), T
0
, as follows:

(t ): t =

(q
0
)(x s(T
0
)) +T
0
, (3.4)

1
+
(t ): t =

(q
1
)(x s(T
0
)) +T
0
.
Since is convex and q
1
< q
0
(then

(q
1
) <

(q
0
) < 0) we have

1
(t ) <
+
1
(t ) < s(t ), T
0
< t < T
0
+,
where
s(t ) = s(T
0
) +
q
1
(q
1
)
(t T
0
). (3.5)
Denoting by T
1

, T
1
+
respectively, the times in which
1

,
1
+
reach the surface x = 0, that
is,
T

1
= T
0

(q
0
)s(T
0
), T
+
1
= T
0

(q
1
)s(T
0
), (3.6)
with T
1

< T
1
+
. Then by the classical theory of rst order equations (see [3]) we have that in
the triangle
1

1
=
_

1
(t ) x
+
1
(t ), T
0
< t < T

1
,
0 x
+
1
(t ), T

1
< t < T
+
1
,
(3.7)
q(x, t ) is dened as the unique solution of the following equation

(q) =
t T
0
x s(T
0
)
in
1
. (3.8)
Shock Propagation in a One-Dimensional Flow 127
It follows immediately that in
1
, q
1
< q < q
0
and q is decreasing in x. Moreover, for
T
0
< t T
1

, q is constant outside
1
so that
q
s
(t ) = q
1
for T
0
< t T
1

. (3.9)
Therefore, denoting by
2

the characteristic starting from s(T


1

), T
1

, i.e.

2
(t ): t =

(q
1
)(x s(T
1

)) + T
1

, (3.10)
and by T
2

the time in which


2

reaches x = 0, i.e.
T

2
= T

(q
1
)s(T

1
), (3.11)
we have that
q q
1
in D
1
, (3.12)
D
1
=
_

1
+
(t ) x s(t ), T
0
< t < T
1

1
+
(t ) < x <
2

(t ), T
1

< t < T
1
+
,
0 x
2

(t ), T
1
+
< t < T
2

,
where of course (see (3.5))
s(t ) = s(T
0
) +
q
1
(q
1
)
(t T
0
), T
0
< t < T
1

. (3.13)
As for the interval T

1
< t < T
+
1
we have, according to (3.7) and (3.8), that q
0
(t ) is dened
through the equation

(q
0
(t )) =
t T
0
s(T
0
)
, T
1

< t < T
1
+
, (3.14)
so that
q
1
< q
0
(t ) < q
0
, lim
t T
1
+
q
0
(t ) = q
1
, T
1

< t < T
1
+
.
Therefore, according to the denition of q
s
(3.3), q
s
is the solution of
p
+
0
= H(q
s
(t )) G(q
0
(t )), (3.15)
with q
0
(t ) given by (3.14).
Then we have
lim
t T
1
+
q
0
(t ) = q,
with q satisfying (3.15) with q
0
(t ) = q
1
, that is by Lemma 3.1 q = q
2
.
Since q
0
(t ) = q
1
, for T
1
+
< t < T

2
(see (3.12)), we have (still from (3.3)):
q
s
(t )) q
2
, T
1
+
< t < T
2

. (3.16)
Let us remark that, since q
s
(t ) is given by (3.15) in T
1

< t < T
1
+
, also s(t ) is given in the
same interval as
s(t ) = s(T

1
) +
_
t
T

1
q
s
()
(q
s
())
d, T

1
t T
+
1
. (3.17)
128 E. Comparini and M. Ughi
In particular, s(T
+
1
) is given by
s(T
+
1
) = s(T

1
) +
_
T
+
1
T

1
q
s
(q
s
)
d. (3.18)
Starting from x = s(t ), T

1
t T
+
1
, we can solve the Equation (1.2) along the characteristics
in a strip bounded on the right by the characteristic starting from (s(T
+
1
), T
+
1
) on which q has
the value q
2
. Let us denote it by
+
2
, i.e.

+
2
(t ): t =

(q
2
)(x s(T
+
1
)) +T
+
1
,
+
2
>

2
(3.19)
and by T
+
2
the time in which it reaches x = 0, i.e.
T
+
2
= T
+
1

(q
2
)s(T
+
1
). (3.20)
Hence, q
0
(t ) is given for T

2
< t < T
+
2
, and it decreases from q
1
to q
2
.
Then one can repeat step by step the previous argument (since

(q) is decreasing and the


characteristics enlarge toward x = 0) and dene q(x, t ) and s(t ) (increasing) for any t > 0.
A typical feature of this solution is that it is constant on innitely many strips (of the type of
D
1
(3.12)). In fact, since q
s
is constant on intervals of the type (3.9) and (3.16), q has to be
constant on the characteristic starting from the free boundary in that time intervals.
To be more precise let us dene the following sequences, starting from the sequence {q
i
}
of Lemma 3.1
T

i+1
= T

(q
i
)s(T

i
), T
+
i+1
= T
+
i

(q
i+1
)s(T
+
i
), i = 1, 2, . . . (3.21)
while for i = 0 we have the denition (3.6) (Let us remind that the quantities involved in the
previous equation can be calculated explicitly).
Clearly we have from Lemma 3.1
T
+
i
< T

i+1
, T

i
< T
+
i
, i = 1, 2, . . .
Then we have the following.
THEOREM 3.1. In the assumptions (2.1) and (2.2)(ii) and given the sequences (3.1) and
(3.21), q(x, t ) and s(t ) are continuous and
Q < q(x, t ) q
0
,
Q
(Q)
t < s(t ) <
q
0
(q
0
)
t,
as for q
s
(t ) we have
q
s
(t ) =
_

_
q
0
, 0 < t < T
0
,
q
1
, T
0
< t T

1
,
q
i+1
, T
+
i
t T

i+1
, i = 1, 2, . . .
(3.22)
Moreover, the sequences T

i+1
T
+
i
, T
+
i
T

i
are increasing as i + so that
lim
i
T

i
= , lim
i
T
+
i
= +.
Shock Propagation in a One-Dimensional Flow 129
Figure 3. q(x, t ).
Proof. The rst part comes from the previous argument repeated step by step and from
Lemma 3.1. The second part, concerning the sequences of T

i
comes immediately from the
following estimates, in which we use the fact that is convex so that

(q
i+1
) <

(q
i
) < 0
T

i+1
T
+
i
= T

i
T
+
i1

(q
i
)(s(T

i
) s(T
+
i1
))
> T

i
T
+
i1
(3.23)
T
+
i+1
T

i+1
= T
+
i
T

(q
i+1
)s(T
+
i
) +

(q
i
)s(T

i
)
> T
+
i
T

(q
i
)(s(T
+
i
) s(T

i
))
> T
+
i
T

i
. (3.24)
Let us nally remark that q
s
(t ) is not constant in the intervals T

i
< t < T
+
i
i = 1, 2, . . . ,
actually in such intervals q
s
(t ) decreases from q
i
to q
i+1
.
While in the case p
0
increasing, there was an innite sequence of discontinuities for q
s
in
the present case of p
0
decreasing we have that q
s
has only one discontinuity at t = T
0
and is
continuous afterwards. The jump of p
0
(t ) gives an innite sequence of discontinuities only on
q

s
(t ).
For Cauchy problem of hyperbolic equations this is a typical feature because of the entropy
condition, so that an increasing jump gives a shock wave, while a decreasing jump gives a
rarefaction wave.
In Figures 3 and 4 we can see the solution q(x, t ) and the domain (with the characteristic
lines

i
,
+
i
, dividing the zones where q is constant from the zones where q decreases).
4. General Case (with p
0
(t ) p
0
(t ) p
0
(t ) Decreasing)
We assume here p
0
(t ) is decreasing t > 0.
From the paper [4], we have existence in a small interval of time [0, T
0
] with q
s
(t ) =
q(s(t ), t ) decreasing in t . This implies that (q
s
(t )) is increasing so that the characteristic
130 E. Comparini and M. Ughi
Figure 4. Domain.
lines tend to diverge towards x = 0. Precisely the relation between q
s
(t ) and q
0
(t ) = q(0, t )
is given by (3.3), which we rewrite here for the sake of clearness
H(q
s
) = p
0
+G(q
0
(t )), H(q
s
) = F(q
s
) +G(p
s
), (4.1)
F, G given in (2.4) and (3.2).
The equation of the characteristic lines starting from x = s(), t = is
t =

(q
s
()(x s()) + . (4.2)
Hence the time in which the above characteristic reaches the xed boundary x = 0 is given
by the following:
=

(q
s
())s() +. (4.3)
Hence, we have
d
d
= 1
d
d
(

(q
s
())s()). (4.4)
In the case p
0
(t ) decreasing we have from the result of [10] that for 0 < t < T
0
q
0
(t ) is
decreasing. Then from (4.1) q
s
is decreasing in the same interval so that from (4.4) we have
d/d > 1 (which means that the characteristics diverge towards x = 0). Therefore, there can
be no shock in the region
R
1
= {0 < x <
1
(t ), 0 < t < T
1
},
where

1
(t ): t =

(q
s
(T
0
))(x s(T
0
)) +T
0
, (4.5)
T
1
= T
0

(q
s
(T
0
))s(T
0
) (4.6)
(i.e.
1
is the characteristic starting at s(T
0
), T
0
and T
1
is the time in which
0
reaches the
surface x = 0).
Shock Propagation in a One-Dimensional Flow 131
We can then retrieve q
0
(t ) for T
0
< t < T
1
from q
s
() in 0 < < T
0
, solving the equation
along the characteristics. Since q
s
is decreasing in that interval so is q
0
(t ) for 0 < t < T
1
. We
can then dene q
s
(t ) for T
0
< t < T
1
from (4.1), and q
s
is again decreasing. Then again there
are no shocks in the whole region
R = {0 < x < s(t ), 0 < t < T
1
},
and the solution is classical there.
Repeating this argument up to a generic time T we get existence in large for the problem.
So we have proved the following.
THEOREM 4.1. If p
0
0 then there exists a classical solution of the problem (1.1) for any
T > 0.
Let us remark that in this case s(t ) is concave.
Remark 4.1. The situation is quite different if p
0
(t ) is increasing (see Section 2). In this
case q
s
(t ) is increasing for 0 < t < T as q
0
(t ), hence from (4.2) the characteristics get closer
one to the other and eventually we can have shocks in R
1
dened in (4.5), unless d/d > 0,
0 < < T
0
.
Also if there are no shocks in R
1
one can easily nd examples of p
0
(t ) growing fast enough
so that q
s
(still dened from (4.1) for t > T
0
) increases in such a way to form a shock in the
region R
2
= R \ R
1
at a certain time T

, T
0
< T

< T
1
. If this happens one has to redene
a solution with shock and of course the relation (4.1) does not hold anymore and one has to
redene the relation between q
s
(t ) and q
0
(t ) for t > T

. This can be done and will be the


subject of a following paper [2]. Let us only remark here that the main problem to continue
the solution arises when the shock line originated for t = T

reaches the surface x = 0, say


at T =

T . Then (see the example of Section 2) one has that q
s
(t ) has to be discontinuous at
t =

T so that a new shock line arises at the point s(

T ),

T . Then one has to prove existence


again with a xed point argument for t >

T . Anyhow even if only one shock is present at each
time, one will have an innite sequence of shock lines starting at the free boundary x = s(t )
and proceeding to the surface x = 0 as the example of Section 2 shows clearly. The physical
interpretation is the same as the one given in Section 2.
Acknowledgement
Work partially supported by the Italian MURST National Project Problemi non lineari. . .
References
1. Bear, J., Dynamics of Fluids in Porous Media, America Elsevier, New York, 1972.
2. Comparini, E. and Ughi, M., On the existence of shock propagation in a ow through deformable porous
media, (To appear).
3. Evans, L.C., Partial Differential Equations, Berkeley Mathematics Lecture Notes, 1994.
4. Fasano, A. and Tani, P., Penetration of a wetting front in a porous medium with ux dependent hydraulic
parameters, In: K. Cooke et al. (eds.), SIAM, 1995.
132 E. Comparini and M. Ughi
5. Green, W. and Ampt, G., Studies on soil physics. The ow of air and water through soils, J. Agric. Sci. 4
(1911) 124.
6. Ta-tsien, Li and Wen-ci, Yu, Boundary Value Problem for Quasilinear Hyperbolic Systems, Duke University
Mathematics Series V, 1985.
7. Ta-tsien, Li, Global Classical Solutions for Quasilinear Hyperbolic Systems, Masson, Wiley, 1994.
8. Ta-tsien, Li and Yan-chun, Zhao, Global shock solutions to a class of piston problems for a system of
one-dimensional isentropic ow, Chin. Ann. Math. 12B (1991) 495499.
9. Mancini, A., A discontinuous in time stabilized Galerkin approach for an hyperbolic system with a free
boundary, (To appear).
10. Tani, P., Fronti di Saturazione in Mezzi Porosi con Caratteristiche Idrauliche Variabili, Tesi, Univ. Firenze,
1994.
11. Terracina, A., A free boundary problem for scalar conservation laws, SIAM J. Math. Anal. 5 (1999)
9851009.

Вам также может понравиться