| Academic Open Internet Journal |
Volume 12, 2004 |
WAVE TYPE SOLUTIONS FOR THE SYSTEMS OF COMPRESSIBLE
ADIABATIC FLOW THROUGH POROUS MEDIA
Galina
St. Panayotova
Burgas, Bulgaria
Abstract
:
In this paper we study the problem for existence of wave type solution for the quasi-linear system of compressible adiabatic flow through porous media. They are often considered in a variety of physical problems, where one encounters source terms that are balanced by internal forces.
It
turns out that the problem for the existence of such solutions reduces to the
problem of investigation of systems of Frobenius type, which is equivalent to a
system of ordinary DEs and can be treated by standard methods.
1991 AMS Subject Classification
35L40, 35L65
Key words and phrases: quasi-linear systems of PDEs, simple waves, Pfaff system
1.Introduction
In
1-dimensional porous media, the motion of compressible adiabatic flow can be
modeled by the compressible Euler equations with frictional damping terms. For
smooth solutions, we have the following balance laws:
(1)
vt – ux = 0,
ut + p(v,s)x
= -au, a>0,
st = 0.
Here,
v>0 is the specific volume, u is the velocity, s stands for entropy, p
denotes the gas pressure with pv(v,s)<0 for v>0
and a>0 is a constant. (See[4].)
In this case we investigate the
existence wave type solutions for the above model of hyperbolic system.
Definition
1. (see e.g.[2]) We say that a solution (u(t,x),v(t,x),s(t,x))
of the system (1) is constructed by means of Riemann invariants if it is of the
form
(2)
u = U(R(t,x)),
v = V(R(t,x)), s = S(R(t,x)),
where
U(R), V(R) and S(R) are functions of a single variable and R(t,x)
is a suitable function, called Riemann invariant .
We also say that solutions constructed by means of Riemann invariant are wave type solutions.
Note that the classical wave solutions are special case of the wave
type solutions and that the class of
the wave type solutions is the minimal class of solutions, which contains the
classical wave solutions and is invariant with respect to the changes of
variables.
Such
solutions are called simple waves when the systems (1) is homogeneous and simple states when (1) is non-homogeneous.
Lemma 1. [3]. The functions u, v and s
can be represented in the form (2) if and only if
utvx – uxvt
= 0
stvx – sxvt
= 0.
Note
that if the condition of Lemma1 is satisfied and if u is nondegenerated, i.e.if Ñu ¹ 0, then u is
a Riemann invariant .
3.The
main result
Theorem 1. (u,v,s) is a solution of the system (1), constructed by means of
Riemann invariant, exactly when (u,v,s) is a solution of the system
(3)
vt – ux = 0,
ut + p(v,s)x
= -au, a>0,
st = 0,
utvx
– uxvt = 0,
stvx
– sxvt = 0.
Proof. The proof of this theorem follows directly from Definition1 and Lemma1.
Let
ux = y and
K = K(u,v,s) = -au + p(v,s)x .
Now (3) can be represented in the form
(4)
ux = y
ut = K
vx = y2/K
vt = y
sx = 0
st = 0.
Therefore, s = const. In the case of isentropic flow where s = const.,(1)
takes the form
vt – ux = 0,
ut + p(v,s)x
= -au, a>0.
For
the investigation of (4) we will apply a method described in [6] and used in
[3], [5] and [7].
The associated with (4) Pfaff system
in R5 has the form
(5)
w1(dz) = du – Kdt – ydx
w2(dz) = dv – ydt – y2/Kdx
where dz
= (dt,dx,dy,du,dv).
Theorem
2. 1) The system (5)
has a solution if and only if
there exist vector fields h1, h2Î R5 such
that wi (hk) = 0 (I,k = 1,2) and
[h1,h2] belongs to the linear hull
of h1 and h2 .
2)
If FI (I=1,2,3) denote
functionally independent solutions of the system hiF=0 (I=1,2) (whose existence
follows from the condition 1) in view of Frobenius’Theorem ), then the system
of equations FI = CI (I=1,2,3) ,
where CI are suitable constants determined by the initial functions
yo = yo(t,x) , uo = uo(t,x)
and vo = vo(t,x),
then namely the pair (u,v) is a solution of system (4).
Proof. Theorem 2 is a particular
case of the classical theory of Pfaff’s systems.
So our initial problem reduces to the problem of finding suitable vector
fields h1 and h2 with properties, described in the
conditions 1) of Theorem 2 . This is a particular case of the general Pfaff
problem.
Consider system (5) as an algebraic
system of equations with respect to coordinates of the field dz.
Since its rank equals 2, the set of its solutions form a linear space (
depending on the point z = (t,x,y,u,v) ), called distribution ;
denote this distributions by q(z).
Definition
2 [6]. Every involutive
subdistribution qi(z) of the
distribution q(z) is called resolving
distribution for system (5).
Comparing this definition with
condition 1) of Theorem 2, we conclude that in order to reach our goals we must
find a resolving distribution of dimension 2 for system (5).
The following three linearly
independent vector fields
(6)
x1 = (0,0,1,0,0), x2 = (0,1,0,y, y2/K), x3 = (1,0,0,K,y)
satisfy the Pfaff system (5), i.e. wi (xk) = 0 (i = 1,2; k = 1,2,3).
The linear hull of xj (j = 1,2,3) determines a
three-dimensional distribution q(z). If we choose a pair of linearly independent fields h1, h2Îq(z), then their linear hull determines a two-dimensional
subdistribution q1(z)Îq(z). Thus if [h1, h2] Î q1(z), then q1(z) is involutive and from the classical Frobenius’Theorem it follows
that q1(z) is completely integrable subdistribution of q(z). Therefore the system hiF =0 (i=1,2)
possesses three functionally independent solutions Fk (k=1,2,3) . So our first task is to build
a basis consisting of a pair of vector fields h1, h2, generating an involutive
two-dimensional subdistribution q1(z) of q(z).
The vector fields x1 , x2 , x3 defined above form a basis of the
distribution q(z). Then the following theorems give us a way to find a pair of suitable
vector fields h1(z), h2(z).
Theorem 3. There exists exactly one vector field
h1 , satisfying the system
(7)
wi (dz) = 0; ¶w1 (xj,dz) = 0 (I = 1,2; j
= 1,2,3)
and
exactly one vector field h2 , satisfying the system
(8)
wi (dz) = 0; ¶w2 (xj,dz) = 0 (I = 1,2; j
= 1,2,3)
If
the system (5) has 2-dimensional resolving distributions q1(z) , then h1Îq1(z) and h2Îq1(z).
Proof.
For a proof of Theorem 3 see [5] and [6].
Theorem 4. If the system (5) has a two-dimensional resolving distribution q1(z) , then the vector fields
(9)
h1 = ( -y, K , 0, 0, 0 )
h2 = ( 0, 1, ry/s, y, y2/K
)
where x1 , x2 , x3 are defined
by (6), belong
to q1(z) and s = K – yKy,
r = y(yKv
+ KKu)/K.
Proof. We have [x1 , x2 ] = ¶yx2 = (0, 0, 0, 1, 2y/K - y2Ky
/K2) ;
[x1 , x3] = ¶yx3 = (0, 0, 0, Ky, 1 ) ;
[ x2 , x3] = ¶x2x3 - ¶x3 x2 = (0, 0, 0, r, yr /K).
Hence using the identity ¶w(X,Y) = Xw(Y) – Yw(X) – w([X,Y]), we obtain:
¶w1(x1 , x2) = -¶w1([x1 , x2]) = -1;
¶w1(x1 , x3) = -¶w1([x1 , x3]) = -Ky ;
¶w1(x2 , x3) = -¶w1([x2 , x3]) = -r ;
¶w2(x1 , x2) = -¶w2([x1 , x2]) = -2y/K + y2Ky /K2;
¶w2(x1 , x3) = -¶w2([x1 , x3]) = -1 ;
¶w2(x2 , x3) = -¶w2([x2 , x3]) = -yr/K.
Therefore,
solving (7) and (8), we get:
h10 = rx1 - Kyx2+ x3 ; h20 =(yr/K)x1 + x2 + (-2y/K + y2Ky
/K2)x3 .
According
to Theorem 3, h10 and h20 belong to q1(z).
We
define h1 = yh10 -h2 ; h2 = h10 + (K/y)h20 , which can be written in
coordinate basis as (9).
If [h1, h2] Îq1(z), then q1(z) is involutive and from
Frobenius’ theorem it follows that q1(z) is a completely integrable subdistribution of q(z).
The proof of Theorem 4 is complete.
Theorem
5. The subdistribution q1(z)Îq(z) defined as a linear hull of the vectorial
fields hi(x) (i =1,2) is involutive and therefore the system (1)
has simple wave solutions.
Proof. The subdistribution q1(x) spanned by the pair of vectorial fields h1 and h2 is invilutive if and only if
there exists a linear dependence between h1, h2 and [h1, h2] i.e. if the rank of the matrix M º (h1, h2 , [h1, h2] ) equals 2. Indeed [h1, h2] = (1, -Ky, -s/y, 0, 0 ) . Therefore the rank of the M
is 2.
The proof of Theorem 5 is complete.
Having in mind the classical
Frobenius Theorem we conclude that wave type solutions of the problem (1) do
always exist. The Frobenius system is:
(10)
h1F = Ft - ayFy - auFu + yFv = 0
h2F = -yFt - auFx = 0
Further
we continue with the special case when the system (1) is of the form
(11)
vt – ux = 0
ut = -au,
The
fields h1 and h2 in this case are
h1 = ( 1, 0, yKu, K, y )
h2 = ( -y, K , 0, 0, 0
)
They
lead to the Frobenius system
(12)
h1F = Ft - ayFy - auFu + yFv = 0
h2F = -yFt - auFx = 0
which
has the following three functionally independent solutions Fi = Ci (i=1,2,3), where CI are suitable constants determined by
the initial values.
y + av = C1 and y = C2u
Therefore we have C2u + av = C1
. The second integral gives us a
connection between y and u, i.e. u = eC2X+C(t) , where C(t) is an arbitrary differentiable
function of t. For v we obtain
of (11) v = (-1/a) eC2X. The solutions (u,v) is a
general wave type solutions for the system (11).
References
[1]. Grundland A. “Riemann invariants. In: Wave
phenomena: Modern theory and applications” , North-Holand Math. Studies 97,
Elsevier, New York, 1984.
[2]. Jeffrey A. “Equations of evolution and waves” In: :Wave phenomena
modern theory and applications” ,North-Holand Math. Studies 97, Elsevier, New
York, 1984.
[3]. Kolev D. “Simple states of
non-linear electric field model in R2.
Mathematika Balkanika, 12 (1998), 315-320.
[4]. Pan R. ”Boundary effects and large time behavior for the
system of compressible adiabatic flow through porous media” Michigan Math.J.49,
2001, 519-540.
[5]. Panayotova G. ”Simple waves for certain
hyperbolic systems of PDEs” “International
Journal of Differential Equations and Applications” Volume 6 No.3 2002.
[6]. Tabov
J. “Simple
waves and simple states in R2” Jornal of Math. Analysis and
Applications”, 214(1997), 613-632.
[7]. Tabov J. and Kolev D. “Solvability of quasi-linear systems
in R2.with time-depending coefficients”. Compt. rend. de
l’Acad. bulg. des sci.,49(1996), 9-10, 35-38.
Technical College - Bourgas,