1. Introduction
In the paper [2] the authors applied the viscosity-flux approximation method coupled with the maximum principle to obtain the a-priori time-independent L∞ estimates for the approximation solutions of the following compressible polytropic gas dynamics system with an outer force:
with bounded measurable initial data
where the pressure function takes P = ργ, γ > 1 denotes the polytropic exponent, ρ is the density of gas and u is the velocity. The source α(x, t) is a function of the space variable x and the time variable t. Based on the compactness framework from the compensated compactness theory [11], [9], [4], [5], the pointwise convergence of the viscosity-flux approximation solutions and the global existence of the bounded entropy solutions for the Cauchy problem (1)-(2) were proved under the condition
where T(t) ∈ L1 (0, +∞), X(x) ∈ L1 (−∞, +∞).
In this paper, we study the isothermal gas, which is corresponding to the case of γ = 1. System (1) first appeared in [12]. When α(x, t) = 0, γ = 1, the study of the homogeneous system
has a long history. Under the Lagrangian coordinates, system (4) is equivalent to the system
where v = is the specific volume.
The first large data existence theorem for the Cauchy problem (5) with locally finite total variation initial data away from the vacuum (v = ∞) was obtained in [10] by using the Glimm’s scheme method (cf. [1]). The ideas of compensated compactness developed in [9], [11] were used in [3] to establish a global existence theorem for the Cauchy problem (4) with arbitrary large initial data including the vacuum (ρ = 0), with the use of the viscosity method.
To study the Cauchy problem (1)-(2), following [6], we consider the parabolic system
with initial data
where δ > 0 denotes a regular perturbation constant, and ϵ > 0 is the viscosity coefficient. We mainly have the following theorem in this paper,
Theorem 1.1. I. Suppose α(x, t) is a suitable smooth function and it satisfies (3), where |X(x)|L1(R) + |T(t)|L1(R+) ≤ . Let the initial data (2) satisfy
where
are the Riemann invariants of (4)and M > 1 is a constant.Then, for fixed ϵ, δ, the Cauchy problem (6) and (7)has a global solution (ρϵ,δ, uϵ,δ), ρϵ,δ ≥ 2δ > 0, satisfying
II. There exists a subsequence of (ρϵ,δ(x, t), uϵ,δ(x, t)), which converges pointwisely to a pair of bounded functions (ρ(x, t), u(x, t)) as δ, ϵ tend to zero, and the limit is a weak entropy solution of the Cauchy problem (1)-(2).
Remark. To construct the approximate solutions of the Cauchy problem (1)-(2), besides adding the classical viscosity parameter ϵ to the right-hand side of (1), we approximate the flux by adding a term depending on another parameter δ > 0 which vanishes when δ = 0. An obvious advantage of this kind of approximation on the flux functions is to obtain the positive lower bound ρ ≥ 2δ > 0, directly from the first equation in (6), which grantees that the term ρu2 = is well defined. Moreover, for any weak entropy-entropy flux pair (η(ρ, m), q(ρ, m)) of system (4), we can easily prove that
with respect to the viscosity solutions (ρϵ,δ, mϵ,δ) (cf. [7]).
2. Proof of Theorem 1.1
In this section, we shall prove Theorem 1.1. Multiplying (6) by and , respectively, we have
and
where
are two eigenvalues of the approximation system (4), and m = ρu denotes the momentum and (w, z) is given by (9).
Letting
where T(t), X(x) are given by (3), and z = B(x, t) + v, we have from (13) that
or
where ϵ1 > 0 is a suitable small constant, a(x, t) = u − ρx and b(x, t) = X(x). Similarly, letting
and w = C(x, t) + s, we have from (2.1) that
where c(x, t) = u + ρx and d(x, t) = X(x).
First, we can choose ϵ = o(ϵ1) (if necessary, we may smooth X(x) by a mollifier as the authors did in [2]) such that the following three terms on the left-hand side of (17) and (19)
and
Second, by using the maximum principle to the first equation in system (6), we have the a-priori estimate ρϵ,δ ≥ 2δ.
Finally, by the calculations, we may obtain the some inequalities about the variables v and s.
Lemma 2.1.
where
and
Proof of Lemma 2.1. First, at the points (x, t) where ρ ≤ 1, the following terms on the left-hand side of (17)
Second, at the points (x, t) where ρ > 1, we have ≥ 1 − ϵ2 > 0 for a small ϵ2 > 0, and X(x) ln ρ = X(x)( (v + s) + M). Then,
Therefore, the first inequality in (22) is proved. Similarly, we may prove the second inequality in (22). Lemma 2.1 is proved.
Since b2(x, t) ≤ 0, d2(x, t) ≤ 0, under the conditions given in (8), it is clear that v(x, 0) ≤ 0, s(x, 0) ≤ 0, so, we may apply the maximum principle (cf. Lemma 2.4 given in [8]) to (22) to obtain the estimates v(x, t) ≤ 0, s(x, t) ≤ 0, and so the estimates in (10).
From the estimates in (10), we can use the Riemann invariants (9) to obtain the following uniformly bounded estimates on (ρϵ,δ(x, t), mϵ,δ(x, t)) directly:
for a suitable positive constant M1, which is independent of ϵ, δ and the time t. The existence result of the Cauchy problem (6)-(7) follows by applying the standard theory on the parabolic systems. Part I of Theorem 1.1 is proved. Furthermore, we may select a subsequence of (ρϵ,δ(x, t), mϵ,δ,µ(x, t)), which converges pointwisely to a pair of bounded functions (ρ(x, t), m(x, t)) as δ, ϵ tend to zero by using the compactness framework given in [3], and prove that the limit (ρ(x, t), m(x, t)) is an entropy solution of the Cauchy problem (1)-(2), so Theorem 1.1 is proved.