Abstract

The existence and uniqueness results for the Tricomi problem of Chaplygin’s hodograph equation are shown, in the case that the domain considered is close to the parabolic degenerate line, by adopting the energy integral methods and choosing judiciously suitable multipliers.

1. Introduction

In this paper we consider the Tricomi problem of the following second-order linear partial differential equation. Considerwhereand is a constant. Equation (1) is of elliptic type for and , hyperbolic type for and , and parabolic degenerate on the line . We are interested in this equation because it is actually an equivalent form of Chaplygin’s hodograph equation (with as the unknown). (In this paper we will use the subscripts like and to denote the partial derivatives and )where the function (called sonic speed in gas dynamics) is given by the Bernoulli law [1, page 23] with being a positive constant, , and the adiabatic exponent for polytropic gas. One can easily show that, by taking and , (3) is transformed to (1), with (cf. [2, page 72]).

The significance of Chaplygin’s hodograph equation (3) lies in the fact that it is the hodograph transform of the following compressible Euler equations of isentropic irrotational flows:where is the density of mass of the flow and is the velocity of the flow along the coordinates of the Euclidean plane. Since in this case the sonic speed , then is a function of given by the Bernoulli law. Some fundamental problems in gas dynamics, such as detached shocks in supersonic flow past blunt bodies and subsonic jets (cf. [1, 3]), could be considered more favorably by using hodograph equation (3) (or (1)) rather than Euler equations (5), because the latter are generally a quasi-linear mixed elliptic-hyperbolic system, which is still far beyond the ability of present-day analytical tools to study.

In a previous work [2], the authors have studied a mixed boundary value problem of (3) in the sonic circle , with an artificial Dirichlet boundary condition on part of the sonic line , to understand the regularity and behavior of solutions of (3) in the elliptic region and near the degenerate line. Now, we continue our project in this paper to investigate the Tricomi problem of (1), that is, to find a function satisfying (in certain sense to be specified later) (1) in a planar domain which is simply connected, containing a segment of the -axis, and bounded by the characteristic curves (by definition, a characteristic curve of (1) satisfies equation for ) and lying in the lower half plane and a Jordan curve lying in the upper half plane , with Dirichlet conditions on and (see Figure 1). Here emanates from the origin and intersects the horizontal line at a point , where is sufficiently small. The characteristic curve emanates from the point and intersects the -axis at a point . The arc has two endpoints and . The Dirichlet conditions on and are, respectively,where is the arc-length parameter of the boundary curve so that the point moves counterclockwise on as increases. Then the outward unit normal along is given by Note that one can require to be piecewise smooth except at the point , where at best the curve is . (We thank a referee for pointing out this fact. Here as usual we use to denote the Hölder space of -times continuously differential real-valued functions on whose th order derivatives are all Hölder continuous with the exponent .) Let be a given function in the standard Sobolev space . The functions and are the traces of on and , respectively. Then it is obvious that their union belongs to

It is well known that the Tricomi problem was firstly proposed and studied by Tricomi in [4] for the now so-called Tricomi equation , by using singular integral equations and the matching technique. Tricomi’s study of this problem was mainly motivated to understand second-order mixed elliptic-hyperbolic type equations from a purely mathematical point of view. Later it was discovered that the Tricomi equation may be considered in certain sense as a simple approximation near the sonic line of Chaplygin’s hodograph equation in transonic aerodynamics (cf. [5]) and the Tricomi problem is physically relevant to determining some flow field in transonic flows, such as the detached bow shock and the mixed subsonic-supersonic flow ahead of a blunt body [6]. More general linear mixed elliptic-hyperbolic equations and more general formulations of boundary conditions (such as generalized Tricomi problem, Frankl problem, and generalized Frankl problem) were also considered. For example, Morawetz [7] proved the uniqueness for smooth solutions using Noether’s theorem on conservation laws for the equation Rassias [8] studied weak solutions for the equation Osher [9] showed the existence for Lavrentiev-Bitsadze equation . Aziz and Schneider [10] investigated the existence of weak solutions of the Gellerstedt problem and the Gellerstedt-Neumann problems for the equation Lupo et al. [11] proved existence of weak solutions for Tricomi problem with closed Dirichlet boundary conditions. Lupo et al. [12] considered the existence, uniqueness, and qualitative properties of weak solutions to the degenerate hyperbolic Goursat problem on characteristic triangles for linear and semilinear equations of Tricomi type. See also, for example, [1316] for works on the nonlinear Tricomi problems. We recommend the introduction in the monograph [17] for a review of the status of mixed-type equations around the 1970s. Morawetz [18] also reviewed the existence and uniqueness theorems for mixed-type equations and their applications to transonic flows, and Chen [19] introduced more recent progress.

However, to the best of our knowledge, there is not any result on the Tricomi problem of Chaplygin’s hodograph equation (1), which is relevant to many physical problems in transonic aerodynamics. So we will devote this work to establishing some basic properties of such problems. The main result is the following theorem.

Theorem 1. There is a positive constant determined only by such that if the domain is contained in the strip , then the Tricomi problem (1) and (8) has a quasi-regular distributional solution. Furthermore, the solution is unique in .

For the definition of quasi-regular distributional solution, see Definition 2. The constant is given in (77).

Our proof depends on the classical energy methods, or the method of multipliers (see [8, 20, 21]). Besides the method of singular integral equations, these seem to be the only general way to study well-posedness of mixed-type equations. (However, see also [22] for regularity of solutions of Tricomi equation by using the methods from harmonic analysis.) Although the idea of energy method is rather simple, it is usually very technical to choose appropriate multipliers to a physically interesting equation, like (1), as shown in this paper.

We remark that there is another type of mixed elliptic-hyperbolic equations, firstly studied by Maria Cinquini-Cibrario, now called Keldysh type, whose canonical form is (see [23, page 11]). An up-to-date review of studies of Keldysh-type mixed equations was presented in [24]. It is possible now to study directly many boundary value problems of quasi-linear Keldysh-type equations; for example, see [5, 25, 26] for studies of steady continuous subsonic-supersonic flows in (approximate) de Laval nozzles.

The rest of the paper is organized as follows. In Section 2 we will define a quasi-regular distributional solution to our Tricomi problem and show that it satisfies the equation and boundary conditions in the ordinary sense if it is a classical solution. We will establish the uniqueness of classical solutions in Section 3. Finally, in Section 4, the existence of a quasi-regular distributional solution is proved by the dual method in functional analysis.

2. Definition of Solutions

Denote the linear operator by with It is a formal dual operator of .

Definition 2. Let and and for some . A function is a quasi-regular distributional solution of the equationsubjected to boundary conditions (8), iffor all .

Now we show that a quasi-regular distributional solution satisfies (1) and boundary conditions (8) in the classical pointwise sense if it belongs to . In fact, using integration by parts and (15), we get, for all , thatChoosing particularly that , all the three boundary integrals vanish, and (16) is reduced to Since is dense in , we getand hence almost everywhere in

Next, by employing (16) and (18), we have, for all , Since , it follows that Therefore, for all . Taking that vanishes in a neighborhood of and , we infer that , and furthermore, for all , there holdsRecall that , and we have Hence we get on , orwhere is a normalizing factor. Thus, we have by using (24). Since on and is a characteristic curve, then and . Thus, we haveby using (22). Since is an arbitrary function, we see .

3. Uniqueness of Classical Solutions

Assume that are two solutions of Tricomi problem (1) and (8), and takeThen solvesWe will show that in .

Set where , , and are sufficiently smooth functions to be determined (cf. Remark 4). Since we obtain that The goal is to show that all integrals , , , and are nonnegative by choosing suitable functions , , and .

One observes that the integral ifand the integral if the following conditions hold in :

By using (28), we haveSince on , it follows that Then Recall that and , we haveWe also note that and henceTherefore, we getby using (37) and (39). It follows thatby using (34). Since and on , we have Thus So by using (43) provided that

Next, observe that the integral if whereis a quadratic form of and . Since , we have which implies that, by similar analysis as in Section 2, we can setwith being a normalizing factor. Thus, we obtain that Since on and is characteristic, then and . Thus, we have provided that

Also, since is characteristic, we infer that . Moreover, we have Since , we have by using (46), and then , provided thatRecall that on and on ; then (53) is equivalent to

Therefore, by (32), (33), (44), (54), and (51), we summarize the requirements on the multipliers , , and as follows:

Our task below is to find sufficient conditions such that inequalities (55a)–(55f) hold. Set to be the elliptic region and set to be the hyperbolic region. We will actually choose where can be taken as and is to be . What is left is to choose announced in Theorem 1 sufficiently small (depending only on that appeared in the coefficients of (1)), as computations shown below.

Elliptic Region . First of all, we specify to meet the requirement of (55f). Thus, remember that in , and inequalities (55a)–(55c) are reduced to Thus, if , (55a) is transformed toNext, we choose . Then , , and (59) is simplified as It is easy to see that this holds for sufficiently small , provided that , which is exactlyBy fixing , a sufficient condition for this inequality is to take a small (depending only on ) and then require that

Hyperbolic Region . Now we choose to satisfy (55e). Then and (55c) becomes Therefore, we must choose Note that, by using (2), direct computation yields for , where is a small positive constant determined by . Hence (66) is well defined.

Next, we still choose in . Since condition (62) is valid as required, then, as shown above, (55a) holds automatically. Hence, (55d) becomes That is,By fixing , this is valid for with being a small positive constant. Here we still used continuity and the facts that and on .

In order to get (55b), we need to have Since (66) implies that we only need to guarantee that In fact, by using (68), we have It is obvious that Direct computation yields thatfor as desired. Here is a small positive constant determined by .

Finally we see that the positive constant should be chosen so thatThis finishes the choice of multipliers and we proved that , , , and . (To guarantee existence claimed in Section 4,   might need to be chosen further smaller, according to the construction of multipliers in Section 4.1, but anyway it is in essence determined by the parameter that appeared in (1)).

Finally, observe that (62) actually guarantees the stronger property that Hence, implies that as desired.

Remark 3. Note that in the above we have chosen and to be only continuous across the degenerate line . This is harmless for our earlier computations since, by applying integration by parts separately in and and then summing up, the resultant line integrals on are cancelled.

Remark 4. There are some other ways to choose the multipliers. For example, we may set where and ().
The other way is to choose where and is sufficiently large.
However, in both cases, as before, we need to be quite close to the line . So the restriction on smallness of required in Theorem 1 is still not removed.

4. Existence of Quasi-Regular Distributional Solutions

In this section, we firstly indicate how to obtain a priori estimate for our Tricomi problem and then use this estimate to show the existence of a quasi-regular distributional solution by a dual method in functional analysis.

4.1. A Priori Estimate

We now prove that

Similar to the analysis in the previous section, we have Since , it follows thatObserving that , or , and hence on , we have Then Remember and , and we get

Next, using we may writeHenceforth, using (83)–(88), provided thatSince and on , we have Thus, using (90) and , provided that

The integral is nonnegative if where is a quadratic form with respect to and . Similarly to get (24), we have Since on and is a characteristic curve, then and . Thus, we have provided that

Since is characteristic, then . Moreover, we have on . Because , we see (then ) provided thatSince on , then (99) is equivalent toNote that on , so (99) or (100) is equivalent to

Therefore we conclude, with the further help of Young’s inequality, that provided that (92), (97), and (101) hold. (Here and are two positive constants that can be chosen arbitrarily small.) Therefore, we have the a priori estimate (81), if (92), (97), (101), and are valid for some multipliers , , and . Here and are two positive constants.

As a matter of fact, we can choose the functions , , and such that to guarantee that conditions (92), (97), (101), and (103a)–(103d) hold, if and were chosen similarly as in Section 3, and are taken appropriately small. The verification is very similar to that in the previous section and therefore we omit the details.

4.2. The Proof of the Existence

By our assumption, there exists a function such that and . Next, take and then satisfies

Next, we show that there is a quasi-regular distributional solution of Tricomi problem (106).

In fact, let be the range (image) of the operator defined on . For , we define a linear functional on byHere we consider as a linear subspace of . Using estimate (81), we haveThus, the functional is bounded.

Since is a linear subspace of , by Hahn-Banach theorem, there exists a linear functional as an extension of that preserves the operator norm.

Thus, by Riesz representation theorem, there is such that Therefore, Take . Then, for all , we have Therefore, is a quasi-regular distributional solution of Tricomi problem (106) by Definition 2.

Finally, by using (105), it is obvious that is a quasi-regular distributional solution of (1) with boundary conditions (8). This finishes the proof of Theorem 1.

Conflict of Interests

The authors declare that there is no conflict of interests regarding the publication of this paper.

Acknowledgments

M. Xu is supported in part by National Nature Science Foundation of China (NNSFC) under Grants no. 11001132 and no. 11571020. L. Liu (the corresponding author) is supported by NNSFC under Grants no. 11101264 and no. 11371250. H. Yuan is supported by NNSFC under Grant no. 11371141. This paper is supported by Science and Technology Commission of Shanghai Municipality (STCSM) under Grant no. 13dz2260400. The authors thank referees for helpful comments and valuable suggestions.