Research Plan

Yong Jung Kim

RESEARCH INTERESTS

applied mathematics, hyperbolic conservation laws, correlation of convection-diffusion and reaction terms in scalar conservation laws, piecewise self-similar solutions, numerical schemes

GENERAL INTRODUCTION

My research has focused on regularized conservation laws. To avoid the blowup of the solution of nonlinear conservation laws, a weak solution is considered as a global solution under appropriate admissibility criteria, for example, the entropy condition. Regularizing the problem by adding small viscosity is another way to resolve the situation. Regularized problems are actually realistic since they are models of physical phenomena.
When the zero-viscosity limit of a regularized problem is considered, the viscosity plays the corresponding role of the admissibility criteria, i.e., choosing the meaningful solution. But, if the long time behavior is considered, the viscosity plays a more significant role deciding the global structure of the solution. So understanding of the correlation of convection-diffusion terms in scalar equations has become an interesting project to me. If a reaction term is added, its analysis becomes harder but we can observe physically and numerically interesting issues.
Self-similarity has been one of my key words. The solution of a Riemann problem of a conservation law is given as a function of self-similarity variable $\xi=x/t$. For Cauchy problems this self-similarity can be recovered only locally by considering piecewise constant solutions. We suggest to approximate the solution with piecewise self-similar functions which have characteristic lines of `$(x-x_k)/t=\text{constant}$' piecewise. Then the self-similarity is global and the time discretization is not necessary.

PREVIOUS RESEARCH PROJECTS

Isentropic gas dynamics
In the thesis the Riemann problem for the one dimensional isentropic gas dynamics has been considered in Eulerian coordinates, [1]. As suggested by Dafermos ([D], Section 9.5), and followed in many articles, we consider the system with a self-similar viscosity, and the problem is transformed to ODEs in terms of the variable $\xi=x/t$,
(\u-\xi)\rho'&+ \rho \u' = 0,\cr
(\u-\xi)\rho \u'+& p(\rho)' = \eps \u'',\cr      (1)
\rho(\pm\infty)=\rho_\pm&,\u(\pm\infty)=\u_\pm.
The problem is considered without size restrictions on the data. So our biggest concern is the appearance of the vacuum state. For the existence we apply the Leray--Schauder type degree theory which is applied under a-priori estimates. We also study the structure of solution and provide a good understanding of the structure near vacuum which is not possible in Lagrangian coordinates.

CURRENT RESEARCH PROJECTS

Project 1, $L^1$ solutions of convection-diffusion-reaction equations
This project is to survey the correlation of convection, diffusion and reaction terms in scalar equations.

  • Burgers equation (a joint work with A.E. Tzavaras, UW-Madison)
    In this study we focus on the effect of viscosity on the large time behavior of the viscous Burgers equation,
     u_t+uu_x=\mu  u_{xx},\quad  u(x,0)=u_0(x),     (2)
    
    which causes the metastability of the problem. For a fixed small viscosity $\mu >0$ the solution $u(\cdot,t)$ evolves into a diffusion wave of a single hump structure (see [H]). On the other hand, for a fixed large time $t>>1$, the solution converges to an N-wave as $\mu \to 0$. To capture the general mechanism for the behavior effectively we consider self-similar variables $s=\ln(t),\xi=x/\sqrt{t}$ and $w(\xi,s)=\sqrt{t}u(x,t)$, and transform the Burgers equation to
    w_s+{1\over2}(w^2-\xi w)_\xi=\mu w_{\xi\xi}.      (3)
    
    It allows to estimate the time of evolution from an N-wave to the final stage of a diffusion wave. Standard hyperbolic theories of $L^{1}$-contraction and Oleinik inequality for the self-similar Burgers (3) are considered. We show how to use a special Lyapunov function to establish the asymptotic in time profile (for $\mu$ fixed) of a diffusion wave.
    Finally, using a set of similarity variables and a variant of the Cole-Hopf transformation, we obtain an integrated Fokker-Planck equation. The latter is solvable and provides an explicit solution of the viscous Burgers in a series of Hermite polynomials. This format captures the long time - small viscosity interplay, as the diffusion wave and the diffusive N-waves correspond respectively to the first two terms in the Hermite polynomial expansion, see [2].

  • Convection-diffusion equations (a joint work with Wei-Ming Ni, U of MN)
    The phenomena of metastability in Burgers can be naturally phrased quantitatively for general convection-diffusion equations. To make the issue clearly consider 3 Cauchy problems,
              u_t+f(u)_x=0,                (4)
              u_t=\mu  u_{xx},             (5)
              u_t+f(u)_x=\mu  u_{xx},      (6)
    
            |A\sin(x), -\pi\le x\le 0,
    u(x,0)=-|B\sin(x), 0\le x\le \pi,      (7)
            |    0,      otherwise,
    
    where $B>A>0$. Let $A(t)$ and $B(t)$ be the area of the negative part and the positive part of solutions at time $t$. It is well known that if the flux of (4) is convex, $A(t),B(t)$ are invariant variables for the solution of (4,7). On the other hand one of them converges to zero for the solution of (5,7). To answer how $A(t)$ and $B(t)$ evolve for the problem (6,7) it is needed to understand the correlation of the convection and the diffusion terms closely.
    For the case $A(t)\to0$, we propose to estimate time $T=T(\eps ,\mu )>0$ such that
    A(t)\le\eps,\quad \text{for all } t\ge T(\eps  ,\mu ), (8)
    
    for given $\mu,\eps>0$. Our first result of this project comes from posing this question to the heat equation. We show that $T(\eps,\mu)=O(\ln(\eps)/\mu)$, and the convergence to diffusive N-waves has been studied, and it is transfered to Burgers under the Cole-Hopf transformation, see [5]. We are pursuing the idea for a general case.

  • Contaminant transport in porous media (a joint work with C.J. van Duijn, CWI-Amsterdam)
    The techniques developed in previous convection-diffusion problems are mostly focused to the sign changing solutions. Here we consider positive solutions of
    (u+u^p)_t+u_x=u_{xx},  u(x,0)=u_0(x),  p>0.     (9)
    
    This model has been suggested in [DDG]. This topic is in progress.

  • Roll waves(a joint work with Shi Jin, UW-Madison)
    Burgers equation is considered with a linear source term,
    u_t+uu_x=u,\quad  u(x,0)=u_0(x)     (10)
    
    with mean zero periodic initial data, which is a simple model of roll waves. It is easily verified that the total mass over a period $M(t)=\int_0^L u(x,t)dx$ satisfies $M(t)=e^tM(0)$. So the total mass increases exponentially if the initial total mass is not identically zero. In this case even a small rounding off error of numerical computations may grow up and eventually breaks down the solution. That is the reason for the instability observed in [JK]. Analytic behavior of (10) is considered together with numerical solutions in [3]. We suggest interface method to provide numerically stable solutions for the problem.

  • A generalized Burgers equation
    An example to see the correlation of convection, diffusion and reaction terms is
    u_t+f'(u)u_x=\mu  u_{xx}+f'(u).      (11)
    
    Numerical solutions of this example is studied in [SW]. I have interest for the case with highly oscillating initial data. In the case we can observe different rolls of these three terms clearly. This topic is in preparation.

    Project 2, Piecewise self-similar solutions($L^1$ solutions for inviscid problems)

    Both of the conservation law and the Riemann initial data are invariant under the dilation $(x,t)\to(ax,at),a>0$, and hence the solution of the Riemann problem turns out to be a function of a single variable $\xi=x/t$. This observation has been the basis of various schemes devised for general Cauchy problems.
    In a numerical scheme piecewise constant solutions are considered and the scheme basically solves Riemann problems locally in time and space. Note that the solution of conservation laws with piecewise constant initial data is not piecewise constant anymore. So in a scheme there should be a process to rearrange the solution into a piecewise constant function. The front tracking method (see [D], chapter 14) avoids this issue by approximating the nonlinear flux with a piecewise linear function.
    These approaches are based on the self-similarity of the Riemann data. In Project 2 we focus on the self-similarity of the conservation law itself. Suppose that the solution of a scalar equation,
    u_t+f(u)_x=0, u(\cdot,0)\in L^1(\Bbb R),   (12)
    
    is a function of $\xi=x/t$. Then a characteristic line should be a line that passes through the origin, $f'(u)=x/t$. Roughly speaking (see [4] for detail), a piecewise self-similar solution has a structure,
    f'(u)=(x-z_k)/(t+t_k),s_{k-1}(t)\le x\le s_k(t), k=1,...,N, (13)
    
    and zero outside. If initial data are given as this form, the solution of (12) keeps this structure. This is the way to fully recover the self-similarity of conservation laws.
    The artificial shocks $s_k(t)$ at a given time $t>0$ is decided from an equation derived from the same area rule. After finding the shock place we can solve the solution (13),
    u(x,t)=(f')^{-1}((x-z_k)/(t+t_k)),s_{k-1}(t)\le x\le s_k(t),k=1,...,N.(14)
    
    (14) is valid if $f'$ is invertible, and it is under the hypothesis
            f''(u)\ne 0.          (H)
    
    So in the scheme $N$ algebraic equations are solved to find shock places instead of doing time marching process of other schemes. So the number of operation is of order $N$, not $N^2$. It is also possible to generalize the structure to have constant parts and handle them exactly. In the figure, numerical solutions of these two cases are given. Several unique properties of the scheme is discussed in [4].
    General case without Hypothesis (H) is also considered. If the flux satisfies $f''(u)\ge 0$, it can be handled in a similar way. The case of real interest is the case of phase change, for example, the Buckley-Leverett equation. In the case convex envelope should be used instead of $f$ in (14) and the method developed under (H) is applied with some modification. The phase change case is under progress.

    RESEARCH PLAN IN FUTURE

    Regularized problems of hyperbolic conservation laws have been my main research topics so far and it will be for many more years. After the study of scalar convection-diffusion equations, Project 1, I am planning to return to systems modeling gas dynamics or magnatohydrodynamics. From the rotational model of MHD,
    v_t+((v^2+w^2)v)_x\;=\;\eps v_{xx},
      w_t+((v^2+w^2)w)_x=\eps w_{xx},       (15)
    
    we expect to observe similar relation between convection and diffusion terms as the scalar case. Basically systems are different from scalar case since there are interactions between the elements. The inviscid case of the model is also interesting from the view point of piecewise self-similar solutions, Project 2. Then (13) corresponds to relations between eigen-values and eigen-vectors. The focus of the project should be how to transfer the properties enjoyed by self-similarity solutions of scalar equations to solutions of the system. The study of the general hyperbolic system with respect to self-similar solutions is the final goal of Project 2.
    Another direction of my future research is to study multi-dimensional problems of scalar conservation laws. From the view point of numerical schemes the approach of Project 2 has several significant benefits. First the scheme does not contain any numerical viscosity. Second it takes operations of order $N$ only. Since computational time is an issue for multidimensional problems, this scheme is useful for the cases. In particular, for the problems which are sensitive on numerical viscosity, this scheme may helpful. I am considering geophysical problems as the continuation of the project. \noindent{\bf Figure\quad Two numerical solutions based on piecewise self-similar functions. \it The number of operations is of order $N$ (not $N^2$) with $N$ meshes. It is clearly observed that the $\Delta x\to0$ limit of the solution always passes through artificial shocks (a) or constant parts (b). This gives a pointwise error estimate from a single computation.} \endinsert
    References

  • [1] Yong-Jung Kim, A self-similar viscosity approach for the Riemann problem in isentropic gas dynamics and the structure of the solutions, (Quarterly of Applied Mathematics, to appear) (28 pp)
  • [2] Yong-Jung Kim and A.E. Tzavaras, Diffusive N-waves and metastability in Burgers equation, (Submitted to Siam Journal of Mathematical Analysis) (26 pp)
  • [3] Shi Jin and Yong-Jung Kim, On the computation of roll waves, (Submitted to M2AN, ``Mathematical Modeling and Numerical Analysis") (18 pp)
  • [4] Yong-Jung Kim, Piecewise Self-similar Solutions and a Numerical Scheme for Scalar Conservation Laws (Submitted to Siam Journal on Numerical Analysis) (19 pp)
  • [5] Yong-Jung Kim and Wei-Ming Ni, Convergence rate to N-waves of solutions of the heat equation and Burgers equation (manuscript in preparation) (20 pp)
  • [6] Yong-Jung Kim, Behavior of piecewise self-similar solutions of scalar conservation laws with phase change, (In progress)
  • [Da] C.M. Dafermos, Hyperbolic conservation laws in continuum physics, Grundlehren der Mathematischen Wissenschaften, vol. 325. Springer-Verlag, 2000 xvi+443 pp.
  • [DDG] C.N. Dawson, C.J. van Duijn and R.E. Grundy, Large time asymptotics in contaminant transport in porous media, SIAM J. Appl. Math. 56 (1996), no. 4, 965--993
  • [H] E. Hopf, The partial differential equation $u\sb t+uu\sb x=?u\sb {xx}$. Comm. Pure Appl. Math. 3, (1950). 201--230.
  • [JK] Shi Jin and M.A. Katsoulakis, Hyperbolic Systems with Supercharacteristic Relaxations and Roll Waves, SIAM J. Appl. Math. 61, 273-292, 2000
  • [SW] X. Sun and M.J. Ward, Metastability for a generalized Burgers equation with applications to propagating flame fronts, European J. Appl. Math. 10(1999), no. 1, 27--53