For a large class of piecewise polynomial subspaces $S^h$ defined on the regular mesh, $W^{1,∞}$-interior estimate $\|u_h\|_{1,∞,Ω_0}$ ≤ $c\|u_h\|_{-s,Ω_1}$, $u_h\in S^h{Ω_1}$ satisfying the interior Ritz equation is proved. For the finite element approximation $u_h$ (of degree $r-1$) to $u$, we have $W^{1,∞}$-interior error estimate $\|u-u_h\|_{1,∞,Ω_0}$)≤$ch^{r-1} (\|u\|_{r,∞,Ω_1}+\|u\|_{1,Ω}$). If the triangulation is strongly regular in $Ω_1$ and $r=2$ we obtain $W^{1,∞}$-interior superconvergence.

In this paper the idea of "singularity-separating" presented in [10] is used to solve a two-dimensional phase-change problem. A difference scheme with second-order accuracy everywhere, including the region near the boundary between two phases, is constructed for the above problem. Through the computation it is shown that the singularity-separating method, whose accuracy is high, is efficient for two-dimensional phase-change problem.

In this paper we give a lower bound of the separation $sep_F(A,B)$ of two diagonalizable matrices A and B. The key to finding the lower bound of $sep_F(A,B)$ is to find an upper bound for the condition number of a transformation matrix Q which transforms a diagonalizable matrix A to a diagonal form. The obtained lower bound of $sep_F(A,B)$ involves the eigenvalues of A and B as well as the departures form the normality $\delta_F(A)$ and $\delta_F(B)$. 

An order interval secant method is given. Its rate of convergence is faster than that of order interval Newton method in [1]. The existence and uniqueness of a solution to nonlinear systems and convergence of the interval iterative sequence are also proved.

A general approach of deferred correction procedure based on linear multistep formulas is proposed. Several deferred correction procedures based on backward differentiation formulas, which allow us to develop L-stable algorithms of order up to 4 and $L(\alpha)$-stable algorithms of order up to 7, are derived. Preliminary numerical results indicate that this approach is indeed efficient.

The aim of this paper is the study of the convergence of a finite element approximation for a variational inequality related to free boundary problems in non-steady fluid flow through porous media. There have been many results in the stationary case, for example, the steady dam problems, the steady flow well problems, etc. In this paper we shall deal with the axisymmetric non-steady porous flow well problem. It is well know that by means of Torelli's transform this problem, similar to the non-steady rectangular dam problem, can be reduced a variational, inequality, and the existence, uniqueness and regularity of the solution can be obtained ([12, 7]). Now we study the numerical solution of this variational inequality. The main results are as follows:
1. We establish new regularity properties for the solution $W$ of the variation inequality. We prove that $W \in L^\infty(0, T; H^2(D))$, $γ_0W\in L^\infty(0, T; H^2(T_n))$ and $D_1γ_0W\in L^2(0, T; H^1(T_n))$ (see Theorem 2.5). Friedman and Torelli [7] obtained $W\in L^2(0, T; H^2(D))$. Our new regularity properties will be used for error estimation.
2. We prove that the error estimate for the finite element solution of the variational inequality is $$ ( \sum^N_{i=1}\| W^1 - W^1_h \|^2_{H^1(D)}\Delta t)^{1/2} = O(h+\Delta t^{1/2})$$ (see Theorem 3.4). In the stationary case the error estimate is $\|W-W_h\|_{H^1(D)} = O(k)$ ([3,6]).
3. We give a numerical example and compare the result with the corresponding result in the stationary case.
The result of this paper are valid for the non-ready rectangular dam problem with stationary or quasi-stationary initial data (see [7], p.534).

In this article the high order approximation of the one-way wave equations are discussed. The approximate dispersion relations are expressed in explicit form of sums of simple fractions. By introducing new functions, the high order approximations of the one-way wave equations are put into the form of systems of lower order equations. The initial-boundary value problem of these systems which corresponds to the migration problem in seismic prospecting is discussed. The energy estimates for their solutions are obtained.