FAN Mingzhi(樊明智)
( School of Science,Xuchang University,Xuchang 461000, China)
Abstract:A low order H1-Galerkin mixed finite element method (for short FEM) is proposed for a class of nonlinear dispersion dissipative wave equations with the simplest bilinear rectangular element and zero order Raviart-Thomas(for short R-T) element.By use of the interpolation operator instead of the traditional Ritz projection,together with integral identity technique,the superclose properties of u in H1-norm and →p in H(div,Ω)-norm are deduced for the semi-discrete and fully-discrete schemes,which improve the results in the associated existing literature.Finally,numerical results are provided to confirm the validity of our theoretical analysis.
Key words:Nonlinear dispersion dissipative wave equations; H1-Galerkin MFEM;Superclose property;Semi-discrete and fully-discrete schemes
Consider the following dispersion-dissipative wave equations[1]
whereΩis a convex polygonal domain in R2with bounded boundary∂Ω,J=(0,T] is the time interval,X=(x,y),u0(X),u1(X) are known smooth functions,f(u) is a Lipschitz continuous function.
The nonlinear dispersion-dissipative wave equations (1.1) are usually used to diserible the propagation problems of lenghways-wave in nonlinear elastic rods and ion-sonic of space transformation by weak nonlinear effect,and there have been appeared some studies.For example,[1] discussed this problem which initial boundary value condition;[2-4] studied solitary wave solution,[5-6] proved existence and asymptotic behaviour of the global solution,respectively;[7] investigated mixed covolume methods;[8] derived convergence results for semi-discrete and the backward Euler fully-discrete schemes;although,[9] deduced the superclose properties and the superconvergence results of conforming linear triangular FEM,however.Unfortunately,there is no consideration on Crank-Nicolson fully-discrete scheme of equations (1.1),up to now.
In recent years,a lot of studies have been devoted to mixed FEMs for partial differential eauation[10-21].For example,[10-11] established the general theory of Mixed FEMs;[12]developed a new FEM called expand Mixed FEMs.However,the pair of the approximating spaces has to satisfy the LBB consistency condition on the approximating spaces.[13]proposed theH1-Galerkin Mixed FEMs,and has been applied to different PDEs.For exemple parabolic partial integro-differential equations[14],second order hyperbolic equations[15],hyperbolic type integro-differential equation[16],Sobolev equations[17]and pseudo-hyperbolic equations[18],etc.
At present work,we mainly study the superclose properties of the lowest orderH1-Galerkin mixed FEM for equations (1.1) under semi-discrete and fully-discrete schemes.the bilinear FE and the zero order R-T FE spaces are taken as approximation spaceV hfor the original variableuandfor the stress=∇ut,respectively.Firstly,we derive the superclosed properties with orderO(h2) foruinH1-norm and forinH(div,Ω) for semidiscrete scheme.Secondly,we obtain the superclosed results with orderO(h2+Δt)/O(h2+(Δt)2) foruinH1-norm and for→pinH(div,Ω) under backward Euler and Crank-Nicolson fully-discrete schemes,respectively.The results of this paper for semi-discrete scheme and Euler fully-discrete scheme is one order higher than the results with respect tohin[18].This proposed mixed FE scheme has a smallest total degrees of freedom (about 3NP,NP denotees the numbers of the nodal points in the subdivision).
Throughout this paper,hand Δtdenote the mesh size and the time step,respectively.Cdenotes a general positive constant which does not depend onhand Δt,andCmay represent different values in different places.
We reurite the problem (1.1) into the following system
In this section,we will derive the error estimates between the exact solution an its approximation solution for fully-discrete scheme.
Let 0=t0<t1<...<tN-1<tN=Tbe a given partition of the time interval [0,T]with step length Δt=T/N,tn=nΔt,n=1,2,···,N,for some positive integerN.Next,we introduce some notations as follows
Tab.6.1 Numerical results for Euler scheme at t=0.1
Tab.6.2 Numerical results for Euler scheme at t=0.5
Tab.6.3 Numerical results for Euler scheme at t=1.0
Tab.6.4 Numerical results for Crank-Nicolson scheme at t=0.1
Tab.6.5 Numerical results for Crank-Nicolson scheme at t=0.5
Tab.6.6 Numerical results for Crank-Nicolson scheme at t=1.0