Numerische Mathematik
c Springer-Verlag 1999
Electronic Edition
Fully discrete finite element approaches for time-dependent Maxwell’s equations
P. Ciarlet, Jr1, Jun Zou2,?
1 Ecole Nationale Sup´erieure des Techniques Avanc´ees, 32, boulevard Victor, F-75739 Paris Cedex 15, France; e-mail: [email protected]
2 Department of Mathematics, The Chinese University of Hong Kong, Shatin, N.T., Hong Kong; e-mail: [email protected]
Received February 3, 1997 / Revised version received February 27, 1998
Summary. A fully discrete finite element method is used to approximate the electric field equation derived from time-dependent Maxwell’s equa- tions in three dimensional polyhedral domains. Optimal energy-norm error estimates are achieved for general Lipschitz polyhedral domains. Optimal L2-norm error estimates are obtained for convex polyhedral domains.
R´esum´e. On r´esout, dans un domaine poly´edrique, les ´equations de Maxwell temporelles. Une m´ethode par ´el´ements finis discr`ete en temps et en espace est propos´ee pour calculer le champ ´electrique. Une estimation d’ordre op- timal est obtenue pour l’erreur en norme-´energie dans le cas g´en´eral. Pour la normeL2, on obtient une estimation optimale dans le cas d’un poly`edre convexe.
Mathematics Subject Classification (1991): 65N30, 35L15
1. Introduction
Many problems in sciences and industry involve the solutions of Maxwell’s equations, for example, problems arising in plasma physics, microwave de- vices, diffraction of electromagnetic waves. In this paper, we are interested in the numerical solution of time-dependent Maxwell’s equations in a bounded polyhedral domain in three dimensions. In the literature, one can find a great deal of work on numerical approximations to time-dependent Maxwell’s
? The work of this author was partially supported by Hong Kong RGC Grant No. CUHK 338/96E
Correspondence to: J. Zou
equations and also analyses on the convergence of numerical schemes for stationary Maxwell’s equations and related models. We refer readers to Raviart [21], Assous et al [5], Hewett-Nielson [16], Degond-Raviart [11], Ambrosiano-Brandon-Sonnendr¨ucker [2] and Ciarlet-Zou [8], etc. But to our knowledge, it seems that there are few existing works on the convergence analysis for semi-discrete or fully discrete numerical methods for the time- dependent Maxwell systems. In [18], Monk obtained error estimates for a semi-discrete finite element approximation to the time-dependent Maxwell’s equations using N´ed´elec’s elements, from which our current paper was initi- ated. Furthermore, in [17] Makridakis-Monk proposed a fully discrete finite element scheme and obtained the error estimates under strong regularities on the solutions. This scheme involves solving coupled non-symmetric and indefinite linear algebraic systems of both electric and magnetic fields.
The purpose of the current paper is to analyse the convergence of a simple fully discrete finite element scheme for the electric field equation derived from Maxwell’s equations by eliminating the magnetic field. The scheme is a fully discrete version of the semi-discrete scheme studied in [18], and it is constructed in a way that involves only solving a symmetric and positive definite linear algebraic system. One of our major interests here is to investigate the convergence order of the fully discrete scheme without making use of strong regularities on the solutions, which is certainly of practical importance. Under appropriate assumptions on the regularity of the continuous solutions, we derive for the concerned fully discrete scheme the optimal energy-norm error estimates for general polyhedral domains and optimalL2-norm error estimates for convex polyhedral domains.
We now introduce the Maxwell’s equations to be considered in the paper.
LetΩbe a bounded Lipschitz continuous polyhedral domain inR3,E(x, t) andH(x, t)the electric and magnetic fields respectively. Then Maxwell’s equations can be formulated as follows:
εEt+σE−curlH =J inΩ×(0, T), (1)
µHt+ curlE= 0 inΩ×(0, T), (2)
whereε(x)andσ(x)are the dielectric constant and the conductivity of the medium respectively, whileµ(x)andJ(x, t)are the magnetic permeability of the material inΩand the applied current density respectively. Here, the subscripttdenotes the time derivative. It is assumed that these coefficients are piecewise smooth, real, bounded and positive, that is, there existε0>0 andµ0>0such that, for allx∈Ω,
ε0 ≤ε(x), µ0≤µ(x), and 0≤σ(x).
(3)
Moreover, these coefficientsε(x), µ(x) andσ(x)may be discontinuous.
We assume that the boundary of the domainΩis a perfect conductor, that
is,
E×n= 0 on ∂Ω×(0, T).
(4)
We supplement Maxwell’s equations with the initial conditions E(x,0) =E0(x) and H(x,0) =H0(x), x∈Ω.
Instead of solving the coupled system (1)-(2) with both the electric and magnetic fields as unknowns, we eliminate the magnetic fieldH, by taking the time derivative of (1) and using (2), to obtain the second order electric field equation:
εEtt+σEt+ curl (1
µcurlE) =Jt, in Ω×(0, T), (5)
with the boundary condition still being (4) but the previous initial conditions being replaced by
E(x,0) =E0(x) and Et(x,0) =E1(x), (6)
whereE1(x) =ε−1(J(x,0) + curlH0(x)−σ(x)E0(x)).
Remark 1.1 We have implicitly assumed that the electromagnetic field is generated by a current with densityJ, without any charge density: i.e. the medium is locally electrically neutral, anddivJ = 0. In the more general case, the charge conservation equation reads:
ρt+ divJ = 0, whereρis the charge density.
Therefore, ifσ = 0, we derive from (1) and (2) that div (εE) =−ρ, and div (µH) = 0,
when these relations hold for the initial data. In this case, one may con- sider a saddle-point approach, like in Raviart [21] or Ciarlet-Zou [8], where Darwin’s model of approximation to Maxwell’s equations was studied.
We end this section with the introduction of some notations used in the paper. We define
H(div ;Ω) ={v∈(L2(Ω))3; divv∈L2(Ω)}, H(div 0;Ω) ={v∈H(div ;Ω); divv= 0},
H(curl ;Ω) ={v∈(L2(Ω))3; curlv∈(L2(Ω))3}, Hα(curl ;Ω) ={v∈(Hα(Ω))3; curlv∈(Hα(Ω))3},
H0(curl ;Ω) ={v∈H(curl ;Ω); v×n= 0onΓ},
where α is a nonnegative real number. H(div ;Ω), H(curl ;Ω) and Hα(curl ;Ω)are equipped with the norms
||v||0,div =
||v||20+||divv||201/2 ,
||v||0,curl =
||v||20+||curlv||201/2 ,
||v||α,curl =
||v||2α+||curlv||2α1/2 .
Here and in the sequel of the paper,k · k0will always mean the(L2(Ω) )3- norm (orL2(Ω)-norm if only scalar functions are involved). And in general, we will usek·kαand|·|αto denote the norm and semi-norm in the Sobolev space(Hα(Ω) )3(orHα(Ω)if only scalar functions are involved). We refer to Adams [1] and Grisvard [15] for more details on Sobolev spaces.Cwill always denote a generic constant which is independent of both the time step τ and the finite element mesh sizeh.
2. Fully discrete finite element schemes
We consider discretizing the electric field Cauchy problem (5)-(6) by the implicit backward difference scheme in time together with N´ed´elec’s finite elements in space.
Let us first triangulate the space domainΩand assume thatThis a shape regular triangulation ofΩwith a mesh sizehmade of tetrahedra. An element ofTh is denoted byK, and the diameters of K and its inscribed ball are denoted byhK andρK respectively. As usual, we leth = maxK∈ThhK. As the triangulation is shape regular, we havehK/ρK≤C(cf. Ciarlet [6]).
We then introduce the following N´ed´elec’sH(curl ;Ω)-conforming finite element space
Vh ={vh ∈H(curl ;Ω); vh|K ∈(P1)3, ∀K∈ Th}
whereP1 is the space of linear polynomials. It was proved in N´ed´elec [20]
that any function v in Vh can be uniquely determined by the degrees of freedom in the moment setME(v)on each elementK ∈ Th. HereME(v) is defined as follows:
ME(v) ={ Z
e(v·τ)q ds; ∀q ∈ P1(e) on any edgeeofK}, whereτ is the unit vector along the edgee.
From [3], Lemma 4.7, we know that the integrals required in the definition ofME(v)make sense for anyv∈Xp(K), withp >2, where
Xp(K) ={v∈(Lp(K))3; curlv∈(Lp(K))3; v×n∈(Lp(∂K))3}.
Thus we can define, for any v ∈ H1/2+δ(curl ;Ω)3 with δ > 0 (which implies thatcurlv ∈ (Lpδ(K))3 andv ∈ (Lpδ(∂K))3 for somepδ > 2 which depends onδ), an interpolationΠhvofvsuch thatΠhv ∈ Vh and Πhvhas the same degrees of freedom (defined by ME(v)) as von each K ∈ Th.
In order to take the boundary conditionE×n= 0on∂Ωinto account, we define a subspace ofVh:
Vh0={vh∈Vh; vh×n= 0 on∂Ω}.
This can be done simply by zeroing the degrees of freedom which correspond to the boundary edges.
Next we divide the time interval(0, T)intoMequally-spaced subinter- vals by using nodal points
0 =t0 < t1 <· · ·< tM =T
withtn = nτ, and denote the n-th subinterval by In = (tn−1, tn].For a given sequence{un}Mn=0⊂L2(Ω)or(L2(Ω))3, we introduce the first and second order backward finite differences:
∂τun= un−un−1
τ , ∂τ2un= ∂τun−∂τun−1
τ .
For a continuous mappingu: [0, T]→L2(Ω)or(L2(Ω))3, written asu∈ C(0, T; (L2(Ω))3)subsequently, we defineun=u(·, nτ)for 0≤n≤M. Using the above notation, our fully discrete finite element approximation to the electric field equations (5)-(6) is formulated as follows:
E0h =ΠhE0, E0h−E−1h =τ ΠhE1, (7)
and forn= 1,2,· · ·, M, findEnh ∈Vh0such that (ε∂τ2Enh,v) + (σ∂τEnh,v) + (1
µcurlEnh,curlv) = (∂τJn,v),
∀v∈Vh0. (8)
Obviously, for eachn = 1,2,· · ·, M, it is clear that, by Lax-Milgram theorem, the system (8) has a unique solutionEnhas its left-hand side defines a symmetric positive definite bilinear form inH(curl ;Ω)with respect to Enh. In addition, as (8) is symmetric and positive definite, it can be solved by the well-known conjugate gradient method.
Remark 2.1 Instead of the first order backward difference in time used in the fully discrete scheme (7)-(8), one can also use some second order difference
approximation in time, e.g. the Crank-Nicolson scheme. In this case, the whole discrete system can be taken as the following:
E0h =ΠhE0, E1h−E−1h = 2τ ΠhE1, (9)
and forn= 0,1,· · ·, M −1, findEn+1h ∈Vh0such that (ε δτ2Enh,v) + (σ δ2τEnh,v) + (1
µcurl ¯Enh,curlv) = (δ2τJn,v),
∀v∈Vh0. (10)
whereδτ2un= (un+1−2un+un−1)/τ2, u¯n= (un+1+un−1)/2,δ2τun= (un+1−un−1)/(2τ). Note that the scheme preserves the symmetry and the positive definiteness. The first unknownE1hcan be solved by using the initial approximation in (9) and (10) forn = 0, and the resultant linear system is also symmetric and positive definite. With this scheme we can achieve similar convergence results as obtained in the paper, see Remark 4.3.
3. Interpolation properties
This section is devoted to some basic approximation properties of the finite element interpolantΠhdefined in Sect. 2, which will be needed in the later error estimates for the finite element scheme (7)-(8). First of all, we know the following properties ofΠh: for anyu∈(H2(Ω))3,
ku−Πhuk0 ≤C h2|u|2, (11)
while for anyu∈H1(curl ;Ω), we have
kcurl (u−Πhu)k0≤C hkcurluk1. (12)
The estimate (11) can be found in Girault [13] (Theorem 3.1) and N´ed´elec [20] (Proposition 3). The estimate (12) was proved by Monk [18] (Lemma 2.3).
The estimates (11) and (12) stand for functions which are appropriately smooth, i.e. for functions in (H2(Ω))3 or H1(curl ;Ω). But usually the solutions of the Maxwell system considered in the paper may not have such kind of regularity, especially when the domainΩ is not convex and only Lipschitz continuous. Next we are going to present some approximation properties of the interpolantΠhunder weak assumptions on regularity. We first show a similar result to (12) but for theL2-norm. Comparing with the estimate (12) for thecurl operator, we lose one error order. Similar results were obtained in [12] for a different finite element (see Remark 3.3).
Lemma 3.1 We have
ku−Πhuk0 ≤C h||u||1,curl, ∀u∈H1(curl ;Ω).
The proof of the lemma is omitted, since it can be inferred from that of Lemmas 3.2 and 3.3 (see [9] for a detailed proof).
Lemma 3.2 We have, for1/2< α≤1,
ku−Πhuk0 ≤C hα||u||α,curl, ∀u∈Hα(curl ;Ω).
Remark 3.1 α > 1/2 is needed for the definition of the moments in ME(v, φ).
Proof. For any elementK ∈ Th, letx=BKxˆ+bKbe the affine mapping betweenKand the reference elementKˆ, and we define (cf. N´ed´elec [20]),
u(x) = (B∗K)−1u(ˆˆ x) or u(ˆˆ x) =BK∗u(x), (13)
whereBK∗ is the transpose of the matrixBK. LetΠˆ be the interpolant on the reference elementK, thenˆ
ku−Πhuk2L2(K)≤ |BK| k(BK∗ )−1k2kˆu−Πˆukˆ 2L2( ˆK). (14)
Throughout the paper,|A|means|det(A)|for any square matrixA. Let us now boundkˆu−Πˆukˆ L2( ˆK). For that, leteˆ(respectivelyFˆ) be any edge (respectively face) ofKˆ. Forp >2andp0 such that1/p+ 1/p0 = 1, on any edgeeˆofKˆ we define
||ˆv||Meˆ = sup
φ∈Pˆ 1(ˆe)3
|Meˆ(ˆv,φ)|ˆ
||φ||ˆ W1−1/p0,p0(ˆe)
(15)
where Meˆ(ˆv,φ) =ˆ R
ˆ
e(ˆv·τˆ) ˆφds. Using the norm equivalence in finite dimensional spaces, we have
kΠˆukˆ L2( ˆK) ≤CX
ˆ e⊂Kˆ
kΠˆukˆ Meˆ =CX
ˆ e⊂Kˆ
kˆukMˆe
≤Cn
kcurl ˆdukLp( ˆK)+ X
Fˆ⊂Kˆ
kˆu×nkˆ Lp( ˆF)
o,
where the last inequality is obtained by integration by parts and the standard extension and lifting techniques (cf. Lemma 4.7 of [3]). This implies kˆu−Πˆukˆ L2( ˆK)≤Cn
kcurl ˆdukLp( ˆK)+kˆukL2( ˆK)+ X
Fˆ⊂Kˆ
kˆu×nkˆ Lp( ˆF)
o
≤Cn
kcurl ˆdukHα( ˆK)+kˆukHα( ˆK)
o. (16)
As the left hand side does not change when replacinguˆ byuˆ plus any constant, we have
kˆu−Πˆukˆ L2( ˆK)≤Cn
kcurl ˆdukHα( ˆK)+ inf
ˆ
p∈P0( ˆK)3kˆu+ ˆpkHα( ˆK)
o.
Note
|w|ˆ Hα( ˆK)= Z
Kˆ
Z
Kˆ
||w(ˆˆ x)−w(ˆˆ y)||2
||ˆx−y||ˆ 3+2α dˆxdˆy 1/2
,
it is clear that|wˆ + ˆp|Hα( ˆK) = |w|ˆ Hα( ˆK) for allpˆ ∈ P0( ˆK)3. From this point, one can easily adapt the proof of Theorem 14.1 in [6] to obtain the norm equivalence in the quotient spaceHα/P0. Then one has
kˆu−Πˆukˆ L2( ˆK) ≤Cn
kcurl ˆdukHα( ˆK)+|ˆu|Hα( ˆK)
o. (17)
There remains to bound the right-hand side in (17). Noting that||x−y|| ≤
||BK|| ||BK−1(x−y)||, we deduce
|ˆu|2Hα( ˆK)≤ ||BK||5+2α|BK−1|2|u|2Hα(K). Similarly we have (see [9] for details)
||curl ˆdu||2L2( ˆK) ≤C||BK||4|BK−1| ||curlu||2L2(K), and
|curl ˆdu|2Hα( ˆK) ≤C||BK||7+2α|BK−1|2|curlu|2Hα(K). This with (17) shows (for||BK||small)
kˆu−Πˆukˆ 2L2( ˆK)
≤C max(||BK||5+2α|BK−1|2,||BK||4|BK−1|)||u||2Hα(curl ;K). Using the bounds onBK and the shape regularity ofTh, we get from (14) that
ku−Πhuk2L2(K)≤C h2αK||u||2Hα(curl ;K).
2
Remark 3.2 The following lemma is an improvement on the results obtained in N´ed´elec [20] (Propositions 1 and 2) and Monk [18] (Lemma 2.3), where only integersα≥1were considered.
Lemma 3.3 For1/2< α≤1, we have
kcurl (u−Πhu)k0 ≤Chα|curlu|Hα(Ω), ∀u∈Hα(curl ;Ω).
Proof. We follow N´ed´elec [20] for the notation used below. LetΘh be the H(div 0;Ω)-conforming space of degree0:
Θh ={v∈H(div 0;Ω); v|K∈(P0)3, ∀K∈ Th}, and letrhbe the corresponding interpolant toΘh with the moment set:
MF0 (v) ={ Z
F(v·n)q dσ; ∀q∈ P0(F) on any faceF ofK}.
There are four degrees of freedom attached to this finite element, but as divv= 0by definition, their sum (withq= 1) is equal to 0.
We can show
rh(curlu) = curl (Πhu).
(18)
Indeed,curl (Πhu)∈(P0)3,div (curl (Πhu)) = 0and Z
F curl (Πhu)·ndσ= Z
FcurlF(Πhu)dσ= Z
∂F Πhu·τ ds
= Z
∂F u·τ ds= Z
FcurlF(u)dσ
= Z
Fcurlu·ndσ.
Hence
kcurl (u−Πhu)k0=k(curlu)−rh(curlu)k0, (19)
this with the following result (20) gives the lemma.
Next, we show that for any elementKand1/2< α≤1, kw−rhwkL2(K)≤Chα|w|Hα(K),
∀w∈H(div 0;Ω)∩Hα(Ω)3. (20)
To prove this, we replace the degrees of freedom inMF0 (v)(cf. [19] or [20]) by
{ 2
|BK| Z
F(v·n)qdσ; ∀q|K ∈(P0)3, K∈ Th}.
Then the following transformation
x=BKxˆ+bK, w(x) =BKw(ˆˆ x), preserves the interpolation and divergence, i.e.
ˆ
rw(ˆˆ x) =rhw(x), div ˆdw(ˆx) = divw(x), whererˆis the reference interpolant onKˆ.
Now consider the moment MF0ˆ( ˆw,φ) =ˆ R
Fˆ( ˆw·n) ˆˆ φ dˆσ. Noting that divw= 0impliesdiv ˆdw= 0, we have by integration by parts
|MF0ˆ( ˆw,φ)|ˆ = Z
Kˆ
div ˆdwφ dˆˆ x+ Z
Kˆwˆ ·grad ˆdφ dˆx
= Z
Kˆ wˆ ·grad ˆdφ dˆx
≤C||w||ˆ Lp( ˆK)||φ||ˆ W1−1/p0,p0( ˆF)
(21)
wherep0 is again the conjugate number ofp andφˆthe extension by zero fromW1−1/p0,p0( ˆF)intoW1−1/p0,p0(∂K)ˆ combined with a lifting operator fromW1−1/p0,p0(∂K)ˆ ontoW1,p0( ˆK).
Using (21), we can bound
||w||ˆ M0ˆ
F = sup
φ∈(Pˆ 0( ˆF))3
|MF0ˆ( ˆw,φ)|ˆ
||φ||ˆ W1−1/p0,p0( ˆF)
bykwkˆ Lp( ˆK). This with the norm equivalence in finite dimensional spaces gives
kˆrwkˆ L2( ˆK)≤C X
Fˆ∈Kˆ
kˆrwkˆ M0ˆ
F =C X
Fˆ∈Kˆ
kwkˆ M0ˆ
F
≤Ckwkˆ Lp( ˆK)3 ≤Ckwkˆ Hα( ˆK), that implies
kwˆ −rˆwkˆ L2( ˆK)≤Ckwkˆ Hα( ˆK). (22)
As replacingwˆ bywˆ plus any constant does not change the left hand side, we come to
kwˆ −rˆwkˆ L2( ˆK) ≤C|w|ˆ Hα( ˆK). Thus we finally derive
kw−rhwk2L2(K)≤ kBKk2|BK| kwˆ −rˆwkˆ 2L2( ˆK) ≤Ch2αK|w|2Hα(K), this proves (20).2
Remark 3.3 All the results of this paper are also valid for certain other first order,H(curl ;Ω)-conforming, N´ed´elec’s elements, e.g. the element defined in [19], i.e. each element function has the formvh|K ∈ R1(K) = {aK+bK×x,(aK,bK)∈R6}, with the related degrees of freedom. The crucial step for the validity is to establish Lemmas 3.1-3.3 for this element.
In fact, Lemma 3.1 was established in ([12], Theorem 3.2). Lemmas 3.2- 3.3 can be extended to include this first order element by means of similar
techniques as used in the present paper. Note, in addition, that higher order finite elements are not considered here as we do not assume, for the solutions, a stronger regularity thanHα(curl ;Ω)with1/2< α≤ 1in Subsect. 4.1 (energy-norm error estimates), or than H2(Ω) in Subsect. 4.2 (L2-norm error estimates).
4. Finite element error estimates
We are now going to derive the error estimates for the fully discrete finite element scheme (7)-(8) both in the energy-norm and theL2-norm. Through- out this section,En andEnh will denote the solutions of the electric field equations (5)-(6) and the finite element approximation (8) at timet=tn.
For the error analysis, we need the solution functionEto be defined also in the interval[−2τ, T]in terms of the time variablet. This can be done by extendingEwith some regularity from the time interval[0, T]to the interval [−2τ, T]. So we shall always implicitly assume that E is well defined in terms of time variableton the interval[−2τ, T]. Furthermore, to achieve the optimal energy-norm error estimates for the concerned fully discrete finite element scheme, we introduce an important projection operatorPh : H0(curl ;Ω)→Vh0defined by
a(Phu,v) =a(u,v), ∀v∈Vh0 (23)
wherea(u,v)is the scalar product associated withk · k0,curl. Obviously, Phis well-defined inH0(curl ;Ω).
By the definition of the projection Ph in (23), we easily see that, for α >1/2,
ku−Phuk0,curl ≤ ku−Πhuk0,curl,
∀u∈H0(curl ;Ω)∩Hα(curl ;Ω).
(24)
Later on, we will need the following identity Xk
m=1
(am−am−1)bm =akbk−a0b0− Xk m=1
am−1(bm−bm−1) (25)
and the following estimates forB=H1(curl ;Ω)orB = (Hα(Ω))3 with α≥0,
k∂τunk2B≤ 1 τ
Z tn
tn−1kut(t)k2Bdt, ∀u∈H1(0, T;B), (26)
k∂τ2unk2B≤ 1 τ
Z tn
tn−2kutt(t)k2Bdt, ∀u∈H2(0, T;B), (27)
k∂τunt −∂τ2unk2B≤Cτ Z tn
tn−2kuttt(t)k2Bdt, ∀u∈H3(0, T;B).
(28)
4.1. Energy-norm error estimates
This subsection is devoted to the estimate on the energy-norm error for En−Enh. For the purpose, we first analyse the errorηhk = Ekh−PhEk, for1 ≤k≤n. Once we have estimates forηnh, we can easily get the error estimates forEn−Enhby the triangle inequality, the projection properties (24) and the interpolation properties discussed in Sect. 3.
We conduct our analysis only for the constant coefficients case, i.e., we assumeε(x),µ(x)andσ(x)are all constants. It is straightforward to extend the analysis to the non-constant or elementwise constant case by simply keeping these coefficients inside the integrals or norms and bounding them by taking their maximum or minimum values if necessary.
To analyseηkh, we multiply the equation (5) byv/τ ∈Vh0and integrate then the resultant overΩin space and overIkin time to obtain
ε(∂τEkt,v) +σ(∂τEk,v) + 1 τµ
Z
IkcurlEdt,curlv
= (∂τJk,v),
∀v∈Vh0. (29)
Now subtracting (29) from (8) and making some rearrangements, we have ε(∂τ2ηhk,v) + 1
µ(curlηhk,curlv) +σ(∂τηhk,v)
=ε
∂τ(Ekt −∂τPhEk),v +σ
∂τ(Ek−PhEk),v + 1
τµ Z
Ikcurl (E−PhEk)dt,curlv
, ∀v∈Vh0. Then takingv=τ∂τηhk=ηkh−ηk−1h above and usinga(a−b)≥a2/2− b2/2, for any real numbersaandb, yield
στk∂τηhkk20 + ε
2k∂τηkhk20−ε
2k∂τηk−1h k20 + 1
2µkcurlηhkk20− 1
2µkcurlηk−1h k20
≤ τσ
(∂τEk−Ph∂τEk), ∂τηkh +τε
∂τ(Ekt −∂τEk), ∂τηkh +τε
∂τ2Ek−Ph∂τ2Ek), ∂τηhk +1
µ Z
Ikcurl (E−Ek)dt,curl∂τηkh +τ
µ
curl (Ek−PhEk),curl∂τηhk
≡:X5
i=1
(I)i. (30)
Next, we will estimate (I)ifori= 1,2,3,4,5one by one.
First for (I)1, using Cauchy-Schwarz inequality, we have (I)1 ≤ 1
2στk∂τηkhk20+1
2στk∂τEk−Ph∂τEkk20
≤ 1
2στk∂τηkhk20+C σ τ
k∂τEk−Πh∂τEkk20,curl
(by (24))
≤ 1
2στk∂τηkhk20+C σ τ h2k∂τEkk21,curl (by (12) and Lemma 3.1)
≤ 1
2στk∂τηkhk20+C σ h2 Z
IkkEtk21,curl dt (by (26)).
For the estimation of (I)2, by writing ∂τ(·)k into the integral of form R
Ik(·)tdtand using Cauchy-Schwarz inequality, we easily come to (I)2≤ 1
2τεk∂τηkhk20+C ετ2 Z tk
tk−2kEtttk20dt.
The estimate for (I)3 is achieved using the same technique as used for (I)1,
(I)3≤ 1
2ετk∂τηkhk20+C ε h2 Z
IkkEttk21,curldt.
To analyse (I)4, we use Green’s formula and the boundary condition to derive
(I)4 = 1 µ
Z
Ik
curl curl (E−Ek), ∂τηhk dt
=−1 µ
Z
Ik
Z tk
t
curl curlEt, ∂τηhk dt0dt,
then by Cauchy-Schwarz inequality we obtain (I)4≤ 1
2τεk∂τηhkk20+ τ2 2εµ2
Z
Ikkcurl curlEtk20dt.
Finally, we estimate (I)5. By the definition ofPhin (23), we have (I)5= τ
µ
curl (Ek−PhEk),curl∂τηkh
=−τ µ
Ek−PhEk, ∂τηkh ,
then applying the Cauchy-Schwarz inequality and (12), (24) and Lemma 3.1, we come to
(I)5 ≤ τ
µkEk−PhEkk0k∂τηhkk0
≤ τ
µ2k∂τηhkk20+C τ h2kEkk21,curl.
This completes all the estimates for (I)i in (30). Now summing both sides in (30) overk = 1,2,· · ·, nand making use of the previous estimates for (I)i (1≤i≤5), lead to
ε
2k∂τηhnk20+ 1
2µkcurlηnhk20 ≤C m0(E)(τ2+h2) +d(ηh0) +C τXn
k=1
k∂τηhkk20+kcurlηhkk20 , (31)
whereCis a constant depending on the coefficientsε,σandµ, andm0(E) is an a priori bound ofEof the following form
m0(E) = max
0≤t≤TkE(t)k21,curl + Z T
0 (kEttk21,curl +kcurl curlEtk20)dt +
Z T
−τkEtttk20dt, whiled(ηh0)is the initial error
d(η0h) = ε
2k∂τη0hk20+ 1
2µkcurlηh0k20, which can be analysed as follows:
First by the definitions ofE0hand the projectionPh, we have kcurlηh0k0 =kcurl (ΠhE0−PhE0)k0 =kcurlPh(ΠhE0−E0)k0
≤ kΠhE0−E0k0,curl ≤C hkE(0)k1,curl.
Then for the first term in d(η0h), by definition ofηh0,E0h andE−1h , we know
∂τηh0 =τ−1(τΠhEt(0)−PhE(0) +PhE(−τ))
=τ−1Ph(E(−τ)−E(0) +τEt(0)) +Ph(ΠhEt(0)−Et(0)), using the property of the projectionPh, we derive
k∂τηh0k0 ≤τ−1kE(−τ)−E(0) +τEt(0)k0,curl +kΠhEt(0)−Et(0)k0,curl
≤C τ sup
(−τ,0)kEttk1+C hkEt(0)k1,curl.
Therefore, we get the estimates for the initial errord(ηh0):
d(ηh0)≤C τ2 sup
(−τ,0)kEttk21+C h2 kE(0)k21,curl +kEt(0)k21,curl . Then substituting this into (31) and applying the well-known discrete Gron- wall’s inequality, we conclude that
1≤n≤Mmax
k∂τ(Enh−PhEn)k20+kcurl (Enh−PhEn)k20
≤C m0(E) (τ2+h2).
Finally, applying the triangle inequality to
∂τEnh−Ent = (∂τEnh−Ph∂τEn) + (Ph∂τEn−∂τEn) + (∂τEn−Ent) and
Enh−En= (Enh−PhEn) + (PhEn−En), we have proved the following energy-norm error estimates
Theorem 4.1 Let E andEnh be the solutions of the electric field equa- tions (5)-(6) and the finite element approximation (7)-(8) at time t = tn, respectively. Assume that
E∈H2(0, T;H0(curl ;Ω)∩H1(curl ;Ω) )∩H3(0, T; (L2(Ω))3).
Then we have
1≤n≤Mmax
k∂τEnh −Entk20+kcurl (Enh−En)k20
≤C(τ2+h2) whereCis a constant independent of both the time stepτ and the meshsize h.
Remark 4.1 The error estimate in Theorem 4.1 is optimal both in terms of time step sizeτ and mesh sizehas we have used only the H1(curl ;Ω)- regularity in space andH3(0, T)-regularity in time.
If the solutionEhas no so much regularity in space as in Theorem 4.1 (cf.
Costabel [10], Assous et al [4]), we then have the following weaker error estimates
Theorem 4.2 Let E andEnh be the solutions of the electric field equa- tions (5)-(6) and the finite element approximation (7)-(8) at time t = tn, respectively. Assume that for some1/2< α <1,
E ∈H2(0, T;H0(curl ;Ω)∩Hα(curl ;Ω) )∩H3(0, T; (L2(Ω))3).
Then we have
1≤n≤Mmax
k∂τEnh−Entk20+kcurl (Enh−En)k20
≤C(τ2+τ2h2(α−1)+h2α).