Special Issues

Convergence and quasi-optimality of L2norms based an adaptive finite element method for nonlinear optimal control problems

  • This paper aims at investigating the convergence and quasi- optimality of an adaptive finite element method for control constrained nonlinear elliptic optimal control problems. We derive a posteriori error estimation for both the control, the state and adjoint state variables under controlling by L2-norms where bubble function is a wonderful tool to deal with the global lower error bound. Then a contraction is proved before the convergence is proposed. Furthermore, we find that if keeping the grids sufficiently mildly graded, we can prove the optimal convergence and the quasi-optimality for the adaptive finite element method. In addition, some numerical results are presented to verify our theoretical analysis.

    Citation: Zuliang Lu, Fei Huang, Xiankui Wu, Lin Li, Shang Liu. Convergence and quasi-optimality of L2norms based an adaptive finite element method for nonlinear optimal control problems[J]. Electronic Research Archive, 2020, 28(4): 1459-1486. doi: 10.3934/era.2020077

    Related Papers:

    [1] Zuliang Lu, Fei Huang, Xiankui Wu, Lin Li, Shang Liu . Convergence and quasi-optimality of L2norms based an adaptive finite element method for nonlinear optimal control problems. Electronic Research Archive, 2020, 28(4): 1459-1486. doi: 10.3934/era.2020077
    [2] Ying Liu, Yanping Chen, Yunqing Huang, Yang Wang . Two-grid method for semiconductor device problem by mixed finite element method and characteristics finite element method. Electronic Research Archive, 2021, 29(1): 1859-1880. doi: 10.3934/era.2020095
    [3] Xiu Ye, Shangyou Zhang . A stabilizer free WG method for the Stokes equations with order two superconvergence on polytopal mesh. Electronic Research Archive, 2021, 29(6): 3609-3627. doi: 10.3934/era.2021053
    [4] Mingxia Li, Dongying Hua, Hairong Lian . On P1 nonconforming finite element aproximation for the Signorini problem. Electronic Research Archive, 2021, 29(2): 2029-2045. doi: 10.3934/era.2020103
    [5] Wenya Qi, Padmanabhan Seshaiyer, Junping Wang . A four-field mixed finite element method for Biot's consolidation problems. Electronic Research Archive, 2021, 29(3): 2517-2532. doi: 10.3934/era.2020127
    [6] Liupeng Wang, Yunqing Huang . Error estimates for second-order SAV finite element method to phase field crystal model. Electronic Research Archive, 2021, 29(1): 1735-1752. doi: 10.3934/era.2020089
    [7] Guifen Liu, Wenqiang Zhao . Regularity of Wong-Zakai approximation for non-autonomous stochastic quasi-linear parabolic equation on RN. Electronic Research Archive, 2021, 29(6): 3655-3686. doi: 10.3934/era.2021056
    [8] Jianguo Huang, Sen Lin . A C0P2 time-stepping virtual element method for linear wave equations on polygonal meshes. Electronic Research Archive, 2020, 28(2): 911-933. doi: 10.3934/era.2020048
    [9] Shuhao Cao . A simple virtual element-based flux recovery on quadtree. Electronic Research Archive, 2021, 29(6): 3629-3647. doi: 10.3934/era.2021054
    [10] Hyung-Chun Lee . Efficient computations for linear feedback control problems for target velocity matching of Navier-Stokes flows via POD and LSTM-ROM. Electronic Research Archive, 2021, 29(3): 2533-2552. doi: 10.3934/era.2020128
  • This paper aims at investigating the convergence and quasi- optimality of an adaptive finite element method for control constrained nonlinear elliptic optimal control problems. We derive a posteriori error estimation for both the control, the state and adjoint state variables under controlling by L2-norms where bubble function is a wonderful tool to deal with the global lower error bound. Then a contraction is proved before the convergence is proposed. Furthermore, we find that if keeping the grids sufficiently mildly graded, we can prove the optimal convergence and the quasi-optimality for the adaptive finite element method. In addition, some numerical results are presented to verify our theoretical analysis.



    Since the pioneer work in adaptive finite element methods was proposed by Babu˘ska and Rheinboldt [2], adaptive finite element methods have been applied successfully in engineering and scientific computations. The adaptive finite element method is based on the error information obtained by the computer to determine whether the solution is accurate enough. Hence the soul of adaptive finite elements is the a posteriori error estimation.

    When Dörfler [15] presented a marking strategy aiming at electing the set of elements for refinement, based on the error indicators which was controlled by the control, the state and adjoint state, adaptive finite element algorithm was put on the stage of academic research. He provided a fineness assumption on the initial grid Th0 which was used to prove the reduction of energy errors while in later investigations, Morin, Nochetto and Siebert [30] removed the assumption. Moreover, they proposed the interior node property in order to obtain the proof for convergence of adaptive finite element methods [31]. More pioneering works on adaptive finite element methods see literatures [5,3,9,13,16,17,22], in which the linear elliptic optimal control problem was mainly been investigated.

    Convergence and quasi-optimality are the two key factors of adaptive finite element methods. It was noteworthy that Mekchay and Nochetto [29] extended the convergence result of Morin, Nochetto and Siebert [31] for general second order linear elliptic partial differential equations by introducing a novel concept that was the total error which was the sum of the energy errors adding the oscillations. This provided a valuable empirical basis for future scholars' work on convergence analysis. Meanwhile, Binev, Dahmen and Devore [4] firstly presented the property of optimality. Later, a large number of scholars participate in the study of the property. For example, Carstensen and Hoppe [6] proposed convergence and quasi-optimality which were established for the Raviart Thomas finite element method. Gong and Yan [20] considered the convergence analysis of adaptive finite element method for elliptic optimal control problems with pointwise control constraints.

    As to our best knowledge, nonlinear optimal control problems have gradually penetrated into many fields of scientific research and engineering technology. Chen and Lu [11] investigated adaptive fully-discrete finite element methods for semilinear parabolic quadratic boundary optimal control problems. Gaevskaya, Hoppe, Iliash and Kieweg [17] developed an adaptive finite element method for a class of distributed optimal control problems with control constraints, and found requirement h01 on the initial grid Th0 is not restrictive for the convergence analysis of adaptive finite element for nonlinear problems. Chen, Gong, He and Zhou [10] studied an adaptive finite element method for a class of a nonlinear eigenvalue problems that may be of nonconvex energy functional and prove the convergence of adaptive finite element approximations.

    Leng and Chen [23] proved the convergence and the quasi-optimality of an adaptive element method with integral control constraints while we extend the result of [23] to a nonlinear optimal control problem with integral control constraint on L2norms in this paper. We follow the idea of [27] to derive reliable and efficient posteriori error estimations and the idea of [19] to prove the posteriori upper and the global lower bound of the errors, in which the bubble function matters. Moreover, a contraction for an adaptive finite element method is obtained based on a mild assumption on initial mesh Th0 which can be seen in [10,14,20,21,23,24,25,29,31]. Furthermore, we propose the optimal convergence rate. However, the quasi-optimality is the best obstacle mission for us to prove. Therefore, we continue to use the idea of [14,21,23]. Finally, we provide some numerical experiments to verify our theoretical analysis.

    Here are some notations will be used in this paper. Let Ω be a bounded Lipschitz domain in R2 and Ω denote the boundary of Ω. We use the standard notation Wm,q(ω) with norm m,q,ω and seminorm ||m,q,ω to express the standard Sobolev space for ωΩ. Moreover, we will omit the subscription if ω=Ω. For q=2, we denote Wm,2(Ω) by Hm(Ω) and m=m,2. Also for m=0 and q=2, we denote W0,2(ω)=L2(ω) and 0,2,ω=0,ω. For m=0 and q=, we denote W0,(ω)=L(ω) and ||||0,,ω=maxω||=||||,ω. Additionally, we observe that H10(Ω)={vH1(Ω):v=0 on Ω}. Beyond that, let c and C are the constants which independent of grids size, then we use AB to represent cABCA. In addition, (,) denotes the L2 inner product.

    The rest of our paper is arranged as follows. In Section 2, we give what the optimal control problems we want to investigate and some basic notations must be used. Then the a posteriori error estimation is obtained and an adaptive algorithm is proposed in Section 3. In Section 4, we use quasi-orthogonality and discrete local upper bound to prove the convergence of the adaptive finite element method and so is the quasi-optimality for details in Section 5. In the end, some numerical simulations is given to verify our theoretical analysis.

    In this section we first introduce some basic notations, and then we show what the nonlinear optimal control problem we discussed about.

    Th is a regular triangulation of Ω such that ˉΩ=TThˉT. T is an element of Th. Let Th0 be the initial partition of ˉΩ into disjoint triangles. By newest-vertex bisections for Th0, we can obtain a class T of conforming partitions. For Th,˜ThT, we use Th˜Th to indicate that ˜Th is a refinement of Th and hT=|T|1/2. According to [14], the continuous piecewise linear mesh function is defined by hTh. Moreover, hTh(z) is the average of the hT over all TTh for any vertex z of Th with zT. Then we have the following properties via keeping the meshs level low enough [21].

    Lemma 2.1. [21] For some constants c and C and fixed constant μ, there holds

    chThTh|TChT, (1)
    ||hTh||μ, (2)

    where all grids satisfied above are denoted by Tμ.

    In this paper we mainly enter into meaningful discussions with the following nonlinear optimal control problem:

    minuUad{12yyd20+α2u20},Δy+ϕ(y)=f+u,in Ω, (3)
    y=0,on Ω, (4)

    where ydL2(Ω), Uad={v:vL2(Ω), Ωv0} is a closed convex subset of U=L2(Ω) and ϕ()W2,(R,R) for any R>0, ϕ(y)L2(Ω) for any yH1(Ω), ϕ0. Let V=H10(Ω), we give the weak formulation to deal with state equation, namely, find yV such that

    a(y,v)+(ϕ(y),v)=(f+u,v), vV,

    where

    a(y,v)=Ωyvdxand|||v|||=a(v,v).

    Then the nonlinear optimal control problem can be restated as follows

    minuUad{12yyd20+α2u20}, (5)
    a(y,v)+(ϕ(y),v)=(f+u,v), vV. (6)

    It is well known [26,27] that the nonlinear optimal control problem has at least one solution (y,u), and that if a pair (y,u) is the solution of the optimal control problem, then there is a co-state pV such that the triplet (y,p,u) satisfies the following optimality conditions:

    a(y,v)+(ϕ(y),v)=(f+u,v), vV, (7)
    a(q,p)+(ϕ(y)p,q)=(yyd,q), qV, (8)
    (αu+p,vu)0, vUad. (9)

    Since the coercivity of a(,), we define a solution operator S:L2(Ω)H10(Ω) of (4) such that S(f+u)=y and let S be the adjoint of S such that S(yyd)=p. Suppose Vh is the continuous piecewise linear finite element space with respect to the partition ThT. For ThT, we define Uh as the piecewise constant finite element space with respect to Th. Let Uhad={vhUh:Ωvh0}L2(Ω). Then we derive the standard finite element discretization for the nonlinear optimal control problem as follows:

    minuhUhad{12yhyd20+α2uh20}, (10)
    a(yh,v)+(ϕ(yh),v)=(f+uh,v), vVh. (11)

    Similarly the nonlinear optimal control problem (10)-(11) has at least one solution (yh,uh), and that if a pair (yh,uh) is the solution of (10)-(11), then there is a co-state phVh such that the triplet (yh,ph,uh) satisfies the following optimality conditions:

    a(yh,v)+(ϕ(yh),v)=(f+uh,v), vVh, (12)
    a(q,ph)+(ϕ(yh)ph,q)=(yhyd,q), qVh, (13)
    (auh+ph,vhuh)0, vhUhad. (14)

    Based on [12,14,21,27], we have the following Lemmas in order to derive a L2norms posteriori error estimation for both the control, the state and adjoint state variables.

    Lemma 2.2. [21] Suppose that Ω is convex such that for any fL2(Ω), (3)-(4) have at least one solution y=S(f+u)H2(Ω)H10(Ω) and

    ||y||2C||f+u||0,

    and apparently the assumption is valid for S.

    Lemma 2.3. [12] Assume that Ω is convex, here holds

    ϕ(y)ϕ(p)=˜ϕ(y)(py)=ϕ(py)+˜ϕ(y)(py),

    for any (y,p)V and where

    ˜ϕ(y)=10ϕ(y+s(py))ds,˜ϕ(y)=10(1s)ϕ(p+s(py))ds,

    are bounded functions in ˉΩ.

    Lemma 2.4. [14,27] For all vH1(Ω), ThT and TT, we have

    ||v||0,TΩCh1/2T||v||0+Ch1/2T|v|2. (15)

    In this section, we will recall a residual-based a posteriori error estimation for nonlinear elliptic equations. For the model problem that we studied in Section 2, a reliable and efficient a posteriori estimation will be obtained. In the end of this section, an adaptive finite element algorithm will be introduced.

    Here we define some error indicators. η() are error indicators and osc() represent the data oscillations. For ThT, TTh, we define

    η21,Th(ph,T)=h2Tph20,T,η22,Th(uh,yh,T)=h4Tf+uhϕ(yh)20,T+h3T(yh)n20,TΩ,η23,Th(yh,ph,T)=h4Tyhydϕ(yh)ph20,T+h3T(ph)n20,TΩ,osc2Th(f,T)=h4TffT20,T,osc2Th(yhyd,T)=h4T(yhyd)(yhyd)T20,T,

    where uhUhad, yh,phVh, (yh)n denotes the jump of yh, and n denotes the outward normal oriented to TΩ, and where fT is L2-projection of f onto piecewise constant space on T and fT=Tf|T|. For ωTh, we have

    η21,Th(ph,ω)=Tωη21,Th(ph,T),osc2Th(f,ω)=Tωosc2Th(f,T).

    Similarly, we have η22,Th(uh,yh,ω), η23,Th(yh,ph,ω) and osc2Th(yhyd,ω).

    Lemma 3.1. Let ThTμ under the conditions of Lemma 2.2, we have

    ||yyh||0C(||uuh||0+|||hTh(yyh)|||),||pph||0C(||uuh||0+|||hTh(yyh)|||+|||hTh(pph)|||),

    for sufficiently small μ.

    Proof. Suppose that yh,phV are intermediate variables satisfying the equations as follows

    a(yh,v)+(ϕ(yh),v)=(f+uh,v), vV, (16)
    a(q,ph)+(ϕ(yh)ph,q)=(yhyd,q), qV. (17)

    Employing the Galerkin orthogonality, the approximation properties, Lemma 2.2, and ||hT||μ, there exist similar results, which resemble to Lemma 3.1 in [21], for nonlinear elliptic optimal control problems with sufficiently small μ that

    ||yhyh||0C|||hTh(yhyh)|||,||phph||0C|||hTh(phph)|||.

    It has been proved |||yyh|||C||yyh||1C||uuh||0 in the Theorem 3.1 of [28] for nonlinear elliptic optimal control problems. Then associated with the Lemma 4.4 in [7], we deduce similar conclusions for nonlinear optimal control problems that

    |||yyh|||C||uuh||0. (18)
    |||pph|||C(||uuh||0+||yyh||0). (19)

    By using the triangle inequality, we have

    ||yyh||0||yyh||0+||yhyh||0,||pph||0||pph||0+||phph||0.

    In connection with what we discussed above, the triangle inequality and Lemma 2.1, it is easy to prove the prevenient results in Lemma 3.1.

    Now we are in the position to derive a posteriori error estimation for both the control, the state and adjoint state variables.

    Theorem 3.2. Let (u,y,p)Uad×H10(Ω)×H10(Ω) be the solution of (7)-(9) and (uh,yh,ph)Uhad×Vh×Vh be the solution of (12)-(14). Then we have a posteriori error upper bound

    ||uuh||20+||yyh||20+||pph||20c(η21,Th(ph,Th)+η22,Th(uh,yh,Th)+η23,Th(yh,ph,Th)),

    and apparently a global lower bound

    C(η21,Th(ph,Th)+η22,Th(uh,yh,Th)+η23,Th(yh,ph,Th))||uuh||20+||yyh||20+||pph||20+osc2Th(f,Th)+osc2Th(yhyd,Th),

    where c and C only depend on the shape regularity of Th.

    Proof. In view of Lemma 7.3.1 in [27], we similarly derive that

    ||uuh||20Cη21,Th(ph,Th)+C||phph||20, (20)

    where ph is the solution of (3.2). Then we shall to deal with ||phph||20.

    Let ξ=phph, and ξI=πhξ where πh is the standard Lagrange interpolation operator of ξ, then it follows from Lemma 2.1, Lemma 2.3 and (8) that

    ||phph||20=a(ξ,(phph))+(ϕ(yh)(phph),ξ)=((ξξI),(phph))+(ϕ(yh)phϕ(yh)ph,ξξI)+(ξI,(phph))+(ϕ(yh)phϕ(yh)ph,ξI)+((ϕ(yh)ϕ(yh))ph,ξ)=TThT(yhydϕ(yh)ph)(ξξI)TThT[(ph)n](ξξI)ds+(yhyh,ξI)+((ϕ(yh)ϕ(yh))ph,ξ)
    CTThh2T||yhydϕ(yh)ph||0,T|ξ|2,T+CTΩh3/2T(T[(ph)n]2)1/2||ξ||2,TΩ+||yhyh||0||ξ||0+C||ϕ(yh)ϕ(yh)||0||ph||0||ξ||0,CTThh4TT(yhydϕ(yh)ph)2+CTΩh3T[(ph)n]2+||yhyh||20+Cδ||ξ||22,

    where ϕ()W2,(Ω), the embedding ||v||0,C||v||2 and ||ph||0C have been adopted and δ is positive. Then choosing δ=12C, we obtain

    ||phph||20Cη23,Th(yh,ph,Th)+C||yhyh||20. (21)

    Similarly there is going to be proved by letting ˜ξ=yhyh that

    ||yhyh||20=a((yhyh),˜ξ)+(ϕ(yh)ϕ(yh),˜ξ)=((yhyh),(˜ξ˜ξI))+(ϕ(yh)ϕ(yh),˜ξ˜ξI)=TThT(f+uhϕ(yh))(˜ξ˜ξI)TΩT[(yh)n](˜ξ˜ξI)CTThh4TT(f+uhϕ(yh))2+CTΩh3TT[(yh)n]2+Cδ||˜ξ||20,

    where ϕ()W2,(Ω) have been applied and hence we have

    ||yhyh||20Cη22,Th(uh,yh,Th). (22)

    It is easy to derive the expect upper bound by combining with (20)-(22).

    Next we are going to deduce the global lower error bound through the standard bubble function [1,19]. Similar to Lemma 3.7 in [19], it can be similarly proved that there exists polynomial wTH20(T) such that

    Th4T((yhyd)Tϕ(yh)ph)2=Th4T((yhyd)Tϕ(yh)ph)wT, (23)

    and apparently

    ||wT||20,TCT((yhyd)Tϕ(yh)ph)2, (24)
    ch2T||wT||20,T|wT|22,TCh2T||wT||20,T. (25)

    Then it follows from (23) and (24) that

    Th4T((yhyd)Tϕ(yh)ph)2=Th4T((yhyd)Tϕ(yh)ph)wT,=Th4T(yhydϕ(yh)ph)wT+Th4T((yhyd)T(yhyd))wTI1+I2.

    Combining with (2.8) and Lemma 2.4, there holds

    I1=Th4T(php)ΔwT+Th4T(yhy)wT+Th4Tϕ(y)(pph)phwT+Th4T((ϕ(y)ϕ(yh))phwTC(||pph||20,T+||yhy||20,T+h4T||ϕ(y)||0,T||pph||0,T||ph||0,T||wT||0,,T+h4T||ϕ(y)ϕ(yh)||0,T||ph||0,T||wT||0,,T)+C(h2T||wT||20,T+h4T|wT|22,T)C(||pph||20,T+||yhy||20,T)+Cδh4T||wT||20, (26)

    where ϕ()W2,(Ω) have been used, the embedding ||v||0,,TC||v||2,T and the property ||ph||C have been adopted.

    Similarly, we have

    I2CTh4T((yhyd)T(yhyd)2+Cδh4T||wT||0,T. (27)

    Hence by using the Cauchy inequality with the help of (25)-(27), we obtain

    Th4T((yhyd)Tϕ(yh)ph)2C(||pph||20,T+||yhy||20,T+Th4T((yhyd)T(yhyd))2)+Ch4TTh4T((yhyd)Tϕ(yh)ph)2.

    Then it brings about

    Th4T((yhyd)ϕ(yh)ph)2C(Th4T((yhyd)Tϕ(yh)ph)2+Th4T((yhyd)T(yhyd))2)C(||pph||20,T+||yhy||20,T+Th4T((yhyd)T(yhyd))2). (28)

    Then we need to use the new bubble functions defined in [19] to both in deal with the jump. Similar to [18,19], it can be similarly proved that there exists polynomial wTH20(T) such that

    Th3T[(ph)n]2=Th3T[(ph)n]wT, (29)

    and apparently

    ||wT||20,TΩCTh3T[(ph)n]2, (30)
    ch2T||wT||20,TΩ||wT||22,TΩCh2T||wT||20,TΩ. (31)

    And then it follows from (29) and(30) that

    Th3T[(ph)n]2=Th3T[(ph)n]wT=T[(ph)n(p)n]wT=Th3T(pph)ΔwT+Th3T(yydϕ(y)p)wTI3+I4.

    Similarly, it can be deduced that

    I3Cδh3T||wT||20,TΩ+C||pph||20,TΩ.

    Combining with (13) and Lemma 2.4, there holds

    I4=Th3T(yhydϕ(yh)ph)wT+Th3T(yyh)wT+Th3Tϕ(y)(pph)wT+Th3T˜ϕ(y)(yyh)phwTC(||yyh||20,TΩ+Th3T(yhydϕ(yh)ph)2+h3T||ϕ(y)||0,TΩ||pph||0,TΩ||wT||0,,TΩ+h3T||˜ϕ(y)||0,TΩ||yyh||0,TΩ||ph||0,TΩ||wT||0,,TΩ)+Cδ(h3T||wT||20,TΩ+h3T|wT|22,TΩ)C(||yyh||20,TΩ+||pph||20,TΩ+Th3T(yhydϕ(yh)ph)2)+Cδh3T||wT||20,TΩ.

    Hence by using the Cauchy inequality with the help of (29)-(30), we have

    TΩTh3T[(ph)n]2C(||pph||20+||yyh||20)+CTThTh4T(yhydϕ(yh)ph)2, (32)

    where ϕ()W2,(Ω) and wTH20(Ω) have been used.

    In connection with (28) and (32), it is easy to get that

    η23,Th(yh,ph,Th)=TThTh4T(yhydϕ(yh)ph)2+TΩTh3T([ph]n)2Cphp20+Cyyh20+CTThTh4T((yhyd)(yhyd)T)2=Cphp20+Cyyh20+Cosc2Th(yhyd,Th).

    It can also be deduced that

    η22,Th(uh,yh,Th)=TThTh4T(f+uhϕ(yh))2+TΩTh3T([yh]n)2Cyhy20+Cuuh20+CTThTh4T(ffT)2=Cyhy20+Cuuh20+Cosc2Th(f,Th).

    Above-mentioned results tell the proof of Theorem 3.2 is accomplished.

    Theorem 3.2 gives a reliable and efficient posteriori error estimations for the sum of the L2norms errors for the control, the state and the co-state variables. Then we introduce an adaptive finite element algorithm to explain what we mainly investigate in this paper.

    Algorithm 3.1. Adaptive finite element algorithm for nonlinear optimal control problems:

    (o) Given an initial mesh Th0 and construct finite element space Uh0ad and Vh0. Select marking parameter 0<θ1 and set k:=0.

    (1) Solve the discrete nonlinear optimal control problem (12)-(14), then obtain approximate solution (uhk,yhk,phk) with respect to Thk.

    (2) Compute the local error estimator ηThk(T) for all TThk.

    (3) Select a minimal subset Mhk of Thk such that

    η2Thk(Mhk)θη2Thk(Thk),

    where η2Thk(ω)=η21,Th(ph,ω)+η22,Th(uh,yh,ω)+η23,Th(yh,ph,ω) for all ωThk.

    (4) Refine Mhk by bisecting b1 times in passing from Thk to Thk+1 and generally additional elements are refined in the process in order to ensure that Thk+1 is conforming.

    (5) Solve the discrete nonlinear optimal control problem (12)-(14), then obtain approximate solution (uhk+1,yhk+1,phk+1) with respect to Thk+1.

    (6) Set k=k+1 and go to step (2).

    In this section, we will do our best to demonstrate the convergence while we first give some properties which take vital significance to the proof of the convergence and even the quasi-optimality for the error indicators and the data oscillations before we begin to show the convergence analysis.

    Lemma 4.1. Let (u,y,p)Uad×H10(Ω)×H10(Ω) be the solution of (7)-(9) and (uh,yh,ph)Uhad×Vh×Vh be the solution of (12)-(14). Then we have a posteriori upper bound

    ||uuh||20+|||hTh(yyh)|||2+|||hTh(pph)|||2C(η21,Th(ph,Th)+η22,Th(uh,yh,Th)+η23,Th(yh,ph,Th)),

    where C only depends on the shape regularity of Th.

    Proof. By applying (18), the triangle inequality, and Lemma 2.1, we obtain

    |||hTh(yyh)|||2|||hTh(yyh)|||2+|||hTh(yhyh)|||2(||hTh|||||yyh|||)2+|||hTh(yhyh)|||2C||uuh||20+|||hTh(yhyh)|||2. (33)

    Similar to the proof of Lemma 3.3 in [21], we deduce that

    |||hTh(yhyh)|||2η22,Th(uh,yh,Th). (34)

    Analogously, the following conclusions can be drawn

    |||hTh(pph)|||2C||uuh||20+C||yyh||2+|||hTh(phph)|||2, (35)

    and apparently

    |||hTh(phph)|||2η23,Th(yh,ph,Th). (36)

    It is obvious to get the expected result in Lemma 4.1 via using (33)-(36), and the upper bound in Theorem 3.2.

    Next, we gives a stability result for error indicators which is can be found in Lemma 3.4 in [21], Lemma 4.1 in [23], and even Proposition 3.3 in [8], and so on.

    Lemma 4.2. For TTh, ThT, let uh1,uh2Uhad, yh1,yh2Vh and ph1,ph2Vh, we have

    η1,Th(ph1,T)η1,Th(ph2,T)C(|||hTh(ph1ph2)|||T+||hTh||,T||ph1ph2||0,T), (37)
    η2,Th(uh1,yh1,T)η2,Th(uh2,yh2,T)C(h2T||uh1uh2||0,T+|||hTh(yh1yh2)|||ωT+||hTh||,T||yh1yh2||0,ωT), (38)
    η3,Th(yh1,ph1,T)η3,Th(yh2,ph2,T)C(h2T||yh1yh2||0,T+|||hTh(ph1ph2)|||ωT+||hTh||,T||ph1ph2||0,ωT), (39)
    oscTh(yh1yd,T)oscTh(yh2yd,T)C(|||hTh(yh1yh2)|||T+||hTh||,T||yh1yh2||0,T), (40)

    where ωT denotes the patch of elements that share an edge with T.

    Proof. We first prove (37) while (40) can be just proved similarly. Consulting the literatures [1,21,25], namely the trace inequality, there exists TTh,ThT such that

    ||v||0,TC(h1/2T||v||0,T+h1/2T||v||1,T), (41)

    for arbitrary vH1(Ω). In connection with the inverse estimates and (41), we have

    ||[(ph1ph2)]n||0,TΩCh1/2T|||ph1ph2|||ωT. (42)

    Recalling (1) in Lemma 1, we know that

    hT|||ph1ph2|||ωTC||hTh(ph1ph2)||0,ωTC(|||hTh(ph1ph2)|||ωT+||(ph1ph2)hTh||0,ωT). (43)

    Recalling the definition of η1,Th(ph,T), we employ the triangle inequality to calculate for TThk that

    η1,Th(ph1,T)η1,Th(ph2,T)+h3/2T||[(ph1ph2)]n||0,TΩ. (44)

    Then it is easy to derive the desired result (37) by adopting (42)-(44).

    Next we are to prove (38) while (39) can be proved similarly. We calculate while applying the inequality (15) in Lemma 2.4 for the edge TT to obtain that

    h3/2T||(yh1yh2)||0,TTh3/2T(||(yh1yh2)T||0,TT+||(yh1yh2)T||0,TT)Ch3/2T(h1/2T||(yh1yh2)||0,T+h1/2T|(yh1yh2)|0,T+h1/2T||(yh1yh2)||0,T+h1/2T|(yh1yh2)|0,T)C(|||hTh(yh1yh2)|||0,TT+||(yh1yh2)hTh||0,TT. (45)

    Recalling the definition η2,Th(uh,yh,T), Lemma 2.3 and Lemma 3.1, we adopt the triangle inequality to calculate for TThk that

    η2,Th(uh1,yh1,T)η2,Th(uh2,yh2,T)+(h3T||[(yh1yh2)]||0,TΩ+h4T||ϕ(yh1)ϕ(yh2)||20,T)12η2,Th(uh2,yh2,T)+(h4T||Δ(yh1yh2)||20,T+h3T||[(yh1yh2)]||0,TΩ+h4T||ϕ(yh1)||0,T||yh1yh2||20,T)12η2,Th(uh2,yh2,T)+(h3T||[(yh1yh2)]||0,TΩ+h4T||uh1uh2||20,T+h4T|||yh1yh2|||T)12. (46)

    Then it is easy to deduce the expected result (38) by connecting (45) into (46).

    Lemma 4.3. For ThT, let MhTh be the set of marked elements and let ˜ThT be the refinement of Th so that we have

    η21,˜Th(˜p,˜Th)(1+σ)[η21,Th(p,Th)λη21,Th(p,Mh)]C(1+1σ)(T˜Thh4T||uh˜uh||20,T+|||h˜Th(yh˜yh)|||), (47)
    η22,˜Th(˜uh,˜yh,˜Th)(1+σ)[η22,Th(uh,yh,Th)λη22,Th(uh,yh,Mh)]C(1+1σ)(T˜Thh4T||uh˜uh||20,T+|||h˜Th(yh˜yh)|||2+||h˜Th||2||yh˜yh||20), (48)
    η23,˜Th(˜yh,˜ph,˜Th)(1+σ)[η23,Th(yh,ph,Th)λη23,Th(yh,ph,Mh)]C(1+1σ)(T˜Thh4T||uh˜uh||20,T+|||h˜Th(ph˜ph)|||2+||h˜Th||2||ph˜ph||20), (49)
    osc2Th(yhyd,Th˜Th)2osc2˜Th(˜yhyd,Th˜Th)2C(|||hTh(yh˜yh)|||2+||h˜Th||2||yh˜yh||20), (50)

    where uhUhad,˜uhU˜had,yh,phVh,˜yh,˜phV˜h,σ(0,1] and λ=123b2.

    Proof. We just prove (49) and (50). The proofs of (47) and (48) are similar with (49). Employing the Young's inequality with parameter σ and (39), we obtain

    η23,˜Th(˜yh,˜ph,~Th)η23,˜Th(yh,ph,~Th)C(1+1σ)(TThh4Tuh˜uh20,T+|||h˜Th(ph˜ph|||2+||h˜Th||2||ph˜ph||20)+δη23,~Th(yh,ph,~Th). (51)

    For a marked element TMhTh, let ˜ThT={T˜Th:TT}. Note that the jump [p]=0 as uUhadU˜had, and y,pVThV˜Th. According to the definition of hT, we can obtain hT=|T|12(2b|T|)122b2hT, inferring that

    T˜ThTη23,˜Th(y,p,T)23b2η23,Th(y,p,T).

    For TThMh, we deduce that

    η23,˜Th(y,p,T)η23,Th(y,p,T),

    which can use to derive that

    η23,˜Th23b2η23,Th(y,p,Mh)+η22,Th(y,p,ThMh)=η23,Th(y,p,Th)λη23,Th(y,p,Mh). (52)

    Then adding (52) into (51) and rearranging the terms can obtain the expected result (49).

    Next, for arbitrary TTh˜Th by using (50) in Lemma 4.2, we have

    osc2Th(yhyd,T)2osc2Th(˜yhyd,T)2C(|||hTh(yh˜yh)|||T+||hTh||,T||yh˜yh||0,T), (53)

    where the Young's inequality have been applied and oscTh(yyd,T)=osc˜Th(yyd,T). Then summing over TTh˜Th for (53), we can easy to derive the desired result (50).

    In order to facilitate computation, we introduce the following new notation

    e2hk=|||hThk(yyhk)|||2+|||hThk(pphk)|||2,E2hk=|||hThk+1(yhk+1yhk)|||2+|||hThk+1(phk+1phk)|||2,η2Thk(ω)=η21,Thk(phk,ω)+η22,Thk(uhk,yhk,ω)+η23,Thk(yhk,phk,ω),osc2Thk(ω)=osc2Thk(f,ω)+osc2Thk(yhkyd,ω),

    for ωThk. As to the proof of the convergence, one of the main obstacle is that there do not have the orthogonality while it is vital of the proof for the convergence. Thus getting back to the second place we transfer proof of the quasi-orthogonality. The latter is popularly adopted in the adaptive mixed and the nonconforming adaptive finite element methods [24]. Apparently it is true for the following basic relationships with Thk,Thk+1T and ThkThk+1 that

    |||hThk+1(vvhk+1)|||=|||hThk+1(vvhk)||||||hThk+1(vhk+1vhk)|||2a(hThk+1(vvhk+1),hThk+1(vhk+1vhk)), (54)

    so that we obtain the quasi-orthogonality in Lemma 4.4 by estimating the last term of (54).

    Lemma 4.4. For any ϵ>0,Thk,Thk+1T, there holds

    (1ϵ)|||hThk+1(yyhk+1)|||2(1+ϵ)|||hThk(yyhk)|||2+|||hThk+1(yhk+1yhk)|||2
    Cϵ(μ2+h20)(||uhuhk+1||20+||yhyhk||20+||yyhk+1||20), (55)
    (1ϵ)|||hThk+1(pphk+1)|||2(1+ϵ)|||hThk(pphk)|||2+|||hThk+1(yhk+1yhk)|||2Cϵ(μ2+h20)(||uhuhk+1||20+||phphk||20+||pphk+1||20), (56)

    where h0=maxTTh0hT.

    Proof. We just prove (55) while (56) can be proved in a similar way. Let yhk+1 satisfying (16) with f+uhk+1, then we have

    a(hThk+1(yyhk+1),hThk+1(yhk+1yhk))=a(hThk+1(yyhk+1),hThk+1(yhk+1yhk))+a(hThk+1(yhk+1yhk+1),hThk+1(yhk+1yhk)). (57)

    Similar to the proof of Lemma 4.3 in [23], we can estimate the second term of the right side of (57) as shown below

    a(hThk+1(yhk+1yhk+1),hThk+1(yhk+1yhk))ϵ8(|||hThk+1(yhk+1yhk+1)|||2+|||hThk+1(yhk+1yhk|||2)+C(1+1ϵ)μ2(||yhk+1yhk+1||20+||yhk+1yhk||20).

    Next we will subdivide the proof. By applying (2), (18), and the triangle inequality, we have

    |||hThk+1(yhk+1yhk+1)||||||hThk+1(yyhk+1)|||+|||hThk+1(yyhk+1)|||μ||yyhk+1||0+|||hThk+1(yyhk+1)|||+|||hThk+1(yyhk+1)|||C(μ+h0)||uuhk+1||0+|||hThk+1(yyhk+1)|||,

    in which we use the same way to derive that

    |||hThk+1(yhk+1yhk)|||C(μ+h0)||uuhk+1||0+|||hThk+1(yyhk)|||.

    It follows from the triangle inequality that

    ||yhk+1yhk+1||0C(||uuhk+1||0+||yyhk+1||0),

    and apparently

    ||yhk+1yhk||0C(||uuhk+1||0+||yyhk||0).

    Then in connection with what we discuss above to deduce that

    a(hThk+1(yhk+1yhk+1),hThk+1(yhk+1yhk))ϵ4(|||hThk+1(yyhk+1)|||2+|||hThk+1(yyhk)|||2)+Cϵ(μ2+h20)(||uuhk+1||20+||yyhk||20+|||yyhk|||2). (58)

    For the first term of the right side of (57), we obtain

    a(hThk+1(yyhk+1),hThk+1(yhk+1yhk))=((yyhk+1)hThk+1,(hThk+1(yhk+1yhk)))+(hThk+1(yyhk+1),(hThk+1(yhk+1yhk)))ϵ4(|||hThk+1(yyhk+1)|||2+|||hThk+1(yyhk)|||2)+Cϵ(μ2+h20)||uuhk+1||20. (59)

    Contacting (57), (58) and (59) to gain

    a(hThk+1(yyhk+1),hThk+1(yhk+1yhk))Cϵ(μ2+h20)(||uuhk+1||20+||yyhk||20+|||yyhk|||2)+ϵ2(|||hThk+1(yyhk+1)|||2+|||hThk+1(yyhk)|||2). (60)

    Because of hThk+1hThk, we have the following result

    |||hThk+1(yyhk)||||||hThk+1(yyhk)|||+μ||yyhk||0|||hThk(yyhk)|||+2μ||yyhk||0,

    such that

    |||hThk+1(yyhk)|||2(1+ϵ)|||hThk(yyhk)|||2+4μ2(1+1ϵ)||yyhk||20. (61)

    Then we can obtain (55) by combing with (54), (60), and (61).

    Theorem 4.5. Let (u,y,p)Uad×H10(Ω)×H10(Ω) be the solution of (7)-(9) and (uh,yh,ph)Uhad×Vh×Vh be the solution of (12)-(14) generated by the adaptive finite element algorithm 3.1. Then exist γ>0 and α(0,1] depending only on the shape of regularity of initial Th0,b,Ω and the marking parameter θ(0,1] such that

    e2hk+1+γη2hk+1(Thk+1)α(e2hk+γη2hk(Thk)), (62)

    apparently providing h01 and sufficiently small μ.

    Proof. Taking Lemma 3.1, Lemma 4.3, and Lemma 4.4 into account, we can easy to derive that

    e2hkCη2Thk(Thk), (63)
    (1ϵ)e2hk+1(1+ϵ)2e2hkE2hk+Cϵ(μ2+h20)(||uuhk+1||20+||yyhk||20+||yyhk+1||20+||pphk||20+||pphk+1||20), (64)
    η2Thk+1(Thk+1)(1+σ)(η2Thk(Thk)λη2Thk(Mhk))+C(1+1σ)[(μ2+h20)(||uuhk||20+||uuhk+1||20+||yyhk||20+||yyhk+1||20+||pphk||20+||pphk+1||20)+E2hk]. (65)

    Simplifying (64) and (65) by employing Lemma 3.1 and Lemma 4.1, we have

    (1ϵ)e2hk+1(1+ϵ)2e2hkE2hk+Cϵ(μ2+h20)(η2Thk(Thk)+η2Thk+1(Thk+1)), (66)
    η2Thk+1(Thk+1)(1+σ)(η2Thk(Thk)λη2Thk(Mhk))+C(1+1σ)[(μ2+h20)(η2Thk(Thk)+η2Thk+1(Thk+1))+E2hk]. (67)

    Multiplying (67) with γ1=1C(1+σ1) and adding the result to (66), we can easy to deduce that

    (1ϵ)e2hk+1+γ1η2Thk+1(Thk+1)(1+ϵ)2e2hk+γ1(1+σ)(η2Thk(Thk)λη2Thk(Mhk))+γ2(μ2+h20)(η2Thk(Thk)+η2Thk+1(Thk+1)),

    where γ2=1+Cϵ. Applying the marking strategy in Algorithm in 3.1 and (63), we have

    (1ϵ)e2hk+1+[γ1γ2(μ2+h20)]η2Thk+1(Thk+1)((1+ϵ)2γ3γ1(1+σ)λθC)e2hk+[γ1(1+σ)(1(1γ3)λθ)+γ2(μ2+h20)]η2Thk(Thk),

    where γ3(0,1). By simple calculation and letting

    γ=γ1γ2(μ2+h20)1ϵ,α1=11Cγ3γ1(1+σ)λθϵ(3+ϵ)1ϵ,α2=γ1(1+σ)(1(1γ3)λθ)+γ2(μ2+h20)γ1γ2(μ2+h20),

    where α1(0,1) if choosing ϵ,σ small enough and where α2(0,1) if electing μ,h0 sufficiently small, then here holds

    e2hk+1+γη2Thk+1(Thk+1)α1e2hk+α2γη2Thk(Thk),

    which tells a contraction property, namely (62), in Algorithm 3.1 if selecting α=max{α1,α2}.

    Theorem 4.6. Let (u,y,p)Uad×H10(Ω)×H10(Ω) be the solution of (7)-(9) and (uh,yh,ph)Uhad×Vh×Vh be the solution of (12)-(14) generated by the adaptive finite element algorithm 3.1 with the other conditions being same with Theorem 4.5, then there holds

    uuhk20+yyhk20+pphk200ask.

    Proof. It follows from Lemma 4.1 and Theorem 4.5 that we can obviously get

    ||uuhk||00ask.

    Then combining with Lemma 3.1, it is distinct to get the desired result in Theorem 4.6.

    In this section, we consider the quasi-optimality for the adaptive finite element method. Firstly we give the notations interpretation. For any Th,˜ThT, let #Th be the number of elements in Th, and Th˜Th be the smallest common conforming refinement of Th and ˜Th satisfying [15,30,31] the property

    #(Th˜Th)#Th+#˜Th#Th0. (68)

    According to [14,21,23,24], we need to define a function approximation class

    As:={(u,y,p,yd)L2(Ω)×H10(Ω)×H10(Ω)×L2(Ω):|(u,y,p,yd)|s<+},

    where

    |(u,y,p,yd)|s:=supN>0NsinfThTNinf(uh,yh,ph)Uhad×Vh×Vh{uuh20+yyh20+pph20+osc2Th(Th)}12,

    and

    TN:={ThT:#Th#Th0N0}.

    We need a local upper bound for the distance between two nested solutions consulting [8] in order to illustrate the quasi-optimality of an adaptive finite element method due to the errors here can only be estimated by using refined element indices without buffer.

    Lemma 5.1. Let (u,y,p)Uad×H10(Ω)×H10(Ω) be the solution of (7)-(9). Given sufficiently small μ, let ThTμ and Th˜ThT, (uh,yh,ph)Uhad×Vh×Vh and (˜uh,˜yh,˜ph)U˜had×V˜h×V˜h be the solution of (12)-(14) on Th and ˜Th. Then there holds

    ||uh˜uh||20+||yh˜yh||20+||ph˜ph||20Cη2Th(Rh), (69)

    where Rh:=RTh˜Th is the subset of elements that are refined from Th to ˜Th.

    Proof. According to the optimal condition (14), we obtain

    (αuh+ph,vuh)0, vVh,(α˜uh+˜ph,v˜uh)0, vV˜h,

    thus getting

    α||uh˜uh||20=(αuh,uh˜uh)(α˜uh,uh˜uh)(˜phph,uh˜uh)+(αuh+ph,uh˜uh). (70)

    For (˜phph,uh˜uh) of (70), we have

    (˜phph,uh˜uh)=(S˜Th(S˜Th(f+˜uh)yd)STh(STh(f+uh)yd),uh˜uh)=(S˜Th(S˜Th(f+˜uh)yd)S˜Th(S˜Th(f+uh)yd),uh˜uh)+(S˜Th(S˜Th(f+uh)yd)STh(STh(f+uh)yd),uh˜uh)=(S˜Th(S˜Th(˜uhuh)),uh˜uh)+(S˜Th(S˜Th(f+uh)yd)STh(STh(f+uh)yd),uh˜uh)=(S˜Th(˜uhuh),S˜Th(uh˜uh))+(S˜Th(S˜Th(f+uh)yd)STh(STh(f+uh)yd),uh˜uh)(S˜Th(S˜Th(f+uh)yd)STh(STh(f+uh)yd),uh˜uh).

    In order to estimate the right side of the above inequality, we suppose that φH10(Ω) is the solution of the following equation

    a(φ,v)=(S˜Th(S˜Th(f+uh)yd)STh(STh(f+uh)yd),v), vH10(Ω).

    By applying the duality arguments, we can gain that

    ||S˜Th(S˜Th(f+uh)yd)STh(STh(f+uh)yd)||20=a(φ,S˜Th(S˜Th(f+uh)yd)STh(STh(f+uh)yd))=a(φξThφ,S˜Th(S˜Th(f+uh)yd)STh(STh(f+uh)yd))+(S˜Th(f+uh)STh(f+uh),ξThφφ)+(S˜Th(f+uh)STh(f+uh),φ), (71)

    in which ξTh is the standard Lagrange interpolator onto Vh. For the first term of right side of (71), we have

    a(φξThφ,S˜Th(S˜Th(f+uh)yd)STh(STh(f+uh)yd))=a(φξThφ,S˜Th(S˜Th(f+uh)yd)S˜Th(STh(f+uh)yd))+a(φξThφ,S˜Th(STh(f+uh)yd)STh(STh(f+uh)yd))C(||φ||2||S˜Th(f+uh)STh(f+uh)||0+||h1Th(φξThφ)||0||hTh(S˜Th((STh(f+uh)yd)STh(STh(f+uh)yd)))||0)C||φ||2(||S˜Th(f+uh)STh(f+uh)||0+||hTh(S˜Th((STh(f+uh)yd)STh(STh(f+uh)yd)))||0). (72)

    Similar to Lemma 2 in [14], we infer that

    ||S˜Th(f+uh)STh(f+uh)||0Cη2,Th(uh,yh,Rh), (73)
    ||hTh(S˜Th((STh(f+uh)yd)STh(STh(f+uh)yd)))||0Cη3,Th(yh,ph,Rh). (74)

    By using the similar way, we deduce that

    (S˜Th(f+uh)STh(f+uh),ξThφφ)Cη2,Th(uh,yh,Rh), (75)
    (S˜Th(f+uh)STh(f+uh),φ)Cη2,Th(uh,yh,Rh). (76)

    Combining (72)-(76) and Lemma 2.2 with H2-regularity, we derive that

    (˜phph,uh˜uh)C(η2,Th(uh,yh,Rh)+η3,Th(yh,ph,Rh)). (77)

    Next we are going to estimate the second term on the right side of (70). Assume that

    (αuh+ph,uh˜uh)(αuh+ph,vh˜uh),vhUhad.

    Then we set vh=πTh˜uh for which πTh is L2projection onto P0(Th), thus obtaining

    (αuh+ph,vh˜uh)T=((IThπTh)(αuh+ph),(πThITh)(vh˜uh)),

    for arbitrary TTh and where ITh is the identical L2projection onto P0(Th). Owing to TThRhTh such that (πThITh)(˜uhuh)=0, then here holds

    (αuh+ph,vh˜uh)Cη1,Th(ph,Rh)||uh˜uh||0. (78)

    In connection with (70), (77) and (78), we infer that

    ||uh˜uh||0C(η1,Th(ph,Rh)+η2,Th(uh,yh,Rh)+η3,Th(yh,ph,Rh)). (79)

    Employing (73)-(76) and the triangle inequality, we deduce that

    ||ph˜ph||0=||STh(STh(f+uh)yd)S˜Th(S˜Th(f+˜uh)yd)||0||STh(STh(f+uh)yd)S˜Th(S˜Th(f+uh)yd)||0+||S˜Th(S˜Th(f+uh)yd)S˜Th(S˜Th(f+˜uh)yd)||0C(||uh˜uh||0+η2,Th(uh,yh,Rh)+η3,Th(yh,ph,Rh)). (80)

    It is similar to Lemma 2 in [14] that we infer that

    ||yh˜yh||0=||STh(f+uh)S˜Th(f+˜uh)||0||STh(f+uh)S˜Th(f+uh)||0+||S˜Th(f+uh)S˜Th(f+˜uh)||0C(||uh˜uh||0+η2,Th(uh,yh,Rh)). (81)

    To sum up, the proof is finished by adopting (79)-(81).

    Next lemma tells the Dörfler property on the set Rh=RThkTh in order to bound the number of marked elements.

    Lemma 5.2. We assume that the marking parameter θ(0,θ) with θ=C1+4C and let (uh,yh,ph)Uhad×Vh×Vh and (uhk,yhk,phk)Uhkad×Vhk×Vhk be the solution of (12)-(14) on Th and Thk. There exists a constant δ=12(1θθ) such that

    e2Thk+osc2Thk(Thk)δ(e2Th+osc2Th(Th)), (82)

    where e2Th=||uuh||20+||yyh||20+||pph||20, similarly for e2Thk.Then there holds

    η2Th(Rh)θη2Th(Th).

    Proof. Combining with (82) and the upper bound in Theorem 4.5, we can obtain

    C(12δ)η2Th(Th)(12δ)(e2Th+osc2Th(Th)e2Th2e2Thk+osc2Th(Th)2osc2Thk(Thk).

    Employing the triangle inequality and the Young's inequality, here holds

    ||uuh||202(||uuhk||20+||uhuhk||20),||yyh||202(||yyhk||20+||yhyhk||20),||pph||202(||pphk||20+||phphk||20).

    Thus obtaining the result

    η2Th(Th)2η2Thk(Thk)2Cη2Th(Rh), (83)

    with Lemma 5.1. By applying the dominance property which is similar to Remark 2.1 in [8], we infer that

    osc2Th(T)η2Th(T), (84)

    for TRh and apparently

    osc2Th(f,T)η2Thk(f,T), (85)

    for TThRh. For TTh˜Th, we can get the following result by employing (50) of Lemma 4.3 and the inverse estimates

    osc2Th(ThRh)2osc2Thk(ThRh)2C(||yhyhk||20+||phphk||20). (86)

    Then it can be derived via using (83), (84) and (86) that

    C(12δ)η2Th(Th)(1+4C)η2Th(Rh), (87)

    which tells the proof.

    Lemma 5.3. Assume that the marking parameter θ satisfies the the conditions in Lemma 5.2. Let (u,y,p)Uad×H10(Ω)×H10(Ω) be the solution of (7)-(9) and (uhk,yhk,phk)Uhkad×Vhk×Vhk be the solution of (12)-(14) generated by Algorithm 3.1. Then the number of marked elements MhkThk satisfies

    #MhkCβ12sδ12s|(u,y,p,yd)|1ss(e2Thk+osc2Thk(Thk))12s,

    if (u,y,p,yd)As for μ being small enough.

    Proof. Let ν2=δβ12s(e2Thk+osc2Thk(Thk)), where δ is defined in Lemma 5.2 and β is to be defined as follows. Then there exists a ThνT and (uhν,yhν,phν)Uhνad×Vhν×Vhν such that

    #Thν#Th0C|(u,y,p,yd)|1sν1s, (88)
    ||uuhν||20+||yyhν||20+||pphν||20+osc2Thν(f,Thν)+osc2Thν(yhνyd,Thν)ν2, (89)

    for any (u,y,p,yd)As. Next we suppose Thϵ=ThνThk is the smallest common refinement of Thν and Thk, and let (uhϵ,yhϵ,phϵ) be the solution of (12)-(14). Then we give the inequalities as follows in order to deduce the expect result we wanted

    e2Thϵ+osc2Thϵ(Thϵ)β(e2Thν+osc2Thν(Thν), (90)

    where there are some notations been defined by

    e2Thν=||uuhν||20+||yyhν||20+||pphν||20,osc2Thν(Thν)=osc2Thν(f,Thν)+osc2Thν(yhν,Thν),

    and apparently e2Thϵ and osc2Thϵ(Thϵ) can be defined similarly. Just as obviously, here holds

    ||vvhν||20=||vvhϵ||20+||vhϵvhν||20+2(vvhϵ,vhϵvhν), (91)

    for all vUad,vhνUhνad, and vhϵUhϵad. Applying the Young's inequality and (91), we have

    (uuhϵ,uhϵuhν)=(uuhν,uhϵuhν)+(uhνuhϵ,uhϵuhν)(uuhν,uhϵuhν)||uuhν||20+14||uhϵuhν||20, (92)

    and apparently in the same way

    (yyhϵ,yhϵyhν)=||yyhν||20+14||yhϵyhν||20, (93)
    (pphϵ,phϵphν)=||pphν||20+14||phϵphν||20. (94)

    Hence combining (91)-(94) to get

    ||uuhϵ||20+||uhϵuhν||20+||yyhϵ||20+||yhϵyhν||20+||pphϵ||20+||phϵphν||206(||uuhν||20+||yyhν||20+||pphν||20). (95)

    For all TThν, assuming that ThT:={TThϵ:TT}, then we derive that

    TThT||ff2T||20,T=TThT(Tf2(Tf)2|T|)=Tf2TThT(Tf)2|T|Tf2TThT(Tf)2|T|CTf2(Tf)2|T|=C||ffT||20,T,

    which tells that

    osc2Thϵ(f,Thϵ)Cosc2Thν(f,Thν). (96)

    By using the same way, we have

    osc2Thϵ(yhϵyd,Thϵ)Cosc2Thν(yhνyd,Thν). (97)

    Therefore, we can gain (90) via employing (95)-(97). Based on the definition of ν2 and (89), we deduce that

    e2Thϵ+osc2Thϵ(Thϵ)βν2=δ(e2Thk+osc2Thk(Thk)).

    According to Lemma 5.2, we find that the subset RThkThϵ verifies the marking property for θθ, then we infer that

    #Mhk#RThkThϵ#Thϵ#Thk#ThνTh0. (98)

    Hence we can obtain the desired result by contacting with (88), (98), and the definition of ν2.

    Next we derive a equivalent property of the indicator dominates oscillation by concluding Theorem 3.2, Lemma 3.1, and Lemma 4.1 as follows

    e2Thk+osc2Thke2hk+γη2Thk(Thk), (99)

    which is of vital importance for the proof of quasi-optimality.

    Theorem 5.4. Assume that Th0 satisfies the condition (b) of Section 4 in [32]. Let (u,y,p)Uad×H10(Ω)×H10(Ω) be the solution of (7)-(9) and (uhk,yhk,phk)Uhkad×Vhk×Vhk be the solution of (12)-(14) generated by Algorithm 3.1. Then there holds

    #Thk#Th0C|(u,y,p,yd)|1ss(e2Thk+osc2Thk(Thk))12s,

    provided h01 if (u,y,p,yd)As for μ being small enough.

    Proof. Similar to the Theorem 4 in [14], we infer that

    #Thk#Th0Ck1i=0Mhk. (100)

    Employing (100) and Lemma 5.3 to get that

    #Thk#Th0Ck1i=0MhkCχk11=0(e2Thk+osc2Thk(Thk))12s, (101)

    where χ=β12s|(u,y,p,yd)|1ssδ12s. Combining with (99), (101), and Theorem 4.5, we deduce that

    #Thk#Th0Cχ(e2Thk+osc2Thk(Thk))12ski=0αik,

    which yields the proof of Theorem 5.4.

    In this section, we firstly present the adaptive iteration method where the purpose is to provide empirical analysis for our theory.

    Algorithm 6.1. Given an initial control u0hUhad, then for k=1,2,, seek (ykh,pkh,ukh) such that

    a(ykh,wh)+(ϕ(ykh),wh)=(f+uk1h,wh), whVh,a(qh,pk1h)+(ϕ(ykh)pk1h,qh)=(ykhyd,qh), qhVh,(αukh+pk1h,vhukh)0, vkhUhad,

    and apparently

    ukh=1α(Phpkh+max(0,ˉpkh)),

    where Ph is the L2-projection from L2(Ω) to Uh and ˉpkh=Ωpkh|Ω|.

    Example 1. We consider the nonlinear optimal control problem subject to the state equation

    Δy+y3=f+u,Δp+3y2p=yyd,

    where we choose α=0.5 and Ω=[0,1]×[0,1] and apparently exact solution

    u=1α(max(0,ˉp)p),p=(sin(πx1)+sin(πx2)),y=sin(πx1)sin(πx2).

    By simple calculation we have Ωpdx=4π which satisfies uUad.

    In terms of the same error and actuarial accuracy, the adaptive refinement process saves time than the uniform refinement process. In Figures 1-2, we provide the profiles of the exact state variables, the numerical state variables, the exact co-state and the co-state on adaptively refined grids with θ=0.3 and 15 adaptive loops for Example 1 generated by Algorithm 3.1 and then we plot the adaptive girds after 5 steps and 13 steps with θ=0.3 and 15 adaptive iterations for Example 1 in Figure 3. It is easy to observe that lager gradients exist in some certain regions but the solutions are smooth. Moreover, the grid refinement at the center of the domain, while the solutions may have lager gradients near the boundary.

    Figure 1.  The profiles of the exact state and the numerical state on adaptively refined grids with θ=0.3 and 15 adaptive loops for Example 1.
    Figure 2.  The profiles of the exact state, the numerical state, the exact co-state variables and the co-state variables on adaptively refined grids with θ=0.3 and 15 adaptive loops for Example 1.
    Figure 3.  The adaptive girds after 5 steps and 13 steps with θ=0.3 and 15 adaptive loops for Example 1 generated by Algorithm 3.1.

    In Figure 4, we plot the profiles of the numerical state and the co-state variables on uniformly refined grids (θ=1) and 15 adaptive loops for Example 1 generated by Algorithm 3.1, and then the adaptively refined triangulations after 5 adaptive iterations with θ=0.4 and uniformly refined triangulations after 5 uniform iterations of 15 loops are performed in Figure 5. Obviously, it is not difficult to find that the uniformly refined grids seem to have a better encryption effect, but in connection with Figures 1-2 and Figure 4 to observe that the adaptive finite element method may deliver even much smaller errors compared to uniformly refined method. Combining Figure 3 and Figure 5, the adaptive iteration is more efficient and effective than the uniform iteration whereas the uniform iteration needs to be solved at a higher cost of grids, thus increasing iteration time. So the adaptive finite element method has some advantages in the numerical approximation process.

    Figure 4.  The profiles of the numerical state and the co-state variables on uniformly refined grids (θ=1) and 15 adaptive loops for Example 1.
    Figure 5.  The adaptive girds after 5 steps with θ=0.3 and the uniform refinement (θ=1) after 5 steps for Example 1 generated by Algorithm 3.1.

    In Figure 6, we show the convergence history of the total error estimate indicators, where we plot the adaptive triangle iterations of 15 adaptive loops with the coefficient θ=0.3 and θ=0.4, even more there provides a convergence of the total error estimation indicators for the uniform triangle iterations (θ=1). We can see an error reduction with slope -1 that is the optimal convergence rate what we expect via applying linear finite elements from the upper two pictures in Figure 6. Meanwhile, we give the comparisons of the error estimations, for which we know that the optimal second order convergence for the error reductions of error estimations.

    Figure 6.  The error estimations of adaptively refined grids with θ=0.3, θ=0.4, and the error estimates of uniformly refined grids for Example 1.

    Example 2. We consider the same nonlinear optimal control problem as Example 1 with α=0.1, Ω=(0,1)×(0,1), and apparently the exact solution

    y={5×1010e1m,m<0,0,m0,p={7×1010e1m,m<0,0,m0,

    where m=(x10.2)2+(x20.6)20.04 and absolutely Ωudx>0 via simply calculating.

    In Figure 7, we plot the numerical state and the co-state on adaptively refined grids with θ=0.3 and 27 adaptive loops for Example 2 generated by Algorithm 3.1. Obviously, we can observe the lager gradients concentrate on the certain regions as the adaptive finite element method may deliver much smaller errors compared with the uniform refinement. In Figure 8, we present the adaptive grids after 15 and 25 adaptively refined by choosing D¨orfler's marking parameter θ=0.3 with 27 adaptive loops. Apparently, the grids gather around the regions where there exist much lager gradients. Therefor they just validate the phenomenon in Figure 7. In addition, the grides focus on the points as f and yd have singularities near these points. Hence when dealing with singular points, adaptive encryption has better effect.

    Figure 7.  The profiles of the numerical state and the co-state on adaptively refined grids with θ=0.3 and 27 adaptive loops for Example 2.
    Figure 8.  The adaptive girds after 15 steps and 25 steps with θ=0.3 for Example 2 generated by Algorithm 3.1.

    We give the comparisons of convergence history of Example 2 in Figure 9. The left plot in Figure 9 is adaptively refined with D¨orfler's marking parameter θ=0.3 and 27 adaptive loops while the right one is uniform refinement (θ=1). With the optimal L2norms convergence we desired, we can see the errors reduction for adaptive refinement. Moreover, the reduced orders only can be found in uniform refinement because of the singularity of the solutions.

    Figure 9.  The error estimations of adaptively refined grids with θ=0.3 and the error estimations of uniformly refined grids for Example 2.

    The authors express their thanks to the referees for their helpful suggestions, which led to improvements of the presentation.



    [1] A posteriori error estimators in finite element analysis. Comput. Methods Appl. Mech. Engrg. (1997) 142: 1-88.
    [2] Error estimates for adaptive finite computations. SIAM J. Numer. Anal. (1978) 15: 736-754.
    [3] A convergent noncomforming adaptive finite element method with quasi-optimal complexity. SIAM J. Numer. Anal. (2010) 47: 4639-4659.
    [4] Adaptive finite element methods with convergence rates. Numer. Math. (2004) 97: 219-268.
    [5] Convergence analysis of a conforming adaptive finite element method for an obstacle problem. Numer. Math. (2007) 107: 455-471.
    [6] Error reduction and convergence for an adaptive mixed finite element method. Math. Comput. (2006) 75: 1033-1042.
    [7] Error estimations for the numerical approximation of semilinear elliptic control problems with finitely many state constraints. ESAIM Control Optim. Calc. Var. (2002) 8: 345-374.
    [8] Quasi-optimal convergence rate for an adaptive finite element method. SIAM J. Numer. Anal. (2008) 46: 2524-2550.
    [9] An adaptive finite element algorithm with reliable and efficient error control for linear parabolic problems. Math. Comput. (2004) 73: 1167-1193.
    [10] Adaptive finite element approximations for a class of nonlinear eigenvalue problems in quantum physics. Adv. Appl. Math. Mech. (2011) 3: 493-518.
    [11] Error estimates of fully discrete mixed finite element methods for semilinear quadratic parabolic optimal control problem. Comput. Methods Appl. Mech. Engrg. (2010) 199: 1415-1423.
    [12] (2015) High Efficient and Accuracy Numerical Methods for Optimal Control Problems. Beijing: Science Press.
    [13] Convergence and optimal complexity of adaptive finite element eigenvalue computations. Numer. math. (2008) 110: 313-355.
    [14] Convergence and quasi-optimality of an adaptive finite element method for controlling L2 errors. Numer. Math. (2011) 117: 185-218.
    [15] A convergent adaptive algorithm for Poisson's equation. SIAM J. Numer. Anal. (1996) 33: 1106-1124.
    [16] An adaptive finite element method for linear elliptic problems. Math. Comput. (1988) 50: 361-383.
    [17] Convergence analysis of an adaptive finite element method for distributed control problems with control constrains. Inter. Ser. Numer. Math. (2007) 155: 47-68.
    [18] L2 norm equivalent a posteriori error for a constraint optimal control problem. Inter. J. Numer. Anal. Model. (2009) 6: 335-353.
    [19] Adaptive finite element approximation for a constrained optimal control problem via multi-meshes. J. Sci. Comput. (2009) 41: 238-255.
    [20] Adaptive finite element method for elliptic optimal control problems: Convergence and optimality. Numer. Math. (2017) 135: 1121-1170.
    [21] W. Gong, N. Yan and Z. Zhou, Convergence of L2-norm based adaptive finite element method for elliptic optimality control problem, arXiv: 1608.08699.
    [22] Convergence of adaptive finite elements for optimal control problems with control constraints. Inter. Ser. Numer. Math. (2014) 165: 403-419.
    [23] Convergence and quasi-optimality of an adaptive finite element method for optimal control problems on L2-errors. J. Sci. Comput. (2017) 73: 438-458.
    [24] Convergence and quasi-optimality of an adaptive finite element method for optimal control problems with integral control constraint. Adv. Comput. Math. (2018) 44: 367-394.
    [25] Adaptive finite element approximation for distributed elliptic optimal control problems. SIAM J. Control Optim. (2002) 41: 1321-1349.
    [26] J.-L. Lions, Optimal Control of Systems Governed by Partial Differential Equations, Springer, Berlin, 1971.
    [27] (2008) Adaptive Finite Element Methods for Optimal Control Governed by PDEs. Beijing: Science Press.
    [28] A posteriori error estimates for control problems governed by nonlinear elliptic equation. Appl. Numer. Math. (2003) 47: 173-187.
    [29] Convergence of adaptive finite element methods for general second order linear elliptic PDEs. SIAM J. Numer. Anal. (2005) 43: 1803-1827.
    [30] Data oscillation and convergence of adaptive FEM. SIAM J. Numer. Anal. (2000) 38: 466-488.
    [31] Convergence of adaptive finite element methods. SIAM Rev. (2002) 44: 631-658.
    [32] Optimality of a standard adaptive finite element method. Found. Comput. Math. (2007) 7: 245-269.
  • This article has been cited by:

    1. Liming Guo, Chunjia Bi, Convergence and quasi-optimality of an adaptive finite element method for nonmonotone quasi-linear elliptic problems on L2 errors, 2023, 139, 08981221, 38, 10.1016/j.camwa.2023.03.005
  • Reader Comments
  • © 2020 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Metrics

Article views(3162) PDF downloads(255) Cited by(1)

Figures and Tables

Figures(9)

Other Articles By Authors

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog