李氣發(fā),謝資清,,陶 霞
(1.貴州師范大學(xué)數(shù)學(xué)與計(jì)算機(jī)科學(xué)學(xué)院,中國(guó) 貴陽(yáng) 550001;2.湖南師范大學(xué)數(shù)學(xué)與計(jì)算機(jī)科學(xué)學(xué)院,中國(guó) 長(zhǎng)沙 410081)
Volterra積分微分方程廣泛應(yīng)用于物理、生物、控制論等自然科學(xué)領(lǐng)域.該類方程中的積分項(xiàng)反映了實(shí)際過(guò)程中的記憶或反饋性質(zhì),使得它與傳統(tǒng)的微分方程有著本質(zhì)的區(qū)別.如何快速、高效而準(zhǔn)確地求解這類問(wèn)題是科學(xué)計(jì)算中的重要問(wèn)題.
早期關(guān)于積分微分方程的數(shù)值方法主要是差分方法[1].近年來(lái),Hesthven[2],Tang[3-4], Guo[5]和Wang[6]等在用譜方法求解積分微分方程方面做了大量的工作,使得這方面研究逐漸引起了學(xué)者們的關(guān)注.事實(shí)上,譜方法具有“無(wú)窮階”的收斂性,即如果原方程解無(wú)限光滑,那么適當(dāng)?shù)淖V方法求得的近似解將以N-1的任意冪次速度收斂于精確解.特別是譜方法適合于求解非常規(guī)則而幾何區(qū)域維數(shù)非常大的問(wèn)題.
最近,Tang[14]等采用節(jié)點(diǎn)上的譜配點(diǎn)法求解帶光滑核的第二類積分方程.在此基礎(chǔ)上,Xie[15]等采用譜和偽譜Jacobi-Galerkin方法求解第二類Volterra積分方程,并進(jìn)行了嚴(yán)格的收斂性分析.他們最近又進(jìn)一步證明了M-條件,以及Galerkin譜方法求解上述積分方程的超幾何收斂性.
本文采用譜Legendre-Gelerkin方法求解第二類Volterra積分微分方程,在一定的假設(shè)條件下,證明了解滿足M-條件,并證明了譜Legendre-Galerkin方法在L2和L∞意義下的超幾何收斂性質(zhì).事實(shí)上其結(jié)果相比文獻(xiàn)[17]中的譜Petrov-Galerkin方法能夠達(dá)到更高的收斂率.
首先考慮具有如下形式的一般線性Volterra積分微分方程(VIDEs):
(1)
y(0)=y0,
(2)
事實(shí)上,方程(2)等價(jià)于
(2′)
(3)
(4)
(5)
(6)
(7)
首先討論譜Legendre-Galerkin方法的數(shù)值實(shí)現(xiàn).令PN是定義在[-1,1]上的次數(shù)不超過(guò)N的多項(xiàng)式構(gòu)成的空間,Lj(x)是j次Legendre多項(xiàng)式,j=0,1,…,N.即有PN=span{L0(x),L1(x),…,LN(x)}.
譜Legendre-Galerkin方法為:對(duì)(7)式尋求uN∈PN,vN∈PN,使得
(vN,wN)-(uN,wN)-(SuN,wN)=(g,wN),?wN∈PN,
(8)
(uN,wN)-(KvN,wN)=(u0,wN),?wN∈PN,
(9)
(10)
將式(10)分別代入(8)和(9),并取wN=Li(x),i=0,1,…,N,可得
(11)
(12)
定義正交投影算子ΠN:L2(I)→PN,使得對(duì)任意的u∈L2(I)都有(ΠNu,wN)=(u,wN),?wN∈PN.
最后,考慮方程(7),為使解滿足M-條件,需要作以下假設(shè):
令D={(x,τ):-1≤x≤1,-1≤τ≤x},I=[-1,1].
(A1)g(x)和k(x,τ)分別在I和D上充分光滑,且k(x,τ)=k(x-τ).
引理2(M-條件)如果條件(A1)和(A2)成立,則存在常數(shù)c和M≥max{r0,r1},使得方程(7)的解u和v滿足‖u(n)(x)‖L∞(I)≤cMn,‖v(n)(x)‖L∞(I)≤cMn,n=0,1,2,….
證由文獻(xiàn)[17]中定理3.1可直接得到.
根據(jù)式(8),(9)和投影算子ΠN的定義,(8),(9)等價(jià)于
vN-uN-ΠNSuN=ΠNg,
(13)
uN-ΠNKvN=u0.
(14)
定理1滿足方程(8)和(9)的譜Legendre-Galerkin解存在且唯一.
證設(shè)uN,vN為(8)和(9)的譜Legendre-Galerkin解.由于解空間是有限維,且問(wèn)題(8)和(9)與(13)和(14)等價(jià),為此只要證明(13)和(14)中,當(dāng)g=0,u0=0時(shí),uN=0,vN=0即可.為此考慮
vN-uN-ΠNSuN=0,
(15)
uN-ΠNKvN=0.
(16)
將(16)代入(15)得
vN-ΠNKvN-ΠNSuN=vN-KvN+(KvN-ΠNKvN)-SuN+(SuN-ΠNSuN)=
vN-KvN+(KvN-ΠNKvN)-S[ΠNKvN]+(SuN-ΠNSuN)=0.
(17)
即有
vN=KvN+S[ΠNKvN]-(KvN-ΠNKvN)-(SuN-ΠNSuN).
(18)
對(duì)上式的第2項(xiàng),根據(jù)k(x,τ)的光滑性,并運(yùn)用Dirichlet公式
(19)
可得
|S[ΠNKvN]|=|S[KvN-(KvN-ΠNKvN)]|≤|S(KvN)|+|S(KvN-ΠNKvN)|≤
(20)
由(18),(20)可得
(21)
根據(jù)([6]p.239,Lemma 3.7)有
(22)
令H1=SuN-ΠNSuN,H2=KvN-ΠNKvN,由(21),(22)及引理4可得
C(‖H1‖+‖H2‖).
(23)
根據(jù)引理3有
CN-1‖uN‖.
(24)
同理可得
‖H2‖≤CN-1‖vN‖.
(25)
由(23)~(25)可得,當(dāng)N足夠大時(shí)
‖vN‖≤CN-1‖uN‖.
(26)
由(16),(22)和(25)可得
‖uN‖≤‖KvN‖+‖KvN-ΠNKvN‖≤C‖vN‖+CN-1‖vN‖≤C‖vN‖.
(27)
從(26),(27)可以看到,當(dāng)N充分大時(shí),uN=0,vN=0.從而存在唯一性得證.
證分別將(7)中的第1,第2式與(13),(14)相減得
v-vN-(u-uN)-(Su-ΠNSuN)=g-ΠNg,
(28)
(u-uN)-(Kv-ΠNKvN)=0.
(29)
Su-ΠNSuN=Su-ΠNSu+ΠNS(u-uN)=Su-ΠNSu+S(u-uN)-{S(u-uN)-ΠNS(u-uN)}=
(-g+v-u)-ΠN(-g+v-u)+S(u-uN)-{S(u-uN)-ΠNS(u-uN)}=
-g+ΠNg+(v-ΠNv)-(u-ΠNu)+Se-(Se-ΠNSe).
(30)
同理可得
(31)
將(30),(31)分別代入(28),(29)得
(32)
(33)
(34)
對(duì)(34)第2個(gè)等號(hào)右邊第3項(xiàng),運(yùn)用公式(19)可得
(35)
由(34),(35)及k(x,τ)的光滑性可知,
(36)
由(36),(22)和引理4知,
‖J3‖+‖J4‖).
(37)
由假設(shè)(A1)和(A2),(7)的解滿足M-條件.根據(jù)引理3,有
(38)
(39)
將(38),(39)代入(37)得到
(40)
由(33),(22)可得
(41)
另一方面,由(36)和引理4有
C(‖J1‖L∞‖+‖J2‖L∞+‖J3‖L∞‖+‖J4‖L∞‖).
(42)
再由引理3有
(43)
(44)
將(43),(44)代入(42)式并整理得
(45)
又由(33),(43)得
(46)
考察式(11),(12),采用譜Legendre-Galerkin方法求解算例.表1為譜Legendre-Galerkin方法近似解及近似導(dǎo)數(shù)解的L2和L∞誤差,分別與圖1(a)、(b)相對(duì)應(yīng).從誤差數(shù)據(jù)和圖中可以看出,該方法具有超幾何收斂性.
表1 譜Legendre-Galerkin方法的誤差
圖1 (a)譜Legendre-Galerkin法近似解的誤差 (b)譜Legendre-Galerkin法近似導(dǎo)數(shù)解的誤差 Fig.1 (a) The error of approximate solution for the spectral (b) The error of approximate derivative solution for the spectral Legendre-Galerkin method Legendre-Galerkin method
參考文獻(xiàn):
[1] BRUNNER H. A survey of recent advances in the numerical treatment of Volterra integral and integral-differential equations[J]. J Comput Appl Math, 1982,8(3):76-102.
[2] HESTHVEN J S, GOTTLIED S, GOTTLIEB D. Spectral methods for time-dependent problems[M].New York:Cambridge University Press, 2007.
[3] SHEN J, TANG T. Spectral and high-order methods with applications[M].Beijing:Science Press, 2006.
[4] SHEN J, TANG T, WANG L L. Spectral method:algorithms, analysis and applications[M].New York:Springer, 2011.
[5] GUO B Y. Spectral methods and their applications[M].Singapore:World Scientific, 1998.
[6] GUO B Y, WANG L L. Jacobi interpolation approximations and their applications to singular differential equations[J].Adv Comput Math, 2001,14(3):227-276.
[7] BERNARDI C, MADAY Y. Handbook of Numerical Analysis:Spectral methods[J].Techniques of Scientific Computing(part 2), 1997, 5:209-485.
[8] GOTTLIEB D, ORSZAG T A. Numerical analysis of spectral methods:theory and applications[M].Philadelphia:SIAM, 1997.
[9] KARNIADAKIS G E, SHERWIN S J. Spectral/hp element methods for CFD[M].New York:Oxford University Press, 1999.
[10] TREFETHEN L N. Spectral methods in matlab[M].Philadelphia:SIAM, 2000.
[11] TAO X, XIE Z Q, ZHOU X J. Spectral petrov-Galerkin methods for the second kind volterra integro-differential equations[J].Numer Math Theor Meth Appl, 2011,4(2):216-236.
[12] WEI Y X, CHEN Y P. Legendre spectral collocation methods for pantograph Volterra delay-integro-differential equations[J].J Sci Comput, 2012,53(3):672-688.
[13] ZHANG Z. Superconvergence of spectral collocation andp-version methods in one dimensional problems[J].Math Comput, 2005,74(252):1621-1636.
[14] TANG T, XU X, CHEN J. On spectral methods for Volterra type integral equations and the convergence analysis[J].J Comput Math, 2008,26(6):825-837.
[15] XIE Z Q, LI X J, TANG T. Convergence analysis of spectral Galerkin methods for Volterra type integral equations[J].J Sci Comput, 2012,53(2):414-434.
[16] BRUNNER H. Collocation methods for Volterra integral and related functional equations[M].Cambridge:Cambridge University Press, 2004.
[17] 陶 霞.求解Volterra積分微分方程的高階方法[D].長(zhǎng)沙:湖南師范大學(xué), 2012:33-36.