AXnL(,)(1,2,)Yn=.又对n2xX,{}Ax在Y中收敛.n求证AXL(,)Y,使得A强xxxxnn=+suppxx(n)n1xxn=1收敛到A,且AAlim.n1314p证明xX,x={})保证*a)收敛,kkk$q求证{}a.又若Ax=limAxnn,{}Ax在kn%pfxa:*),求证f作为kkY中收敛,{Ax}在Yn上的线性泛函,有中有界,即%1fa=(||).*qqkk=1supAxn<%(xX)pn1证x={)k},令%由共鸣定理2.3.15,M>0,fx,;=*()kkk=1s.t.AMn.(1.)于是nnfxnk,=*()kk=1+Ax=limAxlimAxMxAppnn%nn%fn()=L,(K,)且limn%fx,,=fx.由习题2.3.7,nLX,Y,()并且AAlimn.p+n%f().2.3.8设1<
(=argkk(f(n)p11联合qf=(.x,1+=,pqqf(nq(n)q1i一方面fx,=*((ekkkk=11nn2.3.9证x={)},qq1iik=**((eekk=(;%kkkkk==11令fx,;=*()kk另一方面,k=1n1-.-np()qpq1=fx,.=()fx,()nnfx()==f/0/(()q1pfnk0*k*kk=1121k=1211+11f()=L,,(k)且limnnnpqnq-.qq-.n%***((((kkkff/0/0=(fxk,,=fx.由习题2.3.7,nkkk===1111212+1且f().(f.%下面证明((={}.qk1718k$2.3.10用Gelfand引理证明共鸣定1设e={0,0,1,0,0,},则k理.(kk=f(e.)$px()=supAxAWe1k=pxMx()AxMx((kk=fff()eek=={}(k,()AMAW.且(f.又%2.3.11设XY,是Banach空间,nnAXL(,)Y是满射.求证如果在fxnk,**()ksup(k)ksup(kxkk==1111knknY中yy,则c>0与n0fnksup((%xxn0使Axnn=y,且1knxcy.nn证明由习题2.3.7,设NA()=={xX|Ax0,}考fflim(n%n%虑映射A:XNA()(f.Y,[x]XNA(),%f==((sup.%kf(k1Ax[]=Ax,xx[].证明A%1920
3单射、满射.再由yn0xxnn456733x,且Ax=Ax4567nnn3[][]定义yAxnn,Ax=Ax33Ax2,AxY1Y3则有xCy,CA=2.nn推出A有界.由Banach逆算设yyn&0,记11子定理,4567xA==yxA,,4567ynn001111ANL,(YX()A).不妨假45456767xxA=yAyAyynn00n0设y0=0,yn0,记31取xx004567,满足4567xAnn=y,3xx2,456700114567xA=yAy.xx34567,满足nnnnn33于是,取xxnn4567,使得xxnn002.45456767xx于是xx2,4567便有nn3333xxx+xCyy+CyxCy,其中nn00n00nn1CA=2.Cy+2.Cyn02122再想办法将y折合到y射.0n上去.条件A满射,yY,yy&,N,x3X,使得n00nN>0Ax33=yAx[]==Ax3y.A满射.1yyyyyyy2.0nn0002n1.4.175()3Ax[]=Ax33Ax2,Ax[]于是对nN>0,xn按YY上面取法,xC35.ynnA有界.nN,取xx34567,由Banach逆算子定理,0nn1满足xx32,4567则有ANL,(YX()A).nn1Ax3==Ax4567ynnn设yn0,记4567xAnn=y,11134567xA=yAy.xxAnn224567ynn=CynnnAx[]=00Ax=(x[x])注意到这个结论与要证的结果十分1xNA()[x]=.A单类似,其中A相当于C.2324
3下面要做的事就是将4567xnxx002,4567中的[],去掉,过河拆桥.3xx4567,满足nn取xx34567,使得33nnxx2.45456767xx于是nn00131xxA224567y33nnnxxCyyCA,2.=nn003yn0xxnn4567333x,且Ax=Ax4567xCynnn00333定义yAx,xxnnn0133则有xC3y,CA=2.yA==xA,xynnnn00设yyn&0,记3333xxx+xCyy+Cy11nn00n004567xA==yxA,,4567ynn00Cy+2.Cyn011145456767xnnx00=AyAyAyyn0再想办法将y折合到y0n上去.3yy&,N,取xx4567,满足n0000nN>02526xaTxxD()Tyyyyyyy12.0nn0002n3于是对nN>,x按(3)RT()在Y中闭的充分必要0n上面取法,xC35.y条件是a>0,nnnN,取xx3,dxNT(,())aT(xDT()).45670nn3其中dxC(,)表示x到X的子满足xx2,4567则有nn集C的距离.3Ax==Ax4567ynnn证明13xxAnn224567ynn=CyxNTn()xxn(1)2.3.12设XY,是Banach空间,TxxnTxn=00是闭线性算子,0.=TxxNT()即得DT()XRT,()Y,求证(1)NT()是X的闭线性子空NT()闭间;(2)()RT()是B空间,(2)NT()={0},RT()在Y中TDTRT:()()闭的充分必要条件是a>0,使单射、满射,由逆算子定理知2728
1TRL((T),.X)(>0,s.t.Ty1(y(yRT()).于(3)注意到XNT()是B空间.是xX,考虑T:XNT()Y.令yT=x,即有xT(x.(){[]()()}(8)DT=xXNT|,xDTRT()ynyxDn(T),TxTx[]=.s.t.yT=xy,nn显然NT()=[],()()由所给的不等式,RT=RT.如果T是闭算子,用(2)的结果,xx(TxTxxX即得结论.nmnm,s.t.xx.于是nxxT下面证明T是闭算子.就看nyT=xyRT().Txyn即证得RT()闭.2930RT()闭:RT()闭(T单射)(2):(>0,s.t.[xT]([x],即0[xDT](),(())dxNT,.(TxDT()4567x[xnyTx=[].2.3.13设axy(,)是Hilbert空间Tx4567yH上的一个共轭双线性形式,满足n(1)M>0,使得|(,)|axyMxy;9(2)>0,使得2T()|(,)|axyx.xDnn()Txx,xDT,yTx+求证:fH,!yH,f24567xxn0使得Tx=Tx4567yaxy(,)f=fx()xH,nn3132
而且yf连续依赖于f.,.xH3取xyy=,便有ff证明根据Lax-Milgram定理2.3.17,必存在唯一的有连续逆的(2)233330,=ayyyy(ffff)yyffyyff=连续线性算子AHL(),s.t.axy(,,)=(xAy).又根据Riesz22.3.14设;是R中边界光滑+表示定理,对fH,11的有界开区域,(:;R有界zH,使得fx()=(xz,,)对ff可测并满足此z,求解方程2f0,<((0fL(;).规定1Ay=zff1,y=Azffx()=(xzfxAy,faxy,f.auv(,,)=!;(<0,ptx()=limtxnn(3)令Xx={},00==txtlimpx().nfxfx()===()fx()px()n000000==11pxy(+=)lim(xynn+)nfx00()===fx0()0px()()0px0.xy+=+p(x)p(y)limnnlim.xX,是否有nn02.4.3令fxpx0()()?即=1R,1fxpx?Xxf=={}()(),xfx00()(0)000000=10显然是正确的.当<0,==px()()00px.12fx00()===fx(00)fx(0)||px(0)supxfnn,=supfx,nn==pxpx(0)()=0,求证:为xx=kkE,了存在fX,满足k=1定义fM,nfx(),1jj==cj,2,,n,必须且仅fx0()=kkC,k=1须对,,,K,有12nnn特别是fxC0(kk)=,并由充分性||jcMjjxj.假设,jj==11()56nfX,使得ifx0()=kkCMxf0MxEk=1"fi=1fi"fi=d再根据Hahn-Banach定理,"ifX,使得#fMi()ii=00fx()j=()ji&,""fxfxxE()=0(),"fxdi()ii=$#ffM=.$0fx()=1.ii2.4.7设xx,,,x是线性赋范12n2.4.8空间X中线性无关元,求证M是极大线性子空间的充分且ff,,,fX,使得12n必要条件是,M是线性真子空间,fx()==%ij,1,2,,.nijij并且11求证=(),但()&.xMX有0证X={xM|R.1}(0Mxdij==span{},,i()xiMi,思路:对照命题2.4.101jnji&证如果M是线性真子空间,并则di>0.且xM0X有1由推论2.4.7,对Mi,di>0,X={xM0|R.}(那么78
[]iixMX,xx[],fex()=Refex()1xM1,R,使得2.4.10xxx=+01证由Ascoli定理,存在实线性连[xx]=+=+,-.01x+,-.x0故续泛函gx()及0>0使得dimX(M)=1.gx()<<0g(x0)如果dimX(M)=1,那么supgxgx()<(0)xExMX,[xM]X,0(1)1R,[x]=+,-.x0令fx()=gxi()+gi(x),则+,-.xx0=[]xxMxx0{M2.4.9(1)1supfx()=0,0efxi()supRefex(i)!xExE则有||//910fx()supfx()<0,使得Bxd(1,,)C=+xy2(1.)C便与xC2的假设矛盾.12.4.13xC,3%>0,使得证0Bx(0,.%)C这样,yBx(0,,%)都有zx=+2(1.)yC根据定理2.4.15,存在"xx1=+20(1)xfX,R,1使得#zx=+()1y$2fy()fzyMzBxd()(,,()),zx10=(1)(yx)dx()=(xM,.)zx10=(11)yx<()%1314supfy()inffz()fx()=+
405表示点yMzBxd(),x=(45,)到通过原点的直线(即=inffxdy()yB(),1含有零点的超平面)=inf+,fxdfy()()0-.Hxf=={(45,|)f(x)=+=
4050}yB(),1的距离,即=fxdx()()supfy()0yB(),1fx()=+=
405(xH,,f)2fxdxf()().x=(45,R).f2取f1=f即为所求.2,4.14设M为R中的闭2注我们先说明这个定理的几何意凸集,x=(45,R),为确定起2见,不妨假定fx()>0.那么,实义:设XR,=fX2x=(45,R),数值fx()supfz()便是由zMfx()=+
4050点x到平行于直线H的f其中(
0,)是由f确定的rM的支撑直线Hf22实数对.如果f=+=
01,(r是某个实常数)的距离则根据平面解析几何知识,r(xH,f).由此可见,本题的结果1516
说明:当fX,f=1,xM时等式成立.使得点x在M的支撑直线xMX,1>0,Hr0上的投影y恰好是MzM,使得f010中,对点x的最佳逼近元时,这xzdx1<+()1.个距离"6sup#7fx()supfz()sup{fxfz()()1}(xH,r0)达到最大.ffXX$8zMf0ff==11证记dx()=infzMxz.sup{fxz(1)}fX如果xM,左边:f=1dx()=0.sup{fxz1}<+dx()1.右边:一方面对fX,fXf=1fx()supfz()0,蕴含右边0.zM根据1>0任意性,有另一方面,又存在fX,"6sup#7fx()supfz()dx().fM()=0,f=1,蕴含右边fX$8zMf=10,另一方面,根据习题2.4.13,必故右边也=0.因此当ffX,=1,使得001718dx()fx00()supfz(),zM所以dx()"6=sup#7fx()supfz().并且右fX$8zMf=1端上确界在f达到.019
xABx==2.6.1设X是B空间,求证:1Bx=x=B.即B单L(X)中的可逆(有有界逆)算子集射.是开的.BA=IyX,Bx=y1证明设A,AL(X,)有解x==ABxAy.即B满射.1考虑当>0充分小时,是否BAI==BBI111BB==BAA.有(AI+)L(X.)注意到2.6.2设A是闭线性算于,1AIA+=+(IA,)根据引理1n,,p(A)两两互异.又1设x是对应于的本征元2.2.6,当A1<时,ii11(i1=,2,,n)求证x,x,,x(IA+)L(X)故当12n1是线性无关的.11<1时,(IA+)A证明用反证法.令x为第一个mL(X,)而可由它的前面m1个向量线11()AI+=+(IAA11)性表出的向量,即m1L(X.)xx=(1)mkkk1=这里用到,A,BL(X,)且x,x,,x线性无关,对(1)112m1AB==BAIB=A.事实上,两边施以IA得到m12nnnAxxA1rAl==()==imA1.nm10=()mmkIAx=()mkIAx=0,k1=()m1IAx0A=x0x0===()xkmkk0,(IAx0)=k1=x,x,,x12m1线性无关kk1=0k(Z)kmk()==0k1,2,,m1()(1)mk==0k1,2,,m1.()=11==1,k10,2102,于是(1)x0=.与x为特征=1同理=nmmn0n,n0,向量矛盾.由此可见,如果=0,则022.6.3在双边l空间上,考察右推=0z(Z)x0=.n移算于A:如果0,(0x,=nn,,+1,10,,1,,,n1n,l+++2222<+++<+,n0nnyA==x(,,nn+1,,,10,1,,,,n1nnn===1n1即得Ax=xA=1,34
++nn+122+112n+22n<+000x()nn=(0,0,,0,,0,),Ax()=(0,0,,0,1,0,n1==n10矛盾.xA(n)n1+因此只能=00n()n=0z(Z)x0=.于是x=(0,0,,,1,0,)(nn)()(A?)=.(xAx,z0)=pzz0=(nZ)再证nn+1(A?)=.RIA()=2.(2)r即证RIA()={}.(2)与(1)完全类似,同理可得设zRIA,()则对z.=2再证c(A)={=1.}要证x,((IAx,)z0=RIA()+()z0=kk1kk=nRIA()=()n2x=(0,0,,0,1,0,),56先看=1.222取N,s.t.<,x,nnN1=+(IAxyyxx)=kkk=1(kZ).(jN)j特别对令y,=jk0=!0j()>+N1y=(,0,0,,0,1,0,,0,)2,但是y=(,0,0,,,,,,0,0)yN10,1,Nxx1=,012xxxx012===x00=为了证明yRIA,()即xx0=x201证x,2s.t.==xxx0=211(IA)xyyxxk==(Z)kkk1矛盾.由此可见,yRIA.()即有2j(jN)RIA().再证注意到y,=j2!0j()>+N1RIA()=.设()xxkk1k=(kN.)=,,nn+1,,,10,1,,,,n1nl78
xx0kk1=>(kN1xx+)k=N1+kN1.yI=(A)xkkk=1k1kk1yxx=yxx=kkk1kkk1令N1+当=1时,()kNj22x,k=jk1=+显然={k}x.0k()"N+1={}22y!k2xx={k},并满足重复上面证明即可.yxxk.=(Z)kkk12从而(IAxy,)=即2.6.4在l空间上,考察左推yRIA.()1Ac().移算子对于一般的=1,可以化归A:xx,(1,2,xn1,x,n)(x,,x2n1,x,n=1,情况.求证:(A)={C|<1;}p910cp(A)={C|=1;A}()=(A)cA.>=1A时,#(A.)证明(1)>1时,(2)(A)={C|<1.}p#(A.)记D={C|<1.}2xx=(1,x2,,x,n1xn,l)对于D,数列{n2}l0.y==Ax(x,,x,x,),2n1n即222A1(,,,)=(,,)=(1,,,,)(Ax)12==x,Ax23()x,,Ax()k=x,k+12(A,)而(1,,,)p便是相应的特征向量.2222Ax=xx=xA1.nnn2==n1反之,设Ax=x,x,又2xl2x=(0,1,0,,0,),0.则Ax=(1,0,,0,)0AxA1"0=,A1=当nnnx0xAxx,x,==(n1++n2)01112
1<1D.出x,即(IA.)(3)显然,非零分量个数有限的y在$c(A){C|=1.}RIA()中.事实上,设y的非C={C|=1.}零分量个数为K,取记K2x=()xx,,x1,2n1n,x,lxy1j=,j=1先看=1,yIAx=()yxx=kkk+1kyxx112=xxykk1+=1j()=1,2,,K2j1=xl.yxx=223!x0k=()>K.k+1yxx=334注意到非零分量个数有限的y在yxx=k1k1k22l中稠密,故有RIA()=l.yxx=kkk+12RIA()l.例如kk12yl={j},但是yRIA.()yxx=xxy=,j1=j1k1++k11jj1==j1k事实上,按xx,=1求得利用这个公式,我们可以从y求k+11jj=113142的xx,={}使得x.2.6.5在L0,()上,考察微分算kk1=k故xl.2子于是1A().cdx1对于适合=1的一般,可以A:xt()dt,DA()=H0,()化归=1情形.事实上,求证(1)(AR)={Ce<0};p(IAxy)=xxkk+1k=y(2)c(AR)={Ce=0};xxykk+1k==(k1,2,.)(3)(A?)=.kk1++k1rdefdefdbittxy(1)(ee)=,令kk,k()1,2,,dtkk=kk==+1=+aib,则有当Re=a0<,时,=(k1=,2,.)此即kk+1ktatibt2eeeL0=&(,,)化归=1情形.于是dtt2总结起来,我们有dt(ee)=L(0,)t1p(A)={C|<1,}eH0(,.)c(A)={C|=1,}p(A.)即得r(A.)=%p(AR)={Ce<0}.1516
dx?(2)dtbix=ydx(0()dtxy()""d=0()xDA()y("dbitbitdt(xe)=ye
x,yC(0,),0bitbitt"bixc=+eeyed(0("")((dxxy)("")d=00dtbitt"bixecyed=+((0("")),((ydx("")dx""yd0()=00dtn"当y(")span{"e}2d时,yL(")(0,),且当"((0dtyxd()(")"+0""xyd0()=dcn=(+1)!时,使得(0(dtyy(")+"""())x()d=0t2"dyy0()+"()=cydL0+(0("")(,)dtRe>0n""xDA,()因为{"e}构"yC()=eC=0.2()成L0,()的完全系,所以"y0=.Re0>n"2RA(IL0,)=2()span{"e}在L0,()中稠2(A)密.即RAbiL0,.()=()由此rRe0(A)推出
b,RbiA().即得rcRe>0(A))c(AR)={Ce=0}.r*Re0()A+(3)当Re=a0>时,r(A?)=.r1718计算细节:xxntibn(tib)Lxn()==((00teedttedtxn1+(tib)Lxtedn1+()=(0tn1xib+xn(tib)=xee++(n1te)(0dtn1xib+=xee++(n1Lx)()n(tib)Lxn!enn()=+Px()(tib)Lxn1()=++()!ePx()n1++n12Lxn!L0,n()()19
2.3.1设X是Banach空间,X0U(,1)XX0中的开单位球.是X的闭子空间,映射下面证明UB(
,1)=(,1.):/XXX0,定义为xB(,1)x<1[x]x<1:[xxxX],其中[]x表
xxU=[](),1示含x的商类.求证是开映
BU(,1)(,1)射.反之,证法1用开映射定理,只需证明满射.事实上,[xU](,1)[x]<1xx[],[x]XX,0任取xx[],使得则有xX,xx=[].xx<1,Bx(
1),[]=x.UB(
,1)(,1)证法2不用开映射定理.教材p94,定理2.3.8的证明中的2.3.2设XY,是Banach空间.(1)为了证T是开映射,必须且仅UXL(,)Y,设方程Ux=y对每须>0,s.t.一个yY有解xX,并且TB(,1)U(,).取=1.m>0,使得并设Uxmx,.xX1B(,1)X中的开单位球;求证:U有连续逆U,并且121?Um1/.xRA(()),证明由条件,U是满射,且是单20=|(Axx,)|mxx=,射.所以根据Banach定理,?1故有(RA())={}.UYL(,,X)yYy,1=,1所以RA()是稠的.设{y}设Uyx=,则n1==yUxmxmUy=1是RA()中的基本列,并设11Ax=y,UU=supy1/m.nny=1则由Axmx{x}是基n2.3.3设H是Hilbert空间,本列.AHL()并且m>0,使得2|(Axx,)|mx,xH.1xxHyAnn00=xAnxRA().求证AHL().证明由条件,xH,RA()是闭的.RARA()=()2=H即A是满射.mx|(Axx,)|xAxAxmx所以根据Banach定理,1AHL().所以A是单射.2.3.4设XY,是线性赋范空间,34
D是X的线性子空间,A闭算子ADY:是线性映射.求证:(2)如果A连续,又Y完(1)如果A连续,D是闭集,则备,那么根据定理2.3.12(B.L.T),A是闭算子;A能一地延拓到(2)如果A是连续且是闭算子,则D上成为连续线性算子Y完备蕴含D闭;AAA,|,=AA=.本题D(3)如果A是一一的闭算子,则A1也是闭算子;还有一个条件A是闭算子,(4)如果X完备,A是一一的闭下面证明D闭.设算子,RA()在Y中稠密,并且xDxxnn,.则有A1连续,那末RAY()=.Ax=AxAx,于是因为Ann(1)如果A连续且D是闭的,则A是闭算子,闭算子;所以xDAnn(),xxxDxxnn,设,AxnyDxD(A)AxAxnAy,=Ax56XA1xD,且Ax=Ax.11A(3)如果A是单射的闭算子,则A(2)1也是闭算子.RADA()=()设yRAnn(),,yyxDAnn()xx闭.xAyx=1yA=xyY().RAnnnnRA()因为A是闭算子,所以最后RA()()Y.==RA()xDA,yA=x2.3.5用等价范数定理证明1yRA(),xAy=.i.e.([0,1],C)不是Banach空间,其11A是闭算子.1中ff=!|()|tdtfC[0,1].0(4)如果X完备,A是单射的闭证明用反证法.假如算子,RA()在Y中稠密,1(C[0,1,])是B空间,并且A连续,那末11RA()=Y.ff=max(tf).!0(t)dtfmax(t01tt01(3)1A是单射的闭算子A也是闭算子.78
是比强的范数,用等pxx()1+212px()+px(),xx12X1价范数定理,与等价,;1(4)当xx时,px()()px.即M>0,s.t.nnfMf即求证:M>0,使得11px()MxxX.maxft()Mftdt!()001t证明令(fC[0,1.])令$xx=+supp()x,1x=110Mt(t1)x是X上的完备范数,然ft()=M1M1=112M0()Mt1后用等价范数定理.矛盾.所给的条件(4),有两处发挥作用.2.3.6Gelfand引理.设X是其一是证明p()=0:1Banach空间,pXR:满足&xp0,0()limp(11x)=limp((1)px()0xX;00nnnn%%(2)其二是证明(X,x)完备时,1px()""=px()">0,xX;从(3)910yex=i(y=1i(pex()=py()suppy()=suppx()xxnm100xxnm,(X,x)yx==11完备,xX,使得i(xxn.suppex()suppx();()xx==11"0,1,NN,i(yex="ii((i(xxnm1<2(nN>),px()=peex()(=pey)m%y=1ppp(xxnm)<""(xxn)lim(xxnm)i(xxnm22xxnxxnmsuppey()m%y=1(nN>)i(i(suppx()suppex().suppx((()=suppex()xx==11xx==11i(==((suppex()suppx()&xp0,0()limp(1x)xx==1100nn%((x=11==lim1px()0p()=0.x((R).n01n%注xX,x=1下面证明(X,x)完备.11112
xxnmn100xxm,(X,x)""完备,xX,s.t.xxxxnn+22=+(1)xxnxx.n3""(0,1,)NN,<22<"(nN>).xx<"(nN>),nm12根据等价范数定理,M>0,使得xMxpxMx().1p(xxnm)<"xx2nm注:存在某个线性空间上的强、弱m%ppxxnnlimxxm"()xnnxx()xm2两个范数,m%使弱范数完备而强范数不完备.见()nN>反例p36,12."xxAXnL(,)(1,2,)Yn=.又对n2xX,{}Ax在Y中收敛.n求证AXL(,)Y,使得A强xxxxnn=+suppxx(n)n1xxn=1收敛到A,且AAlim.n1314p证明xX,x={})保证*a)收敛,kkk$q求证{}a.又若Ax=limAxnn,{}Ax在kn%pfxa:*),求证f作为kkY中收敛,{Ax}在Yn上的线性泛函,有中有界,即%1fa=(||).*qqkk=1supAxn<%(xX)pn1证x={)k},令%由共鸣定理2.3.15,M>0,fx,;=*()kkk=1s.t.AMn.(1.)于是nnfxnk,=*()kk=1+Ax=limAxlimAxMxAppnn%nn%fn()=L,(K,)且limn%fx,,=fx.由习题2.3.7,nLX,Y,()并且AAlimn.p+n%f().2.3.8设1<
(=argkk(f(n)p11联合qf=(.x,1+=,pqqf(nq(n)q1i一方面fx,=*((ekkkk=11nn2.3.9证x={)},qq1iik=**((eekk=(;%kkkkk==11令fx,;=*()kk另一方面,k=1n1-.-np()qpq1=fx,.=()fx,()nnfx()==f/0/(()q1pfnk0*k*kk=1121k=1211+11f()=L,,(k)且limnnnpqnq-.qq-.n%***((((kkkff/0/0=(fxk,,=fx.由习题2.3.7,nkkk===1111212+1且f().(f.%下面证明((={}.qk1718k$2.3.10用Gelfand引理证明共鸣定1设e={0,0,1,0,0,},则k理.(kk=f(e.)$px()=supAxAWe1k=pxMx()AxMx((kk=fff()eek=={}(k,()AMAW.且(f.又%2.3.11设XY,是Banach空间,nnAXL(,)Y是满射.求证如果在fxnk,**()ksup(k)ksup(kxkk==1111knknY中yy,则c>0与n0fnksup((%xxn0使Axnn=y,且1knxcy.nn证明由习题2.3.7,设NA()=={xX|Ax0,}考fflim(n%n%虑映射A:XNA()(f.Y,[x]XNA(),%f==((sup.%kf(k1Ax[]=Ax,xx[].证明A%1920
3单射、满射.再由yn0xxnn456733x,且Ax=Ax4567nnn3[][]定义yAxnn,Ax=Ax33Ax2,AxY1Y3则有xCy,CA=2.nn推出A有界.由Banach逆算设yyn&0,记11子定理,4567xA==yxA,,4567ynn001111ANL,(YX()A).不妨假45456767xxA=yAyAyynn00n0设y0=0,yn0,记31取xx004567,满足4567xAnn=y,3xx2,456700114567xA=yAy.xx34567,满足nnnnn33于是,取xxnn4567,使得xxnn002.45456767xx于是xx2,4567便有nn3333xxx+xCyy+CyxCy,其中nn00n00nn1CA=2.Cy+2.Cyn02122再想办法将y折合到y射.0n上去.条件A满射,yY,yy&,N,x3X,使得n00nN>0Ax33=yAx[]==Ax3y.A满射.1yyyyyyy2.0nn0002n1.4.175()3Ax[]=Ax33Ax2,Ax[]于是对nN>0,xn按YY上面取法,xC35.ynnA有界.nN,取xx34567,由Banach逆算子定理,0nn1满足xx32,4567则有ANL,(YX()A).nn1Ax3==Ax4567ynnn设yn0,记4567xAnn=y,11134567xA=yAy.xxAnn224567ynn=CynnnAx[]=00Ax=(x[x])注意到这个结论与要证的结果十分1xNA()[x]=.A单类似,其中A相当于C.2324
3下面要做的事就是将4567xnxx002,4567中的[],去掉,过河拆桥.3xx4567,满足nn取xx34567,使得33nnxx2.45456767xx于是nn00131xxA224567y33nnnxxCyyCA,2.=nn003yn0xxnn4567333x,且Ax=Ax4567xCynnn00333定义yAx,xxnnn0133则有xC3y,CA=2.yA==xA,xynnnn00设yyn&0,记3333xxx+xCyy+Cy11nn00n004567xA==yxA,,4567ynn00Cy+2.Cyn011145456767xnnx00=AyAyAyyn0再想办法将y折合到y0n上去.3yy&,N,取xx4567,满足n0000nN>02526xaTxxD()Tyyyyyyy12.0nn0002n3于是对nN>,x按(3)RT()在Y中闭的充分必要0n上面取法,xC35.y条件是a>0,nnnN,取xx3,dxNT(,())aT(xDT()).45670nn3其中dxC(,)表示x到X的子满足xx2,4567则有nn集C的距离.3Ax==Ax4567ynnn证明13xxAnn224567ynn=CyxNTn()xxn(1)2.3.12设XY,是Banach空间,TxxnTxn=00是闭线性算子,0.=TxxNT()即得DT()XRT,()Y,求证(1)NT()是X的闭线性子空NT()闭间;(2)()RT()是B空间,(2)NT()={0},RT()在Y中TDTRT:()()闭的充分必要条件是a>0,使单射、满射,由逆算子定理知2728
1TRL((T),.X)(>0,s.t.Ty1(y(yRT()).于(3)注意到XNT()是B空间.是xX,考虑T:XNT()Y.令yT=x,即有xT(x.(){[]()()}(8)DT=xXNT|,xDTRT()ynyxDn(T),TxTx[]=.s.t.yT=xy,nn显然NT()=[],()()由所给的不等式,RT=RT.如果T是闭算子,用(2)的结果,xx(TxTxxX即得结论.nmnm,s.t.xx.于是nxxT下面证明T是闭算子.就看nyT=xyRT().Txyn即证得RT()闭.2930RT()闭:RT()闭(T单射)(2):(>0,s.t.[xT]([x],即0[xDT](),(())dxNT,.(TxDT()4567x[xnyTx=[].2.3.13设axy(,)是Hilbert空间Tx4567yH上的一个共轭双线性形式,满足n(1)M>0,使得|(,)|axyMxy;9(2)>0,使得2T()|(,)|axyx.xDnn()Txx,xDT,yTx+求证:fH,!yH,f24567xxn0使得Tx=Tx4567yaxy(,)f=fx()xH,nn3132
而且yf连续依赖于f.,.xH3取xyy=,便有ff证明根据Lax-Milgram定理2.3.17,必存在唯一的有连续逆的(2)233330,=ayyyy(ffff)yyffyyff=连续线性算子AHL(),s.t.axy(,,)=(xAy).又根据Riesz22.3.14设;是R中边界光滑+表示定理,对fH,11的有界开区域,(:;R有界zH,使得fx()=(xz,,)对ff可测并满足此z,求解方程2f0,<((0fL(;).规定1Ay=zff1,y=Azffx()=(xzfxAy,faxy,f.auv(,,)=!;(<