S'fff.DMS al OPI'B..t.TIOBL J<,WJ.TIQJS 'bT Gl.LBDI .Glf.HWl .liAOlW.DB A. msxa 1n partial tulf1llaent of tl\e requlnaent• tor tb.e ctecr• ot JQSlliR or SCBJIOI lu• 1955 TPI&OID! Redacted for Privacy rT}f*rx# sf httmtl*r I3 0bntlt of lb,Jor Redacted for Privacy ErrO u( ernou-&mt oil lhtbmttcr Redacted for Privacy Srfmt o# $c&ool 0m&E*tr CfiElttr Redacted for Privacy !ru tf Smihrtl Schpal htr tlmlr tr trd tr th*,Cn*rt Sahillf,t "as·• lll!RODUO TIOB • , • • • • • • • • • • • • • • • • 1 LDJIAR SYstEMS. Y!L"!OllS. AID M.U'iUCIS ) DS«!R!!f SPAOI S 0~ SY8f'JRS cr~· ••••• • • • • • • • • • • • • • • • • ••••••••••••••••••• OPDAS!IONAL !!~UA.fiOJrS IXAKPlil8 • • • • • • • • • •• • • • 9 u • • • • • • • • 14 •• ~ • 36 • • • • • • • • • • • • • • •· • .• • • Moat proil•• equa~lon ui~ ln applied aa.thematlca lnYolva tol:vtag a.u. Ol" WTitea ot eqQAti·oDa. lqu.a.tloaa _,. be o~ J1J1!LDV ttpaa. bat the tba.Pl•at u4 •oat bQ.>Onant la the linear equation. If a. aacl b pP1teaeat lalom qaatit1•• aDd x S.e the u.ab.own, thell u equ­ t1on vh1oh l'educae to the fo'ft u • b 11 called a b1 one UDlmon. A'JIT equat)1on in oue to th1a foa ia .aid to b• thaA llae&l' tnea. Some UJllmo1m whiCh 1.• not l'lduct.lla ~n-U.aeu. great D1"1at7 of fol"'la, .a.n.d ~ 11~ ~tlon lon-lhea~ equattona have & genel"all.T lii10h ao" cUtf1GUlt to aol•• «JAJ~Plee of non-U.neu- f1Cl\1&t1ona ~c (l) ~ - ):r • 0 • (2) (3) ain c. + • -4a • 0 • .Ji + 5 • 0 • lqu.tloa (1) baa thl'.. aoluUona. (2) baa t.nfb.dteq JllllV' aolu­ tlona, aDd (3) • • no Jel.u.t1ona, 1t ve real uuai•~•~ ar. aeaklng a l"oot 41110IIg the The existence of a aolmtioa for a given eqwation 4ePtDda to a large extent on the razage ot 'faluta which we all.ov the UllknoVD to .baTe. l'o!! uuapl$, U' x io tion }x • 5 at no aolut.iQJl• tor z, then lf a ;t.;equl.~ted lt, howne, we penait fltaotloaal T&laaa th1t tqU&tion haa the aDd b are ~tlob.al t .o be an lntegd-1 theh the eqUA­ ~ot x • J• uaaba?a• then ax • ' Et l8 w14ot that &awav• baa a lolution 2 which is a rational. number. provided that =0, no solution uiets unless b f1es the equation. a is not zero. and then every value o:f x In thia ease, we 8aJ that x eat1e­ is a.:rbiira:a. is not the aa.me si tua.tion a.s in e.mmple (2) above. = 0, If a Thia Equation (2) has 1nf'1niteq ma.ny eolutions 1 but these are a aeries of epeeif1c mlllbel!'e: (2) 1a .at.1 ea.tisfied by all values of s • The baste number eystem which will be employed throaghout thie paper is the complex number field. the tt.Pe a + bl, where a &nd. A Untar equation ax = b, baa the unique solution x * b a• b A CoJIIPlex llUlllber is a.rrt ntlllbel" of al!'e r-eal numbers axd where a and b unleaa a • o. x o ~ is arb1traJ7. As before.. if a • o, If this CoDd1• Note that the solution ts generally a. complex UllDiber. The cpmplg coGwm:to of a+ bi .J=i . a.re complex llWilbere, then b mat be zero in ol"der that a solution exist. t1on is aat1ef1ed, then x 1 • te the number a - bi • 3 Only the simplest mathematical problems lead to a single e~tion in one unknown. More complex problema Will require the colution of a If all the equat ions a.re linear set of equations in sever-al unknowns. in ea.eh of the unknowns, this set is called a linEar system of equa­ tions. Sa.ch a qste.m mtq' be wr1 tten in the following fol't!l: (2.1) • • • TJ'I..ie 1e a S7atem of m • • • e~tions • • • • • • • n unknowns. in ;,• ••• • ~, and the known quanti ti ea are the a 1j ~e unknowne and the are 71 • !he systeJil (2.1) can be expreaud in a mu.oh shorter f'ol"'m b1' using the ~tion notation: n t .1-1 a 1SXJ • 71 , (1 = 1, ••• , m) • This e;zpression mea.ne precisely the same thing as (2.1). ao~utioA m of (2.2) is 8Zf3' set of values tor ;,, ... • eqUation• are satisfied simu.ltaneouall'. are complex 1 complex llWilbers. and y- n"Wnbef's~ *b. A Btlch that all Ve shall su.ppose that a 1J and that the eolut1on must be a set of Another Ya7 of dealing with linear qatems is to use the concepts of vectors and matrices. We shall ti'eat these topics by following definitions . Defip.itio:q. 2r:l (2. the pp .l~) An ;n=dimep.aiop} complg yeq3;.ot is ot n complex numbers, maJ.d.Jlg w~itten (~ •••• M ordered aet • ~) ~ x • 'fu& totality- of such Yectora tor- a given n. is called an n-dimeneiona.l eomple.x· vector apace. The nwnber z is called the 1-th component of vectol" x • 1 The zero vector 0 21\1 (O, ••• , 0) is the vector all ot whose com­ ponentt are eeroe. EerG Note that the symbol 0 is used to represent the vector as well as the complex l'l1llllbe1' eero. ~f!:g,&Uon 2.2 The .mam of tWfi vectors x =("1,,. ... • ~) and is the veeto:t x + 1 • (;_ + r 1 • ••• • 2n + Tn) • ot a. vector x 07 a eompl.e x nwilbel' a is the vector 7 w (71 , ••• , 711 ) The a:Jt' • vqjy,s» xa • (~ • •• • • a.xn). Two vectors x and y are tml!i if and on.l¥ if x ... 7 = o. A. vector apace with the above p~perties ts called a },(inee;r veetora x and y IPM,•• is the complex nuiber - where 7s, 1e the complex conJugate ot 11 • The equations (2.2) are ea14 to define a lineM> homogeneous tl'a.nJJ.. formation of the n-dimensional vector x into the m-dimensional veeter 1' , . Defi~tj.on 2,J; is the :rectaJ.~gU.~ !fhe W;i& of the transfol"ma.tion defined by (2,2) .e:Jr.ta'3' • If m • n, is called a sgwe .A etrix of order n • We can now Write the aystem (2,2) 1n the foi"m ~s qmbolte notation meatus that which t.ranef'orms x equation ax • b l!tttn!ttga into '3'• A is thought of as something !he sim1lar1t;r betWe$Il {2 •.3) and the is evident. ~·5 !fhe Jfill of two matrices A ie the matr1% A+ B = (&1.1 + b13 ) = (a.1j) and J3 = (blJ) which t~anaforms tne't'f x into .Ax+Bx. The m:®U$ of a mtrb:t by a complex nwnber a. is the matrix aA = (cr.a1J) which transforms eve1'7 x into In the rest of this chapter ve shall Go( Ax) • as~e that m =n • In the linear ay.atelll (2.2) • we can al.WaJ'I su.ppoae m • n by d.ef'iniag a1.1 -= 0 and r1 • 0 tor 1 n~tmensional >m• Tim• we &ball be concerned only with •ectors and n-th order square matrices. 6 for i.J =l, ••• , n • The te~o mat~1x transforms every vector x into the tero vector. The 3mU maj;ri% is for i r/: J • fhe I = (& 1 J)' vhere unit matrix carries x &11 • l, and into 1tseli': 8ij Ix =0 =x for every x • D,!fip.&.tion 2.] a vector. fhb proflu.et o·f a matrix A and a vector x 1a In one case, we have (i • l, ••• , n) • • where the quantity in parentheses is the 1-th component of the vector Ax • This defines pt"eciselr what is meant bf the tranoforme.tion "We may also have a product with the vector on the lett: xA • Genel!&lq, xA .Detf.n1t1Qn 2~§ I ~.~ \t=l 1 1J'\J/ , x a (J • l, ••• , n) • Ax • The DtQd.ugt of matrix A = (a1 j) by matrix B • (b 13 ) is the matrix u ~~1 •n~J which transforms every x Dt(lA1~1oQ 2e9 into A(k) • If a unique matrix B exists such that 1 All • :BA. • I • is called the invtt'l of A and is written A-l • .B Detini~iQP, ~.lQ The ma.tri:.r: of A. (a1Jh A. (aij) is the A' 0 (aJ 1 ) is called the tran&Pi'! of A; COJl,.1ggatt and r =i' ie the Hermitiy. gopJyate of A • We are now prepared to diseu.es in some detail the solution of the equation A:% • 7• dependi~ Here there are three eases, on the character of matrix A • First , 1! A has an. inverse A-l• then we have • or z • A-ly • This gives the uniqu.e solution x which Mti$fiea the equation. Second. if A • vector o, y = o, there. is no tolution unless x is a solution. ~hese cases are like those for the simple Bu.t. tor Ax • 7 • there is a third case, since a equation ax • b • matrix 1JJB.'Y not be zero and yet haTe no inverse. to be s1Dgal.ar. onlY if y If A is a. singular matrix, aatisfies certain restrictions. some of the x1 are a.rbitra.ry. completely. (2, pp .6-7) . mopiM (z, y) 2,~ = 0 tor and. then evef7 fying the equation zA =o s a. solution will exist If these ~e satisfied, The tollorillg theorem 82;plains thi il The syetem Ax • y all vectors atoh a matrix is said has a solution if and only it ~ttch that sA ;:J 0 • A.;ey z satis­ can be witten as a linear combination ot 8 z' a; tor emmple, certain linearly independent d ~ n • l B ' ••• • I • Iiov 7 m.st sat1ef'7 the d linear homogeneous equations (zk, 7) • o, k = l, ••• , d • E:xaotl.y d of the components ot x can be chosen arb1tra.ril)-, and the remaining n - d detel'mined. and d n .. d fh.e integer d 1 s the ~ will then be 1a oaJ.led the yfeet of system Ax • 1' , • fhe proof o-f this theorea wlll not be g1ven. since it follows veey eloeeq the proof of theorea involving inverses. 5.1, without the restrictio.n 9 In definition 2.1. the idea of an wa.a introduced. ~imensional vector space Now we shall extend t he concept of •apace". :First, we may consider veetQra with a denumerable infinity of components; the totality of such vectors is an 1nt1n1te-dimensional vector space. A further extension might be to vectors w1 th a non-denumer­ able 1Dfin1ty of components. of a cont~ous variable Ve can treat such veetors as tnnct1ona a. !he totality of c~rtain interval a ~ a ~ b ~otions 1• called a gunction In general, an abstract linear space defined on a snace• S consists of a set of elements which have t he following properties: (l) It x and y aze elaenta of s, the defined and is in and comma.tative. x +0 that (2) If x a x S • ih1a operation is associative A zero element 0 e:xists in S 8\lch tor all x in S • ~ is defined and u operation has the property that a. • 0 or x • 0 • or both • If x and y if x + 1 ia is in S and a. is 8.1'1¥ complex IIWnbe1t, the product ca • (3) SWI'l belong to is in S • 1'hia <a= 0 S , then x = y if alld onl.7 i:t if aJJd onlf X - "¥ a 0 • Another operat ion is often def ined tor abstract spaces. operation is called the inner product !hie (x, y), which is a complex 10 ~he l'lWilbel!' aasoeia.ted with x and 7 • inner product must oa.tis:t)' the relation (x, (3.~) where (y, x) y) ~ (y, x) , denotes the comple'x conjugate of (y. z) • !he ieer product for vecto:r epa.ees was given in definition 2.3 • For function epaCeJs, the illl'ler product is =Jr b (3.2) (x, y) x(a) y(s) ds • a Here it iB assumed tba.t a aiagl.e :real variable to x(s) o , and y(a) are complex tu.nctions ot and. all elements x and y belonging are integrable on the interTal. a ~ e ~ b • S tn abstract spaces, one utu.alq d.oee not define a product in the o'i'dinaq aen&e. which a product Tbat is. we do not consider -q is a.a element of the spe.C)ft. nm.l.~ipliea.tion In the next aec­ tion, we ahe.ll, howev-er. consider transformations in abstract spaces. in. u A transformation in an abstract spaoe element x 1n y S a.nothel' element in ~elates S s. to each Su.oh a transformation may be written tn the following notation: (4.1) (x. y in S) • !rhe a,ymbol tra.nsfol"mation. At called an We shall su.ppoee that 1a, tor ea.eh x * 'Phe in S 1a used to rep.tteeen.t the that A is a single-valued operatott; ta a unique element of S • h Jltf1N.t1on l}.,J. fhe eveey- x o-!erato~, .am QJ.letttar: O 1 s the operator which catrles into the zero element of s. 0% =0 for every x • iiMtiy QRmtgr I caniea eveey x into i tselt. Ix • z tor all x in S • PefiJ1U1oa ~,2 .An operator L is said to ·ne a4d1t1t.t 1f L(x + y) ~ Lx + ~ tor i l l :r. and 1 which is continu.otu is a A!Detu: 1n Pl?ttato;:. S • An addi t ive operator The notio.n ot continuity involves topological eoneepte which do not eoneern us here. lf·e shal l hereafter use the telfm linea.:t* operator, although we ba.ve not ade­ qu.a.tel.7 defined it, Dtfiaition 4,1 ~he euro of two ope~tore operator A + B WhiCh transforms :z: into of two operators A and B A and B ia the Ax + llx • !b;e proQllQt ia the operator AB whieh can1ee x 12 into A(lb:) • It can readily be shown that addition of oper-ators is aasooiat1ve and commt1tative~ plication is that mu.ltipl1catlon is associative; and tba.t mu.lti• distributi~e over addition. Def1n,jlon 4.4 If a unique operator B exists ~~ that AB • BA. • 1 • it is ea.lled the invel"ie of A a.nd is written B = A•l. J'or a linear opera.tor L , the invnse, lt it exists, will be lt is asSWDed that the 1nTersee of all U.n-.r opera­ d.enoted by M • tors considered here ue s1Dgle • valued and linear. De1'1nit,12A is denoted by 4.5 The Rumttiap. gon.1m;ate of a linear operator L V and is defined by the l'ela.tion (L'#x, 1') • (x, Iq) • which mat be eati sfied for all x and 1 in S , In a Tector apace, linear Qperatore are matrices. For function epac•a, operators are represented by integrals: f .b (4.2) I.(a, t) x(t) dt ";r(a), (a~ 1 ~b) a !he Hermitian conjUgate of an integral operator ia J b 11 J. x" I. (t, a) x(t) dt , a The operation xL may be defined (& ~ • ~ I>) • , 13 :iL• f b x(t) L(t, e) dt, In abstract epac:e., we mq consider (o < c b) a • to be a linear Lx • '3' equation in Vh.1ch L and .,- are known and x L (a If is to be found. bas an inverse M , the equation ha.a the uniqu..e solution x • M7 • li' · L • 0 • the equation has no· solution unless eveey x in S ts a. solution. stUl he.Te DO inverse. y =0 ., and then However • .1. might not be seo and It seems ltkeq that !heol:!em 2.1 would here as 1t does for vector epa.c.ea. a~ Aa atated for abstract apace••· it is written in the followtng form: '£!l!iiQlj.lf 2.1• The equation Lx • 7 (s. 7) • 0 tor all s su.oh the.t 1L baa a solution 1f and oal7 it =o • A theor81l' aimila.r to this baa been proved for i:ltegral. equations of Fredholm t;ype ami secolld kind (1. pp.l01-1oq). onl¥ those llne&l' opemto~a We shall conaicler vh1oh aa.U&f7 this theol!'em •. SYSm.tS OF OPERAT IOIAL E(tT1M'IONS Now we shall be concerned with problems which involve more than one apace. equation s1 11'1rst, su.ppose we haTe two spaces L;, =~ (;_ in ~ • 22 s2 ) 1n mation which carries each element of s1 sa.y ~ ~ that operator L $pace s1 !he detiniti.ona of' section 4 appq s2 and ~ Then the indicates a transfor­ s2 into a.n el&ment of apace s2 W• • • in an obViO\ls way to the present situation, and will not be restated- Note that the identit,­ operatoll' X elw~e maps a epace into itself. 1t is an operator which mapa spa.()e M , L1]_ the eolut1on of =~ s2 It L bas an inverse Let LJi be a If lin~ • In this case, ie ;, • M ~ • Su.ppoee we have two aets of linear space&: t 1 • • • • , I• • . m ~ n • s1 into Oona1der elements x1 ~-• ... , in Xs, lb and. and YJ 1n rJ • operator which maps apace X1 into space TJ • tJi ha.D an inverse._ t .t is written MJl and is a. ltnea:t opettator which maps YJ into x1 • ll:e contider the linear 81"Btem. of opera­ tional equations ~hi• qstem is aa.1d to have a solution for given Ljl and 7 J it set of elements li.J.• ••. , ~ are satiatie!ld s1mltaneousq. & exists w.ch that all m equation• By defining 31 = 0 and 7J • 0 tor· L J > • , ve TlliJ:¥ su.ppose m • n • It 1s asawned that th1s has been done. J'or convenience, we sha.ll d.e t1ne vectors and matrices fo"tJ 15 operational e~tiona. = The vectgr x • {x1 } {:1)_ 9 . . . • ~J is an n-dbaenaional vector whose 1-th component is an element of apace Definition 5.1 Det1n1.t iqn 5.2 whose !the ..uzg, ttctQ£ ie 0 = {0} ::: {o, ... , 0} • 1-th component is the sel'o element of the The inner product of the tTPe x17 1 Dtf1nit1on (x, 7) 1-th space. will not be defined because product 1 are undefined. 513 The matriJ L e (Lji) nents are linear operators. is a matrix whose compo­ Hereafter, the ~bol L without anb­ aor1pte will be used to denote this matrix. Def1n1 t ion 5 .)t. rhe ptoduct (Jf matrix L a.nd veeto:tt x is the :following vector: Th1e is a veeto~ whose j•th component 1a an element of space YJ • lilv1dentl3', qstem (5,1) 1s equ1nlent to the equation Lz • 1 • Dtf1ait1on 5.~ The· yector OJorator zP = {zP} 1 • {zP 1 , , •, • z:} , where zP 1a ~ 1s an operator which m&ps 16 D!fa,p,itlcm. 5•6 The prgduct of vector operator zP o.n.d matrix L 1a the vector Zp L • {~" zP j J-1 x, · } ji whose 1-th component is an operator which maps sp.aee x1 Dttinitipn 5,7 !he inner Rrodu.ct of vector operator v ector y 7j zP with is theelement of apace YP defined by ( zP • where into YP • zP T> = is the oomp1ez eon~ate of It seema as if Theorem 2.1 t ~ yJ • lhould be valid for w,retem (5.1) • Unfortunate]$, the attempt to prove this has not been su.cceasft11. A more specialised theorem will be proved in this section, vh:lle sec­ tiou 6 is deToted to a. diacaaaion of the general ease. !he statement of the special theorem is rather involved. this reason, we shall prove it before stating it. determine aome oondit1ons on y o• then obviously there is no solution unleae y equations 7.1 • o, y fhe pro'blea 1s to for the ex1 etence of a solution to In the firet place, if all the L • 31 saying this is that J'or mst eatisfy the (j aa 1. ••• , n). n (1, .1 • 1, •••• n) , =0 • Another way ot linear homogene'Ous If this condition is tulfillecl, 17 Xi , .... , then ~ has detect d • n are arbitra17. We that the system Lx • 7 in this case. It my happen that non$ of tbe are not all cero. proved. Btq LJi bae an inverse, and yet they This is the case for which the theorem baa not been In the proof Which followa, this case will be ign.oJ"ed. We shall assume at eaoh etep that either all the operators under con­ a1del'&t1on a~e sero. or that at 1ea.et one ol them bns e.n illTerae. If one of the Ljl The tlrat ~tion has an invuaa. renu.mbet' so that 1 t is ~l in e,ystem (5.1) ta From this we get n where Mu 2). • Mu T1 is the inverse ot L.u • - i~ Mu_ If this is 8'1.lbat1tut&d into the .other n-1 t.,_i :;_ ' equations, the re8\llt is !hia reduce& to S.t ,.). • .....51 L 31 .. L K. L.. .fl ~ J.l ~1 (1. J = 2, ••• • n) , • (J = 2, •••, .n) • Now (5.3) becomes (5.4) (J ~ 2. ••• , n) • sratem (5.4) 1& very aimila.t" to the original qstem (5.1) • 1 tji = 0 , (1, J • 2, ••• , n), the n-1 linear homo• If all geneous equations ma.st be aatiet1ed in order that qatem (5.1) bs.ve a solution. Ret'• ea.n be chosen a.rbi trari]¥ and the defect c1 • n - 1 • 1 1 If one of t he L Ji , SAY' L22 , has an inverse, then x2 , • • • , xn (J =J, ••• , n) • Let Y~ ~ ~~ "t~ "• S,ttem (5.7) become• • T!2. ~ ~ 112 J,4fl' ( tJ~ a 3. . . . . . . .a) . . 19 (J =3. ••• • n) • Not-e that ~ is a. 11llear homogene.ua function of y~ ud 1 y2 • which in tm'n are lineal' homogeneou.e functions of the original llow it all L~1 • 0 , the n-2 e&tisfied in o?d.er that ue arbi tre.r,- and d n have a solution. y~ c o must be, Rere '=; , ••• , x11 = .n-a • 2 I£ one of the (5.1) equations Ilji bas @on inverse , we contime this process for stepa or until a. ayatem is obtained where eveq operatoi' is sero. Iiecall tbat we are asswninc at each step that all operators &l"e eero or that one of them has an inverse. After k steps. we have (5.9) and (5.10) (J =k + l, ••• , n) , where (!).U) Bote that k 7j k-1 1s a llJ1e&l' homogeneous function of the 7 J • k:l-2 which in turn are linear homogeneous :t\mctions c£ the y J , etc., ao that ~ te au.ch a. :t"uncUon Ae befo:r., if all ~ . Y'J • of the original k LJi • 0 , we have the n-k equations • 0 (J • lt + 1, ••• , n) whose aat1efa.ot1on is neeessar,r and. sufficient 1n ord.e~ that system (5.1) ha.ve a solution. ~ , 1 ••• , xJl are arb1tl"&r7 and the defect It we an able te continue fo'r' n•l d. =n - In this case k • ateps we have • n-l If Lnn = o, y mu.st sat1sfy the one e<:{l.lation linear homogeneous equs.tio:n in the y j • at'"b1trar1ly and we have 4 It '=D-1 has =n - .Also, n-1 '~n llh nuv -= O, a be chosen l • a.n inverse, then • (5.13) • • • • • • • • • • • • Jl ~ • ~l 71 .. 1~ Mu. L:!.t ;, • 21 ~s case 1s equivalent to d. = 0 since ther-e are no equations which y must satisfy and none of the : 1 a.:te a.rbitraq. We have shown that y mat satisfy a certain system of d. linear homogeneous. eq'UI'l\.tiona in order that Lx =y have a solution. Now we want to prove that these equations have the :form (p • n - 4. + 1. ••• .. n) 1 whe~e the geneous system zPL =0 lfeUnit.ign 5.8 11; • Ip w where If J 1' zP (zP• f) are solutions of the homo­ • fhe g§ltf!: QPttator is defined aa follows: is the identity operator of space IP YP • p, '11e the so~o ope a.tor whteh maps space YJ ser>o element of o. • 1nto the rp • Consider the . system zPL • 0 • !his may be Written as n ~ (5.14) J*l z~ LJl =o (1 • 1.. • • • • n ) • a ayatem of equations cons1at1D& eolel.)r of operators, (5.14) 1s a aet of op~tora ~~ the n left members If eve17 :LJi e~ =0 • (J • 1••• • • n) A solution ot Which make each of to the sero operator. zP ia arb1traey, Choose zP • {A~} • ( p • 1. ... • • n) • Nov n t J-1 z~ 7 J • !his is the case where d • n n t J-1 A~ 7 J and the n # yP (p = 1, • • • , a) • equations which 7 mnat sstiaf,1 are (5.16) (zP, 7> • T.p o (p ~ 1, ••• , • n) , a.s vas ahown bef'ore. Lu_ If has e.n inverse Mu , we .ha.ve (5.18) System (5.18) can be writt·e a (5.19) vlth (1 • 2, ••• , n) , 1 ~Jt aa before. If all l Ljt• 0 , ;p z2 , ••• , ...J) ~ a:re arbl tra:ry • Choose zP r; tzi, ~. ~ , ... , A!} (p r.: 2, . . . .. n) • Then !his is ~he caae a. = n • l , and we tind that the n...l equations 23 (zP, 'i) • 7P ... Lpl ' \1 r 1 • 0 &8 (p • 2, •• • • n) a.re the same (5.5). .# If one o"' the ,.l "'jt , aq L122 , baa an inverse, then • •• • n) • _~f eTeq Lit =- 0 • then Z~ , • • • , Z: a1"& arbi tra.ey. Cho~•• 80 tba.t and llow Jtter k steps , we have ! oP -- ~ ~ ., j=t*l zP4 (-yr-Jrl -...r-nl) -~ ~ ~ (r = 1• ••• • k ) • 2lf. and n .t J-r'+l (5.26) zPj ., 1•1 __!...1 Let Hi • ... Lji Mit Yi into YJ • • * o (1 =r + l, •• • • n) • fh1e 1s an operator whieh mapa space With this notation, we can write zP r (5.27) If all L~t Lr .11 Jl = .1-1*1 z1:+1 , • 0, then zP xJ E ••• , (5.25) in the form (r = 1, ••• , k) • -' r z: 1181' be chosen &:rbit.i"&rU.,-. Define zP (5.28) = z~. tzi,, •.. • ~· t.~( J • ( p We !ll\I.Bt evaluate zi• ••• • ~~ . uabg (5.27) ~- (5.29) n t .1*k+l zPnt• J n (5.}0) ~l = ;k z~ ~l • n t .1-k+l k + 1, ••• =k + 1, ,n)} , ••• • n) • • AP~ = J -1t llll "11: zt a:,_l + ~1 = ~ ~1 + ~1 • ln ord.e r to eontime with this process. it is necessaey to use a much more condensed notation. Let ue define the symbol Jf! \1 the following recursion formula: , ) ( 5.31 .,;ok ~ • -'Dk J+2 _n'k j+a -'Dk uJ+l trJ+l ~j + ~+2 HJ + ••• + Bj+a HJ + ••• ·lf~·~. 25 Tl:m.a ~k is defined 1n terms .o f Rj , (a • J+l, ••• , k. :p). space y J into yp • Note that H~!. nf (e =1, •••• k-j) and is an operator which maps ~0 give a bette~ idea as to jllst what the a!k ar-e, the first thre& of them are written out ill .full below; {5.32) With thia notat ion. w& have • • • • • • z~. • ...- Jl ~ J-k.-a+l z~ ttJ, 8. Iii ~ • • • ••• We shall use indnetion to prove that (a • o, where 1. ••• , k-1). t <k • !hen zt- =~8 8 tor A#swme that it is trne tor s • 0, 1, •••• t-1, 26 zP lt-t • !1 t Jak-t+l zP uJ J ~t JJPk vK-t+l ;;pk ~t+2 rP.= oJ • lc.-t+l ~t + "1t-t+2 ~ + ••• + J ~:t + ••• ·rf~t+~t * Je-t iJ?k ~1e • We ha:.-e proved ths.t oompl.et•s the indu..otion . ~. (5.33) =~~ (a= o, 1, ••• • ~1) • Renee, (5· 3\t) .... wh ere (zP -) . t:Pk • ., • J. (p =k+l. the same as t he r; (5.35) How ~ ••• • n) • =o, ~'1 -aPk ut>k . + '""2 7'2 + ••• + ~ ,.k + '1p • 0 • We want to show t ba.t equ.at t ons (5.34) are which we had beto~e. The r~eion relation for ta ....r 'I' .....1 Y~8 • 1 ra-1 +. .t1. 8 1 (• = 1, 2, ••• , k,• 1! = .,...1, •• .,a) • It we aP.Pl7 the recutta1on relation • 7pk-3. + uPk ,~3 ~ k a.aa!n. we get + ~k . ..kl-. . 3 + uPk ~3 -"'JI!o-1 •.Jt-l. ~ 7.k.-2 • J.aeuae lor 1n4uct1on that Then by (5.35), we have + uk-e+l .•z-...1 ) '"»-• + 7&-a <(.. + ~ ~. + ••• + ~.1 u::; + ·~~· + + 1 ~.. ~:+1> ~=-1 The induction is complete. ~• 'Tp + ~k 7k + ~l Nov for Y'k:-1 + • • • + • • k ~J (5.36) we have in ~k 11 Ytt-j + • • • + = O, (p • k + 1, ••• , n), which la u:actl7 the same system of equations (zP• '7) • o. a.s This ie the cas where d fba.a, we have proved that (zP, ~ = o, (p =k + 1, =n • k • baa a solution it and only it :Lx • 7 zP •••• n). where the are as defined above. Z be rmy solution of Let such z. !hen certainly implies that solution x, ZL • o • (zP. 'f) • o, (p • k + 7> • 1, ••., == (Z 1 t'*'i ) • (ZL, i) • o tor all 0 for all n) • which Converseq, i:f' Lx L.x • "3' ha.s a aolu·Uon. (z. y) Suppose (z, z * 7 .baa a. such that Bde completes the proof of the fo1loW1ng theorem: WOBJ!iM 5.1 fhe system (z. f) • 0 Lx • 7 has a solution if and only 1t for aJ.l Z SllCh that ZL • o, provided that at each step 1n the elimlnation process. either all the operators are zero, or a.t least one of them has an inverse. . The . tl:teorem is proba.blr atill tru.e without the restriction on the existence of inverses. some examples of the 11 e lhall consider in the next section in•between • ca e. We lhall aolve two t:aaplea 1a wbloh x ad y at'• twO""''Ua.-.. ~ ,.(o l-1 (1a) (~) ~ + 00 (6.1) ( oo\o) (~) ~ -1 + o.) ~ 0 (2 1)(xi) 4 QO (oo oo) (~) fhia 1a a quem where· none ot the I.Jl h&ve ue not aU hro. fheoteua (6.1) r.acs.a to ~ . ( • 0) Q . 0 0 (ti) ~ • (ri) r~ an . inver.s~, iUl4 a\111 t.b.er 5·1 uat not include this • • · fU qatea JO !his is equ.ivalent to \ 0 0 • .. xi • liov xi Then either ~ ~~ ls u.n1qo.ely determined b7 the fou.:t>th ot these equations . xi be detel'm1ned. or ~ ma;y be chosen arbitrarily and the othe~ will Also,. we must ha.Te 7~ c 7i =0 if a. solution is to exist. Thus, the neces~ a.nd autficient concl1t1on for the enstenoe of a eolution to (6 .1) is (6.4) If' this is eattsfied, the solution is }l (6..5) ~ • .. ~ ~ • ......1tl'&l"'' ~ • ubUfat7 ~ ..l ~ • 'i + 2 2 7e • 2xi J.et ua conat.dw i1Pt the e,retea IL • • o • Thia 1a equS;yala.t \o (6.6) Ol" t (6.7) ·z1 1 ~1 + 12 L 2l • 1~2 + 12lt22 * o 0 Ia • (6.8) ( ~ ~) · • z~ z~ SJ'nem 6.7 1• tma 2 ( sh zta)(: :) z!21. (6 ..9) zl 22 z11 + z2 21 (~ ~)(: :) . c2~ ~ ~ (00) •0 •10 22 :l (: :). 0 fhl• t'educu to •O {6.10) 23: .·;1 1 ) 1 'a + 0 ( 0) • 0 0 0 zi1•zln• 0 zia • ~ • o , ~~ t • zl-22 • :zfl • ~~ Z'2: • •• e.J~b1 i1"8l7 ( ~21 z • G) • o 3) z1 71 + z2 72 ('· 7) • -(: ~~'h) (~1) ·(~ :) (i) -z2 21 (r.X~ • ~) ~ ·(~) 1t (a. f) tha. nidtD.t}¥ ~ • ~ • + l 7,a + ta to equal we mll&t r:) (~ ~ (~) •ero tor- baTe ~ • 0 , then IL • 0 ,f t.ll ,f • tor all Z z • such that ZL o • Convet"-aeJ.7, nch thU o• 111 it ZL • 0 • !hi• 1a thlla ta• aa.me oo.Ut1on aa (6.4) • hno•, oample 6.1 aat1•t1e• theor• ;.1. It &ppeu'a 'YflrT ls.tcel7 that the theorem would bOld tof! all qeteu where the ma11'1eea and the ~ and 7 J ate ttnU• d1aena1onal ~ 1 are t1n1te ~ecto~a. Wh•ther it would be ttue tor othai" ayat•• ia a.n interesti,ng q'l.le&tion, JiltHfH' 6.2 tb1a .xaaple ei1Dg\ll&l', 11 oae in vh1oh all the LJi ret the "f•tem has a. "U1qu.e aol11tJ.on. ~. (: !) ~ -(~ :) Lla .L22 Let . (~ ~) . (; .:) Ah Z: aDd 1 • • aa 1n e:turple 6.,1 • !fhe qate Lx • 7 t'edncea to ext+xi+~+~- ti ~~+~+~+~• T~ (6.14) • 1 1 1 2 1 1 l 0 0 0 ..-2 0 3 -1 2 .. • 4 • lroa thla, we knov that qa\ea (6.14) baa a uniqu.e solution for arb1~l!aq y • llhf..a 1& ttla;e e"f'e·n though none ot the L.1~ baa an 35 It v.e oomp11to tot thia oaae, Z eertainl¥ ( Z, i) • o 5.1 appU.ta 1.~ ZL aa 1110.1t th• preoediDg emm.ple. we find that ••to in ol'der that be tot! ar'bl tt"aq y aDd v to exa.aple U. • 0 ~ fhen find that theore• 6.2 • 2hia wcgeata the tollovlnc corollat7 to theorem 5.1: M2HatRX .6 e.J. 1'ho qnem Lx • 7 haa a aola.tloa tffl 7 1t aDd only ~ Su:ppose the. t 1Q11•a that ZL .tr 9 r X« • haa &tbi tl!!alT Z ~ 0 • solu.U.on to'I' ubitrar, Z be a:q noto.r BU.ch that U. • 0 • then {z. y) • o, 1 • Let by taeor• 5.1, tor arbitr&l'T 7 • lbt thia 1apl1es I • 0 • Now np,oae that ZL • arblttU7 7 and b • 7 ~ z, o ill;pUea Z • o. !'hen (I• f) • 0 tor •t1atria6 ZL • o. Then, b1 theora 5 .1. baa a aolution for ar-bltr&l'7 7 • !b.ia cotollary 111 of eour••• tu.OJeot to the aame restriction a.a theor• 5.1 • BDlLIOGRAPB.Y l. Courant, R. and D. Hilbert. Methoden der Mathema.Uechen Physik. 2d. ed. Vol. l. Berlin, Julius Springer, 1931. Q69P. 2. !opicte in Hilbert spa.ee theoey. Notes for graduate seminar. Corvallis, Oregon State College. l95Q-51. (Mimeographed).