Chapter 3 Vector Spaces In Chapter 2 we saw that the set of all plane (space) vectors forms a commutative group relative to addition and, further, relative to scalar multiplication it satisfies certain properties (Properties 2.2.7). All these properties of vectors are so fundamental in mathematics that whenever any system satisfies them we give it a special name, namely, vector space. The precise definition of a vector space follows. 3.1 VECTOR SPACES 3.1.1 Definition A nonempty set V is called a real vector space (or a real linear space or a real linear vector space) if the following axioms are satisfied : VSI There is a binary operation ‘+-” defined on V, called ‘addition’. ‘YS2. There is a scalar multiplication defined on V. (This means, to every real number « and every element u of V, we can associate @ unique element of V and denote it by au.) VS3 Addition and scalar multiplication satisfy the following : (a) Vis a commutative group for addition (i.e. G1, G2, G3, and G4 hold for addition in V). (b) au + y) = au + av and (@ + B)u = au + Bu for all real nymbers a, 8 and all u,v € Ve (c) a(Bu) = (of)u = (au) for all real numbers «, B and all ely, (d) lu = uforallu € V. 3.1.2 Remark A complex vector space is defined analogously by using complex numbers instead of real numbers in Definition 3.1.1. 313 Remark The real or complex numbers used for scalar multiplication in the definition of a vector space are called scalars. Throughout we shall deal with only real vector spaces and complex vector spaces.{64 | VECTOR SPACES y, and V, have addition and om m seo hr aonb ‘i 5 3 hol tion defined on 4) of YS: , pls 223); and that (b), _ ae ee fiopertics of plane ve V, and V, (Properties 2. "abstracted to provide Hie genera sting ves we ners use the word ‘vector’ to a y ee uN Jn other words, elements of a i pe is eo sae caer Plane vectors and space vectors are only sp have not used u, v, w for - ena a ee We shall consistently use th ae W, ... for vectors, and the Greek letters a, B, y, ... for ! However, we use the boldface 0 to denote the zero vector space to distinguish it from the scalar ‘zero’. Clearly, V, and V, are vector spaces. There are many other exe of vector spaces. We shall note a few ofthem here. In fact, the ©: should be able to visualise several vector Spaces before we go deeper the subject. Visualisation of a ‘vector space involves five steps : @) Consider a nonempty set V. (i) Define a binary operation o; (iii) Define scalar multiptica (iv) Define equality in 7, Vv) Check whether (a) 5 ssdioaaatiene mutation i Sa ae mtive to the Bxample3.1. Let'V, be the au) gd ne Thus, an element of Y, would be of Oe form (gee® OF Fea numb His ate real numbers. Define addis n, scalar in V,, as follows : If y = Oy x, xa) elements of V,, then : Mand nV and call it ‘addition’, tion on V as described in VS2. Had? «+5 Xq), where the y nultiplication, and equality ee) are two Addition uty =, +, ote that w+» © yy, 7 te ec ordered n-tuple of real » because the right-hand «: right-hand side of : Bumbers. This is called Coordinate io (1) is an For @ real scalar a, ay jg defineg 9. dition.) a ox) Tass a) Scalar multiplication Equi ‘ate n-tuple (ax,, ax,, the ordered oy = (sy, 0X, «5 xy), which is again in V,. (This is called coordinatewise scalar m, ; ulti) BUD at = Coe x.) ancy — Oe ns ge equal ifx, = y,,7 9 Dy veay M fe Since addition, scalar multiplication, @) Plication ) "© defined ag and equality in py, have been iat3.1 VECTOR sPAcEs / 65 defined, what remains is to check whether (a) to (d) of VS3 are satisfied. We shall leave most of this checking to the reader. However, in order to give him an idea of what such checking involves, we shall now go through some of the details. Commutativity of addition Let ¥ = (Xt) Xa 4, Xn) and ¥ = (J, Js «5 Yn) be two members of V,. Then WAY = (Xt Yt Xa + Yay oes Xn + In) and VM = (yt Xp IaH Xap ves Yn + Xn) The real numbers x, + y, and y, + x, are equal, since the commutative law of addition holds for real numbers. As this argument is true for every K, it follows from the definition of equality that u+v = vy + u. The alert reade would have now noted that, fundamentally, it is the commutative law of addition among real numbers that gives the commuta- tive law of addition in V,,. If this fact is remembered, we can check that u+v = v + weven mentally. Existence of additive identity We have to determine an element 0, which has the property 0+u =u. Im this case we find that 0 = (0, 0, .... 0), because OOO (5 00) x)! xan ess, XQ) for all (xy, xy. Xa) E Vn. Existence of additive inverses Here we have to determine an element —u for the vector u € V, such that —u + u = 0 = (0,0,...,0). Ifu = (x1, %, ~., %,) then we can easily see that —u = (—xy, —Xay 24) —Xn)= We assume that the reader can check the remaining properties. Once (a) to (d) of VS3 are checked, it follows that V,, for the addition and scalar multiplication defined in (1) and (2), is a vector space. f It may be noted that, by Theorem 1.6.3, 0 = (0, 0, ..., 0) is the unique zero of the space and —u = (—Xq, —%y, ..., —%n) is the unique negative OF w= (x, Xp, «.., Xn) Thus, V, is a real vector space, since we have used only real scalars. Can we use complex scalars in this case? No. For, suppose « is a Complex number. Then av = (#x;, %%, ..., a%,) is not in V,, because the numbers ax, are complex and V, contains only n-tuples of real numbers; 80 scalar multiplication is not defined. See, however, Problem 2. The special cases n = 2.andn = 3 of Example 3.1 give us the vector Spaces V, and V,. The special case n = 1 gives the space V;, which is nothing but the space of real numbers, where addition is the ordinary Addition of real numbers and scalar multiplication is the ordinary muiti- ion of real numbers.3.1 VECTOR SPACES / 67 mentioned, He should check the axioms in each case until he reaches the stage where he can quickly say whether the axioms are satis- fied or not. In any case no axiom should be taken for granted. In all cases the zero element of the space and the negative of any given element of the space should be cleat in the reader's mind. Example 3.6 Let (ca, b) stand for the set of all functions defined on [a, b] and having derivatives of all orders on [a, 5]. This is a real vector space for the usual operations. It is called the space of infinitely differenti- able functions on [a, 8). We conclude this article with a theorem giving certain immediate conse- quences of the definition of a vector space. Recall that, in the definition of a vector space, we have property (d) of VS3, namely, 1” = w for all u € V. It should be noted that this apparently trivial axiom is crucial to the development of the theory of vector spaces. _317 Theorem In any vector space V, (a) 20 = 0 for every scalar a. (b) Ou = 0 for every u € V. © (iu = —u for every we V. Proof: (a) «0 = «(0 + 0) (by G2) = 0 + 0 (by VS3(b)). Adding —(a0) to both sides, we get 0 = —(«0) + (a0 + a0) —(40) + (20) + 20 (by G1) ita) (by G3) 2 (by G2). (b) Ou = (0 + Ou = Ou + Ou (by VS3(b)). Adding —(0u) to both sides, we get D 0 = —(Ou) + (Ou + Ou) = (—(Ou) + (Ou)) + Ow (by G1) 2 = 0+ Ou (by G3) = Ou (by G2). (c) (—1)u+u = (—l)u+ lu (by VS3(d)) p = (-1 + lu (by VS3(c)) =u=0 (by (b)). So by G3 and the uniqueness of the negative, (—1)u is the negative of u, he (—l)u = —u fl Tt is convenient to write v — v for u + (—y).3.2 suBsPaces / 69 for space Vi (—1)(—u) = wfor alu € V. porto space V, ty oe ny for all uvev. peter apace Vy —u = v ae (+ 9) forall uy € Vi or space V, —(—w) = u for all ue V. Ris a vector space, e scalars are complex numbers, then $(/) is a complex vector G, the set of complex numbers considered as a complex vector pifa- 1+ i = 0,thena = 0 = 6. 3.2. SUBSPACES jition Let S be a nonempty subset of a vector space V. S is to be a subspace of V if Sis a vector space under the same tions of addition and scalar multiplication as in V. to understand this definition as well as the future concepts in we shall repeatedly consider the concrete cases of Vz and V,. Euclidean plane. Take any straight line S through the origin P on this straight line can be considered as a vector OP of — ‘The sum of two such vectors OP and OQ, where P and Q both again a vector OR, where R lies in S. Similarly, a scalar multi- automatically satisfied in S. So S is a vector space under the stions as in V». Thus, S isa subspace of V2. In other words, every +h the origin is a subspace of V2. ‘same manner, in V, we find that any plane S through the origin sface of V,. Also, every straight line L through the origin is a sub- f Vy. The following examples further illustrate the concept of a .7 Let L be the set of all vectors of the form (x, 2x, —3x, x) en L is a subspace of Vs. u = (x, 2x, —3x, x) andy = (Y, 2y, —3y, y) are two elements Uy = (x+y, Ax +y) —2 ++ Y au = (ax, 2(ax), —3(0x), @x) f ‘to L. The zero element (0, 0, 0, 0) is also of this form and to L. The negative of wis —u = (—x, —2x, 3x, —%), gain of the same form and hence belongs to L, The other laws ‘and commutativity for addition, distributive laws, and the Iu = u are all true in Z, because elements of L are elementsspace under (bv). fi3.2 SuBsKAuED | re = 0 for any w € Vand therefore for any vy © S. Taking any "we see that, from (b),0 = Ou, € S. Similarly, (—I)u = —# y u € V and therefore also for a given 4 € S. Hence, —4 € yu € S. Thus, G2 and G3 hold in 5. ” xioms G1, G4, and (b), (c), and (d) of YS3 are automatically satisfied nts of S, since these laws already hold for elements of V. Note needs no checking, because these axioms are interrelations of the nts. In the cases of G2 and G3 the checking was necessary, because d to show the existence of certain elements in S. In view of Theorem 3.2.2, whenever we want to prove ta set S for certain operations is a vector space, we try to look as a subset of V, where V is a vector space and the operations of dition and scalar multiplication in S are the same as the opera- ions in V. Once we have done this, it is enough to prove that the operations are closed in S. 3.9 Take G[a, 6] of Example 3.5. G[a, 4] is a subset of |, which is a vector space by Example 3.2, So to prove that j] is a vector space (which we have done in Example 3.5), it is to prove that @ [a, b] is a subspace of Fa, b]. Therefore, by m 3.2.2, we only need to prove that Z[4, b] is closed for addition multiplication. the sum of two continuous functions is continuous and any wultiple of a continuous function is again continuous, we find that tion and scalar multiplication are closed in % [4, 6]. This observa- t only proves that Z[a, 6] is a vector space, but also that it is a pace of F[a, 5]. mark The spaces Za, 6], Gla, 61, Gla, 6), and Pla, b] are subspaces of Fa, 6]. rther, note that Ga, b) is a subspace of Z[a, 6). (6) GIa, b] is a subspace of Z[a, b). Gla, b] is a subspace of F [a, b] for every positive integer n. @) Ya, b] is a subspace of G("(a, b] for every m ) = 0. So ae ea ne a i that the rane $0 vanishes at xo and sop + gE 5. So addition is closed in S.2), whic e, but » is a pla Me eps 3.2.2 can be sepP ou + pv © S for all u,v & S and all scalars a, B. tt Wo = (0X Xa 0 Xn) © Va | x, = 0}. Prove that Wis a sub- W = {1 Xa es Xn) E VE | yxy + angry + oe + Oatn = OF a's are given constants} is a subspace of VE, "4, Which of the following sets are subspaces of Vs? 8) (0%, Xa Xa) | 1%, = 0} : Le) {Gu ay x) 15 = v2} (e) {@%, % Xs) | V2x, = 3%} (@) {G1 %, Xs) | %5 is an integer} 3) (%1, Xp, Xs) | XE + x9 + x <1} (8) (Qa, xy %) | 41 + 2 + %2 > 0} (g) {(%1, Xa Xs) | %1 = V/2%2 and Xs = 3x9} (bh) {(%, 9) Xs) | m1 = 2%, or x; = 3x4} 3X, G9) {Gry Xa, Xs) | 1 — 2% = Xe I 5, Which of the following sets are subspaces of P ? \_®@) pe F | degree of p = 4} vb) Pe P| degree of p < 3} ws {p € # | degree of p > 5} @) {p € P| degree of p < 4 and p’(0) = 0} © Pe Pf\p) = 9. 6. Which of the following sets are subspaces of Z (a, b) ? @ {FE FG») | fr) = %%0 € (4, &) (FE G@ |S) = forall x € @ b) b, © FEGarhiNZo =» @ FEE HIS = BLO © Ee Gad |W) + 3xf") —f') + ¥f@) = 0} ® Ee Gla b)| Sifax = 0}. 7. Prove the statements in Remark 3.2.4. {A If U and W are subspaces of a vector space V such that UC W, then prove that U is a subspace of W. (b) If W is a subspace of a vector space V and U is a subspace of WV, then prove that U is a subspace of V. 9% True or false ? (a) In Vievery infinite subset is a subspace.re fully be more +5 Un) ) y}. Any line” m 2(1,0)" inations is |3.3 SPAN OF A set / 75 Proof: By Theorem 3.2.2, we have only to prove that [5] is closed for addition and scalar multiplication. Let wand v be any two vectors in [S]. Then MU Ny + Ayty +... + Oqtly for some scalars a, some u,’s E S, and a positive integer n, and vy = Biv, + Bare +... + Bam for some scalars B,, some vectors v;’s € S, and a positive integer m. Hence, w+ ¥ = ay +... + Ont, + BY toe +H Bete This is again a finite linear combination of S and so u + vy & [5S]. Similarly, ou = (0%), + ... + (#%q)uy is again a finite linear combination of Sand so itis in [S]. Hence, [S] is a subspace of V. ff 3.3.4 Remark A nontrivial subspace always contains an infinite number of elements. So [S] (4V,) always contains an infinite number of ele- ments. But S itself may be a smaller set, even a finite set. By con- vention, we take [6] = Vo. 5 Theorem Jf S is a nonempty subset of a vector space V, then [S] is the smallest subspace of V containing S. Proof: Clearly, [S] is a subspace by Theorem 3.3.3. It contains 5, because each element u, of S can be written as Iu, i.e. a finite linear combination of S. To prove that [5] is the smallest subspace containing S, we shall show that if there exists another subspace T containing S, then T contains [5] also. So let a subspace T contain S. We have to prove that T contains [S]. Take any element of [S]. It is of the form ot, + atl bu + Gall, Where a's are scalars, 1, t, -..; ¥ are in S, and nis a positive integer. Since SCT, each u, also belongs to T. Since T is a subspace, at, + agtla + ... + dally should also belong to 7. This means that each element of [5] is int Example 3.12 In Vz show that (3, 7) belongs to {(1, 2), (0, 1)], but does not belong to {(1, 2), (2, 4)]- Clearly, (3, 7) belongs to [(1, 2), ©, 1)], if itis a linear combination of (i, 2) and (0, 1), ie. if 3,7) = (1, 2) + 8, 1) = ( 2% +8) for some suitable a and 8. This is possible only if a=3, wt+p=7. () Solving Equation (1), we get o = 3,8 = 1. Thus, (3,7) = 3(1, 2) + 1@, 1). Hence, (3, 7) € [(1, 2), (0 1]. Further, if (3,7) € {(1, 2), (2, 4], then (3, 7) = a(1, 2) + B2, 4) = (a + 28, 2a + 4B) for some suitable « and f. This gives a +28 = 3, 29+ 48 = 7. Q)3.4 MORE ABOUT SUBSPACES / 77 (e) xy-plane and yz-plane in V,? (4) x-axis and the plane x + y = 0 in Vy? 7. True or false? (a) Spanofx +» = Oandx —y = in Vyis Vy (b)_ Vy has a subspace consisting of 7 elements, (©) The span of the set {x, x, x9, x4, ...} is 9. (a) In Vy, [y-axis U z-axis] = Vy. @) Inhy k= Ri + MOL KI EM: @) In ¥§, (1, 2, 0) € [G, —i, 1), (0, 4, —D)L 3.4 MORE ABOUT SUBSPACES Let U and W be two subspaces of a vector space V. Their intersection UE W cannot be empty, because each contains the zero vector of V. Now, if u and y are two vectors of U () W, thenu-+v€ Uand u + v & W, because U and W are subspaces of V and u, v both belong to Uas well as W. Hence,u-+v GU) W. Similarly, if « is any scalar, then au isin both Uand W. Hence, au € UW. This shows that if U and W are subspaces of V, then U (\ Wis also a subspace of V. This result can be generalised to any number of subspaces. More precisely, if U,, U;, ..., U, are n subspaces of V, then their intersection U, 2 Uz A... A Un is also a subspace of V (see Problem 1). We shall use this fact in an interesting way in the following example. Example 3.14 Let W be the set of all vectors (x,, X2, ..., Xn) of V,, satisfy- ing the three equations OX + ayXy +... + AnXn = O q@ Bix + BaxX2 + --- + BaXn Q ait, + Yaka toe + Yate = 0 3) Then W = W, 0) Ws Wz, where W, is the solution set of Equation (1), Ws is the solution set of Equation (2), and W, is the solution set of Equation (3). Each W is a subspace by Example 3.11. So, by the foregoing argument, W is a subspace of V.. Thus, we have proved that the intersection of two subspaces is always a subspace. On the other hand, the union of two subspaces need not be a subspace. For example, take U = x-axis and W = j-axisinV, Uand W are subspaces of Vs. Here (1,0) € U and (0, 1) € W. So both (1, 0) and (0, 1) belong to UU W. But (1,0) + (0,1) =(,1) UUW. This shows that U U W is not a subspace of V;. Thus, U U Wis not in general a subspace. However, we know, by Theorem 3.3.5, that [U U W] is the smallest subspace of V containing i 1Ficure 3.2 ‘The If U and W are two subspaces of a vector space V, then U + Wis a subspace of VandU+W = [UUW]. Proof: Obviously, U + WC[U U WI, because each vector of U+ W lite linear combination of U \) W. We shall now prove that [U U WW] + W. Letv€ [U U W]. Then the argument preceding Defini- 4.1 shows that v is of the form u + w, forsomeu € Uandw © W. vEU+W. Thus, 0+ W= [UU W]. This automatically U + Wa subspace of V. § ‘Remark From Theorem 3.4.2 it follows that U + Wis the smallest subspace of V containing UU W, ie. both U and W. le 3.18 If V = Vs, U = x-axis, and W = y-axis, then U + W is st of all those vectors of V, that are of the form (1, 0, 0) + B(, 1, 0). fore, U-+ W = {(2,8, 0)! % B are scalars}. On the other hand, W] consists of vectors of the form au + fw, where a, 8 are scalars, we W. This gives [UU W] = {al1, 0, 0) + BO, 1, 0) | 4, 8 are scalars} , ‘the same as U + W. This illustrates Theorem 3.4.2. Note the relation arising from this example : [x-axis U y-axis] = 2-axis + y-axis = xy-plane. (5)Gr epxces DIRECT SUM Th in Equation (7) is = © Equation (7) can be rewritten as (xy-plane) @ (¢-axis) The sum in Equation (6) isnot a Any vector (a, b, c) where (a, b,c) © sum of two vector written as the si the yz-plane, in +b 6) can be the other in (10) : (11) the following See‘ 34 MORE ABOUT SUBSPACES / 81 w € W. Then — &twew ty (14) a u-w=w ow : ‘ : (as) of ein why re W. From Equation (15), we find that sa Tah Since Z is a direct sum of U and W. 4 ee oe Heute thatu=u' and w=w’. So nm for Equatios ii 3 © Wis thane, n (13) is possible. In other xis let condition (12) hold. We si . hall now prove that Z is . Uand W. Since Z =U + W, we have only to prove that ible, UM W contain a nonzero vector ». Tht t . Thenv € U, ve +0EU+W with vEU and 0EW. Also v=0+9E € Vand v E W. These two ways of expressing a vector » in ») dict the hypothesis (12). Hence, UW = {0} and Z = e : tion If U is a subspace of a vector space V and v a vector of s {} + U, also written asv + U, is called a translate of U ora parallel of U (through v) or a linear variety. led the base space of the linear variety and v a leader. Note U is not a subspace unless v € U. (Why ?) ation, take the line » = x through the origin in V. Call it U. the point v = (1, 0). The translate v + U of U by vis the line | through the point (1, 0) (Figure 3.3). It can also be obtained g (1, 0) to the vectors in y = . ‘Y vty veo) xX O| =0,0) x (b) (c) Ficure 3.3 i the plane y = 0 in V, and call it U. Con- et, oy sly 1)-+ U isthe set of all points of tv=(1,Therefore, fe parallt a (1,fo= (1 tT-)A BSI = A 0= av) @ a a i fo> la Ber} = moe) 4 Sone aulMa}Op pur 4 puke 7 s}9S UAAI3 dy Jo WONDesIO;UT “s1e[e9s paxy “9 = txtte fof Fata fal A joey dt cee ae “4 Jo sovdsqns ® st suorpuoo mw Bur * x ‘Tx) $10199A JO 198 943 1PY1 BAOIg “y "g Burureju09 4 Jo ssoedsqns Tle JO UoRoassazur oyi si [§] YI DAoId ‘4 Jo Jasqns A}dwauoU ess “¢ AD MOM D AFA Jo aoedsqns esr yA (Q) M Jo soedsqns & st yU A (®) yun enoad “4 aovds 10,99 v Jo sooedsqns ase y pur 9 Jp c -MOTOY ayy Surdysnes “4 DB (“x 1-7 F “A Jo sovdsqns v ose st’ U ="“q U' U9 U igre u eaosd usyL “4 a0uds 409A v Jo sooedsqns u aq “Q *- 8Q TQ PTT ae 1 12S Walqolg ‘0+ AU IO" St 4 Hi a+ 4 wor qUssayrp 9q [HM A mous pouteigo 0 + 4 leltered ay) ‘4 3 4 Jo}an sayLouE oye, am JT “A + 4 jayesed & 198 9m “4 D4 109A v War -ouvjd oy ur syurod: jussoUIP Yanosyp Sout] IyBIes jawed KuEW SurMEp pu wi310 ayy anos) oun ayBlENIS 2uO wWoss Buli2VIs se ores ayn se sUL “QJe sjayreaed Aueus ureigo uvo am ‘2 aovdsqns v wos Suumig yaumay £) “4[98194U09 puv jajpered owes 1994 Om} Jo 2ouRJ94Ip ayD Buojaq 430q Aoy) UY ‘7 0} sBuo}9q s10) wey) shvs wimioryy ay) Jo (4) Wed pw Aismorngo sy (1 *] "Q) PUL od Ut eae (0 ‘0 ‘1) pur (1 ) 104 “uonwaysay ayy anurjuoo APIH A PUR + 1 = gd hRID “A puw C'8'O + oD 40) OU) UF Yat aq A8O+ OD 3 404994 A4999 IHD S488 wiaso9y) 1010's) 401% 94) 891971 “g WON Sonar AFA PING? OM 4S UN ARLIOSED OF Ie) 08 / SADWAWNAS AQUEY amon HeJ0 tesqns UY ouruna:JPAUOU, pur [eIAIN, Jo MoNIUYSP oyy YIM spore 3ONAGNadaGNI ‘JONION3d3 Y +5 Aq paorjdes oq jouuws “4 p]"¢ a[dure *(co (0) J ur Ayorsva rwoUI] B st GuEySUOD Be uonenbs 241 Jo suonnjos jje Jo yas 8 axoyn ‘s — UP. om} Jo uoTun ayy ap *AqaiteA reaul] v 9g JOU Poot S91 “AyUIBA I 8 aq j0u Avur BA JBOUTT OM} JO UNS *hyduisuou st 71 QOHIRA IedUT] -od ‘Ajor1vA avout] & st rer sovdsqns & st *7 U way ‘4 Jo sooedsqns ose“ 1, 1), (1, —1, 1), and 1,1, 1) + 20 b aes ) = 1G, -1,3)=0. are coplanar, ie. one of tains the point (1, —1, 1) » 1,1). Let us verify thi en ate 28, « 8 4g 38) A ool D), fe - la, «+38 —1, eT have onl e «= —1/2and8 = 1/2 = 1, 8 any scalars} ly to cho a + 38 ss @ and 8 such that Hence, by Th by Theorem 3.5.9, the plane through (1, 1, 1) and » 1,1) and (3, —1, 3) com We proceed, w record some simple fa; : 1, we shall rd le facts abou is? ab lineal3.5 LINEAR DEPENDENCE, INDEPENDENCE 191 For, if iy Ye «++» Yn) isa set and vy, = 0, - Ory + Ove Have + Ovid + 1M + OV es + OH isa nontrivial linear combination resulting in the zero vector. 35.11 Fact In a vector space V, if v is a linear combination of Yan Mas oven Van Os ¥ & [ay May oves Yady them (Vy Yay vves Ma} iS LD. For, ¥ € [15 ¥2 +++» Yn] means Vo ayy, + mQVa boas sb Onn fe Ty — Ma" — os Ane = 0. Meaeiract In @ yector space V, if the set (v1, Vey «+1 Mn} is LT and ¥ & [Mp Yay «+> YnJ, them the set {v, Vip Yay «+5 Va} is LI. The proof is left to the reader. In the case where the set consists of just two or three vectors, we can Jook at these facts from the ‘geometric’ point of view of Theorem 3.5.9. For the purpose of understanding, imagine that V = Vs (though this is not necessary for the argument). Since 0 lies in the line through v, {0, v} is always LD by Theorem 3.5.9 (b). Similarly, 0 always lies in the plane through any two Vectors vs, V2. Hence, {0, v1, v4} is always LD by Theorein 3.5.9 (C). The idea contained in Theorem 3.5.9 does not stop with just three vectors. We can in fact work out a general result on the same lines for a linearly dependent set of n vectors. Before we do that, let us look at the theorem once again. Actually, in the proof, we have indicated the following : () fy %) is LD if one of the vectors is a scalar multiple of the other and (ii) {v,, ¥, Vs} is LD iff one of them is a linear combination of ‘the others. In general, we shall now prove in Theorem 3.5.14 that if {v1, Vay...) Yat is LD, then one of the vectors is a linear combination of the others, the 7 converse of which is Fact 3.5.11. Before that, we state the following simple theorem and leave its proof to the reader. ‘A513 Theorem (a) If a set is Lt, then any subset of itis also Ut, and (b) Ifa set is LD, then any superset of it is also LD. , | 381 Theorem In a vector space ¥ suppose {0p Yay «+s Yn} iS.an ordered set of veciors with vy, # 0. The set is LD iff one of the vectors as Vay ooc> Yan SAY Vay belongs t0 the Span Of Vay...» Vass HE: Ve € iy Yay...» Vea] for some k = 25 3, soy Me Proof: Suppose r» € [iy Yo» Ml ‘Then y; is a linear combi- tition of »,. v2, wg. Thus, the set {vs. Ya ---» Vick vx} is LD. Hence, Y Theorem 3.5.13 (b), {V5 Yar ==» Yad iS LD.SS ‘onsider the set; op | yECTOR SPACES sys} igo. C post conversely, suppose that wee Si =n 5, = OW va} + vi} nadie ae 4 is nd by assumption On fs LD. So wey, By Theorem 3.5.9 (a), Si 18 11 an Betas, vous: i Jhoose the first linearly depe' ? down the list and cho eae Then Sy is LD and Sp-1 is Le Since Sy is LD, we have ayy, + Os +o + OEM = 0, with at least one of the a’s not zero. Surely oy, # 0, for if x = 0, S,-1 would become a linearly dependent set contradicting our stateme: that Sy is the first linearly dependent set. Therefore, we can write na Cant Cnt ot = Mies. which means 15 © [Yi ¥3) 0 Meade 55 Corollary A finite subset S = {v, Mat = {V1 Va, ..., Vn} Of a vector space V containi ir Ys es ee é ee vector has a linearly independent subset A suck Proof: We assume y, prove, as we have A=S. If ae vector ye such that y, h $5, by on ray Mee ony ag, Diseatd »,. Th ns Problem 4 (0), Problem Set 3.3) oe 28 the same pan as that of S (6 repeat the foregoin eM. St is Lt, ther eo. th dent subset A such tat a ues finally Toad te eee fe ee a (Why 9 a linearly indepen: If S is u1, then ‘i > there is nothing to len, by Theorem 3.5.14, we have @ 1,1, 1)) is wo and | ered. set eg) 's 10 au lose oncior ike voit ah BO 1), (1, 0, —1. iapsides those belongs to the span o! 5% = (1,9) Sema, ot Batre 4 . Obvi 0), een rte Sib alo, SSS tat 10,1, Ono Sei rer of a +1) eGo, ity 2% 0). aeP 3.5 LINEAR DEPENDENCE, INDEPENDENCE / 93 le 3.33 In Example 3.32 find the largest linearly independent subset whose span is [54] ‘As proved in Example 3.32, (1,0, —1) € ((1, 1, 0), @,1, 0) and pence (15.0, —1) & (C1, 1, 0), (, 1, 1), (1, 1, 19]. Therefore, we discard (1,0, —1)- The span of the remaining set A = {(1, 1, 0), (0, 1,1), A, 1, DF je the same as [S,]. Now we check for the linear independence of A. ‘Suppose a(1, 1, 0) + 6(0, 1, 1) + y(1, 1, 1) = ©, 0, 0) or @+yo+8 +78 +7) = (0,0). This means 2 + y= 0,2+6+y=0,8 + y=0. Solving these equa- tions, we get only one solution, i.e. « = y. Hence, the set is L1- Therefore, A= {(I, 1, 0), (0, 1, 1), (1, 1, 1)} is the largest linearly inde- pendent subset and [4] = [S,]. Finally, let us extend the concept of linear dependence and linear independence to infinite sets. 35.16 Definition An infinite subset S of a vector space V is said to be linearly independent (11) if every finite subset of Sis LI. S is said to be linearly dependent (Lb, if it is not L1. Example 3.34 The subset S = {I, x, x4, 7, of P is LI. For, suppose ayx*t + a,xk + ... + anxk» = 0 with ky, ka, ...» ky being distinct nonnegative integers. Note that. the equality is an algebraic identity, since the right-hand side is the zero polynomial. So, either by giving various values to x or by repeated differentiation of both sides of the identity, we get a, =0=a2=... =a. Problem Set 3.5 1. Which of the following subsets S of V, are Li? @ $= {(1,2, 1), (-1, 1,0). —1, 2} (b) S = {(1, 0, 0), (1, 1, 1), (1, 2, 3)} (©) S={(I, 1, 2), (—3, 1,0), Ul, 1, 1), (1, 2, —3)} (@) S = {(1, 5, 2), (0,0, 1), (1, 1, )} () S = {(1/2, 1/3, 1), (0, 0, 0), (2, 3/4, —1/3)}. 2. Which of the following subsets S of V, are LD? (a) S = {, 0, 0, 0), (1, 1, 0, 0), (1, 1, 1, 1), @, 0, 1, D} e a Ae a 2, 0), (I, 1, 2, 0), (3, 0, 0, 1), (2, 1, —1, 0} = {C1}, 1, 0), (3, 2, 2, 1), (1, 1, 3, —2), (1, 2, 6, —5 ates: mS ), (1, 1, 3, =2), (1, 2, 6, — 5),| 4.6 DIMENSION AND BASIS | 95 (g) A subset of a linearly dependent set can never be LD. (bh) Every superset of a linearly independent set is Lt. () x and |x| are tin @[I, 4). 3.6 DIMENSION AND BASIS | We shall begin this article with the definition of a ‘basis’. 36.1 Definition A subset B cf a vector space V is said to be a basis for Vif (a) Bis linearly independent, and (b) [B)=V, ive. B generates V. Another definition of a basis, which is used by some authors, will be introduced as an equivalent property in Theorem 3.6.8. ‘Example 3.35 Take V=V;. Consider the set B= {i, j, 1, where j= (1, 0,0 f= (0, 1, 0), k = (0, 0, 1). WY The set Bis 11, because Yo ai + Pi + rk = a(l, 0, 0) + 800, 1.0) + 70,0, 1) = 0= (0, 0, 0) implies that (#8, 1) =(0, 0,0), ie. a= 0= g=y. Also B spans Vs, because any vector (X1, X2, Xs) of V, can be written as a linear combination oft, j, and k, namely, (%1, Xa» Xs) = 4 + aj + sak. Hence, B is a basis wr ny it ¥ be seen that the set B, = {(I, 1, 0), (1, 0, 1), (0, 1, D} is also a cr js for Vz, because any vector (%, Xo Xs) E Va can be written as Y, ie ze 7 ei, 1, 0) ee) Me ute, 10 and Fads) + FC, 1) +70, 1, 1) - ©, 0, 0) implies « = 0=f = 7, i.e. Bis el is example shows that a basis for a vector space V need not be Unique. The two properties of a basis B, namely, (i) Bis tt and ({i) B gene- rates V are not entirely unrelated especially when B is finite. For, if a set B of n elements generates V, then we can prove, with some effort, that no linearly independent set can have more than n vectors. This result is the content of the following important theorem. 3.6.2 Theorem In a vector space V if (v1, V2» ++» Yn} generates V and if {0y, Way <-> Wn} iS LI, then m n, is uv, Proof Suppose B= {yy ; = {Uy ty, ..., Uy} be a set of p Vee a , Pasis Theorem 3.6.2. Hence, A ig tp, a TE A ig py 3.6.4 Corollary If V has a bs also has n elements, es Proof: Suppose B, — two bases for V. Then B, Since [B,)=V and B, is 11 y J eh, {B,] =V and B, is 11 we get, by the he ‘ The number of element following definitions, 3. efinition If a vec ., m} spans V and Wm € V, Vay vss Yad and “sis of m elements, then every set of p vectors for V. Let 4 ; then p < n by B, and [8,) = b it ie, Wm} are + by Theorem Semitaae ime theorem, in 2, m < tS in a basis ig ther Sm Th n efore “Unig Since a space V hag the space ‘¢lements in a basis is is dim V. If dim y ina basis said tot : called the dimarul@-dimen ot ® finite TVs ste3.6 DIMENSION AND BASIS | 97 ‘0 = {0}, its dimension jg taken to be zero. It follows from Corollary 3.6.4 that, if a vector space V dimensional, (a) there exist » linearly ind indent vectors in V (b) every set of n + 1 vectors in Vis i) "H 36 Va is 2-dimensional, because (1, 0) and (0, 1) form a basis ‘9 i8 3-dimensional, because V, has a basis of 3 elements, namely, tly ar ch is n-dimensional, for, consider the elements 4 = (1, 0, 0, ..., 0) 4 =) 1)0, 4, 0) €n = (0, 0, 0, ..., 1) @; as the vector, all of whose coordinates are zero except the i-th, is 1. It is easy to see that e,,¢, , e, are 11 and every n-tuple is ar combination of e,, e, ..., en. the set {e,, €2,... , en} is a basis for Vy. The basis {€1, €25 --- &n} the standard basis for V, and will be used in the sequel without further mention. In particular, note that {i, j, k} is the standard basis Tn other words, 3.37 Find the dimension of the space Pn. polyno: in Y, is a linear combination of the functions x". Further, this set is t1. (Why 2) Hence dim f, =n + 1. eorem In an n-dimensional vector space V, any set of n linearly ‘Independent vectors is a basis. oof: Suppose B= {¥1, Voy...» ¥n} is a set of m linearly indepen- ors. To prove that B is a basis we have only to show that B ake VEV. The set {¥,, 2...) Yay} is LD, because V is n-dimen- + Yn, Say ¥, is in the span of all its predecessors. Obviously, this Not be any one of ¥2, V4, ...» Ya, for in that.case we shall be contra- g the linear independence of {v;, V2, ...) Yn}. Hence, v © {0,, Yay «5 Pale re, B spans V. this we have to prove only that the set is Lt, because Vy is 3- _ Suppose"ACES (0, 1,1) = (0, 0, vvecTOR SP 1,-1,1) + rao a aren bt net i+? (0, 0, . ios : a +6=08 = 5 a ut end is therefore a basis | Soa Hence, immedia a=0=h=% cup the theorem referred to imm take Wena ci he definition of a We shall now | m gives rise to t This theores nition 3.6.1. some authors (see Remark 3.6.9). — mm Ina vector space V let i 1 oe Pilon two conditions are equivalent : Ai, Yay vsuy Pn} iS @ linearly ea a (b) If VEY, then the expression vy = %¥, Bi unique. _ Proof: Assume (a). We | ae terms of ¥;, ¥%, ..., Y» is unique. J Be 0 vee and also De ce ial BMH OVe to. boty = or oc + @. = By, = But the v/s are u1, so. — 8, = 0 hence expression () is unique. Conversely, assume (b). Suppo now that aretp. Then there exists a itrivial linear ¢o) Bis + Bate +. + Bava, i which equals the zero vector. But 0, on the other hand, is alre, ‘6 the trivial linear combination : : Oy + Oy +... Thus, we get two different expr (0). Hence, fy, », all i. This gives a, — the ¥,'s are not 11, mbination, say Hence, the followin sion ii i conelusio: bt 3 vector space y 1B) = V and the ere a of elements, Of Bis uni3.6 DIMENSION AND BASIS | 99 Definition Let B = {v,, v4, ..., ¥,} be an ordered basis for ¥. Then a vector v G V can be written as vy = ayy + avy + oe + Onn The vector (2%, 25 +5 a,) is called the coordinate vector of v relative to the ordered basis B. It is denoted by [v]p. X45 yy vey mn are called the coordinates of vector y relative to the ordered basis B. coordinates of a vector relative to the standard basis are simply . the coordinates of the vector. “ [tshould be noted that [lp is unambiguously fixed in view of Remark 3.6.9. pxample 3.39 Find the coordinate vector of the vector (2, 3, 4, —1) of ‘Yq relative to the standard basis for Vj. "The standard basis for V4 is fe, es, &, ey}, Where &: = (1, 0, 0, 0), 4= (0, 1,0, 0), es = (0,0, 1,0), and a =, 0, 0,1). Obviously, 34, —1) = 2 + 3e, + 4ey — ley. So the coordinate vector of. 3,4, —I) relative to the standard basis is (2, 3, 4, —1), Therefore, 3,4, and —1 are the coordinates of the vector (2, 3, 4, —1). " Tf'the reader finds that this example is not very illuminating, he should udy the following example. ample 3.40 Find the coordinates of (2,3, 4, —1) relative to the dered basis B = {(1, 1, 0, 0), (0, 1, 1, 0), (0, 0, 1, 1), (1, 0, 0, 0)} for M.. (he reader can verify for himself that B is a basis for V,.) To get the ordinates of (2, 3, 4, —1) relative to B, we write 93, 4, —1) = a(1, 1, 0, 0) + 800, 1, 1,0) + (0, 0, 1, 1) + 8(1, 0, 0, 0) 2,3,4,-D=@+8a+Bb+y¥,¥)- gives «+ 5=2,2+8=3,0+r=—47 Solving these fons, we get «= —2,8=5, y= —1,and5—4. Hence, the co- inates of (2, 3, 4, —!) relative to the ordered basis B are —2, 5, —1, 4. (—2, 5, —1, 4) =[(2, 3,4, —Dle- ¢ now prove a theorem on the extension of a linearly independent set for a vector space V. [Theorem Let the set {V,, Ys, ««-» Yx} be @ linearly independent subset of an n-dimensional vector space V. Then we can find vectors Ym in V such that the set (Vy, 5 Vi Vets ++» Yn} is a basis of: By Theorem 3.6.2,k Pat DS ve =0- ®) a jertl isrt¢] m & an+ 2 Puma 2 wm @) fertl darth = v, say. y is in U, because the left-hand side of Equation @) isin U. v W, since the right-hand side of Equation (9) is in W. Thus, Y, Therefore, vy can be expressed uniquely in terms of _ (10) r y= E 5M i=l bi Hence, : i be + 5 ym =0- ay i jsrtl et102 / VECTOR SPACES So each of the 8/'S and +s is , 10 in Equation (9), We find tha; Wy) is er But {045 Voy 05 Ves Weta Substituting Yr = Yr#2 = r m Sant 3. Pith =O ( +1 ist t= Again, {¥iy 0:5 Yr Metis cos Hm} i8 LI. h Thus, Equation (8) implies that each scalar inva) LI, which proves (a). To prove (b), let z€U+W. Then 2= 4 w EW. This gives So each of the a's and 2's is ,, ved is zero. Hence, , 4 w, where wEU , r m r ee, 7 in tH Bas (3 z= Bam + 2 Baw HEE Pe implifying expression (13), we for suitable scalars a's, 6,'s, a's, B's. 8 a (1 quality is obvi that 2 [A]. Hence, U + WC [A]. The reverse in ¢ because ACU + W.§ 3.6.15 Corollary If U and W are subspaces of a finite-dimensional vec space V such that U 1 W = {0}, then dim (U ® W) = dimU + dimW . The proof is obvious and is left to the reader. As illustration, take U = the xy-plane and W =the yz-plane in | Clearly, U and W are subspaces of V, and dim U = 2, dim W > U A W = y-axis, whose dimension is 1. By Theorem 3.6.14, we have dim (U + W) = dim U + dim W — dim(U NW). But U + W, in this case, is Vs. So dim V, = (xy-plane) + dim (yz-plane) — dim (y-axi or 3=2+42-1. oa This verifies the result of Theorem 3.6.14. On the other hand, if we take U = xy-pl ” = xy-plane and W UW = 0} andU+W=Y,. Also, dim V, = dim (U @ W) = dim U 4 i or Beet a": This verifies the result of Corollary 3.6.15. 5 Now we give a comprehensive example which illusr, studied in this chapter. Ple which illustrates different ide: Example 3.43 Take V = 9,, the s i . space of most 3. Let U be the subspace of 9, Sa that vanish at x= 1. Let W be the sub olynomials of /: polynomials of P; whose first derivatives ONsisting of thos’ spaces U, W, U 1 W, U + W in terms of _ Study the sub- sion of each of these bases to a basis nd the exten” = z-axis, then46 Dimenston any paste | 103 for an arbitra ; TY polynomi a Then fal. Recall that R stands for the w deatee of p < 4) mm (a + Bx 4 xt LdimVm ds. Aig | ISB LOE Ry. =P EV) pl) = = fe + Bx + yx? 4 5, ehrsen let B+ y 48 =O; mee — y — 8) + 2 BR + rit ty ee TEED that the set {x — ], x2 4 ; ae thi *, mS lye ee 7 P i ec it also forms a basis 108 % anno a is, Rietecin, iy a basis for V = 9, by taking ansital Be bratct: ny a aa Polynomials. Such a polynomial Misa basis for —'9,, pe Theorem 3.6.7, {1, x —1, EV ip) = j Bebe yx? + 8x7 | py r 2y + 35 = 0;0,8,7,5€ R) t x(x* = 23) + B(x? — 32) | a, 1,8 E R}. ; ees by {1, x° — 2x, x* — 3x}. Since this can be it follows that {1, x* — 2x, x* — 3x} is a basis for W and nsion of this basis to a basis for 5 is i ji oo x* — 3x} is a basis for 3. Again, ; = {p € Ps | Pl!) = 0, P'() = fe + Bx + yx? + Bx8|a+bB+y+8=0, F B+ 2y +36 = 0;4,8,7,5€ RB} = (1 — 2x + x*) + 822 - 3x +») |y5E R}- + x* and 2 — 3x + x* are Li. So dimU 1 W =2 and — 3x + x3} is a basis for UN W. Now we look for enot in UW. One such element is x*. Now consider linearly independent set {1 — 2x + x2, 2 — 3x + 3%, x7}. is still not V, because dim V=4. So look for one this span. One such element is 2°. (Why? Can Hence, al 2x + x*,2 — 3x Finally, p+q|peu,ge™ ce D+ee—D +h aa I pyte — 23) + Pa(et — 32) en ge oo Pr PP : + x3, 9°, 4}# — 3x)} spans et = 28), rs vel _ wnaw enh 4 dim W — dim (UV 90, BE yy w= aim U + dim —- Han 3.59 to the ordered set , js We ar tet apaey — 3x}, which is LD, we xt — 2x, meus |b? 1, x — 1) for ‘df anv ame | Oe = We? — 1) — 2x = 1) — 1 8 left with the set iscarded. We are thus pace oor" fe pet — 1, x* = 1, x! — axy, ¥ ats, This is still LD (since dim (U + W) =4). Ag ian 359, there exists a vector that is a linear combination of Sy predees sy “Tecannot be 1 or x — 1 or x* — 1 or x? — 1, since th RMefe Ghemscives.. So let us check whether 2° — 3x ic iereomblistion ofits predecessors. We have x — 3x = 1(x* — 1) + O(2* — 1) — 3(x — 1) — 1). Renee 5x © [I,x —1,x*— 1, x? — 1}. So x? — 3x can be d carded. We are thus left with the set De 531, — 15, Which is obviously a basis for U ++ W, Since dim (U + W) = dim v — 4 we get U + W = V, by Theorem 3.6.13. Problem Set 3.6 1 Which of the following subsets $form a basis for V2 Incase 5 is not a basis for Vs, determine a basis for [S]. (a) S={(1, 2, 3), 3,1, 0), (—2, 1, 3)} MOS = A(111), (1,2,-3), (—1, 0, 1} (©) S = {(0,0, 1), (1, 0, 1), d, =1, 1), (3, 0, 1 @ s={a, 2/5, —),(, 1, 2), (3/4, —1, 1} © S= {(-1, 3/2, 2), (3, 2/3, 3)}. 2. Which of the followin, space V ? @ $= {(1, -1,0, 1), 0,0, 9, 0), V=Y, iB subsets S form a basis for the given vector @ 1,0, 1), 3,2, 1, 0}, ‘1 (b+) s {©, 1, 2, 1), (1,2, -1, 1), eee coe )» 2, ~3, 1, 0), © oe ee Iwe+x—12_ @) S$ ={1, x,(« — 1x, x(x — 1 {l, x, Bx? — 1)/2, (533, V=9, © s= ue — 2, (% — 22, 4, —2, 7,5, Pao.’ — 9, IO Wey a 2, ae —3x)/2, (35x4 Oe 2), = Pshe 3.6 DIMENSION ANE masts / 105 o ae Whe lan, a ehh, g, oh 1 vy e the dimension of the subspace ta ot eae S in * = (1, sin x, sint the dim ; limension of the Subspace [5] of V for each S in e set {(3, —1, 2)} to two different bases for Vs _e B= 1.2.3), 0,1, 2 =2), (1, —3, 4), “detepnie? 2,9), 0d Sy = (1, —2, 9), , eo, V and no proper subset of S spans V, then S is a WE Ps\P"=0},V= 9, PE Ps | P(x) = p'(x,) = 0}, v = MN ela ee ay any n scalars}, V = V,, ae ar ie (x1, Xo, Xa5 i+ % +x, = {(%1, %2, Xa» ay Xs) © yi gene ge =v, PE Pe | Pm) =9,V = % W be two distinct (n — 1)-dimensional subspaces of an mal vector space V. Then prove that dim (UM W) that every 1-dimensional subspace of V, is a straight line sh the origin. that every 2-dimensional subspace of Vs is a plane the origin. that the intersection of two distinct planes through the sa straight line through the origin. oe — 0 6;,1 = 1, 2, ---) and for a fixed k( Yeas Vitae “ts for U of Problem 8(c). i tes of the following vectors of V; relative to the is B = (2, 1,0), 2,1, 1), @ 2 DI: é ) (130) (c) rs Xm %9) () @9,—). me) (e) (—I/2, 11/3, 5)16. 18. a5) Mind the Ordered by (a) wernates of the fol 18 sae FP nd at | i. at Ms “Ind an r “hel he (18). Ordered basis for Ve relati q Mis the | % 2, 1) has the coordin ‘41,2, and Let B, May Y, F + Yay ing polynomials relatiy, axt'gf e to 1, Vecto, = {Uy, yy «5 Uy} be two Ordereg Onal vector space V and fet = Saath eye el ieee ve = Ganda “b tga) He RE Chagas Then find the Coordinates of a vector y € FV relative to the basis p, if its coordinates relative to the basis By are x4 %g) «++» Xn Find the dimension of vP Construct two subspaces 4 and B of V, such that dim 4 dim B= 3, and dim 4 B= 1. Let By = (uy, 4, ..., un} and By = fy, for an n-dimensional vector space V su Un, — Vn} is LD. Then prove that there such that [uly, = (ulp,. True or false 2 (a) Every vector space has a finite basis. (b) A basis can never include the zero vector. ©) _ If two bases of V have one common vector, tl are the same. j (d) Ifevery set of p vectors of V withp > ni set of generators. (e) A basis for V; can be extended to a basi (f) A basis for Vs is @ +7 +k, i+ 4, i} (g) A basis for Py is {1, 2x, (x — 17}. (hb) A set of generators of Ps is {@ = D5 @ = D4 (*- 1, 8 i) PJs is independent of B, ‘ i) 9), (0,0, 0) is a 2, Yay «s+» Yn} be ordered bases ich that {4; — %4, wy — »,, .., exists a nonzero vector u € )chapter 4 tinear Transformations The significance of vector spaces arises from the fact that we can pass from one vector space to another b: q certain property called linearity, ransformations. y means of functions that possess These functions are called linear 4.1 DEFINITION AND EXAMPLES 41.1 Definition Suppose U and V are vector spaces either both real or both complex. Then the map 7: U -+ V is said to be a linear map (transformation, operator), Tuy + uw.) = T(us) + T(us) for all u,, € U a) and T(eu) = aT(u) for all u € U, and all scalars «. (2) A linear map T : U -> U is also called a linear map on U. Whenever we say T: UV is a linear map, then U and V shall be taken as vector spaces over the same field of scalars. 41.2 Remark In Definition 4.1.1 the ‘plus’ in m4 + m denotes the addi- tion in space U, and the ‘plus’ in T(#,) + T(u,) denotes the addition in space V. We shall not elaborate these delicate points in the sequel. A similar remark is true for the two scalar multiplications implied in Equation (2). le 4.1 Define T: Vs —> Vg by the rule T(z, Xz, Xs) = (Xv ¥2, 0) - 'is called the projection of V3 on the x,x2-plane. To prove that it is a map, we have to show that T(x + y) = Tex) + T(y) and T(x) = aT (x) all x, »y © Vand all scalars «. Let x = (1, %2 ¥3) and y = (3, ya, ys). x+y (1 + Vr» %2 + Yas Xs + Ys) (ax, 2X3, 2X3) « i Hl| LINEAR TRANSFORM ATIONS iow, by definition of T, we have Tix + y) = TH On the other hand, T(x) + TO) = Te» =i Ee Id gh the verification4.1 DEFINITION AND EXAMPLES | 109 T: U+ VU and V being vector spaces) by the ‘This isa fifte ca * is called because tor to the zero vector of ¥. a. P Let U be a vector Space. The identity mop ly: U -» UO 5 rule Tou) =u F Define T: Vv, + V, by the rule Tix 2) = (x, —x,) . ) map and is called the reflection in the x-axis (Figure 4.1). a). inear, cation ace in Would ae xix) Ficure 4.1 Define D: G'(a, 6) > & (a, b) by the rule Df)=f', ie derivative of f. The facts that D(/ + g) = (f+ 9)’ =f’ +8 af) = (af) = af’ are elementary but important results in _view of these results, we sce that D is linear. This trans- ‘is called the differential operator. Define J : & (a, b) > R by the rule 6 Sf) = SIG) ax. a pate properties of integral, we find that J is a linear map. will - U + U by the rule T(x) = x + This map T is not linear, because = (x+y) + OF et +0 + w= TH) +70). translation by the vector ty. Suppose 7/0 is a fixed vector of a vector space U. (RR defined by f(x) = x + a (‘a' fixed) sause its graph in the xy- i : if ils4.1 DEPTErTON PCM LE eyHEU Sosa r = = Which is just & verification of the two eee f+ TW) and Ten) = ary rary and all scalars «Let w, ¥ botwe totter of w= Om, +... "Satay ¥ am Bits + ok Buiny a, /the definition hive sive ee Te +) = (4. + By 4 4 (an + By, Se CaM t+ oat) ot Bie, ho, BS G+ Bor to. (eq oa Dy, Tu) + TW). Again,” BACH) Oty, t,t cary = (Gy, Fk an) =oMu). and the theorem is proved. __ Theorem 4.1.5 will be used in the foliowing way : To de- Jinear map, very often we shall be content with listing (u:), .- » T(wn). This means the value of T on a general u is obtained by the process shown in the theorem, namely, if Ue Oy + Oty +... + atta , T(w) = %T (uy) + oT (uy) +... + ay T (Un) « ss of defining T on the basis and extending it to the re- g elements is called linear extension cf T. So we shall simply fine T on a basis and extend it linearly’ or, equivalently, the linear map T by specifying its values on a basis’. Suppose we want to define a linear map T:V,—>V4. in V,, In fact, every ordered pair of vectors in V, will give ‘map T. We shall cite a few in Table 4.1. TABLE 4.1 0,1,0,0) (1,110 (1, 1,0,0) (0, 0, 0, 0) (0, 0, 0, 0) (0, 0,0, 0) (1, 0, 0, 0) ay a ed alatefp TRANSFORMATIONS of these maps (0 any u @ 112 / wine es of w relat sy extend each f M1 depend on the coordinat actual vale basis (1, Oy = Let ge BE Ve a4) L spat 3 Then -y A, = w= a, -) Thus, we get xt) a -—2 (0, 0, 0, 9 + zx ‘Therefore, T Bine zero map. To») == TUL D+ == 0,1,0,.0 +7570, 00,0 =» 0,0). T,(x, y) = Bear, Nene (t,t) oat i a, aay ene WII)» Ti Gay) = et D $2 1,1) +252 ry, —1) 9) (-1, -1, - * ©, 0, 0,0)\ at DEFINITION AND ExAMPLES / 3 Po XPLX) + pci) d by TY) a (FO, fay = 70) PTiP +9 defined by Tp) 'P + F defined by Tiny S 9 + 3x + 7x1 p(x) Bee 2 2 teed by oy) 0) + 90) + 2 rey )T:9 > ode ery d TE "a, b) © (a, 6) dette *Y TU) = aft af + .,, anf, a's are fixed scalars @ 7:7 a, 5), Sa, 8) defined b eet (7x 4. DS + 3x 4 Sf. rmine whether there ex; it does exist give ; Va — Vz such that 22) GAG) ana (2, 1) = (1, 2y Mees > VY, such that 7(2, 1) (2, 1) and TA, 2) = (4, 2) 2¥, -> Vz such that TO, 1) = 3, 4), 7,1) = (2, 2), and 3, 2) = (5, 7) Pa — Va such that 7(0, 1, y= 12, 2, 2) Ps Ps such that 7 4 3) — 1 a, TC) % + 3x*, and 7(x*) — 9 Pa > Ps such that T+ x) =1, T(x) = 3, and Te) 4 VE > VS such that TMi) = (1 + 4,1), @ nonzero linear transformation T : V, > V2, which maps tors on the line x = J onto the origin, a linear tranformation TV. V, which ™maps all the On the line x + » = 0 onto themselves (T 4 D. > VE be defined by 7a, + i, ay + 1B.) = (a, ), rove or disprove that T is linear. ‘ : t ‘at a linear transformation on a I-dimensional vector space is it multiplication by a fixed scalar. m 4.1.4. e? exists a linear transformation T:V, > V, such that ) 0, 0). _ ion is the only linear transformation from Ye ¥ MP) = (3x84 aye (3, 1, 2) and 71, tN); defined as T(p(x)) = x(x) is not a linear transfor- ae rat sts a linear map in the following c: or ») the general formula, ae(% Y) H> (x', +y cos 0, i. * 2 of Problem Set hen 7(0, 0, 0) — from the real vect,,4.2 RANG: NGE AND KERNEL OF A LINFAR MAP / 115 vector (4 4) in Vz and solve the equation Ot ~ 5 + 5) ay, quinmeams %1 — %2 = aand x, 4. <5 I -. sft 4 X= b — x, ys 1 — 0, ) 7 . rence, TI. % *1) = (2, ). This show. BEML RO 10 other words, mer) v, “a eM Yet map. To determine the kernel, we solve the equation TX1, X45 X5) = (x, Mts Nay Xq) = (xy = s) = (0,0). This a. eA t+ 9) = (0, 0). Bives x, = xy ay i.e. all form (%» %1» —%) will be mapped into zero. $6 “se Breo= (2,01, 1, 1) |, any 4 y= N)ilie any soataey me - This is the subspace of Vy generated by (1, i aan Brample 4.13 For the zero map (see Example 4 4) the ares Fe 8 4 j - See Ltsisclearly not an onto map. Oe a Example 4.14 For the identity map ty (sce Example 4.5) the kernel is {0}. The range is U, so the map is onto, Solving these, we get Example 4.15 In Example 4.6 7: y, > Vy is defined by T(x, x2) + @y—%). Here R(T) = ¥, and N(T) = ((0, 0)}. This is an onto map. Example 4.16 In Example 4.7 D: 711%, b) E(w ) is defined by D(f)=f. In this case R(D) @ (a, ), since every continuous function £ on (a, b) possesses an antiderivative and hence D is an onto map. N(D) is the set of all constant functions in E a, b). Example 4.17 In Example 4.8 Gab) Ris defined by SA = I fede. Here the range is the whole of R, since every real number can be obtained as the algebraic area under some curve y = f(x) from a to b, Therefore, it is an onto map. The kernel is the set of all those functions F for which the area under the curve y = f(x) from a to b is zero. It is difficult to say anything more than this about the kernel, We shall now check whether the linear transformations discussed in Examples 4.11 to 4.17 are one-one (cf 1.4.7 for the definition of one-one), fn Example 4.11 N(T) is the x-axis. So all points on the X,raxis go into (0, 0,0). So this map is not one-one. In Example 4,12 MT) =[(1, 1, —1)]. So, many points go into (0, 0,0). This again means Tis not one-one. In Example 4.13 M(T) = U. So the zero map is not one-one, because all elements go into the zero of V. In Example 4.14 the identity map is one-one, because if x # y, then certainly I(x) # I(y). It may be noted that N(J) is the zero subspace ofice. Hence, Tn the same because U fies w=" | *his mea" ™ ge belonss mothing that T is 00° 7 TW) =42 RANGE AND KERNEL oF 4 LINEAR MAP / 117 ={0u}. Sou yo Wyieuay, This Proves that T hen each vector u can be expressed as “1; Way sy tp. The vectors Tu), S1Y5 (Tt), Tt) soeeg Te)] C R(T). ‘oru€ Usuch that T(u) =y, Since StL tly ba Let, TOs + ay, + ob ase) STC) + oT) + on. + cig Trg) « +, Tlun)]. This proves that RT) = [T(u), Teu,), ..., T(un)] « he proof is left to the reader. § conclude this article with a definition, Definition Let T: U->V be a linear map. Then If R(7) is finite-dimensional, the dimension of R(Z) is called the € of T and is denoted by r(7). If N(T) is finite-dimensional, the dimension of N(T) is called the ullity of T and is denoted by n(T). e shall study these concepts at length in § 4.3. Problem Set 4.2 4 ine the range of the following linear transformations. Also nd the rank of T, where it exists. T: Vz V, defined by (x1 2) = (41 + Xa 4) T: V, > V, defined by T(x, X2) = (%1, %1 + 5 %2) T: V, > V, defined by T(x, %2) Xs) = (4%, + X2 + Xa, 2, — Ix, x5) T: Vs a V; defined by T(x1s Xa Xs) = (1 Xa5 2) T:V, + V, defined by TX: Xa Xa Xa) = (%1 — XM + 5, X3— X, © 7: Vy 2 V, defined by T(x: Xx, %3) = Gy 1 + Xap My + Xe Dt %, x) T: Vy V, defined by T(x: 2 Xa» ¥0) = Bx + 2Xy 1 = Xap 4x, — xq x) T: > G defined by Tp)(x) = 2702) T: 9 > # defined by (r(x) = 3P') T: P+ F defined by T(p)(*) =P’) — 2p) T: G0, 1) +E (0, 1) defined by T(/1x) = (2) sin x T: GMO, 1) > Y (0, 1) defined by T A(x) = f'(x)e*. Determine the kernel of the linear transformations of Problem !+ (1). Also find the nullity of 7, where it exists.