Student’s Manual Further Mathematics for Economic Analysis nd 2 edition Arne Strøm Knut Sydsæter Atle Seierstad Peter Hammond For further supporting resources please visit: www.pearsoned.co.uk/sydsaeter Preface This student’s solutions manual accompanies Further Mathematics for EconomicAnalysis (2nd edition, FT ⊂⊃ PrenticeHall,2008). Itsmainpurposeistoprovidemoredetailedsolutionstotheproblemsmarkedwith SM inthetext. TheManualshouldbeusedinconjunctionwiththeanswersinthebook. Insomefewcasesonly apartoftheproblemisdoneindetail,becausetherestfollowsthesamepattern. Attheendofthismanualthereisalistofmisprintsandothererrorsinthebook,andevenoneortwoin the errata list in the preliminary and incomplete version of this manual released in September this year. We would greatly appreciate suggestions for improvements from our readers as well as help in weeding out the inaccuraciesanderrorsthatprobablyremain. OsloandCoventry,October2008 ArneStrøm ([email protected]) KnutSydsæter ([email protected]) AtleSeierstad ([email protected]) PeterHammond ([email protected]) Contents 1 TopicsinLinearAlgebra ............................................................ 1 2 MultivariableCalculus ............................................................. 12 3 StaticOptimization ............................................................... 23 4 TopicsinIntegration .............................................................. 38 5 DifferentialEquationsI:First-OrderEquationsinOneVariable............................. 43 6 DifferentialEquationsII:Second-OrderEquationsandSystemsinthePlane .................. 49 7 DifferentialEquationsIII:Higher-OrderEquations ...................................... 56 8 CalculusofVariations ............................................................. 60 9 ControlTheory: BasicTechniques ................................................... 65 10 ControlTheorywithManyVariables.................................................. 81 11 DifferenceEquations .............................................................. 92 12 DiscreteTimeOptimization ......................................................... 96 13 TopologyandSeparation .......................................................... 102 14 CorrespondencesandFixedPoints .................................................. 105 A Sets,Completeness,andConvergence ................................................ 107 B TrigonometricFunctions .......................................................... 109 Correctionstothebook ........................................................... 112 Version1.0 28102008 879 ©ArneStrøm,KnutSydsæter,AtleSeierstad,andPeterHammond2008 CHAPTER 1 TOPICS IN LINEAR ALGEBRA 1 Chapter 1 Topics in Linear Algebra 1.2 ⎛ ⎞ 1 2 0 1.2.3 Let A = ⎝0 1 1⎠ be a matrix with the three given vectors as columns. Cofactor expansion of 1 0 1 |A|alongthefirstrowyields (cid:6) (cid:6) (cid:6) (cid:6) (cid:6) (cid:6) (cid:6) (cid:6) (cid:6)1 2 0(cid:6) (cid:6) (cid:6) (cid:6) (cid:6) (cid:6)(cid:6)0 1 1(cid:6)(cid:6) = 1(cid:6)(cid:6)1 1(cid:6)(cid:6)−2(cid:6)(cid:6)0 1(cid:6)(cid:6)+0 = 1−2(−1) = 3 (cid:4)= 0 (cid:6) (cid:6) 0 1 1 1 1 0 1 ByTheorem1.2.1thisshowsthatthegivenvectorsarelinearlyindependent. 1.2.6 Part (a) is just a special case of part (b), so we will only prove (b). To show that a , a , ..., a are 1 2 n linearlyindependentitsufficestoshowthatifc ,c ,...,c arerealnumberssuchthat 1 2 n c a +c a +···+c a = 0 1 1 2 2 n n thenallthec havetobezero. Sosupposethatwehavesuchasetofrealnumbers. Thenforeachi = 1, i 2,...,n,wehave a ·(c a +c a +···+c a ) = a ·0 = 0 (1) i 1 1 2 2 n n i Sincea ·a = 0wheni (cid:4)= j,theleft-handsideof(1)reducestoa ·(c a ) = c (cid:5)a (cid:5)2. Hence,c (cid:5)a (cid:5)2 = 0. i j i i i i i i i Becausea (cid:4)= 0wehave(cid:5)a (cid:5) (cid:4)= 0,anditfollowsthatc = 0. i i i 1.3 1.3.1 (a) Therankis1. Seetheanswerinthebook. (cid:6) (cid:6) (cid:6) (cid:6) (b) Theminorformedfromthefirsttwocolumnsinthematrixis(cid:6)(cid:6)1 3(cid:6)(cid:6) = −6 (cid:4)= 0. Sincethisminor 2 0 isoforder2,therankofthematrixmustbeatleast2,andsincethematrixhasonlytworows,therank cannotbegreaterthan2,sotherankequals2. (cid:6) (cid:6) (c) Thefirsttworowsandlasttwocolumnsofthematrixyieldtheminor(cid:6)(cid:6)(cid:6)−1 3(cid:6)(cid:6)(cid:6) = 5 (cid:4)= 0,sotherank −4 7 of the matrix is at least 2. On the other hand, all the four minors of order 3 are zero, so the rank is less than3. Hencetherankis2. (Itcanbeshownthatr = 3r +r ,wherer ,r ,andr aretherowsofthe 2 1 3 1 2 3 matrix.) Analternativeargumentrunsasfollows: Therankofamatrixdoesnotchangeifweaddamultiple ofonerowtoanotherrow,so ⎛ ⎞ ⎛ ⎞ 1 2 −1 3 −2 1 1 2 −1 3 ⎝ 2 4 −4 7⎠ ← ∼ ⎝0 0 −2 1⎠. −1 −2 −1 −2 ← 0 0 −2 1 ©ArneStrøm,KnutSydsæter,AtleSeierstad,andPeterHammond2008 2 CHAPTER 1 TOPICS IN LINEAR ALGEBRA Here∼meansthatthelastmatrixisobtainedfromthefirstonebyelementaryrowoperations. Thelast matrixobviouslyhasrank2,andthereforetheoriginalmatrixalsohasrank2. (cid:6) (cid:6) (cid:6) (cid:6) (cid:6)1 3 0(cid:6) (d) Thefirstthreecolumnsofthematrixyieldtheminor(cid:6)(cid:6)2 4 0(cid:6)(cid:6) = −4 (cid:4)= 0,sotherankis3. (cid:6) (cid:6) 1 −1 2 (cid:6) (cid:6) (cid:6) (cid:6) (e) (cid:6)(cid:6) 2 1(cid:6)(cid:6) = 9 (cid:4)= 0,sotherankisatleast2.Allthefourminorsoforder3arezero,sotherankmust −1 4 belessthan3. Hencetherankis2. (Thethreerows,r ,r ,andr ,ofthematrixarelinearlydependent, 1 2 3 becauser = −14r +9r .) 2 1 3 (f) Thedeterminantofthewholematrixiszero,sotherankmustbelessthan4. Ontheotherhand,the firstthreerowsandthefirstthreecolumnsyieldtheminor (cid:6) (cid:6) (cid:6)(cid:6) 1 −2 −1(cid:6)(cid:6) (cid:6)(cid:6) 2 1 1(cid:6)(cid:6) = −7 (cid:4)= 0 (cid:6) (cid:6) −1 1 −1 sotherankisatleast3. Itfollowsthatthematrixhasrank3. 1.3.2 (a) Thedeterminantis(x +1)(x −2). Therankis3ifx (cid:4)= −1andx (cid:4)= 2. Therankis2ifx = −1 orx = 2. ⎛ ⎞ x 0 x2 −2 (a) By cofactor expansion along the first row, the determinant of the matrix A = ⎝ 0 1 1 ⎠ −1 x x −1 is |A| = x ·(−1)−0·1+(x2 −2)·1 = x2 −x −2 = (x +1)(x −2) If x (cid:4)= −1 and x (cid:4)= 2, then |A| (cid:4)= 0, so the rank of A equals 3. If x = −1 or x = 2, then |A| = 0 and r(A) ≤ 2. Ontheotherhand,theminorwegetifwestrikeoutthefirstrowandthethirdcolumninAis (cid:6) (cid:6) (cid:6) (cid:6) (cid:6)(cid:6) 0 1(cid:6)(cid:6) = 1 (cid:4)= 0forall x,sor(A)canneverbelessthan2. −1 x (cid:7) 2 ifx = −1orx = 2 Conclusion: r(A) = 3 otherwise (b) Alittlecalculationshowsthatthedeterminantofthematrixist3+4t2−4t−16,andifwenotethat thisexpressionhast +4asafactor,itfollowsthatthedeterminantis t3+4t2 −4t −16 = t2(t +4)−4(t +4) = (t2 −4)(t +4) = (t +2)(t −2)(t +4) Thus,ift doesnotequalanyofthenumbers−2,2,and−4,therankofthematrixis3. (cid:6) (cid:6) (cid:6) (cid:6) Ifwestrikeoutthesecondrowandthefirstcolumnofthematrix,wegettheminor(cid:6)(cid:6)5 6 (cid:6)(cid:6) = 5t+14, 1 t +4 which is different from 0 for all the three special values of t that we found above, and thus the rank of thematrixis (cid:7) 2 ift = −4,−2,or2 3 otherwise (c) Thefirstandthirdrowsareidentical,asarethesecondandfourth. Butthefirsttworowsarealways linearlyindependent. Sotherankis2forallvaluesofx,y,z,andw. ©ArneStrøm,KnutSydsæter,AtleSeierstad,andPeterHammond2008 CHAPTER 1 TOPICS IN LINEAR ALGEBRA 3 1.4 1.4.2 (a) Itisclearthatx = x = x = 0isasolution. Thedeterminantofthecoefficientmatrixis 1 2 3 (cid:6) (cid:6) (cid:6)(cid:6)1 −1 1(cid:6)(cid:6) D = (cid:6)(cid:6)1 2 −1(cid:6)(cid:6) = 9 (cid:4)= 0 (cid:6) (cid:6) 2 1 3 Hence, by Cramer’s rule the solution is unique and the system has 0 degrees of freedom. (This agrees withTheorem 1.4.2: since the rank of the coefficient matrix is 3 and there are 3 unknowns, the system has3−3 = 0degreesoffreedom.) (b) By Gaussian elimination (or other means) we find that the solution is x = a, x = −a, x = −a, 1 2 3 andx = a,witha arbitrary. Thus,thereisonedegreeoffreedom. (Wecouldgetthenumberofdegrees 4 of freedom fromTheorem 1.4.2 in this case, too. The minor formed from the first three columns of the coefficient matrix has determinant −3 (cid:4)= 0, so the rank of the coefficient matrix is 3. Since there are 4 unknowns,thesystemhas4−3 = 1degreeoffreedom.) 1.4.3 Thedeterminantofthecoefficientmatrixisa2−7a = a(a−7). Thus,ifa (cid:4)= 0anda (cid:4)= 7,thesystem hasauniquesolution. Ifa = 0 ora = 7, therankofthecoefficientmatrixis2(why?), andthesystem eitherhassolutionswith1degreeoffreedomorhasnosolutionsatall,dependingonthevalueofb. One wayoffindingoutisbyGaussianelimination,i.e.byusingelementaryrowoperationsontheaugmented coefficientmatrix: ⎛ ⎞ ⎛ ⎞ 1 2 3 1 1 −3 1 2 3 1 ⎝−1 a −21 2⎠ ← ∼ ⎝0 a+2 −18 3 ⎠ ← 3 7 a b ← 0 1 a−9 b−3 −(a+2) ⎛ ⎞ 1 2 3 1 ∼ ⎝0 0 −a2 +7a −ab−2b+3a+9⎠ ← 0 1 a−9 b−3 ← ⎛ ⎞ 1 2 3 1 ∼ ⎝0 1 a−9 b−3 ⎠ 0 0 a(7−a) −ab−2b+3a+9 Thisconfirmsthataslongasa (cid:4)= 0anda (cid:4)= 7,thesystemhasauniquesolutionforanyvalueofb. Butif a = 0ora = 7,thesystemhassolutionsifandonlyif−ab−2b+3a+9 = 0,andthenithassolutions with1degreeoffreedom. (Therankofthecoefficientmatrixis2andthereare3unknowns.) Ifa = 0,thenthesystemhassolutionsifandonlyifb = 9/2. Ifa = 7,thenthesystemhassolutionsifandonlyif−9b+30 = 0,i.e.b = 10/3. 1.4.6 (a) |A | = (t −2)(t +3),sor(A ) = 3ift (cid:4)= 2andt (cid:4)= −3. Becausetheupperleft2×2minorof t t At is−1 (cid:4)= 0⎛,the⎞rankofAt canneverbelessthan⎛2,so⎞r(A2) = 2,r(A−3) = 2. x 11 1 (b) Letx = ⎝x2⎠. ThevectorequationA−3x = ⎝ 3⎠isequivalenttotheequationsystem x 6 3 x + 3x +2x = 11 (1) 1 2 3 2x + 5x −3x = 3 (2) 1 2 3 4x +10x −6x = 6 (3) 1 2 3 ©ArneStrøm,KnutSydsæter,AtleSeierstad,andPeterHammond2008 4 CHAPTER 1 TOPICS IN LINEAR ALGEBRA Equation(3)isobviouslyequivalentto(2),sowecanremoveit,andthenweareleftwiththetwoequations x +3x +2x = 11 (1) 1 2 3 2x +5x −3x = 3 (2) 1 2 3 Wecanconsidertheseequationsasanequationsystemwithx andx astheunknowns: 1 2 x +3x = 11−2x (1(cid:9)) 1 2 3 2x +5x = 3+3x (2(cid:9)) 1 2 3 For each value of x this system has the unique solution x = 19x −46, x = −7x +19. Thus the 3 1 3 2 3 vectorequationA x = (11,3,6)(cid:9) hasthesolution 3 x = (19s −46, −7s +19, s)(cid:9) wheres runsthroughallrealnumbers. 1.5 1.5.1 Forconvenience,letA,B,...,Fdenotethematricesgivenin(a),(b),...,(f),respectively. (cid:8) (cid:9) 2 −7 (a) ThecharacteristicpolynomialofthematrixA = is 3 −8 (cid:6) (cid:6) |A−λI| = (cid:6)(cid:6)(cid:6)2−λ −7 (cid:6)(cid:6)(cid:6) = λ2 +6λ+5 = (λ+1)(λ+5) 3 −8−λ sotheeigenvalueso(cid:8)fA(cid:9)areλ1 = −1andλ2 = −5. Theeigenvectorscorrespondingtoaneigenvalueλ x arethevectorsx = (cid:4)= 0thatsatisfyAx = λx,i.e. y (cid:10) 2x −7y = −x ⇐⇒ 3x = 7y forλ = −1 3x −8y = −y and (cid:10) 2x −7y = −5x ⇐⇒ 7x = 7y forλ = −5 3x −8y = −5y (cid:8) (cid:9) (cid:8) (cid:9) 7 1 Thisgivesustheeigenvectors v = s and v = t , wheres andt arearbitraryrealnumbers 1 2 3 1 (differentfrom0). (cid:6) (cid:6) (b) The characteristic equation of B is |B−λI| = (cid:6)(cid:6)(cid:6)2−λ 4 (cid:6)(cid:6)(cid:6) = λ2 −8λ+20 = 0. B has two −2 6−λ complexeigenvalues,4±2i,andnorealeigenvalues. (c) ThecharacteristicpolynomialofCis|C−λI| = λ2−25,andweseeimmediatelythattheeigenvalues areλ = 5andλ = −5. Theeigenvectorsaredeterminedbytheequationsystems 1 2 (cid:11) (cid:11) x +4y = 5x x +4y = −5x 3 ⇐⇒ x = y and ⇐⇒ y = − x 6x − y = 5y 6x − y = −5y 2 ©ArneStrøm,KnutSydsæter,AtleSeierstad,andPeterHammond2008 CHAPTER 1 TOPICS IN LINEAR ALGEBRA 5 respectively,sotheeigenvectorsare (cid:8) (cid:9) (cid:8) (cid:9) 1 −2 v = s and v = t 1 2 1 3 wheres andt arearbitrarynonzeronumbers. (d) ThecharacteristicpolynomialofDis (cid:6) (cid:6) (cid:6)(cid:6)2−λ 0 0 (cid:6)(cid:6) |D−λI| = (cid:6)(cid:6) 0 3−λ 0 (cid:6)(cid:6) = (2−λ)(3−λ)(4−λ) (cid:6) (cid:6) 0 0 4−λ Theeigenvaluesareobviouslyλ = 2,λ = 3,λ = 4,andthecorrespondingeigenvectorsare 1 2 3 ⎛ ⎞ ⎛ ⎞ ⎛ ⎞ 1 0 0 v = s⎝0⎠, v = t⎝1⎠, v = u⎝0⎠ 1 2 3 0 0 1 where s, t, u are arbitrary nonzero numbers. (Note: The eigenvalues of a diagonal matrix are always preciselythediagonalelements,and(multiplesof)thestandardunitvectorswillbeeigenvectors. Butif twoormoreofthediagonalelementsareequal,therewillbeothereigenvectorsaswell.Anextremecase istheidentitymatrixI : all(nonzero)n-vectorsareeigenvectorsforI .) n n (e) ThecharacteristicpolynomialofEis (cid:6) (cid:6) (cid:6)(cid:6)2−λ 1 −1 (cid:6)(cid:6) (cid:6)(cid:6) 0 1−λ 1 (cid:6)(cid:6) = −λ3+λ2 +2λ = −λ(λ2 −λ−2) (cid:6) (cid:6) 2 0 −2−λ Theeigenvaluesaretherootsoftheequation−λ(λ2−λ−2) = 0,namelyλ = −1,λ = 0andλ = 2. 1 2 3 Theeigenvectorscorrespondingtoλ = −1aresolutionsof 1 ⎧ ⎪⎨2x1+x2 − x3 = −x1 (cid:16)x = 1x (cid:16)x = −x Ex = −x ⇐⇒ x + x = −x ⇐⇒ 1 2 3 ⇐⇒ 2 1 ⎪⎩ 2 3 2 x = −1x x = 2x 2x −2x = −x 2 2 3 3 1 1 3 2 ⎛ ⎞ ⎛ ⎞ ⎛ ⎞ 1 1 2 sotheyareoftheformv = s⎝−1⎠. Similarly,v = t⎝−1⎠andv = u⎝1⎠aretheeigenvectors 1 2 3 2 1 1 correspondingtoλ = 0andλ = 2. 2 3 (f) ThecharacteristicpolynomialofFis (cid:6) (cid:6) (cid:6)(cid:6)1−λ −1 0 (cid:6)(cid:6) (cid:6)(cid:6) −1 2−λ −1 (cid:6)(cid:6) = −λ3+4λ2 −3λ = −λ(λ2 −4λ+3) = −λ(λ−1)(λ−3) (cid:6) (cid:6) 0 −1 1−λ The eigenvalues are therefore λ = 0, λ = 1, and λ = 3. By the same method as above we find that 1 2 ⎛ ⎞ 3 ⎛ ⎞ ⎛ ⎞ 1 −1 1 thecorrespondingeigenvectorsarev = s⎝1⎠,v = t⎝ 0⎠,andv = u⎝−2⎠. 1 2 3 1 1 1 ©ArneStrøm,KnutSydsæter,AtleSeierstad,andPeterHammond2008 6 CHAPTER 1 TOPICS IN LINEAR ALGEBRA ⎛ ⎞ ax +ay 1.5.2 (a) X(cid:9)AX = X(cid:9)(AX) = (x, y, z)⎝ax +ay⎠ = (ax2 +ay2 +2axy +bz2) (a1×1matrix), bz ⎛ ⎞ ⎛ ⎞ 2a2 2a2 0 4a3 4a3 0 A2 = ⎝2a2 2a2 0 ⎠, A3 = ⎝4a3 4a3 0 ⎠ 0 0 b2 0 0 b3 (cid:6) (cid:6) (cid:6)(cid:6)a−λ a 0 (cid:6)(cid:6) (b) ThecharacteristicpolynomialofAisp(λ) = (cid:6)(cid:6) a a−λ 0 (cid:6)(cid:6) = (λ2−2aλ)(b−λ),sothe (cid:6) (cid:6) 0 0 b−λ eigenvaluesofAareλ = 0,λ = 2a,λ = b. 1 2 3 (c) From (b) we get p(λ) = −λ3 +(2a +b)λ2 −2abλ. Using the expressions for A2 and A3 that we foundinpart(a),itiseasytoshowthatp(A) = −A3+(2a+b)A2 −2abA = 0. 1.5.4 (a) TheformulainProblem1.9.7(b)yields (cid:6) (cid:6) (cid:6)4−λ 1 1 1 (cid:6) (cid:6) (cid:6) (cid:8) (cid:9) |A−λI| = (cid:6)(cid:6)(cid:6) 1 4−λ 1 1 (cid:6)(cid:6)(cid:6) = (3−λ)4 1+ 4 = (3−λ)3(7−λ) (cid:6) 1 1 4−λ 1 (cid:6) 3−λ (cid:6) (cid:6) 1 1 1 4−λ Hence,theeigenvaluesofAareλ = λ = λ = 3,λ = 7. 1 2 3 4 (b) An eigenvector x = (x ,x ,x ,x )(cid:9) of A corresponding to the eigenvalue λ = 3 must satisfy the 1 2 3 4 equationsystem(A−3I)x = 0. The4equationsinthissystemareallthesame,namely x +x +x +x = 0 1 2 3 4 Thesystemhassolutionswith4−1 = 3degreesoffreedom. Onesimplesetofsolutionsis ⎛ ⎞ ⎛ ⎞ ⎛ ⎞ 1 1 1 x1 = ⎜⎜⎝−01⎟⎟⎠, x2 = ⎜⎜⎝−01⎟⎟⎠. x3 = ⎜⎜⎝ 00⎟⎟⎠ 0 0 −1 Thesethreevectorsareobviouslylinearlyindependentbecauseif ⎛ ⎞ c +c +c 1 2 3 c1x1+c2x2 +c3x3 = ⎜⎜⎝ −−cc1 ⎟⎟⎠ 2 −c 3 isthezerovector,thenc = c = c = 0. 1 2 3 1.5.5 (c) If λ is an eigenvalue for C with an associated eigenvector x, then Cnx = λnx for every natural numbern. IfC3 = C2+C,thenλ3x = λ2x+λx,so(λ3−λ2−λ)x = 0. Thenλ3−λ2−λ = 0,because x (cid:4)= 0. IfC+I didnothaveaninverse,|C+I | = 0. Thenλ = −1wouldbeaneigenvalueforC,and n n sowewouldhaveλ3−λ2−λ = 0,whichisnottrueforλ = −1. Hence−1isnotaneigenvalueforC, andconsequentlyC+I hasaninverse. n ©ArneStrøm,KnutSydsæter,AtleSeierstad,andPeterHammond2008 CHAPTER 1 TOPICS IN LINEAR ALGEBRA 7 1.6 (cid:8) (cid:9) 2 1 1.6.1 (a) LetA = . ThecharacteristicpolynomialofAis 1 2 (cid:6) (cid:6) p(λ) = (cid:6)(cid:6)(cid:6)2−λ 1 (cid:6)(cid:6)(cid:6) = (2−λ)2 −1 = λ2 −4λ+3 = (λ−1)(λ−3) 1 2−λ Thus, the eigenvalues are λ = 1 and λ = 3. The associated eigenvectors with length 1 are uniquely 1 2 determineduptosignas (cid:8) √ (cid:9) (cid:8) √ (cid:9) 1 2 1 2 x = 2√ and x = 2√ 1 −1 2 2 1 2 2 2 Thisyieldstheorthogonalmatrix (cid:8) √ √ (cid:9) 1 2 1 2 P = 2√ 2√ −1 2 1 2 2 2 (ItiseasytoverifythatP(cid:9)P = I,i.e.P−1 = P(cid:9).) Wethereforehavethediagonalization (cid:8) √ √ (cid:9)(cid:8) (cid:9)(cid:8) √ √ (cid:9) (cid:8) (cid:9) 1 2 −1 2 2 1 1 2 1 2 1 0 P−1AP = 2√ 2√ 2√ 2√ = 1 2 1 2 1 2 −1 2 1 2 0 3 2 2 2 2 ⎛ ⎞ 1 1 0 (b) LetB = ⎝1 1 0⎠. ThecharacteristicpolynomialofBis 0 0 2 (cid:6) (cid:6) (cid:6)(cid:6)1−λ 1 0 (cid:6)(cid:6) (cid:19) (cid:20) (cid:6)(cid:6) 1 1−λ 0 (cid:6)(cid:6) = (2−λ) (1−λ)2 −1 = (2−λ)(λ2 −2λ) = −λ(λ−2)2 (cid:6) (cid:6) 0 0 2−λ (use cofactor expansion of the first determinant along the last row or the last column). The eigenvalues are λ = 0 and λ = λ = 2. It is easily seen that one eigenvector associated with the eigenvalue 1 2 3 λ = 0isx = (1,−1,0)(cid:9). Eigenvectorsx = (x ,x ,x )(cid:9) associatedwiththeeigenvalue2aregivenby 1 1 1 2 3 (B−2I)x = 0,i.e. −x +x = 0 1 2 x −x = 0 1 2 0 = 0 Thisgivesx = x ,x arbitrary. Onesetoflinearlyindependenteigenvectorswithlength1isthen 1 2 3 ⎛ ⎞ ⎛ ⎞ ⎛ ⎞ 1 1 0 1 1 √ ⎝−1⎠, √ ⎝1⎠, ⎝0⎠ 2 2 0 0 1 Fortunately,thesethreevectorsaremutuallyorthogonal(thisisnotautomaticallytruefortwoeigenvectors associatedwiththesameeigenvalue),andsowehaveasuitableorthogonalmatrix ⎛ √ √ ⎞ 1 2 1 2 0 2√ 2√ ⎜ ⎟ P = ⎝−1 2 1 2 0⎠ 2 2 0 0 1 ©ArneStrøm,KnutSydsæter,AtleSeierstad,andPeterHammond2008 8 CHAPTER 1 TOPICS IN LINEAR ALGEBRA ItisnoweasytoverifythatP−1 = P(cid:9) and ⎛ ⎞ 0 0 0 P−1BP = ⎝0 2 0⎠ 0 0 2 ⎛ ⎞ 1 3 4 (c) ThecharacteristicpolynomialofC = ⎝3 1 0⎠is 4 0 1 (cid:6) (cid:6) (cid:6)(cid:6)(cid:6)(cid:6)1−3 λ 1−3 λ 04 (cid:6)(cid:6)(cid:6)(cid:6) = 4(cid:6)(cid:6)(cid:6)(cid:6)3 1−λ(cid:6)(cid:6)(cid:6)(cid:6)+(1−λ)(cid:6)(cid:6)(cid:6)(cid:6)1−λ 3 (cid:6)(cid:6)(cid:6)(cid:6) = (1−λ)(cid:19)(1−λ)2 −25(cid:20) (cid:6) (cid:6) 4 0 3 1−λ 4 0 1−λ (cofactorexpansionalongthelastcolumn),andsotheeigenvaluesareλ = 1,λ = 6,andλ = −4.An 1 2 3 eigenvectorx = (x,y,z)(cid:9) correspondingtotheeigenvalueλmustsatisfy x +3y +4z = λx Cx = λx ⇐⇒ 3x + y = λy 4x + z = λz Onesetofunnormalizedeigenvectorsis ⎛ ⎞ ⎛ ⎞ ⎛ ⎞ 0 5 −5 u = ⎝−4⎠, v = ⎝3⎠, w = ⎝ 3⎠ 3 4 4 √ withlengths(cid:5)u(cid:5) = 5,(cid:5)v(cid:5) = (cid:5)w(cid:5) = 5 2,andacorrespondingorthogonalmatrixis ⎛ √ √ ⎞ 0 5 2 − 5 2 10√ 10√ ⎜ ⎟ P = ⎝−4 3 2 3 2⎠ 5 10√ 10√ 3 4 2 4 2 5 10 10 AstraightforwardcalculationconfirmsthatP(cid:9)CP = diag(1,6,−4) = diag(λ ,λ ,λ ). 1 2 3 1.6.5 ForthegivenA,wehaveA2 = 5A−5I.(cid:8)ThereforeA(cid:9)3 = A2A = (5A−5I)A = 5A2−5A = 20A−25I 50 75 andA4 = 20A2 −25A = 75A−100I = . 75 125 1.7 1.7.5 (a) ItisclearthatQ(x ,x ) ≥ 0forallx andx andthatQ(x ,x ) = 0onlyifx = x = 0,soQ 1 2 1 2 1 2 1 2 ispositivedefinite. ⎛ ⎞ 5 0 1 ⎝ ⎠ (b) ThesymmetriccoefficientmatrixofQis 0 2 1 . Theleadingprincipalminorsare 1 1 4 (cid:6) (cid:6) (cid:6) (cid:6) (cid:6) (cid:6) (cid:6) (cid:6) (cid:6)5 0 1(cid:6) D1 = 5, D2 = (cid:6)(cid:6)5 0(cid:6)(cid:6) = 10, D3 = (cid:6)(cid:6)0 2 1(cid:6)(cid:6) = 33 0 2 (cid:6) (cid:6) 1 1 4 Since all the leading principal minors are positive, it follows from Theorem 1.7.1 that Q is positive definite. (AnalternativewaytoseethisistowriteQasasumofsquares: Q(x ,x ,x ) = (x +x )2+ 1 2 3 1 3 (x +x )2+4x2+x2+2x2 isobviouslynonnegativeanditiszeroonlyifallthesquaretermsarezero. 2 3 1 2 3 Butitisnotalwayseasytoseehowtorewriteaquadraticforminthisfashion.) ©ArneStrøm,KnutSydsæter,AtleSeierstad,andPeterHammond2008