arxiv: v3 [math.fa] 23 Sep 2016


 Arleen Robinson
 10 months ago
 Views:
Transcription
1 AN INVERSE PROBLEM FOR A CLASS OF CANONICAL SYSTEMS AND ITS APPLICATIONS TO SELFRECIPROCAL POLYNOMIALS MASATOSHI SUZUKI arxiv: v3 [mathfa] 23 Sep 2016 Abstract A canonical system is a kind of firstorder system of ordinary differential equations on an interval of the real line parametrized by complex numbers It is known that any solution of a canonical system generates an entire function of the HermiteBiehler class In this paper, we deal with the inverse problem to recover a canonical system from a given entire function of the HermiteBiehler class satisfying appropriate conditions This inverse problem was solved by de Branges in 1960s However his results are often not enough to investigate a Hamiltonian of recovered canonical system In this paper, we present an explicit way to recover a Hamiltonian from a given exponential polynomial belonging to the HermiteBiehler class After that, we apply it to study distributions of roots of selfreciprocal polynomials 1 Introduction Let H(a) be a 2 2 symmetric matrixvalued function defined almost everywhere on a finite interval I = [a 1,a 0 ) (0 < a 1 < a 0 < ) with respect to the Lebesgue measure da We refer to a firstorder system of differential equations a d [ A(a,z) 0 1 A(a,z) A(a,z) 1 = z H(a), lim = (11) da B(a, z) 1 0 B(a, z) aրa 0 B(a, z) 0] on I parametrized by z C as a quasicanonical system (on I) A column vectorvalued function (A(,z),B(,z)) : I C 2 1 is called a solution if it consists of absolutely continuous functions and satisfies (11) almost everywhere on I for every fixed z C A quasicanonical system (11) is called a canonical system if (H1) H(a) is a real positive semidefinite symmetric matrix for almost every a I, (H2) H(a) 0 on any subset of I with positive Lebesgue measure, (H3) H(a) is locally integrable on I with respect to da/a, and H(a) is called its Hamiltonian Even if a quasicanonical system (11) is not a canonical system, a matrix valued function H(a) is often called a Hamiltonian in this paper by abuse of language if there is no confusion A number of different secondorder differential equations, like Schrödinger equations and Sturm Liouville equations of appropriate form, and systems of firstorder differential equations like Dirac type systems of appropriate form are reduced to a canonical system Fundamental results on the spectral theory of canonical systems was established in works of Gohberg Kreĭn [5], de Branges [3] and many other authors (see the survey articles Winkler [27], Woracek [28] and references there in for historical details on canonical systems) Note that the variable a in (11) is the variable on the multiplicative group R >0 By the change of variable a = e x, (11) is transformed into the equation for the variable x on the additive group R, and the right endpoint of a for the initial value is transformed into the left endpoint of x for the initial value The transformed equation is the one treated by de Branges, Kreĭn, Kac and others (See the final paragraph of the introduction for the reason why we use a multiplicative variable) The subject of the present paper is an inverse spectral problem for canonical systems In order to state it, we review the theory of the Hermite Biehler class 2000 Mathematics Subject Classification 30C15, 34A55, 34L40 1
2 2 M SUZUKI We use the notation F (z) = F( z) for functions of the complex variable z and denote by C + the (open) upper halfplane {z = x +iy C y > 0} An entire function E(z) satisfying E (z) < E(z) for every z C + (12) and having no real zeros is said to be a function of the Hermite Biehler class, or the class HB for short (This definition of the class HB is equivalent to the definition of Levin [11, 1 of Chap VII] if we replace the word the upper halfplane by the lower halfplane, because (12) implies that E(z) has no zeros in C + We adopt the above definition for the convenience to use the theory of canonical systems via the theory of de Branges spaces) Suppose that (11) is a canonical system endowed with a solution (A(a,z),B(a,z)) Then E(a,z) := A(a,z) ib(a,z) is a function of the class HB for every fixed regular point a I Inparticular, lim aրa0 E(a,z) = 1 and E(a 1,z) is a function of theclass HB Therefore, an inverse problem for canonical systems is to recover their Hamiltonians from given entire functions of the class HB satisfying appropriate conditions Usually such inverse problem is difficult to solve in general, because it is an inverse spectral problem ([8, Section 2], [17, Section 7]) However it is already solved by the theory of de Branges in 1960s ([3]) For example, for fixed I and an entire function E(z) of exponential type belonging to the class HB and satisfying (1+x2 ) 1 E(x) 2 dx < and E(0) = 1, there exists a canonical system (that is, there exists a Hamiltonian H(a) on I) such that A(a 1,z) = 1 2 (E(z)+E (z)), B(a 1,z) = i 2 (E(z) E (z)) for its solution (A(a,z),B(a,z)) (see [17, Theorem 73] with [4, Lemma 33]) Moreover, such canonical system is uniquely determined by E(z) and I under appropriate normalizations More general situation is treated in Kac [7] (see also [27] and [28]) As above, de Bragnes s theory ensures the existence of canonical systems or Hamiltonians for given functions of the class HB, but it does not provide explicit or useful expressions of Hamiltonians In fact, an explicit form of H(a) is not known except for a few examples of E(z) as in [3, Chapter 3], [4, Section 8] and some additional examples constructed from such known examples using transformation rules for Hamiltonians and Weyl functions ([26]) In this paper, we deal with the above inverse problem for a special class of exponential polynomials together with the problem of explicit constructions of H(a), and apply the results to the studying of the distribution of roots of selfreciprocal polynomials Let R := R\{0}, g Z >0 and q > 1 We denote by C a vector of length 2g +1 of the form C = (C g,c g 1,,C g ) R R 2g 1 R and consider exponential polynomials along with two associated functions E(z) := E q (z;c) := C m q imz (13) A(z) := A q (z;c) := 1 2 (E q(z;c)+e q(z;c)), (14) B(z) := B q (z;c) := i 2 (E q(z;c) Eq (z;c)) (15) AbasicfactisthatanexponentialpolynomialE(z) of(13)belongstotheclasshbifand only if it has no zeros on the closed upper halfplane C + R = {z = x+iy C y 0} (see [11, Chapter VII, Theorem 6], for example) Therefore, there exists a Hamiltonian H(a) of a canonical system corresponding to E(z) if E(z) has no zeros on C + R
3 INVERSE PROBLEMS FOR CANONICAL SYSTEMS 3 Beyond such results standing on a general theory, we show that there exists a symmetric matrix valued function H(a) of a quasicanonical system corresponding to an exponential polynomial E(z) if E(z) has no zeros on the real line at least It is constructed explicitly as follows We define two lower triangular matrices E + and E of size 2g +1 by E ± = E ± (C) := C g C (g 1) C g C 0 C 1 C g and define square matrices J n of size 2g +1 by J n = J n (2g+1) := C ±(g 1) C ±(g 2) C g C ±g C ±(g 1) C 0 C (g 1) C g J (n) 0 0 0, J(n) := for 1 n 2g, where J (n) is the antidiagonal matrix of size n with 1 s on the antidiagonal line By using the above matrices, we define the numbers n by n = n (C) := det(e+ (C)+E (C)J n ) det(e + (C) E (C)J n ) 1 1 (16) for 1 n 2g and 0 = 0 (C) := 1 We write n = if det(e + E J n ) = 0 In addition, we define the realvalued locally constant function γ : [1,q g ) R by γ(a) = γ(a;c) := (C) n (C) for q ()/2 a < q n/2 (17) Then, we obtain the following results for the inverse problem associated with the exponential polynomial (13) Theorem 11 Let q > 1 and C R R 2g 1 R Let E(z) be the exponential polynomial defined by (13) Define the square matrix D n (C) := C g C g+1 C g C g 1 C g C g+ C g+n 2 C g C g n+1 C g n+2 C g C g C g 1 C g n+1 C g C g+1 C g+ C g C g n+2 C g C g+n 2 C g C g C g of size 2n Suppose that E(z) has no zeros on the real line and that detd n (C) 0 for every 1 n 2g Then, (1) det(e + ±E J n ) 0 for every 1 n 2g, (2) the pair of functions (A(a,z),B(a,z)) defined in (213) below satisfies a d A(a,z) 0 1 A(a,z) = z H(a) (a [1,q da B(a, z) 1 0 B(a, z) g ), z C) (18)
4 4 M SUZUKI together with the boundary condition A(1,z) A(z) = B(1, z) B(z), lim aրq g A(a,z) = B(a, z) E(0), (19) 0 where A(z) and B(z) are functions of (14) and (15), respectively, and γ(a;c) 1 0 H(a) = H(a;C) := (110) 0 γ(a; C) We mention another way of constructing (γ(a),a(a,z),b(a,z)) in Section 6 The positive definiteness of the Hamiltonian H(a) in (110) is characterized by the following tobeexpected way Theorem 12 Let q > 1 and C R R 2g 1 R Let E(z) be the exponential polynomial defined by (13) and let H(a) be the matrix valued function defined by (110) (1) Suppose that E(z) belongs to the class HB Then H(a) is welldefined and is positive definite for every 1 a < q g Hence the quasicanonical system attached to (18) and (19) is a canonical system and (A(a,z),B(a,z))/E(0) is its solution (2) Suppose that H(a) is welldefined and is positive definite for every 1 a < q g Then E(z) belongs to the class HB As mentioned before, E(z) of (13) belongs to the class HB if and only if it has no zeros on C + R On the other hand, the following conditions are equivalent to each other by 0 = 1 and definitions (16), (17) and (110): (i) H(a) is positive definite for every 1 a < q g, (ii) γ(a) > 0 for every 1 a < q g, (iii) 0 < n < for every 1 n 2g Therefore, we obtain the following corollary: Corollary 13 An exponential polynomial E(z) of (13) has no zeros on C + R if and only if 0 < n < for every 1 n 2g The converse of Theorem 11 is the direct problem for quasicanonical systems (11) with the Hamiltonians of the form (110) It is easier than the inverse problem, because the Hamiltonians have a simple form: H(a) = diag(γ(a) 1,γ(a)), and γ(a) is a (special) locally constant function Theorem 14 Let q > 0, g Z >0, and let γ(a) be a locally constant function on [1,q g ) such that γ(a) = γ n R for q ()/2 a < q n/2 (1 n 2g) Then the quasicanonical system (11) with H(a) = diag(γ(a) 1,γ(a)) on [1,q g ) has the unique solution (A(a, z), B(a, z)) whose components have the forms [ (q m) iz q m) ] iz A(a,z) = α m (n) +(, a a +n [ (q ) m iz ) ] (111) q m iz ib(a,z) = β m (n) ( a a +n for q ()/2 a < q n/2 and 1 n 2g, where α ± m(n) and β m(n) ± are real constants depending only on the set {γ n } of values of γ(a) Therefore, for q ()/2 a < q n/2, E(a,z) := A(a,z) ib(a,z) is the exponential polynomial [ ( ) q m iz ) ] q m iz E(a,z) = (α m (n)+β m (n)) +(α m (n) β m (n))( a a +n Moreover, E(a,0) = 1 and E(a,z) has no real zeros for any fixed 1 a q g In particular, each Hamiltonian of the form (110) with (17) yields an exponential polynomial E(1,z) = A(1,z) ib(1,z) having no real zeros
5 INVERSE PROBLEMS FOR CANONICAL SYSTEMS 5 Theorem 14 does not guarantee that E(1,z) has the form (13) In fact, H(a) on [0,q) with γ(a) = 1 for 1 a < q 1/2 and γ(a) = 1 for q 1/2 a < q yields the constant function E(1,z) = 1, and H(a) on [0,q) with γ(a) = 1 for 1 a < q yields the function E(1,z) = q iz In these examples, E(1,z) does not have the form (13) A sufficient condition for which E(1,z) has the form (13) is the following Theorem 15 Under the notation of Theorem 14, E(1, z) is an exponential polynomial of the form (13) if γ n > 0 for every 1 n 2g and γ 1 1 If the assumption of Theorem 15 is satisfied, it is proved that E(1,z) belongs to the class HB in a way similar to the proof of Theorem 12 (2) in Section 52 On the other hand, γ n > 0 for every 1 n 2g and γ 1 1 if we start from the exponential polynomial of the form (13) by Theorem 12 and (58) below Hence, as a consequence of the above theorems, the exponential polynomials (13) belonging to the class HB are characterized in terms of positive definiteness of Hamiltonians Now we turn to an application of Corollary 13 A nonzero polynomial P(x) = c 0 x n +c 1 x + +c x+c n with real coefficients is called a selfreciprocal polynomial of degree n if c 0 0 and it satisfies the selfreciprocal condition P(x) = x n P(1/x), or equivalently, c 0 0 and c k = c n k for every 0 k n The roots of a selfreciprocal polynomial either lie on the unit circle T = {z C z = 1} or are distributed symmetrically with respect to T Therefore, a basic problem is to finda nice condition on coefficients of a selfreciprocal polynomial for which all its roots lie on T There are quite many results for this problem in literatures For example, see books of Marden[14], Milovanović Mitrinović Rassias [15], Takagi [24, Section 10] and the survey paper of Milovanović Rassias [16] for several systematic treatments on roots of polynomials As an application of Corollary 13, we study roots selfreciprocal polynomials of even degree The restriction on the degree is not essential, because if P(x) is a selfreciprocal polynomial of odd degree, then there exists a selfreciprocal polynomial P(x) of even degree and an integer r 1 such that P(x) = (x+1) r P(x) In contrast, the realness of coefficients is essential We denote by P g (x) a selfreciprocal polynomial of degree 2g of the form g 1 P g (x) = c k (x (2g k) +x k )+c g x g, c 0 0 (112) k=0 and identify the polynomial with the vector c = (c 0,c 1,,c g ) R R g consisting of its coefficients For a vector c R R g and a real number q > 1, we define the numbers δ n (c) (1 n 2g) by δ n (c) := det(e+ (C)+E (C)J n ) det(e + (C) E (C)J n ) where C is the vector defined by C m T m := c g m (1 logq m )T m + m=0 { 1 if n is even, glogq if n is odd, (113) c g m (1+logq m )T m (114) Theorem 16 Let g Z >0, q > 1 Let c R R g be coefficients of a selfreciprocal polynomial P g (x) of the form (112) Define C R R 2g 1 R by (114) Then, for every 1 n 2g, it is independent of q whether det(e + ±E J n ) is zero, and the numbers δ n (c) of (113) are independent of q if det(e + ±E J n ) is not zero Moreover, a necessary and sufficient condition for all roots of P g (x) to be simple roots on T is that 0 < δ n (c) < for every 1 n 2g Remark As a function of indeterminate elements (c 0,,c g ), δ n (c) is a rational function of (c 0,,c g ) over Q m=1
6 6 M SUZUKI Thecriterion of Theorem 16 itself may not benew, becauseitseems that thequantity δ n (c)intheorem16probablyessentially coincidewithquantities inaclassical theoryon the studying of the roots of polynomials in Section 75 The author has not yet thought rigorously whether δ n (c) and the classical quantities indeed have the same meaning However, at least, the theory of the presented paper provides a new bridge between the studying on the roots of polynomials and the theory of canonical systems The author believe that this new connection deserve to be recorded in a literature In order to deal with the case that all roots of P g (x) lie on T but P g (x) may have a multiple zero, we modify the above definition of δ n (c) as follow For a vector c R R g and real numbers q > 1, ω > 0, we define the numbers δ n (c;q ω ) (1 n 2g) by δ n (c;q ω ) = det(e+ (C)+E (C)J n ) 1 if n is even, det(e + (C) E (C)J n ) q gω q gω (115) q gω +q gω if n is odd, where C is the vector defined by C m T m := c g m q mω T m + m=0 c g m q mω T m (116) Theorem 17 Let g Z >0, q > 1 Let c R R g be coefficients of a selfreciprocal polynomial P g (x) of the form (112) Then a necessary and sufficient condition for all roots of P g (x) to lie on T is that 0 < δ n (c;q ω ) < for every 1 n 2g and ω > 0 m=1 Furthermore, the quantities δ n (c) and δ n (c;q ω ) are related as follows: Theorem 18 Let δ n (c) and δ n (c;q ω ) be as above Then lim δ n(c;q ω ) = δ n (c) q ω ց1 as a rational function of c = (c 0,,c g ) over Q Suppose that all roots of a selfreciprocal polynomial (112) lie on T and are simple Then we have where implied constants depend only on c δ n (c;q ω ) = δ n (c)+o(logq ω ) as q ω ց 1, Finally, we comment on the reason why we use a multiplicative variable on R >0 and the right endpoint for the initial value in (11) It comes from the author s personal motivation for the work of this paper: it is the inverse problem for entire functions of the class HB obtained by Mellin transforms of functions on [1, ) This problem was stimulated by Burnol [1], in particular Sections 5 8, and was partially treated in [23] under the number theoretic setting In [1], Burnol often use a multiplicative variable on R >0 and the left endpoint (= + ) corresponds to the initial value In order to apply the method of [1] to the above inverse problem, the author noted on the formulas and X 1 1 f(x)x izdx x = lim X X 1 f(x)x izdx x f(x)x izdx logx/logq x = lim 1 f(q n )q inz q 1+ logq Here exponential polynomials appear in the left hand side of the second formula If we fix X > 1, q > 1 and put 2g = logx/logq, we obtain an exponential polynomial of degree 2g Recall that the results of the paper corresponds the exponential polynomial to the canonical system on [1,q g ) Then, if we imagine the limiting situation q 1 and n=0
7 INVERSE PROBLEMS FOR CANONICAL SYSTEMS 7 X +, it is expected (but not rigorous) that a canonical system corresponds to the Mellin transform 1 f(x)x iz dx x should be a system on [1, ) To realize the above heuristic discussion to the original motivation, the right endpoint may be more useful and convenient than the left endpoint for the initial value, although the left endpoint is useful for the initial value in the usual theory of canonical systems Anyway, the author believe that the above number theoretic aspect of de Branges theory is an important one The paper is organized as follows In Section 2, we describe the outline of the proof of Theorem 11 In Section 3, we prepare several lemmas and notations in order to prove statements of Section 2 and Theorem 11 in Section 4 In Section 5, we prove Theorems 12, 14 and 15 In Section 6, we mention a way of constructing (γ(a),a(a,z),b(a,z)) which is different from the way of Section 2 and 4 In Section 7, we prove Theorems 16, 17 and 18 and compare these with classical results on the roots of selfreciprocal polynomials Acknowledgments The author thanks Shigeki Akiyama for suggesting the book of Takagi ([24, Section 10]) The author also thanks the referee for a number of helpful suggestions for improvement in the article and for careful reading In particular, Theorems 14 and 15 were added by the suggestion of the referee This work was supported by KAKENHI (GrantinAid for Young Scientists (B)) No and No Outline of the proof of Theorem Hilbert spaces and operators Let q > 1 and let L 2 (T q ) be the completion of the space of (2π/logq)periodic continuous functions on R with respect to the L 2 norm f 2 L 2 (T := f,f q) L 2 (T q),where f,g L 2 (T q) := logq 2π I q f(z)g(z)dz andi q = [0,2π/(logq)) Every f L 2 (T q ) has the Fourier expansion f(z) = k Z u(k)qikz with {u(k)} k Z l 2 (Z) and f 2 L 2 (T q) = k Z u(k) 2, where l 2 (Z) is the Hilbert space of sequences {u(k) C k Z} satisfying k Z u(k) 2 < ([18, Chapter 4]) For 0 < a q Z/2 = {q n/2 n Z}, we define the vector space V a := { φ(z) = a iz f(z)+a iz g(z) f, g L 2 (T q ) } of functions of z R As a vector space, V a is isomorphic to the direct sum L 2 (T q ) L 2 (T q ), since a iz f(z) + a iz g(z) = 0 if and only if (f,g) = (0,0) In fact, if one of f and g is zero and a iz f(z)+a iz g(z) = 0, the another one should be zero On the other hand, if f 0, g 0 and a iz f(z)+a iz g(z) = 0, we have a 2iz = f(z)/g(z), and hence a q Z/2 The maps p 1 : (a iz f(z)+a iz g(z)) a iz f(z) and p 2 : (a iz f(z)+a iz g(z)) a iz g(z) are projections from V a to the firstand the second components of the direct sum, respectively In addition, we define the inner product in V a by φ 1,φ 2 = f 1,g 1 L 2 (T q) + g 1,g 2 L 2 (T q), where φ j (z) = a iz f j (z) + a iz g j (z) (j = 1,2) Then, V a forms a Hilbert space with respect to the inner product and is isomorphic to the (orthogonal) direct sum L 2 (T q ) L 2 (T q ) of Hilbert spaces ([9, Chapter 5]) We put X(k) := (q k /a) iz, Y(l) := X(l) 1 = (q l /a) iz (21) for k,l Z and a > 0 We regard X(k) and Y(l) as functions of z, functions of (a,z) or symbols depending on the situation For a fixed 0 < a q Z/2, the countable set consisting of all X(k) and Y(l) is linearly independent over C as a set of functions of z, since the linear dependence of {X(k), Y(l)} k,l Z implies the existence of a nontrivial
8 8 M SUZUKI pair of f,g L 2 (T q ) satisfying a iz f(z)+a iz g(z) = 0 Using these vectors, we have { V a = φ = u(k)x(k)+ } v(l)y(l) {u(k)} k Z, {v(l)} l Z l 2 (Z) k Z l Z and L 2 (T q ) L 2 (T q ), { } p 1 V a = u(k)x(k) {u(k)} k Z l 2 (Z) L 2 (T q ), k Z { } p 2 V a = v(l)y(l) {v(l)} l Z l 2 (Z) L 2 (T q ) l Z φ,φ = u(k)u (k)+ v(l)v (l) (φ,φ V a ), k Z l Z φ 2 = φ,φ = u(k) 2 + v(l) 2 (φ V a ), (22) k Z l Z where φ = k Z u(k)x(k)+ l Z v(l)y(l) and φ = k Z u (k)x(k)+ l Z v (l)y(l) On the other hand, we have φ 2 = logq p 1 φ(z)p 1 φ(z)dz + logq p 2 φ(z)p 2 φ(z)dz (23) 2π I q 2π I q for φ V a, since logq (a ±iz f 1 (z))(a 2π ±iz f 2 (z))dz = u 1 (k)u 2 (k) = f 1,f 2 L 2 (T q) I q k Z for f j (z) = k Z u j(k)q ikz L 2 (T q ) (j = 1,2) Note that, for φ V a, p 1 φ and p 2 φ are not periodic functions of z, but the integrals I p jφ(z)p j φ (z)dz (j = 1,2, φ,φ V a ) are independent of the intervals I = [α,α+2π/logq] (α R) We write φ V a as φ(z) (resp φ(a,z)) to emphasize that φ is a function of z (resp (a,z)) If we regard X(k) and Y(l) as symbols, V a is abstract Hilbert space endowed with the norm defined by (22) and is isomorphic to l 2 (Z) l 2 (Z) For each nonnegative integer n, we define the closed subspace V a,n of V a by { } V a,n = φ n = u n (k)x(k) + v n (l)y(l) {u n(k)} k=0, {v n(l)} l= l2 (Z) k=0 k= and denote by P n the projection operator V a V a,n Then, for the conjugate operator P n := JP nj : V a V a of P n by the involution J : φ(z) φ( z), we have P n φ = JP njφ = u(k)x(k) + v(l)y(l) (φ V a ), k= because J acts on V a as exchanging of coefficients between X(k) and Y(l): Jφ = v(k)x(k) + u(l)y(l) (φ V a ) k= l= Therefore, P n maps V a,n into V a,n for every nonnegative integer n In addition, JP n also maps V a,n into V a,n for every nonnegative n, because l=0 l=0 JP n φ n = v n (l)x(l)+ u n (k)y(k) (24) for φ n V a,n Note that P 0 Va,0 = JP 0 Va,0, since P 0 J V a,0 = 0 by definition k=0
9 INVERSE PROBLEMS FOR CANONICAL SYSTEMS 9 For an exponential polynomial E(z) of (13), we have E (z) = C m q imz = E q (z;c ) for C := (C g,c (g 1),,C g ) R R 2g 1 R by a simple calculation By using E(z) and E (z), we define two multiplication operators E : φ(z) E(z)φ(z), E : φ(z) E (z)φ(z) (25) These operators map V a into V a, because E and E are expressed as E = C m T m, E = C m T m by using shift operators T m : V a V a (m Z) defined by T m v = u(k)x(k +m)+ v(l)y(l m) k= l= Both E and E are bounded on V a, since E g C m T m M and E g C m T m M for M = max{c m g m g} Suppose that E(z) has no zeros on the real line Then the operator E is invertible on V a (Lemma 41(1)) Thus the operator is welldefined We have Θ := E 1 E (26) (Θφ)(z) = Θ(z)φ(z) for Θ(z) := E (z) E(z), φ V a (27) and Θ(V a,n ) V a,n for each nonnegative integer n (Lemma 41(2)) 22 Quasicanonical systems associated with exponential polynomials Under the above settings, a quasicanonical system associated with an exponential polynomial E(z) of (13) is constructed starting from solutions of the set of linear equations (I±ΘJP n )φ ± n = ΘX(0) (φ ± n V a,n, 0 n 2g), (28) where I is the identity operator The set of 4g+2 equations (28) is a discrete analogue of the (right) Mellin transform of differential equations (117a) and (117b) in [1, Section 6] (see also [23, Section 4]) Under the assumption that both I±ΘJP n are invertible on V a,n for every 0 n 2g, that is, (I±ΘJP n ) 1 exists as a bounded operator on V a,n, we define A n (a,z) := 1 2 (I+J)Eφ+ n (a,z), B n (a,z) := 1 2i (I J)Eφ n (a,z) (29) by using unique solutions of (28) Then they are entire functions of z and are extended to functions of a on (0, ) (by formula (45)) Here we define functions A (a,z) and B (a,z) of (a,z) [1,q g ) C by A (a,z) : = A n (a,z) B (a,z) : = B n(a,z) for q ()/2 a < q n/2 (210) The function A (a,z) is discontinuous at a [1,q g ) q Z/2 in general, because A n(q n/2,z) = A n+1(q n/2,z)
10 10 M SUZUKI may not be hold It is similar about B (a,z) However, we will see that α n = α n (C) := A (q()/2,z) A n(q ()/2,z), β n = β n (C) := B (q()/2,z) B n(q ()/2,z) (211) are independent of z for every 1 n 2g (Proposition 48) Therefore, we obtain functions A(a,z) and B(a,z) of (a,z) [1,q g ) C which are continuous for a and entire for z by the modification for 1 n 2g and A n (a,z) : = α 1 α n A n(a,z), B n (a,z) : = β 1 β n B n(a,z) A(a,z) : = A n (a,z) B(a,z) : = B n (a,z) (212) for q ()/2 a < q n/2 (213) Moreover, (A(a, z), B(a, z)) satisfies the system (18) endowed with the boundary condition (19) for the locally constant function γ(a) defined by and γ n = γ n (C) := α 1(C) α n (C) β 1 (C) β n (C) (214) γ(a) := γ(a;c) := γ n if q ()/2 a < q n/2 (215) (Proposition 45 and Theorem 418) This γ(a) is equal to the function of (17) (Proposition 48) Therefore, (A(a, z), B(a, z))/e(0) is a solution of a quasicanonical system on [1,q g ) if E(0) 0 and α n,β n 0 for every 1 n 2g As a summary of the above argument, we obtain Theorem 11 See Section 4 for details In order to prove Theorems 12, 14 and 15, we use the theory of de Branges spaces which is a kind of reproducing kernel Hilbert spaces consisting of entire functions Roughly, the positivity of H(a) corresponds to the positivity of the reproducing kernels of de Branges spaces See Section 5 for details 3 Preliminaries 31 Identities for matrices The following basic facts of linear algebra are used often in later sections Lemma 31 Let A and B be square matrices of the same size Then, A B det = det(a+b)det(a B) B A Lemma 32 Let A, B, C, D be square matrices of the same size If deta 0, detd 0 and det(d CA 1 B) 0, 1 [ A B (A BD = 1 C) 1 A 1 B(D CA 1 B) 1 ] C D D 1 C(A BD 1 C) 1 (D CA 1 B) 1 Lemma 33 (Laplace s expansion formula) Let A be a square matrix of size n Let i = (i 1,i 2,,i k ) (resp j = (j 1,j 2,,j k )) be a list of indices of k rows (resp columns), where 1 k < n and 0 i 1 < i 2 < i k < n (resp 1 j 1 < j 2 < < j k < n) Denote by A(i,j) the submatrix of A obtained by keeping the entries in the intersection of any row and column that are in the lists Denote by A c (i,j) the submatrix of A obtained
11 INVERSE PROBLEMS FOR CANONICAL SYSTEMS 11 by removing the entries in the rows and columns that are in the lists By Laplace s formula for determinants, we get deta = j ( 1) i + j deta(i,j)deta c (i,j), where i = i 1 +i 2 + +i k, j = j 1 +j 2 + +j k, and the summation is taken over all ktuples j = (j 1,j 2,,j k ) for which 1 j 1 < j 2 < < j k < n Proof See [6, Chap IV], for example 32 Definition of special matrices, I We define several special matrices for a convenience of later arguments The matrix E 0 ± is the square matrix of size 8g defined by E 0 ± = E± 0 (C) := e ± 0 (C) 0 0 e± 0 (C), where e ± 0 (C) are lower triangular matrices of size 4g defined by C g C (g 1) C g e ± 0 = C e± 0 (C) := ±(g 1) C (g 1) C g C ±g C (g 1) C g 0 C±(g 1) C±g C ±(g 1) C (g 1) C g 0 0 C ±g C ±(g 1) C (g 1) C g Replacing the column at the left edge of e 0 by the zero column vector, C g C (g 1) Cg e 1 = e 1 (C) := C (g 1) C(g 1) C g C g C(g 1) C g 0 C (g 1) 0 C g C (g 1) C(g 1) C g C g C (g 1) C (g 1) C g We define and E nj = E nj(c) := E 0 J = E 0 J(C) := 0, E 0,1 J = E 0,1J(C) := 0 0 e 1,n (C) e 2,n (C) 0, E n,1 J = E n,1j(c) := 0 e 3,n (C) e 2,n (C) 0
12 12 M SUZUKI by setting e 1,n = e 1,n (C) := e 2,n = e 2,n (C) := C g C (g 1) C g 0 C (g 1) C (g 1) 0 C g C (g 1) C g C g C (g 1) C g C (g 1) C (g 1) 0 C g C (g 1) C g = = 2g n n 2g n 4g n 2g n 2g +n 2g +n, 2g n and e 3,n = e 3,n (C) := C g 0 0 C (g 1) 0 C g 0 C (g 1) C (g 1) 0 0 C g 0 C (g 1) C g = 2g n n 2g 2g +n 2g n for 1 n 2g, where the righthand sides mean the size of each block of matrices in middle terms In addition, we define square matrices e 0 of size 4g by e 0 = e 0 (C) := C g C g 1 C (g 1) C g C g 1 C g C g C (g 1) C g C g 1 C (g 1) C g C g 1 C g C (g 1) C (g 1) C g C g
13 INVERSE PROBLEMS FOR CANONICAL SYSTEMS 13 so that J (8g) E + 0 = 0 e 0 e 0 0 Replacing n columns from the left edge of e 0 by zero column vectors, C g C g 1 C (g 1) C g C g 1 C g C g C (g 1) e 0,n = e 0,n (C) : = 0 Cg 1 C (g 1) C g C g C (g 1) C g = n 4g n 4g We denote by I (m) the identity matrix of size m and by J n (m) of size m: J n (m) = We also use the notation J (n) 0 0 0, J (n) = 1 the following square matrix 1 χ n = t[ ] = the unit column vector of length n For a square matrix M of size N, we denote by [M] տn (resp [M] րn ) the square matrix of size n( N) in the top left (resp the top right) corner, and denote by [M] ցn (resp [M] ւn ) the square matrix of size n( N) in the lower right (resp the lower left) corner: M = [ [M]տn = [M] ցn n [M] ւn n [M] րn 33 Definition of special matrices, II We define the square matrix P k (m k ) of size 2k+2 endowed with the parameter m k and the (2k+2) (2k+4) matrix Q k for every nonnegative integer k as follows For k = 0,1, P 1 (m 1 ) := P 0 := [ m 1 ], Q 0 :=, Q 1 := [ ], ] For k 2, we define P k (m k ) and Q k blockwisely as follows V + k 0 W + P k (m k ) := 0 V k 0 k, Q k := 0 W k 0I k m k 0I k 0 k,k+2 0 k,k+2,
14 14 M SUZUKI where 0I k := 0 k,1 I k, mk 0I k = 0 k,1 m k I k, 0k,l is the k l zero matrix, I k is the identity matrix of size k, 1 0 V + k := 1 1 ( ) k+3 (k+1) if k is odd, 2 1 V k := := := ( k +2 2 ( k +1 ) (k +1) if k is even, 2 ) (k +1) if k is odd, ( ) k+2 (k+1) if k is even, 2 and matrices W ± k are defined by adding column vectors t (1 0 0) to the rightside end of matrices V ± k : W + k := 1 1 ( ) k+3 (k+2) if k is odd, ( ) := k +2 (k +2) if k is even, W k := := Lemma 34 For k 1, we have where ε 2j+1 = detp k (m k ) = { { ( k +1 ( k+2 ε 2j+1 2 j m j+1 2j+1 if k = 2j +1, ε 2j 2 j m j 2j if k = 2j, 2 2 ) (k +2) if k is odd, ) (k+2) if k is even { +1 j 2,3 mod4 1 j 0,1 mod4, ε +1 j 0,1 mod4 2j = 1 j 2,3 mod4 In particular, P k (m k ) is invertible if and only if m k 0
15 INVERSE PROBLEMS FOR CANONICAL SYSTEMS 15 Proof This is trivial for k = 1 Suppose that k = 2j and write P k (m k ) as (v 1 v 2k+2 ) by its column vectors v l At first, we make the identity matrix I k+2 at the leftupper corner by exchanging the columns v (k+5)/2,,v k+1 and v k+2,,v (3k+3)/2 so that detp k (m k ) = det(v 1 v (k+3)/2 v k+2 v (3k+3)/2 v (k+5)/2 v k+1 v (3k+5)/2 v 2k+2 ) Then, by eliminating every 1 and m k under I k+2 of the leftupper corner, we have Ik+2 detp k (m k ) = det 0 k,k+2 Z k Here Z k is the k k matrix Z k = Zk,1 Z k,2 0 j+1,j Z k,3 for which Z k,1 is the j j antidiagonal matrix with 1 on the antidiagonal line and m k j (j +1) Zk,2 m k = Z k,3 m k 2m k (j +1) (j +1) 2mk The above formula of detp k (m k ) implies the desired result The case of even k is proved in a way similar to the case of odd k Lemma 35 We have P k (m k ) 1 Q k = Mk,1 M k,2 M k,3 M k,4 (2k +2) (2k +4) (31) for every 0 k 2g, where M k,1, M k,2, M k,3, M k,4 are (k+1) (k+2) matrices defined by M0,1 M 0, M1,1 M =, 1,2 = M 0,3 M 0, M 1,3 M 1, /m and M k,1 = if k 3 is odd, = if k 2 is even,
16 16 M SUZUKI M k,4 = if k 3 is odd, = if k 2 is even, M k,2 = m k if k 3 is odd, = m k if k 2 is even, M k,3 = m k 2 if k 3 is odd, = 1 2m k 1 1 if k 2 is even Proof According to definitions of P k (m k ), Q k and M k,l (1 l 4), the identity Mk,1 M P k (m k ) k,2 = Q M k,3 M k k,4 is checked by an elementary way
17 INVERSE PROBLEMS FOR CANONICAL SYSTEMS 17 4 Proof of Theorem 11 In this section, we complete the proof of Theorem 11 by showing each statement in Section 2 We fix g Z >0, q > 1 and C R R 2g 1 R throughout this section Lemma 41 Let E be the multiplication operator defined by (25) for E(z) of (13) Suppose that E(z) has no zeros on the real line Then (1) E is invertible on V a Moreover, for the operator Θ of (26), we have (2) Θ(V a,n ) V a,n for each nonnegative integer n Proof (1) The multiplication by 1/E(z) maps a sum of X(k) (resp Y(l)) to another sum of X(k) (resp Y(l)) as well as multiplication by E(z), since 1 E(z) = q igz 2g m=0 C m gq = C m q imz imz for some sequence { C m } of real numbers Therefore, it is sufficient to prove that E is invertible on p 1 V a and p 2 V a We have 1/E(z) L (T q ) by the assumption Therefore, multiplication by1/e(z) definesaboundedoperatore 1 onp i V a withthenorm E 1 = 1/E L (T q) for i = 1,2 (2) We have Θ(z) = 1 on the real line by definition (27) Thus, multiplication by Θ(z) defines a bounded operator on V a with the norm Θ = 1 We can write 2g m=0 Θ(z) = C g mq imz 2g m=0 C m gq = R m q imz (R m R) imz m=0 Therefore, Θ(z)φ n (z) = R m u n (k)x(k +m)+ R m v n (l)y(l m) = m=0 k=0 ( m=0 k=m This shows Θ(V a,n ) V a,n Let us consider the equations ) R m u n (k m) X(m)+ m=g m=0l= 0 m= (+m l= ) R m v n (l m) Y(m) (E±E JP n )φ ± n = E X(0) (φ ± n V a,n, 1 n 2g) (41) instead of (28) They are equivalent to (41) if E is invertible Lemma 42 Let 0 < a q Z/2 Let D n (C) be matrices define in Theorem 11 Suppose that E(z) of (13) has no zeros on the real line and that detd n (C) 0 for every 1 n 2g Then ΘJP n defines a compact operator on V a,n for each 0 n 2g and the resolvent set of ΘJP n Va,n contains both ±1 In particular, I ±ΘJP n are invertible on V a,n and (41) has unique solutions in V a,n for each 0 n 2g Remark If a q Z/2 (0, ), I±ΘJP n may not be invertible Proof It is trivial for n = 0, since P 0 Va,0 = 0 On V a,n for n 1, the HilbertSchmidt norm of ΘJP n is finite In fact, ΘJP n Va,n 2 2 = ΘJP n X(k) 2 + ΘJP n Y(l) 2 k=0 JP n X(k) 2 + k=0 l= l= JP n Y(l) 2 k=0 1 < 1+ l=0
18 18 M SUZUKI by(22)and(24) Therefore, ΘJP n Va,n isahilbertschmidtoperatoronv a,n, andhence it is compact Thus, either λ C\{0} is an eigenvalue of ΘJP n Va,n or an element of the resolvent set of ΘJP n Va,n Assume that ΘJP n φ n = ±φ n Then ΘJP n φ n = φ n Because Θ and p i (i = 1,2) commute (see the proof of Lemma 41(2)), ΘJP n φ n 2 = logq 2π by (23) While, φ n 2 = logq 2π = logq 2π k=0 by (23) Thus, it must be In this case, and Θ(z)p 1 JP n φ n (z) 2 dz + logq Θ(z)p 2 JP n φ n (z) 2 dz I q 2π I q p 1 JP n φ n (z) 2 dz + logq p 2 JP n φ n (z) 2 dz I q 2π I q = u(k) 2 + v(l) 2 E JP n φ n = ±Eφ n = ± l=0 p 1 φ n (z) 2 dz + logq I q 2π k=0 k=0 I q p 2 φ n (z) 2 dz = φ n = u(k)x(k) + v(l)y(l) l=0 C m v(k)x(k +m)+ k=0 C m u(k)x(k +m)± l=0 l=0 u(k) 2 + k=0 l= C m u(l)y(l m) C m v(l)y(l m) v(l) 2 Comparing the coefficients of {X(k)} g k g+ and {X(k)} g k g+ in the equality E JP n φ n ±Eφ n = 0, yields where A ± = A ± t[ u n (0) u n () v n (0) v n () ] = 0, (42) ±C g C g ±C g+1 ±C g C g 1 C g ±C g+ ±C g+n 2 ±C g C g n+1 C g n+2 C g ±C g ±C g 1 ±C g n+1 C g C g+1 C g+ ±C g ±C g n+2 C g C g+n 2 ±C g C g Here deta ± 0 by detd n (C) 0 Therefore (42) has no nontrivial solutions, thus φ n = 0 Consequently, both ±1 are not eigenvalue, and hence they belong to the resolvent set ofθjp n Va,n whichmeansthati±θjp n Va,n existsandboundedonv a,n In the remaining part of the section, we assume that E(z) 0 for z R and detd n (C) 0 for 1 n 2g so that both E±E JP n are invertible on V a,n for every 0 n 2g (cf Lemma 41 and Lemma 42)
19 Let φ ± n = k=0 u± n (k)x(k) + Using coefficient of φ ± n and putting INVERSE PROBLEMS FOR CANONICAL SYSTEMS 19 l= v± n (l)y(l) be solutions of (41) for 0 n 2g v ± n(n) = v ± n(n+1) = v ± n(2g 1) = 0 if 0 n 2g 1, we define the column vectors Φ ± n of length 8g by Φ ± n = t[ u ± n (0) u± n (1) u± n (4g 1) v± n (2g 1) v± n (2g 2) v± n ( 2g)] Substituting φ ± n in (41) and then comparing coefficients of X(k) and Y(l), we obtain and (E + 0 ±E n J)Φ± n = E 0 χ 8g, (Φ ± n V a,n, 0 n 2g) (43) 2 j=0 C g j u ± n(k +j) = 0, for every K 4g and L 2g 1 2 j=0 C g j v ± n(l j) = 0 (44) Lemma 43 Let 0 n 2g We have det(e + 0 ±E nj) 0 if E±E JP n is invertible Proof For fixed n, equations (41) have unique solutions φ ± n, if E ± E JP n are invertible On the other hand, the solutions φ ± n correspond to solutions of the pair of linear equations (43) and (44) Therefore, det(e 0 + ±E nj) 0 On the other hand, Eφ ± n = E X(0) E JP n φ ± n ( = C m (X(m) v ± n (k)x(k +m)+u ± n (k)y(k m))) by (41) Therefore, we can write k=0 g+ Eφ ± n = ) (p ± n (k)x(k)+q± n (k)y(k) k= g (45) for some real numbers p ± n(k) and q n(k) ± Hence Eφ ± n(a,z) are extended to smooth functions of a on (0, ) by the righthand side of (45) We use the same notation for such extended functions We put p ± n(k) = q n(k) ± = 0 for every g+n k 3g 1 if 0 n 2g 1 and define the column vectors Ψ ± n of length 8g by Ψ ± n = t[ p ± n ( g) p± n ( g +1) p± n (3g 1) q± n (3g 1) q± n (3g 2) q± n ( g)] Then, we have by g+ Eφ ± n = ) (p ± n (k)x(k)+q± n (k)y(k) = k= g k=0 C m u n (k)x(k +m)+ Ψ ± n = E+ 0 Φ± n (46) l= C m v n (l)y(l m) Lemma 44 Let 0 n 2g We have p ± n(k)±q n(k) ± = 0 if g +1 k g + or g k g+, and p ± n(g)±q n(g) ± = C g Therefore, (I±J)Eφ ± n = (p ± n(k)±q n(k))(x(k) ± ±Y(k)) k= g+n
20 20 M SUZUKI Proof By definition of matrices, the equality (E 0 + ±E nj)φ ± n = E0 χ 8g is written as e + 0 ±e 1,n ±e 2,n e + Φ± n = E 0 χ 8g (47) 0 On the other hand, e + 0 ±e 0 ±e 0 e + 0 Φ± n = p ± n( g)±q n( g) ± p ± n( g+1)±q n( g ± +1) p ± n(3g 1)±q n(3g ± 1) ±(p ± n(3g 1)±q n(3g ± 1)), ±(p ± n(3g 2)±q n(3g ± 2)) ±(p ± n( g)±q ± n( g)) since E + 0 ±J(8g) E + 0 = e + 0 ±e 0 ±e 0 e + 0 by definition of e 0 and p ± n ( g)±q± n ( g) p ± n ( g +1)±q± n ( g +1) (I (8g) ±J (8g) )E 0 + Φ± n = (I (8g) ±J (8g) )Ψ ± n = p ± n (3g 1)±q± n (3g 1) ±(p ± n (3g 1)±q± n (3g 1)) ±(p ± n (3g 2)±q± n (3g 2)) ±(p ± n ( g)±q± n ( g)) by (46) In addition, e + 0 ±e 0 ±e 0 e + 0 Φ± n = e + 0 ±e 0,n ±e 0 e + 0 Φ± n, since v n(n) ± = v n(n ± +1) = v n(2g ± 1) = 0 for 1 n 2g 1 by definition of Φ ± n Therefore, p ± n ( g)±q± n ( g) p ± n ( g +1)±q± n ( g +1) e + 0 ±e 0,n ±e 0 e + Φ± n = p ± n(3g 1)±q n(3g ± 1) ±(p ± n(3g 1)±q n(3g ± 1)) (48) 0 ±(p ± n (3g 2)±q± n (3g 2)) ±(p ± n( g)±q n( g)) ±
21 INVERSE PROBLEMS FOR CANONICAL SYSTEMS 21 Here we find that 2g rows of both E 0 + e + ±E nj and 0 e 0,n e 0 e + with indices (2g+1,2g+ 0 2,,4g) and n rows of both E 0 + e + ±E nj and 0 e 0,n e 0 e + with indices (8g n+1,8g 0 n+2,,8g) have the same entries Therefore, by comparing 2g rows of (47) and (48) with indices (2g +1,2g +2,,4g), we have p ± n(g)±q ± n(g) = C g and p ± n(k)±q n(k) ± = 0 (g +1 k 3g 1) Similarly, by comparing n rows of (47) and (48) with indices (8g n + 1,8g n + 2,,8g), we have Hence we complete the proof p ± n(k)±q ± n(k) = 0 ( g k g+) We define the column vectors A n and B n of length 8g by A n = A n(c) := (I +J)Ψ + n = (I +J)E 0 + Φ+ n, B n = B n(c) := (I J)Ψ n = (I J)E + 0 Φ n, where I = I (8g) and J = J (8g), and define the row vector F(a,z) of length 8g by F(a,z) := [ X( g) X( g +1) X(g) 0 0 Y(g) Y(g 1) Y( g) ] Then, we obtain by (210) and Lemma 44 Proposition 45 We have a d (49) A n(a,z) = 1 2 F(a,z) A n, B n(a,z) = 1 2i F(a,z) B n (410) da A n (a,z) = zb n for every 0 n 2g (a,z), a d da B n (a,z) = za n (a,z) We provide two lemmas in order to prove Proposition 45 Lemma 46 We have for every 0 n 2g u + n(k) = u n(k) (0 k 4g 1), v n + (k) = v n (k) ( 2g k 2g 1) Proof We have Φ ± n = (E+ 0 ±E nj) 1 E0 χ 8g with E 0 + ±E n J = e + 0 ±e 1,n ±e 2,n e + 0 by (43) and definition of EnJ Applying Lemma 32 to A = D = P := e + 0, B = ±Q := ±e 1,n and C = ±R := ±e 2,n, we have [ Φ ± n = (P QP 1 R) 1 P 1 Q(P RP 1 Q) 1 ] P 1 R(P QP 1 R) 1 (P RP 1 Q) 1 E0 χ 8g This shows Lemma 46, since all 4g entries of E0 χ 8g with indices (4g+1,4g+2,,8g) are zero
22 22 M SUZUKI Lemma 47 We have p + n (k)+q+ n (k) = p n (k) q n (k) ( g k g +) for every 0 n 2g, where p ± n (k)±q± n (k) = 0 if g k g + or g +1 k g + by Lemma 44 Proof By (41), Eφ ± n = E X(0) E JP n φ ± n = C m X(m) Therefore, (I+J)Eφ + n = + On the other hand, (I J)Eφ n = = C m X(m) C m Y(m) C m X(m) C m Y(m)+ C m X(m) C m Y(m)+ C m v n ± (k)x(k +m) k=0 ( ( ( ( C m C m u ± n (l)y(l m) l=0 ) u + n(k)x(k +m)+c m v n(k)x(k + +m) k=0 C m k=0 ) u + n (l)y(l+m)+c m v n + (l)y(l+m) l=0 C m ( l=0 ) u n(k)x(k +m) C m vn(k)x(k +m) k=0 C m C m ( k=0 ) u n(l)y(l+m) C m vn(l)y(l+m) l=0 l=0 ) u + n(k)x(k +m)+c m v n(k)x(k + +m) k=0 C m k=0 ) u + n(l)y(l+m)+c m v n(l)y(l+m) + by Lemma 46 Comparing the righthand sides of the above formulas of (I ± J)Eφ ± n with formulas of (I±J)Eφ ± n in Lemma 44, we obtain Lemma 47 Proof of Proposition 45 By definition of X(k) and Y(l), (45), and Lemma 44, (I±J)Eφ ± n (a,z) = (p ± n (k)±q± n (k))((qk /a) iz ±(q k /a) iz ) k= g+n Therefore, the differentiability of A n(a,z) and Bn(a,z) with respect to a is trivial, and a d da (I+J)Eφ+ n(a,z) = iz (p + n(k)+q n(k))((q + k /a) iz (q k /a) iz ), a d 1 da i (I J)Eφ n(a,z) = z k= g+n k= g+n l=0 l=0 (p n(k) q n(k))((q k /a) iz +(q k /a) iz ) Applying Lemma 47 to the righthand sides, we obtain Proposition 45 by definition (29) Asmentionedbefore, A n (a,z)andb n (a,z)of(29)aresmoothfunctionsofaon(0, ) for every 0 n 2g However A (a,z) and B (a,z) of (210) may be discontinuous at a (0, ) q Z/2 Therefore, as mentioned in Section 2, the next aim is to make modifications so that A (a,z) and B (a,z) become continuous functions of a on [1,q g ) The essential part is the following proposition
23 INVERSE PROBLEMS FOR CANONICAL SYSTEMS 23 Proposition 48 Let 1 n 2g Suppose that E±E JP is invertible on V a, for every q (n 2)/2 < a < q n/2 and E±E JP n is invertible on V a,n for every q ()/2 < a < q n/2 Define α n and β n by (211) Then, we have det(e + +E J 1 ) if n = 1, α n = det(e + ) det(e + +E J 2 ) det(e + ) det(e + 0 ) det(e + 0 +E 1 J) if n = 2, det(e + +E J n ) det(e 0 + +E n 2 J) det(e + +E J n 2 ) det(e 0 + +E J) if 3 n 2g, det(e + E J 1 ) det(e + if n = 1, ) det(e + E J 2 ) det(e 0 + ) β n = det(e + ) det(e 0 + +E 1 J) if n = 2, det(e + E J n ) det(e 0 + E n 2 J) det(e + E J n 2 ) det(e 0 + E J) if 3 n 2g In particular, α n and β n depend only on C We prove Proposition 48 after preparing several lemmas Lemma 49 For every 0 n 2g, we have (411) (412) det(e + 0 +E nj) = det(e + 0 E nj) (413) Proof Suppose that n 1, since EnJ = 0 by definition We have E 0 + ±E n J = e + 0 ±e 1,n ±e 2,n e + 0 A 0 ±B C D 0 = ±B E A = 2g +n 4g 2n 2g +n 2g +n 4g 2n 2g +n by definition of e 1,n and e 2,n, where [ A = [e + 0 ] տ2g+n, B = [e [e + 2,n ] ւ2g+n, D = 0 ] ց2g n [e + 0 ] տ2g n and the righthand side means the size of each block of the matrix of the second line Applying Lemma 33 to the (4g 2g n) columns of E 0 + ± E nj with indices (2g + n+1,,4g), and then applying Lemma 33 to the (4g 2g n) rows of the resulting matrix with indices (2g +n+1,,4g), we obtain det(e 0 + A ±B ±E nj) = det(d)det ±B A By Lemma 31, det This equality shows (413) A ±B = det(d)det(a+b)det(a B) ±B A ]
24 24 M SUZUKI Lemma 410 We have adj(e + 0 ±e 0 J(4g) n )e 0 χ 4g = adj(e + 0 ±e 1 J(4g) n )e 0 χ 4g for every 1 n 2g, where adj(a) = (deta)a 1 is the adjugate matrix of A Proof Let w = t[ C g C g 0 0 ] be the column vector of length 4g Denote by u n,j and v n,j the column vectors of e + 0 ±e 0 J(4g) n e + 0 ±e 0 J(4g) n = [ u n,1 u n,4g ], e + 0 ±e 1 J(4g) n = [ v n,1 v n,4g ] By Cramer s formula, it is sufficient to prove that det[u n,1 u n,j 1 j w un,j+1 u n,4g ] = det[v n,1 v n,j 1 j w vn,j+1 v n,4g ] and e + 0 ±e 1 J(4g) n, respectively: (414) for every 1 j 2g By definition of e 0 and e 1, u n,j = v n,j for every j n, 1 j 4g and u n,n = v n,n +w Therefore, (414) is trivial if j = n If 1 j < n, det[u n,1 u n,j 1 j w un,j+1 u n,4g ] = det[v n,1 v n,j 1 j w vn,j+1 v n,4g ]+det[v n,1 j w n w vn,4g ] = det[v n,1 v n,j 1 j w vn,j+1 v n,4g ]+0 Hence we obtain (414) Lemma 411 We have and for 3 n 2g det(e + 0 ±e 1 J(4g) 1 ) = det(e + 0 ±e 1 J(4g) 2 ) = det(e + 0 ) det(e + 0 ±e 1 J(4g) n ) = det(e + 0 ±e 0 J(4g) n 2 ) Proof We have det(e + 0 ± e 1 J(4g) 1 ) = det(e + 0 ), since e 1 J(4g) 1 = 0 On the other hand, det(e + 0 ±e 1 J(4g) 2 ) = C 4g g = det(e+ 0 ), since e+ 0 is a lower triangular matrix of size 4g and For 3 n 2g, we have e 1 J(4g) 2 = e + 0 ±e 1 J(4g) n = C g 0 0 [e + 0 ±e 0 J(4g) n 2 ] տ4g 1 The 4gth row and the 4gth column of e + 0 ± e 0 J(4g) n 2 are [0 0 C g C g ] and t [0 0 C g ], respectively Therefore, det([e + 0 ±e 0 J(4g) n 2 ] տ4g 1) = C 1 g det(e+ 0 ±e 0 J(4g) n 2 ) Hence, we obtain the equality of Lemma 411
25 Lemma 412 For 1 n 2g, we have det(e + 0 ±E n INVERSE PROBLEMS FOR CANONICAL SYSTEMS 25 J) = C8g 2n g det([e 0 + +E 0 J(8g) n ] տn )det([e 0 + E 0 J(8g) n ] տn ) (415) Proof By the proof of Lemma 49, det(e + 0 ±E n Here J) = C4g 2n g det([e + 0 ] տ2g+n +[e 2,n ] ւ2g+n)det([e + 0 ] տ2g+n [e 2,n ] ւ2g+n) [e + 0 ] տ2g+n ±[e 2,n ] ւ2g+n = [E + 0 ±E 0 J(8g) n ] տ2g+n, since [e + 0 ] տ2g+n = [E + 0 ] տ2g+n and [e 2,n ] ւ2g+n = [E 0 J(8g) n ] տ2g+n by definition of matrices Applying Lemma 33 to the 2g columns of [E + 0 ± E 0 J(8g) n ] տ2g+n with indices (n+1,n+2,,2g +n), we have det[e + 0 ±E 0 J(8g) n ] տ2g+n = C 2g g det([e+ 0 ±E 0 J(8g) n ] տn ), since all entries above [E 0 + ±E 0 J(8g) n ] ց2g in [E 0 + ±E 0 J(8g) n ] տ2g+n are zero and [E 0 + ± E0 J(8g) n ] ց2g is lower triangular with diagonal (C g,,c g ) Therefore, we obtain (415) Lemma 413 We have det(e + 0 ±E 0 J(8g) n ) ( 1 (u ± n ()+v± n (0))) = det(e + 0 ±E 0 J(8g) n 2 ) Proof Let w = t[ C g C g 0 0 ] be the column vector of length 8g Denote by u ± n,j the columns of E+ 0 ±E nj By Cramer s formula, det(e + 0 ±E nj)(1 (u ± n()+v ± n(0))) = det(e + 0 ±E nj) det[u ± n,1 n w u ± n,8g ] det[u ± n,1 6g w u ± n,8g ] = det(e + 0 ±E n J) det[u± n,1 n w u ± n,8g ] det[u ± n,1 6g ±w u ± n,8g ] = det(e + 0 ±E nj) det[u ± n,1 n w u ± n,8g ] det[u ± n,1 6g u ± n,6g u± n,8g ]+det[u± n,1 6g u ± n,6g w u± n,8g ] = det[u ± n,1 n w u ± n,8g ]+det[u ± n,1 6g u ± n,6g w u± n,8g ] n 6g 6g n w = det[u ± n,1 w ±w u ± n,8g ] det[u± n,1 u ± n,6g w u± n,8g ] +det[u ± n,1 6g u ± n,6g w u± n,8g ] n 6g = det[u ± n,1 u ± n,6g w Therefore, it is sufficient to prove det(e + 0 ±E 0 J(8g) n 2 )det(e+ 0 ±E n J) = det(e + 0 ±E 0 J(8g) n )det[u ± n,1 u ± n,6g w u± n,8g ] u ± n,6g n w 6g Applying Lemma 33 to the columns (n+1,n+2,,8g), (416) u ± n,6g w u± n,8g ] det(e + 0 ±E 0 J(8g) n 2 ) = det([e+ 0 ±E 0 J(8g) n 2 ] տn 2)det([E + 0 ±E 0 J(8g) n 2 ] ց8g n+2) = C 8g n+2 g det([e 0 + ±E 0 J(8g) n 2 ] տn 2),
26 26 M SUZUKI since all entries above [E 0 + ±E 0 J(8g) n 2 ] ց8g n+2 are zero and [E 0 + ±E 0 J(8g) n 2 ] ց8g n+2 is lower triangular with diagonal (C g,,c g ) By the proof of Lemma 49, det(e + 0 ±E n J) = C4g n g det([e 0 + ] տ2g+n +[e 1,n ] ր2g+n)det([e 0 + ] տ2g+n [e 1,n ] ր2g+n), since [e 1,n ] ր2g+n = [e 2,n ] ւ2g+n We observe that By Lemma 412, [E + 0 ] տ2g+n ±[e 1,n ] ր2g+n = [E + 0 ±E 0 J(8g) n ] տ2g+n (LHS of (416)) = C 16g 2n+2 g det([e 0 + ±E 0 J(8g) n 2 ] տn 2) On the right hand side of (416), we have det([e + 0 +E 0 J(8g) n ] տn )det([e + 0 E 0 J(8g) n ] տn ) det(e 0 + ±E 0 J(8g) n ) = C 8g n g det([e + 0 ±E 0 J(8g) n ] տn ) by applying Lemma 33 to the columns (n+1,n+2,,8g) of E + 0 ±E 0 J(8g) n Therefore, it is sufficient to prove D ± n := det[u ± n,1 u ± n,6g n w 6g u ± n,6g w u± n,8g ] = C 8g n+2 g det([e 0 + E 0 J(8g) n ] տn )det([e 0 + ±E 0 J(8g) n 2 ] տn 2) Subtracting the (6g j)th column from the (n j)th column for 1 j, D ± n = det[v± n,1 v± n, u ± n,6g n w 6g u ± n,6g w u± n,8g ], (417) where v ± n,n j = u± n,n j u n,6g j for 1 j Successively, adding the (n j+1)th column to the (6g j 1)th column for 1 j n 2, D ± n = det[v ± n,1 v± n, n u ± n,6g w v n,6g n+1 v n,6g 2 6g u ± n,6g w u± n,8g ], where v n,6g j 1 ± = u± n,6g j 1 v± n,n j+1 for 1 j n 2 Note that the columns (n+1,,6g n) are not changed Put M ± n := [ v ± n,1 v± n, n u ± n,6g w v n,6g n+1 v n,6g 2 By the deformation of the matrix, we observe that [M ± n ] տ4g = [E + 0 E 0 J(8g) n ] տ4g Therefore, by applying Lemma 33 to the columns (n+1,n+2,,4g), D n ± = C4g n g det([e 0 + E 0 J(8g) n ] տn )det([m n ± ] ց4g) 6g u ± n,6g w u± n,8g] Here we note that [M n ± ] ց2g+n = [E 0 + E 0 J(8g) n 2 ] տ2g+n and that the columns (4g+1,,6g n) are not changed by the deformation of matrix Then, by applying Lemma 33 to the rows (1,,2g n) of [M n ± ] ց2g+n, D n ± = C 6g 2n g det([e 0 + E 0 J(8g) n ] տn )det([e 0 + E 0 J(8g) n 2 ] տ2g+n) Now (417) is proved = C 8g 2n+2 g det([e 0 + E 0 J(8g) n ] տn )det([e 0 + E 0 J(8g) n 2 ] տn 2)
27 Lemma 414 We have for every 1 n 2g Proof If 1 n 2g, we have Therefore, INVERSE PROBLEMS FOR CANONICAL SYSTEMS 27 det(e 0 + ±E 0 J(8g) n ) = C 4g g det(e+ 0 ±e 0 J(4g) n ) = C 6g 1 g E 0 + ±E 0 J(8g) n = det(e + ±E J (2g+1) n ) e + 0 ±e 0 J(4g) n 0 0 e + 0 det(e 0 + ±E 0 J(8g) n ) = C 4g g det(e+ 0 ±e 0 J(4g) n ), since det(e + 0 ) = C4g g Further, if 1 n 2g, we have e + 0 ±e 0 J(4g) n = E + ±E J (2g+1) n 0 [e + 0 ] ց(2g 1) Hence det(e 0 + ±E 0 J(8g) n ) = C 6g 1 g det(e + ±E J n (2g+1) ), and we complete the proof Lemma 415 We have det(e 0 + ±E 1,1 J) = det(e+ 0 ) and det(e 0 + ±E n,1 J) = det(e+ 0 ±E J) for 2 n 2g Proof By the proof of Lemma 49, for det(e 0 + ±E J) = C4g 2n+2 g det(a +B )det(a B ) (418) A = [e + 0 ] տ2g+, B = [e 2, ] ւ2g+ Ontheotherhand,applyingLemma33tothecolumns(2g+n+1,,4g) ofe 0 + ±E n,1 J, and then applying Lemma 33 to the rows (2g +n+1,4g) of the resulting matrix, [ det(e 0 + An ±B ±E n,1j) = C4g 2n g det ], ±B n A n [ where B = [e 2, ] An ±B ւ2g+n Applying Lemma 33 to the first row of ] ±B n A n and then applying Lemma 33 to the (4g +2)th column of the resulting matrix, An ±B det[ ] = C ±B n A g 2 det A ±B n ±B A Hence = C 2 g det(a +B )det(a B ) det(e 0 + ±E n,1j) = C4g 2n+2 g det(a +B )det(a B ) (419) We complete the proof by comparing (418) and (419)
Determinants Chapter 3 of Lay
Determinants Chapter of Lay Dr. Doreen De Leon Math 152, Fall 201 1 Introduction to Determinants Section.1 of Lay Given a square matrix A = [a ij, the determinant of A is denoted by det A or a 11 a 1j
More informationChapter 2:Determinants. Section 2.1: Determinants by cofactor expansion
Chapter 2:Determinants Section 2.1: Determinants by cofactor expansion [ ] a b Recall: The 2 2 matrix is invertible if ad bc 0. The c d ([ ]) a b function f = ad bc is called the determinant and it associates
More informationDeterminants. Beifang Chen
Determinants Beifang Chen 1 Motivation Determinant is a function that each square real matrix A is assigned a real number, denoted det A, satisfying certain properties If A is a 3 3 matrix, writing A [u,
More informationELEMENTARY LINEAR ALGEBRA WITH APPLICATIONS. 1. Linear Equations and Matrices
ELEMENTARY LINEAR ALGEBRA WITH APPLICATIONS KOLMAN & HILL NOTES BY OTTO MUTZBAUER 11 Systems of Linear Equations 1 Linear Equations and Matrices Numbers in our context are either real numbers or complex
More information1 Matrices and Systems of Linear Equations
March 3, 203 66. Systems of Linear Equations Matrices and Systems of Linear Equations An m n matrix is an array A = a ij of the form a a n a 2 a 2n... a m a mn where each a ij is a real or complex number.
More informationSPRING 2006 PRELIMINARY EXAMINATION SOLUTIONS
SPRING 006 PRELIMINARY EXAMINATION SOLUTIONS 1A. Let G be the subgroup of the free abelian group Z 4 consisting of all integer vectors (x, y, z, w) such that x + 3y + 5z + 7w = 0. (a) Determine a linearly
More informationMATRIX ALGEBRA AND SYSTEMS OF EQUATIONS. + + x 1 x 2. x n 8 (4) 3 4 2
MATRIX ALGEBRA AND SYSTEMS OF EQUATIONS SYSTEMS OF EQUATIONS AND MATRICES Representation of a linear system The general system of m equations in n unknowns can be written a x + a 2 x 2 + + a n x n b a
More informationThe Determinant: a Means to Calculate Volume
The Determinant: a Means to Calculate Volume Bo Peng August 16, 2007 Abstract This paper gives a definition of the determinant and lists many of its wellknown properties Volumes of parallelepipeds are
More informationProblem 1A. Suppose that f is a continuous real function on [0, 1]. Prove that
Problem 1A. Suppose that f is a continuous real function on [, 1]. Prove that lim α α + x α 1 f(x)dx = f(). Solution: This is obvious for f a constant, so by subtracting f() from both sides we can assume
More information1. General Vector Spaces
1.1. Vector space axioms. 1. General Vector Spaces Definition 1.1. Let V be a nonempty set of objects on which the operations of addition and scalar multiplication are defined. By addition we mean a rule
More informationUpper triangular matrices and Billiard Arrays
Linear Algebra and its Applications 493 (2016) 508 536 Contents lists available at ScienceDirect Linear Algebra and its Applications www.elsevier.com/locate/laa Upper triangular matrices and Billiard Arrays
More informationFinitedimensional spaces. C n is the space of ntuples x = (x 1,..., x n ) of complex numbers. It is a Hilbert space with the inner product
Chapter 4 Hilbert Spaces 4.1 Inner Product Spaces Inner Product Space. A complex vector space E is called an inner product space (or a prehilbert space, or a unitary space) if there is a mapping (, )
More informatione j = Ad(f i ) 1 2a ij/a ii
A characterization of generalized KacMoody algebras. J. Algebra 174, 10731079 (1995). Richard E. Borcherds, D.P.M.M.S., 16 Mill Lane, Cambridge CB2 1SB, England. Generalized KacMoody algebras can be
More informationMath Camp Lecture 4: Linear Algebra. Xiao Yu Wang. Aug 2010 MIT. Xiao Yu Wang (MIT) Math Camp /10 1 / 88
Math Camp 2010 Lecture 4: Linear Algebra Xiao Yu Wang MIT Aug 2010 Xiao Yu Wang (MIT) Math Camp 2010 08/10 1 / 88 Linear Algebra Game Plan Vector Spaces Linear Transformations and Matrices Determinant
More informationWeek 910: Recurrence Relations and Generating Functions
Week 910: Recurrence Relations and Generating Functions April 3, 2017 1 Some number sequences An infinite sequence (or just a sequence for short is an ordered array a 0, a 1, a 2,..., a n,... of countably
More informationENGR1100 Introduction to Engineering Analysis. Lecture 21
ENGR1100 Introduction to Engineering Analysis Lecture 21 Lecture outline Procedure (algorithm) for finding the inverse of invertible matrix. Investigate the system of linear equation and invertibility
More informationTheorem A.1. If A is any nonzero m x n matrix, then A is equivalent to a partitioned matrix of the form. k k nk. mk k mk nk
I. REVIEW OF LINEAR ALGEBRA A. Equivalence Definition A1. If A and B are two m x n matrices, then A is equivalent to B if we can obtain B from A by a finite sequence of elementary row or elementary column
More informationand let s calculate the image of some vectors under the transformation T.
Chapter 5 Eigenvalues and Eigenvectors 5. Eigenvalues and Eigenvectors Let T : R n R n be a linear transformation. Then T can be represented by a matrix (the standard matrix), and we can write T ( v) =
More informationContinued fractions for complex numbers and values of binary quadratic forms
arxiv:110.3754v1 [math.nt] 18 Feb 011 Continued fractions for complex numbers and values of binary quadratic forms S.G. Dani and Arnaldo Nogueira February 1, 011 Abstract We describe various properties
More informationRecall the convention that, for us, all vectors are column vectors.
Some linear algebra Recall the convention that, for us, all vectors are column vectors. 1. Symmetric matrices Let A be a real matrix. Recall that a complex number λ is an eigenvalue of A if there exists
More informationHASSEMINKOWSKI THEOREM
HASSEMINKOWSKI THEOREM KIM, SUNGJIN 1. Introduction In rough terms, a localglobal principle is a statement that asserts that a certain property is true globally if and only if it is true everywhere locally.
More information5.6. PSEUDOINVERSES 101. A H w.
5.6. PSEUDOINVERSES 0 Corollary 5.6.4. If A is a matrix such that A H A is invertible, then the leastsquares solution to Av = w is v = A H A ) A H w. The matrix A H A ) A H is the left inverse of A and
More informationNotes on Mathematics
Notes on Mathematics  12 1 Peeyush Chandra, A. K. Lal, V. Raghavendra, G. Santhanam 1 Supported by a grant from MHRD 2 Contents I Linear Algebra 7 1 Matrices 9 1.1 Definition of a Matrix......................................
More informationGATE Engineering Mathematics SAMPLE STUDY MATERIAL. Postal Correspondence Course GATE. Engineering. Mathematics GATE ENGINEERING MATHEMATICS
SAMPLE STUDY MATERIAL Postal Correspondence Course GATE Engineering Mathematics GATE ENGINEERING MATHEMATICS ENGINEERING MATHEMATICS GATE Syllabus CIVIL ENGINEERING CE CHEMICAL ENGINEERING CH MECHANICAL
More informationA matrix A is invertible i det(a) 6= 0.
Chapter 4 Determinants 4.1 Definition Using Expansion by Minors Every square matrix A has a number associated to it and called its determinant, denotedbydet(a). One of the most important properties of
More informationTopic 2 Quiz 2. choice C implies B and B implies C. correctchoice C implies B, but B does not imply C
Topic 1 Quiz 1 text A reduced rowechelon form of a 3 by 4 matrix can have how many leading one s? choice must have 3 choice may have 1, 2, or 3 correctchoice may have 0, 1, 2, or 3 choice may have 0,
More informationLinear Algebra review Powers of a diagonalizable matrix Spectral decomposition
Linear Algebra review Powers of a diagonalizable matrix Spectral decomposition Prof. Tesler Math 283 Fall 2016 Also see the separate version of this with Matlab and R commands. Prof. Tesler Diagonalizing
More informationDeterminants An Introduction
Determinants An Introduction Professor Je rey Stuart Department of Mathematics Paci c Lutheran University Tacoma, WA 9844 USA je rey.stuart@plu.edu The determinant is a useful function that takes a square
More informationBoolean InnerProduct Spaces and Boolean Matrices
Boolean InnerProduct Spaces and Boolean Matrices Stan Gudder Department of Mathematics, University of Denver, Denver CO 80208 Frédéric Latrémolière Department of Mathematics, University of Denver, Denver
More informationFunctional Analysis. James Emery. Edit: 8/7/15
Functional Analysis James Emery Edit: 8/7/15 Contents 0.1 Green s functions in Ordinary Differential Equations...... 2 0.2 Integral Equations........................ 2 0.2.1 Fredholm Equations...................
More informationCalculating determinants for larger matrices
Day 26 Calculating determinants for larger matrices We now proceed to define det A for n n matrices A As before, we are looking for a function of A that satisfies the product formula det(ab) = det A det
More informationMATH 1210 Assignment 4 Solutions 16RT1
MATH 1210 Assignment 4 Solutions 16RT1 Attempt all questions and show all your work. Due November 13, 2015. 1. Prove using mathematical induction that for any n 2, and collection of n m m matrices A 1,
More informationMath 18, Linear Algebra, Lecture C00, Spring 2017 Review and Practice Problems for Final Exam
Math 8, Linear Algebra, Lecture C, Spring 7 Review and Practice Problems for Final Exam. The augmentedmatrix of a linear system has been transformed by row operations into 5 4 8. Determine if the system
More informationRefined Inertia of Matrix Patterns
Electronic Journal of Linear Algebra Volume 32 Volume 32 (2017) Article 24 2017 Refined Inertia of Matrix Patterns Kevin N. Vander Meulen Redeemer University College, kvanderm@redeemer.ca Jonathan Earl
More informationMath 121 Homework 5: Notes on Selected Problems
Math 121 Homework 5: Notes on Selected Problems 12.1.2. Let M be a module over the integral domain R. (a) Assume that M has rank n and that x 1,..., x n is any maximal set of linearly independent elements
More informationA matrix is a rectangular array of. objects arranged in rows and columns. The objects are called the entries. is called the size of the matrix, and
Section 5.5. Matrices and Vectors A matrix is a rectangular array of objects arranged in rows and columns. The objects are called the entries. A matrix with m rows and n columns is called an m n matrix.
More informationKernel Method: Data Analysis with Positive Definite Kernels
Kernel Method: Data Analysis with Positive Definite Kernels 2. Positive Definite Kernel and Reproducing Kernel Hilbert Space Kenji Fukumizu The Institute of Statistical Mathematics. Graduate University
More informationLinear Algebra II. Ulrike Tillmann. January 4, 2018
Linear Algebra II Ulrike Tillmann January 4, 208 This course is a continuation of Linear Algebra I and will foreshadow much of what will be discussed in more detail in the Linear Algebra course in Part
More informationMath Linear Algebra Final Exam Review Sheet
Math 151 Linear Algebra Final Exam Review Sheet Vector Operations Vector addition is a componentwise operation. Two vectors v and w may be added together as long as they contain the same number n of
More informationMath 240 Calculus III
The Calculus III Summer 2015, Session II Wednesday, July 8, 2015 Agenda 1. of the determinant 2. determinants 3. of determinants What is the determinant? Yesterday: Ax = b has a unique solution when A
More informationdet(ka) = k n det A.
Properties of determinants Theorem. If A is n n, then for any k, det(ka) = k n det A. Multiplying one row of A by k multiplies the determinant by k. But ka has every row multiplied by k, so the determinant
More informationFundamentals of Linear Algebra. Marcel B. Finan Arkansas Tech University c All Rights Reserved
Fundamentals of Linear Algebra Marcel B. Finan Arkansas Tech University c All Rights Reserved 2 PREFACE Linear algebra has evolved as a branch of mathematics with wide range of applications to the natural
More informationLinear Algebra. Workbook
Linear Algebra Workbook Paul Yiu Department of Mathematics Florida Atlantic University Last Update: November 21 Student: Fall 2011 Checklist Name: A B C D E F F G H I J 1 2 3 4 5 6 7 8 9 10 xxx xxx xxx
More informationGeneralized eigenvector  Wikipedia, the free encyclopedia
1 of 30 18/03/2013 20:00 Generalized eigenvector From Wikipedia, the free encyclopedia In linear algebra, for a matrix A, there may not always exist a full set of linearly independent eigenvectors that
More informationMathematical Methods wk 2: Linear Operators
John Magorrian, magog@thphysoxacuk These are workinprogress notes for the secondyear course on mathematical methods The most uptodate version is available from http://wwwthphysphysicsoxacuk/people/johnmagorrian/mm
More informationCHAPTER 6. Representations of compact groups
CHAPTER 6 Representations of compact groups Throughout this chapter, denotes a compact group. 6.1. Examples of compact groups A standard theorem in elementary analysis says that a subset of C m (m a positive
More informationOptimization Theory. A Concise Introduction. Jiongmin Yong
October 11, 017 16:5 wsbook9x6 Book Title Optimization Theory 01708Lecture Notes page 1 1 Optimization Theory A Concise Introduction Jiongmin Yong Optimization Theory 01708Lecture Notes page Optimization
More informationFormula for the inverse matrix. Cramer s rule. Review: 3 3 determinants can be computed expanding by any row or column
Math 20F Linear Algebra Lecture 18 1 Determinants, n n Review: The 3 3 case Slide 1 Determinants n n (Expansions by rows and columns Relation with Gauss elimination matrices: Properties) Formula for the
More informationLinear algebra and applications to graphs Part 1
Linear algebra and applications to graphs Part 1 Written up by Mikhail Belkin and Moon Duchin Instructor: Laszlo Babai June 17, 2001 1 Basic Linear Algebra Exercise 1.1 Let V and W be linear subspaces
More informationMath 313 (Linear Algebra) Exam 2  Practice Exam
Name: Student ID: Section: Instructor: Math 313 (Linear Algebra) Exam 2  Practice Exam Instructions: For questions which require a written answer, show all your work. Full credit will be given only if
More informationSPRING OF 2008 D. DETERMINANTS
18024 SPRING OF 2008 D DETERMINANTS In many applications of linear algebra to calculus and geometry, the concept of a determinant plays an important role This chapter studies the basic properties of determinants
More informationReview from Bootcamp: Linear Algebra
Review from Bootcamp: Linear Algebra D. Alex Hughes October 27, 2014 1 Properties of Estimators 2 Linear Algebra Addition and Subtraction Transpose Multiplication Cross Product Trace 3 Special Matrices
More informationOn Projective Planes
CUPPSATS 2002:02 TFM, Mid Sweden University 851 70 Sundsvall Tel: 06014 86 00 On Projective Planes 1 2 7 4 3 6 5 The Fano plane, the smallest projective plane. By Johan Kåhrström ii iii Abstract It was
More informationDepartment of Mathematics, University of California, Berkeley. GRADUATE PRELIMINARY EXAMINATION, Part A Fall Semester 2016
Department of Mathematics, University of California, Berkeley YOUR 1 OR 2 DIGIT EXAM NUMBER GRADUATE PRELIMINARY EXAMINATION, Part A Fall Semester 2016 1. Please write your 1 or 2digit exam number on
More informationMath 443 Differential Geometry Spring Handout 3: Bilinear and Quadratic Forms This handout should be read just before Chapter 4 of the textbook.
Math 443 Differential Geometry Spring 2013 Handout 3: Bilinear and Quadratic Forms This handout should be read just before Chapter 4 of the textbook. Endomorphisms of a Vector Space This handout discusses
More information1 Determinants. 1.1 Determinant
1 Determinants [SB], Chapter 9, p.188196. [SB], Chapter 26, p.719739. Bellow w ll study the central question: which additional conditions must satisfy a quadratic matrix A to be invertible, that is to
More informationarxiv: v1 [math.fa] 13 Dec 2016 CHARACTERIZATION OF TRUNCATED TOEPLITZ OPERATORS BY CONJUGATIONS
arxiv:1612.04406v1 [math.fa] 13 Dec 2016 CHARACTERIZATION OF TRUNCATED TOEPLITZ OPERATORS BY CONJUGATIONS KAMILA KLIŚGARLICKA, BARTOSZ ŁANUCHA AND MAREK PTAK ABSTRACT. Truncated Toeplitz operators are
More information1 The linear algebra of linear programs (March 15 and 22, 2015)
1 The linear algebra of linear programs (March 15 and 22, 2015) Many optimization problems can be formulated as linear programs. The main features of a linear program are the following: Variables are real
More informationIr O D = D = ( ) Section 2.6 Example 1. (Bottom of page 119) dim(v ) = dim(l(v, W )) = dim(v ) dim(f ) = dim(v )
Section 3.2 Theorem 3.6. Let A be an m n matrix of rank r. Then r m, r n, and, by means of a finite number of elementary row and column operations, A can be transformed into the matrix ( ) Ir O D = 1 O
More informationDETERMINANTS. , x 2 = a 11b 2 a 21 b 1
DETERMINANTS 1 Solving linear equations The simplest type of equations are linear The equation (1) ax = b is a linear equation, in the sense that the function f(x) = ax is linear 1 and it is equated to
More informationk=1 ( 1)k+j M kj detm kj. detm = ad bc. = 1 ( ) 2 ( )+3 ( ) = = 0
4 Determinants The determinant of a square matrix is a scalar (i.e. an element of the field from which the matrix entries are drawn which can be associated to it, and which contains a surprisingly large
More informationChapter Two Elements of Linear Algebra
Chapter Two Elements of Linear Algebra Previously, in chapter one, we have considered single first order differential equations involving a single unknown function. In the next chapter we will begin to
More informationAn introduction to some aspects of functional analysis
An introduction to some aspects of functional analysis Stephen Semmes Rice University Abstract These informal notes deal with some very basic objects in functional analysis, including norms and seminorms
More informationTHEODORE VORONOV DIFFERENTIABLE MANIFOLDS. Fall Last updated: November 26, (Under construction.)
4 Vector fields Last updated: November 26, 2009. (Under construction.) 4.1 Tangent vectors as derivations After we have introduced topological notions, we can come back to analysis on manifolds. Let M
More informationMATHEMATICS. IMPORTANT FORMULAE AND CONCEPTS for. Final Revision CLASS XII CHAPTER WISE CONCEPTS, FORMULAS FOR QUICK REVISION.
MATHEMATICS IMPORTANT FORMULAE AND CONCEPTS for Final Revision CLASS XII 2016 17 CHAPTER WISE CONCEPTS, FORMULAS FOR QUICK REVISION Prepared by M. S. KUMARSWAMY, TGT(MATHS) M. Sc. Gold Medallist (Elect.),
More informationc c c c c c c c c c a 3x3 matrix C= has a determinant determined by
Linear Algebra Determinants and Eigenvalues Introduction: Many important geometric and algebraic properties of square matrices are associated with a single real number revealed by what s known as the determinant.
More informationTRANSLATION INVARIANCE OF FOCK SPACES
TRANSLATION INVARIANCE OF FOCK SPACES KEHE ZHU ABSTRACT. We show that there is only one Hilbert space of entire functions that is invariant under the action of naturally defined weighted translations.
More informationEE/ACM Applications of Convex Optimization in Signal Processing and Communications Lecture 2
EE/ACM 150  Applications of Convex Optimization in Signal Processing and Communications Lecture 2 Andre Tkacenko Signal Processing Research Group Jet Propulsion Laboratory April 5, 2012 Andre Tkacenko
More informationQuadratic forms. Here. Thus symmetric matrices are diagonalizable, and the diagonalization can be performed by means of an orthogonal matrix.
Quadratic forms 1. Symmetric matrices An n n matrix (a ij ) n ij=1 with entries on R is called symmetric if A T, that is, if a ij = a ji for all 1 i, j n. We denote by S n (R) the set of all n n symmetric
More informationM.6. Rational canonical form
book 2005/3/26 16:06 page 383 #397 M.6. RATIONAL CANONICAL FORM 383 M.6. Rational canonical form In this section we apply the theory of finitely generated modules of a principal ideal domain to study the
More information2 Constructions of manifolds. (Solutions)
2 Constructions of manifolds. (Solutions) Last updated: February 16, 2012. Problem 1. The state of a double pendulum is entirely defined by the positions of the moving ends of the two simple pendula of
More informationProblems in Linear Algebra and Representation Theory
Problems in Linear Algebra and Representation Theory (Most of these were provided by Victor Ginzburg) The problems appearing below have varying level of difficulty. They are not listed in any specific
More informationLIE ALGEBRAS AND LIE BRACKETS OF LIE GROUPS MATRIX GROUPS QIZHEN HE
LIE ALGEBRAS AND LIE BRACKETS OF LIE GROUPS MATRIX GROUPS QIZHEN HE Abstract. The goal of this paper is to study Lie groups, specifically matrix groups. We will begin by introducing two examples: GL n
More informationTOEPLITZ OPERATORS. Toeplitz studied infinite matrices with NWSE diagonals constant. f e C :
TOEPLITZ OPERATORS EFTON PARK 1. Introduction to Toeplitz Operators Otto Toeplitz lived from 18811940 in Goettingen, and it was pretty rough there, so he eventually went to Palestine and eventually contracted
More informationDeterminants by Cofactor Expansion (III)
Determinants by Cofactor Expansion (III) Comment: (Reminder) If A is an n n matrix, then the determinant of A can be computed as a cofactor expansion along the jth column det(a) = a1j C1j + a2j C2j +...
More informationNo books, no notes, no calculators. You must show work, unless the question is a true/false, yes/no, or fillintheblank question.
Math 304 Final Exam (May 8) Spring 206 No books, no notes, no calculators. You must show work, unless the question is a true/false, yes/no, or fillintheblank question. Name: Section: Question Points
More informationChapter 2 Spectra of Finite Graphs
Chapter 2 Spectra of Finite Graphs 2.1 Characteristic Polynomials Let G = (V, E) be a finite graph on n = V vertices. Numbering the vertices, we write down its adjacency matrix in an explicit form of n
More informationLecture 2 INFMAT : , LU, symmetric LU, Positve (semi)definite, Cholesky, SemiCholesky
Lecture 2 INFMAT 4350 2009: 7.17.6, LU, symmetric LU, Positve (semi)definite, Cholesky, SemiCholesky Tom Lyche and Michael Floater Centre of Mathematics for Applications, Department of Informatics,
More informationSpectral radius, symmetric and positive matrices
Spectral radius, symmetric and positive matrices Zdeněk Dvořák April 28, 2016 1 Spectral radius Definition 1. The spectral radius of a square matrix A is ρ(a) = max{ λ : λ is an eigenvalue of A}. For an
More informationAn example of generalization of the Bernoulli numbers and polynomials to any dimension
An example of generalization of the Bernoulli numbers and polynomials to any dimension Olivier Bouillot, VillebonGeorges Charpak institute, France C.A.L.I.N. team seminary. Tuesday, 29 th March 26. Introduction
More information1 Fields and vector spaces
1 Fields and vector spaces In this section we revise some algebraic preliminaries and establish notation. 1.1 Division rings and fields A division ring, or skew field, is a structure F with two binary
More informationEigenvectors and Hermitian Operators
7 71 Eigenvalues and Eigenvectors Basic Definitions Let L be a linear operator on some given vector space V A scalar λ and a nonzero vector v are referred to, respectively, as an eigenvalue and corresponding
More informationCHAPTER VIII HILBERT SPACES
CHAPTER VIII HILBERT SPACES DEFINITION Let X and Y be two complex vector spaces. A map T : X Y is called a conjugatelinear transformation if it is a reallinear transformation from X into Y, and if T (λx)
More informationReview of some mathematical tools
MATHEMATICAL FOUNDATIONS OF SIGNAL PROCESSING Fall 2016 Benjamín Béjar Haro, Mihailo Kolundžija, Reza Parhizkar, Adam Scholefield Teaching assistants: Golnoosh Elhami, Hanjie Pan Review of some mathematical
More informationTrace Class Operators and Lidskii s Theorem
Trace Class Operators and Lidskii s Theorem Tom Phelan Semester 2 2009 1 Introduction The purpose of this paper is to provide the reader with a selfcontained derivation of the celebrated Lidskii Trace
More information2. Complex Analytic Functions
2. Complex Analytic Functions John Douglas Moore July 6, 2011 Recall that if A and B are sets, a function f : A B is a rule which assigns to each element a A a unique element f(a) B. In this course, we
More information1 Compact and Precompact Subsets of H
Compact Sets and Compact Operators by Francis J. Narcowich November, 2014 Throughout these notes, H denotes a separable Hilbert space. We will use the notation B(H) to denote the set of bounded linear
More information4. Linear transformations as a vector space 17
4 Linear transformations as a vector space 17 d) 1 2 0 0 1 2 0 0 1 0 0 0 1 2 3 4 32 Let a linear transformation in R 2 be the reflection in the line = x 2 Find its matrix 33 For each linear transformation
More informationw T 1 w T 2. w T n 0 if i j 1 if i = j
Lyapunov Operator Let A F n n be given, and define a linear operator L A : C n n C n n as L A (X) := A X + XA Suppose A is diagonalizable (what follows can be generalized even if this is not possible 
More informationREPRESENTATION THEORY WEEK 7
REPRESENTATION THEORY WEEK 7 1. Characters of L k and S n A character of an irreducible representation of L k is a polynomial function constant on every conjugacy class. Since the set of diagonalizable
More informationCOMMUTING PAIRS AND TRIPLES OF MATRICES AND RELATED VARIETIES
COMMUTING PAIRS AND TRIPLES OF MATRICES AND RELATED VARIETIES ROBERT M. GURALNICK AND B.A. SETHURAMAN Abstract. In this note, we show that the set of all commuting dtuples of commuting n n matrices that
More informationTOPOLOGICAL COMPLEXITY OF 2TORSION LENS SPACES AND ku(co)homology
TOPOLOGICAL COMPLEXITY OF 2TORSION LENS SPACES AND ku(co)homology DONALD M. DAVIS Abstract. We use kucohomology to determine lower bounds for the topological complexity of mod2 e lens spaces. In the
More informationThe MatrixTree Theorem
The MatrixTree Theorem Christopher Eur March 22, 2015 Abstract: We give a brief introduction to graph theory in light of linear algebra. Our results culminates in the proof of MatrixTree Theorem. 1 Preliminaries
More informationAN ELEMENTARY PROOF OF THE SPECTRAL RADIUS FORMULA FOR MATRICES
AN ELEMENTARY PROOF OF THE SPECTRAL RADIUS FORMULA FOR MATRICES JOEL A. TROPP Abstract. We present an elementary proof that the spectral radius of a matrix A may be obtained using the formula ρ(a) lim
More informationMath 594, HW2  Solutions
Math 594, HW2  Solutions Gilad Pagi, Feng Zhu February 8, 2015 1 a). It suffices to check that NA is closed under the group operation, and contains identities and inverses: NA is closed under the group
More informationExtra Problems for Math 2050 Linear Algebra I
Extra Problems for Math 5 Linear Algebra I Find the vector AB and illustrate with a picture if A = (,) and B = (,4) Find B, given A = (,4) and [ AB = A = (,4) and [ AB = 8 If possible, express x = 7 as
More information* 8 Groups, with Appendix containing Rings and Fields.
* 8 Groups, with Appendix containing Rings and Fields Binary Operations Definition We say that is a binary operation on a set S if, and only if, a, b, a b S Implicit in this definition is the idea that
More informationMath 225 Linear Algebra II Lecture Notes. John C. Bowman University of Alberta Edmonton, Canada
Math 225 Linear Algebra II Lecture Notes John C Bowman University of Alberta Edmonton, Canada March 23, 2017 c 2010 John C Bowman ALL RIGHTS RESERVED Reproduction of these lecture notes in any form, in
More informationMTH 102A  Linear Algebra II Semester
MTH 0A  Linear Algebra  056II Semester Arbind Kumar Lal P Field A field F is a set from which we choose our coefficients and scalars Expected properties are ) a+b and a b should be defined in it )
More informationRepresentation theory and quantum mechanics tutorial Spin and the hydrogen atom
Representation theory and quantum mechanics tutorial Spin and the hydrogen atom Justin Campbell August 3, 2017 1 Representations of SU 2 and SO 3 (R) 1.1 The following observation is long overdue. Proposition
More information