Optimal Hankel norm approximation for infinite-dimensional systems

Size: px
Start display at page:

Download "Optimal Hankel norm approximation for infinite-dimensional systems"

Transcription

1 Optimal Hankel norm approximation for infinite-dimensional systems A.J. Sasane and R.F. Curtain Department of Mathematics, University of Groningen, P.O.Box 800, 9700 AV Groningen, The Netherlands. Keywords: Infinite-dimensional systems, optimal Hankel norm approximation. Abstract The optimal Hankel norm approximation problem is solved under the assumptions that the system Σ(A, B, C) is an exponentially stable infinite-dimensional system with bounded input and output operators. An explicit parameterization of all solutions is obtained in terms of the system parameters A, B, C. 1 Introduction An important question in the design of controllers for large scale systems is whether a model can be simplified without undue loss of accuracy. One approach is to use a low-order optimal Hankel norm approximation; for finite-dimensional systems this was solved in Glover 9 and extended to a class of infinite-dimensional systems in Glover et al. 10. To approximate stable transfer functions G(s) H (C p m ) in the L norm we suppose that G(s) has a compact Hankel operator Γ : L 2 (0, ; C m ) L 2 (0, ; C p ) which is defined by (Γu)(t) 0 h(t + s)u(s)ds u L 2 (0, ; C m ), (1) where h( ) L 1 (0, ; C p m ) denotes the impulse response of the system. Γ then has countably many singular values σ 1 σ 2... and these are also called the Hankel singular values of G. In Glover 9 and Glover et al. 10, it was shown that if h L 1 L 2 and Γ is nuclear (i.e., i1 σ i < ), then there exist finite-dimensional L approximations to G; error estimates were also provided. These approximations were in terms of truncated balanced realizations. These also yielded approximate solutions to the related optimal Hankel norm approximation problem: Find all K( s) H,l (C p m ) such that G + K σ for σ l > σ > σ l+1, in the L norm, where H,l (C p m ) denotes the set of complex p m matrix valued functions X( ) of a complex variable with a decomposition X Ĝ +F, where Ĝ is the matrix transfer function (2) of a system of MacMillan degree at most equal to l, with all its poles in the open right half-plane, and F H (C p m ). It is well-known that (see Adamjan et al. 1) inf G + K σ l+1. (3) K( s) H,l (C p m ) In the same paper the optimal Hankel norm approximation problem is solved for the scalar case, and solutions for other classes of functions G have been obtained in Ball and Helton 2 and Nikol skii 12. A different approach to solving the optimal Hankel norm approximation problem was taken in Ball and Ran 3, Ran 13 and Curtain and Ran 6, which led to explicit formulas for all solutions K( s) in terms of an arbitrary realization of G(s). The starting point was to quote a result from Ball and Helton 2, which states that the optimal Hankel norm approximation problem is equivalent to solving a certain J-spectral factorization problem. Then a solution is constructed from a given realization of G(s). However, if one looks for the result quoted from 2 in these three papers, one realizes that this is not an obvious corollary of the very abstract and general theory in 2. There remains a gap between 2 and the results obtained in 3, 13 and 6. This motivated the self-contained proofs of the suboptimal Nehari extension problem in Curtain and Zwart 7, Curtain and Ichikawa 4 and Oostveen and Curtain 5 which did not rely on the abstract results in 2. The suboptimal Nehari extension problem is a special case of the optimal Hankel norm approximation problem; one seeks all K( s) H (C p m ) satisfying G + K σ for a given σ > Γ. In this paper we provide a self-contained solution to the optimal Hankel norm approximation problem for the class of exponentially stable infinite-dimensional systems with bounded input and output operators and finite-dimensional input and output spaces. A key step in our derivation is to appeal to the recent inertia results for Lyapunov equations from Sasane and Curtain 15. If we specialize our results to the Nehari problem case, we obtain a streamlined version of the proof in Curtain and Zwart 8. We outline the contents of the following sections. In section 2 we introduce some function spaces and some of their properties that will be used in the sequel. In section 3, using the results from Sasane and Curtain 15, we prove the existence of a solution to the key J-spectral factorization problem and in section 4 we prove the existence of a solution to

2 the Hankel norm approximation problem. Finally, we give a complete characterization of all solutions to the optimal Hankel norm approximation problem in section 5. 2 An algebra of transfer functions Before discussing the problem, let us introduce some function spaces which will be used in the sequel. 1. H (C c p m ) denotes the set of complex p m matrix valued functions defined in the closed right half-plane, which are bounded and holomorphic in C + 0, and continuous in C + 0. Hc, with point-wise addition and multiplication, is a commutative ring with identity. 2. H,l c (Cp m ) denotes the set of complex p m matrix valued functions X( ) of a complex variable with a decomposition X Ĝ+F, where Ĝ is the matrix transfer function of a system of MacMillan degree at most equal to l, with all its poles in the open right half-plane, and F H (C c p m ). 3. H,l c (Cp m ) denotes the set of complex p m matrix valued functions X( ) of a complex variable with a decomposition X Ĝ+F, where Ĝ is the matrix transfer function of a system of MacMillan degree equal to l, with all its l poles in the open right half-plane, and F H (C c p m ). 4. S denotes the set of complex valued functions g H c that have a nonzero limit at infinity in C 0 +, finitely many zeros in C 0 +, and they are all contained in the open right half-plane. 5. MH c denotes the set of matrices (of any size) with elements in H c. We now list a few elementary facts concerning elements from this class of transfer functions. Lemma 2.1 Let f( ), g( ) H c. If g S has at most l zeros in C + 0, then fg 1 H c,l (C). Lemma 2.2 If M, N MH c have the same number of columns, M is a square matrix with det(m) S and det(m) has l zeros in the open right half-plane, then NM 1 H,l c (Cp m ). Definition 2.3 Suppose M, N MH c, and have the same number of columns. Then the pair (M, N) is right coprime over MH c if there exist X, Y MH c such that the following Bezout identity holds: XM Y N I s C + 0. (4) Suppose that G H c,l (Cp m ), that the pair (M, N) is right coprime over MH c. If M is such that det M S and G NM 1, we call this a right coprime factorization of G over MH c. Lemma 2.4 If K H,l c (Cp m ), then there exists a right coprime factorization K NM 1, where M is rational, det(m) R (0) has exactly l zeros in C + 0 and they are all contained in C + 0. Lemma 2.5 If K H c,l (Cp m ) and K 1 H c (C p p ), K 2 H c (C m m ), then K 1 KK 2 H c,l (Cp m ). Lemma 2.6 If (N, M) is a right coprime factorization of K H c,l (Cp m ) and V H c (C m m ) is invertible as an element of MH c, then (NV, MV ) is also a right coprime factorization of K. Moreover, any two right coprime factorizations of K H c,l (Cp m ) are unique up to a common right multiplication by an invertible element in MH c. Lemma 2.7 If K NM 1 H,l c (Cp m ), N H (C c p m ), M H c (C m m ), with N, M right coprime, then det(m) has exactly l zeros in C + 0, and they are all contained in the open right half-plane. The following technical lemma will be used in the characterization of solutions. Lemma 2.8 If K( ) MH c,, then given any ɛ > 0, there exists a δ > 0 such that whenever 0 ζ δ, we have K(ζ + j ) K(j ) + ɛ, where denotes the L norm. Proof Let K(s) G(s) + F (s) where G(s) is the matrix rational transfer operator of a system of MacMillan degree, say l, with all its poles in the open right half-plane and F MH. c Let the poles of G be contained in the half-plane C + r for some r > 0. Consider the function θ(s) K(s) (where denotes the standard Euclidean 2-norm, namely P i,j p ij 2 for P C ) defined for s belonging to the infinite strip Ω : {s C 0 Re(s) r}. Clearly, θ( ) is continuous in Ω and holomorphic in the interior of Ω. Using the triangle inequality, it is easy to see that θ( ) is bounded in Ω: s G(s) is bounded in Ω (since all its poles are in C + r ) and s F (s) is bounded in Ω (in fact, in C + 0 ). For any ζ > 0, define M(ζ) sup ω R {θ(ζ +jω)}. Using Theorem 12.8 (page 257, W. Rudin 14), we obtain M(ζ) ( K(j ) ) 1 ζ ζ r M(r) r ( M(r) ( K(j ) ) K(j ) ) ζ r. ( ) ζ Since lim M(r) r ζ 0 K(j ) 1, there exists a δ such that 0 < δ < r, and for any ζ satisfying 0 ζ δ, we have K(ζ + j ) K(j ) + ɛ.

3 3 The existence of a J-spectral factorization A key step in the solution of the Nehari problem was the construction of a solution X(s) to a certain J-spectral factorization provided that σ > Γ. For the optimal Hankel norm problem we have σ l+1 < σ < σ l, but nonetheless, the X(s) factor is precisely the same. The proof of the following theorem is analogous to that of Lemma in Curtain and Zwart 8: Lemma 3.1 Suppose that Σ(A, B, C) is an exponentially stable infinite-dimensional system with B L(C m, Z), C L(Z, C p ) and σ l+1 < σ < σ l. Let X( ) be given by X(s) + σ 2 CLB 0 σ σb Nσ (si + A ) 1 C L C B (5) where N σ (I σ 2 L B L C ) 1. It follows that 1. X( s) MH c is invertible in MH c, and X(s) 1 0 σ 1 σ 2 CLB B (si + A ) 1 Nσ C σ 1 L C B (6) 2. If G(s) C(sI A) 1 B, then I W (s) : p 0 G (s) 0 σ 2 Ip G(s), 0 where we use the notation F (s) : F ( s), has a J- spectral factorization W (s) X (s) X(s), (7) 0 for s jω, ω R. If σ > Γ, it was shown in Curtain and Zwart 8 that X 11 ( s) belongs to MH, c is invertible in MH c and X11 1 ( s) is the transfer operator of an exponentially stable infinite-dimensional system. In the case that σ l+1 < σ < σ l, X11 1 ( s) still exists, but it is not stable. We can show under the extra assumption that there exists no k N such that σ k 0, i.e., the given system system is truly infinite-dimensional (this will be a standing assumption in the rest of the paper), X11 1 ( s) is the sum of a stable part in H (C c p p ) and an antistable rational part with at most l unstable poles. In order to do so we will use Theorem 1.4 from Sasane and Curtain 15, which we quote below: Theorem 3.2 If P L(Z) is a self-adjoint solution of A P z + P Az C Cz z D(A), (8) C has finite rank, and Σ(A,, C) is exponentially detectable, then P has a pure point spectrum in the open left half-plane, and ν(p ) π(a), where ν( ) denotes the number of eigenvalues in the open left half-plane, and π( ) denotes the number of eigenvalues in the open right half-plane. For σ l+1 < σ < σ l, let N σ : (I σ 2 L B L C ) 1 L(Z). We now show that N σ L B is self-adjoint and it has at most l negative eigenvalues. Lemma 3.3 Suppose that Σ(A, B, C) is an exponentially stable infinite-dimensional system with B L(C m, Z), C L(Z, C p ) and σ l+1 < σ < σ l. Then N σ L B L(Z) is a self-adjoint operator, and ν(n σ L B ) l. Proof Sublemma 3.4 J : I σ 2 Γ Γ L(L 2 (0, ; U)) has a spectrum σ(j) contained in (, δ) (δ, ) for some δ > 0, and σ(j) (, δ) consists of exactly l negative eigenvalues. Proof Γ Γ is compact and has a pure point spectrum with 0 as the accumulation point. Considering the resolvent ((I σ 2 Γ Γ) λi) 1, it is easy to see that J has a spectrum which is a shifted version of the spectrum of σ 2 Γ Γ. Finally, since Γ Γ has a pure point spectrum {σ 2 1, σ 2 2,...} with 0 as the accumulation point, and since σ l+1 < σ < σ l, J has exactly l negative eigenvalues. 0 σ(j). Consider the spectral decomposition of L 2 (0, ; U) into L and L + induced by the self-adjoint operator J. L 2 (0, ; U) L L +, and if v + L +, Jv +, v + 0. Let Z Z Z 0+ be the spectral decomposition induced by the self-adjoint bounded operator N σ L B. We have N σ L B z, z (I σ 2 L B L C ) 1 L B z, z L B z, (I σ 2 L C L B ) 1 z L B (I σ 2 L C L B )z 0, z 0, where z 0 (I σ 2 L C L B ) 1 z. Thus N σ L B z, z B z 0, B z 0 σ 2 B L C BB z 0, B z 0 u, u σ 2 B L C Bu, u, where u B z 0, and so N σ L B z, z (I σ 2 Γ Γ)u, u. (9) Now define Ψ : Z L as follows: If z Z, Ψz : Π L B Nσz, where Π L denotes the canonical projection from L 2 (0, ; U) onto L. Ψ is clearly linear. Moreover, Ψ is injective. For if Π L B Nσz 0, v : B Nσz L + and 0 N σ L B z, z N σ L B z, z (I σ 2 Γ Γ)v, v 0.

4 Thus, Jv, v 0. Using Jv +, v + 2 Jv +, v + Jv +, v +, Jv, Jv 0, which implies that Jv 0. But J is injective, and so v 0. Now B N σz 0 BB N σz 0 L B N σz 0 L C L B N σz 0, and so Nσz z, which means that 1 belongs to the spectrum of Nσ. So 1 { σ2 σ 2 σ } n 2 n N {0}, i.e., σ k 0 for some k N, contradicting the infinite-dimensionality of the system. Thus, dim Z dim L l. Hence ν(n σ L B ) l (From Kato 11, pages , it follows that since the dim Z <, the essential spectrum in the open left half-plane is empty and so dim Z ν(n σ L B )). We now relate the number of negative eigenvalues of N σ L B to the unstable part of X 1 11 ( s). Lemma 3.5 Suppose that Σ(A, B, C) is an exponentially stable infinite-dimensional system with B L(C m, Z), C L(Z, C p ) and σ l+1 < σ < σ l. Let X( ) be given by (5). Then X 11 ( s) 1 H c,l (Cp p ). Proof We have X 11 ( s) I p + σ 2 CL B N σ(si A ) 1 C. (10) It follows from Corollary (Curtain and Zwart 8) that I p σ 2 CL B N σ (si A + σ 2 C CL B N σ) 1 C is the inverse of X 11 ( s). Consider the Riccati equation P A z +AP z +N σ BB N σ z σ 2 P C CP z 0 (11) for all z D(A ). As in the proof of Lemma b (Curtain and Zwart 8), it is readily verified that N σ L B L(Z) is a self-adjoint solution of this Riccati equation. Thus if  : A σ 2 C CL B N σ, then N σ L B L(Z) is a selfadjoint solution of the Riccati equation P Âz +  P z + N σ BB N σz + σ 2 P C CP z 0, (12) for all z D(Â). Consequently, N σl B L(Z) is a selfadjoint solution of the Lyapunov equation B P Âz+ P z Nσ B N σ 1 σ CP σ 1 z z CP D(Â). Moreover, ( Â, B N σ σ 1 CL B N σ ) (13) (14) is exponentially detectable, since 0 σ 1 C L(U Y, Z) and  + 0 σ 1 C B Nσ σ 1 CL B Nσ A σ 2 C CL B Nσ + σ 2 C CL B Nσ A generates an exponentially stable semigroup on Z. Thus applying Theorem 3.2, we obtain that π(â) ν(n σl B ) l. Hence, Z Z + Z ; Â+ 0  0  where Â+ : Z + Z +, dim Z + r l, and Â+ has all its r eigenvalues in the open right half-plane.  : D(Â) Z Z is the infinitesimal generator of an exponentially ( s) can be written as a sum of the transfer function of a system with MacMillan degree at most l with all poles in the open right half-plane and a function in MH. c stable semigroup on Z. Thus, X Existence of a solution to the optimal Hankel norm problem In this section we show the existence of a solution to the optimal Hankel norm approximation problem for the class of exponentially stable infinite-dimensional systems Σ(A, B, C) with bounded input and output operators and finite-dimensional input and output spaces. Although it is known (see Adamjan et al. 1) that ; inf G + K σ l+1, (15) K( s) H,l (C p m ) we only need the inequality stated in Theorem 4.2 below which for completeness is proven in the appendix, together with the following technical lemma. Lemma 4.1 Let S : Z 1 Z 2 be a bounded operator from the Hilbert space Z 1 to the Hilbert space Z 2, with Schmidt vector pairs (v i, w i ): Sv i σ i w i, S w i σ i v i, (16) where σ i σ i+1, w i v i 1, i N. If Ŝ : Z 1 Z 2 is an arbitrary bounded linear map of rank l, then S Ŝ σ l+1. (17) Theorem 4.2 Let G L (jr; C p m ) and K( s) H,l (C p m ). Then inf G + K σ l+1. (18) K( s) H,l (C p m ) As in Curtain and Zwart 8, we construct a solution to the optimal Hankel norm problem using K o ( s) V 12 ( s)v 1 22 ( s), where V ( s) X( s) 1. Lemma V 22 ( s) is invertible as an element of MH,, c and V22 1 ( s) Hc,l (Cm m ). 2. V 12 ( s)v 1 22 ( s) Hc,l (Cp m ).

5 Proof 1. X 1 11 ( s) Hc,l (Cp p ). Using XV 1 V 1 X I, it can be checked that V 22 ( s) 1 X 22 ( s) X 21 ( s)x 1 11 ( s)x 12( s). Thus, V 22 ( s) is invertible as an element of MH,. c Moreover, it follows from Lemma 2.5 that V 22 ( s) 1 X 22 ( s) X 21 ( s)x11 1 ( s)x 12( s) H,l c (Cm m ). 2. V 12 ( s)v22 1 ( s) X 1 11 ( s)x 12( s), and so the result follows from Lemma 2.5. Theorem 4.4 There exists K o ( s) H c,l (Cp m ) such that G + K < σ < σ l. Proof Define K o ( s) : V 12 ( s)v22 1 ( s) H,l c (Cp m ). We know from Lemma that V22 1 ( s) G u +G s, where G s MH, c and G u is a strictly proper rational transfer matrix of a system with all its poles in the open right half-plane. Thus, V22 1 (jω) is defined ω R. Hence, G + Ko Ip G Ko 0 Ip G 0 V 0 V22 1, with s jω, ω R, and so we have (G + K o ) (G + K o ) σ 2 G + Ko G + Ko 0 σ 2 0 V22 1 V Ip G 0 0 σ 2 Ip G 0 V 0 V V V 1 22, where we have used equation (7), the definition of V and Lemma Thus it follows that (G + K o )(jω)u 2 σ 2 u 2 V 1 22 (jω)u 2 (19) for u C m and ω R. Since V 22 (jω)v22 1 (jω) I, u V 22 (jω) V22 1 (jω)u. (20) Since V 22 ( s) MH c, there exists a constant M such that V 22 ( s) M. Since V 22 (jω) is invertible on the imaginary axis, M > 0, and we have u C m and ω R. Hence it follows that G + K o < σ < σ l. (22) In the above theorem, we have constructed a solution K o ( s) H c,l (Cp m ) with at most l unstable poles. In fact, any solution to the optimal Hankel norm approximation problem has an unstable rational part of MacMillan degree exactly l. Corollary 4.5 If K( s) H,l (C p m ) is such that G + K σ < σ l, then K( s) Ĝ +F, where Ĝ has MacMillan degree exactly l, with all l poles in the open right half-plane and F H (C p m ), i.e., K( s) H,l (C p m ). Proof Suppose that K( s) Ĝ1 +F 1, where Ĝ1 has MacMillan degree r and all r poles in the open right halfplane, and F 1 H (C p m ). Since K( s) H,l (C p m ), r l. From Theorem 4.2, it follows that σ l > σ G + K inf K( s) H,r (C p m ) G + K σ r+1. Thus σ l > σ r+1, which implies that l < r + 1, and so l r. Hence l r. Finally we collect more precise information concerning our constructed solution K o ( s) V 12 ( s)v22 1 ( s) to the optimal Hankel norm problem from Theorem 4.4. Corollary 4.6 K 0 (s) V 12 ( s)v22 1 ( s) has the following properties 1. K o ( s) H c,l (Cp m ). 2. (V 12 ( s), V 22 ( s)) is a right coprime factorization over MH c of K o ( s) H c,l (Cp m ). 3. det(v 22 ( s)) has no zeros on the imaginary axis, and exactly l zeros in C + 0. Proof 1. From Lemma 4.3.2, we know that K o ( s) : V 12 ( s)v22 1 ( s) Hc,l (Cp m ), and from the proof of Theorem 4.4, we know that it satisfies G + K < σ < σ l. Thus from Corollary 4.5 above, V 12 ( s)v22 1 ( s) Hc,l (Cp m ). 2. V 12 ( s)v 1 22 ( s) Hc,l (Cp m ), V 12 ( s), V 22 ( s) MH c. Moreover, X 22 V 22 ( X 21 )V 12 I, (23) where X 22 ( s), X 21 ( s) MH c. Hence it follows that (V 12 ( s), V 22 ( s)) is a right coprime factorization of V 12 ( s)v 1 22 ( s). 3. This follows from the above and Lemma 2.7. u M V22 1 (jω)u, (21)

6 5 Characterization of solutions In this section we obtain a nice parameterization of all solutions to the optimal Hankel norm approximation problem. First we prove a few properties that will be used in the proof of the characterization theorem. Lemma 5.1 Suppose that Σ(A, B, C) is an exponentially stable infinite-dimensional system with B L(C m, Z), C L(Z, C p ) and σ l+1 < σ < σ l. Let X( ) be given by (5) and V ( s) : X 1 ( s). Then 1. det(v 22 ( s)) has a nonzero limit at infinity in C V 21 ( s) is strictly proper. 3. X 21 ( s) is strictly proper. 4. det(x 22 ( s)) has a nonzero limit at infinity in C + 0. Proof Since V 22 ( s) σ 1 + σ 3 B (si A ) 1 NσL C B V 21 ( s) σ 2 B (si A ) 1 NσC X 21 ( s) σ 1 B N σ(si A ) 1 C X 22 ( s) σ σ 1 B N σ(si A ) 1 L C B, and A is the infinitesimal generator of an exponential semigroup, the results follow. We now parameterize a family of solutions to the optimal Hankel norm approximation problem. Theorem 5.2 Suppose that Σ(A, B, C) is an exponentially stable infinite-dimensional system with B L(C m, Z), C L(Z, C p ) and σ l+1 < σ < σ l. Let X( ) be given by (5). If Q( s) H (C c p m ) satisfies Q 1, and K( s) : R 1 ( s)r 2 ( s) 1, where R1 ( s) R 2 ( s) : X 1 ( s) Q( s), (24) then K( s) H c,l (Cp m ) and G + K σ. Proof Step 1: We show that V22 1 V 21 < 1, where denotes the L norm. X(s) satisfies equation (7); and so taking inverses, we obtain V V 0 1 Ip G 0 0 σ 2 1 G where s jω, ω R. Considering the (2, 2) block of the above gives V 21 V 21 V 22V 22 σ 2 (25) where s jω, ω R. Lemma shows that V 22 is invertible as an element of L, and thus V 1 22 V 21u 2 u 2 σ 2 V 1 22 u 2. (26) Choosing a M > 1 such that V 22 M, we obtain u 2 V 22 2 V 1 22 u 2 M 2 V 1 22 u 2. (27) Hence V ( ) 1 1 M < 2 1, and so we have V < 1. Step 2: We show that det(v 21 ( s)q( s) + V 22 ( s)) has exactly l zeros in C 0 +, and they are contained in the open right half-plane. By Corollary 4.6.3, we know that s det(v 22 ( s)) has no zeros on the imaginary axis, and exactly l zeros in C + 0. So there exists an ɛ > 0 such that all its zeros are contained in the half-plane C ɛ +. From Corollary 4.6.1, we know that K 0 ( s) : V 22 ( s) 1 V 21 ( s) H,l c (Cp m ), and so it follows from Lemma 2.8 applied to K 0 (s), that a δ > 0 such that ɛ > δ and whenever 0 ζ δ, V 1 22 ( ζ j )V 21( ζ j ) < 1. (28) Fix such a ζ > 0. Consider φ(α, s) : det(αv 21 ( ζ s)q( ζ s) + V 22 ( ζ s)), where α 0, 1. φ(0, ) and φ(1, ) are meromorphic (actually holomorphic in C + 0 ζ/2 ) functions on an open set containing C + 0 at infinity in C + 0. We have: with nonzero limits 1. φ(α, s) : 0, 1 jr C is a continuous function. 2. φ(0, jω) det(v 22 ( ζ jω)), and φ(1, jω) det(v 21 ( ζ jω)q( ζ jω) + V 22 ( ζ jω)) αv 22 ( ζ jω) 1 V 21 ( ζ jω)q( ζ jω) is invertible, since αv 1 22 ( ζ jω)v 21( ζ jω)q( ζ jω) < 1. (29) Moreover, since s det(v 22 ( ζ s)) has no zeros on the imaginary axis, it follows that φ(α, jω) is nonzero for all α 0, 1 and ω R. 4. φ(α, ) is nonzero for all α 0, 1, since det(αv 21 ( s ζ)q( s ζ) + V 22 ( s ζ)) has a nonzero limit at infinity in C + 0. So the assumptions of Lemma A.1.18 (Curtain and Zwart 8, page 570) are satisfied by φ and so the Nyquist indices of φ(0, s) and φ(1, s) are the same. Consequently, the number of zeros are the same (the number of poles in each case is zero, since φ(0, s), φ(1, s) are holomorphic in C + 0 ζ/2 ). But when α 0, det(v 22 ( s ζ)) has l zeros in the closed right half-plane C + 0. Thus the number of zeros of φ(1, s) the number of zeros of φ(0, s) l. But since ζ can be chosen

7 arbitrarily small, det(v 21 ( s)q( s) +V 22 ( s)) has exactly l zeros in C 0 +. Moreover, det(v 21 ( s)q( s) +V 22 ( s)) has no zeros on the imaginary axis, since det(i +V22 1 ( jω) V 21( jω) Q( jω)) 0 for all ω R. Step 3: We show that K( s) H,l c (Cp m ). det(v 21 ( s)q( s) +V 22 ( s)) has a nonzero limit at infinity in C + 0, since Q( s) MHc is proper, V 21 ( s) is strictly proper, V 22 ( s) is proper in C + 0 and det(v 22( s)) has a nonzero limit at infinity in C + 0 (see Lemma 5.1). So by Lemma 2.2 and using Step 2 above, it follows that K defined by K( s) (V 11 ( s)q( s) + V 12 ( s)) (V 21 ( s)q( s) + V 22 ( s)) 1 is a well-defined element of H c,l (Cp m ). Step 4: We show that G + K σ. R 2 (jω) V 21 (jω)q(jω) + V 22 (jω), and since det(v 22 ( s)) has no zeros on the imaginary axis and V 1 22 (jω)v 21(jω)Q(jω) < 1 for all ω R, it follows that R 2 (jω) is invertible for every ω R. Thus with s jω, for all ω R, (G + K) (G + K) σ 2 (R2 1 ) Q V Ip G 0 Ip G Q 0 σ 2 V R 0 I 2 1 m (R 1 2 ) (Q Q )R 1 2. Thus G + K σ. Step 5: We show that K( s) H c,l (Cp m ). Finally, from Corollary 4.5, and Steps 3 and 4 above, it follows that K( s) H c,l (Cp m ). Next we show that any continuous solution K( s) H c,l (Cp m ) to the optimal Hankel norm approximation problem has the form (24). Theorem 5.3 Suppose that Σ(A, B, C) is an exponentially stable infinite-dimensional system with B L(C m, Z), C L(Z, C p ) and σ l+1 < σ < σ l. Let X( ) be given by (5). If K( s) H,l c (Cp m ) and G + K σ, then K( s) R 1 ( s)r 2 ( s) 1, where R1 ( s) X 1 Q( s) ( s) (30) R 2 ( s) for some Q( s) H c (C p m ) satisfying Q 1. Proof Let K( s) H,l c (Cp m ) satisfy G + K σ and suppose it has the coprime factorization K( s) N(s)M 1 (s) over MH, c where N, M MH, c M is rational, and det M R (0) has exactly l zeros in C + 0 and none on the imaginary axis. Define U1 ( s) K( s) X( s) X11 ( s)k( s) + X 12 ( s) X 21 ( s)k( s) + X 22 ( s) (31), and U 1 ( s) U1 ( s) X11 ( s)n(s) + X 12 ( s)m(s) X 21 ( s)n(s) + X 22 ( s)m(s) M(s) (32) Step 1: We show that U 2 (jω) is invertible with U 1 U 1 2 L (jr; C p m ) and U 1 U From (31) we deduce U1 ( s) X( s) For s jω, ω R, we have U 1 U 1 U 2 U 2 U 1 U 2 0 Ip G(s) 1 G(s) + K( s) I p 0 0 U1 and appealing to (7) and (7), we see that for s jω G X 0 0 σ 2. Thus for s jω, U 2 Ip G X 0.,. (33) 1 U 1 U 1 U 2 U 2 (34) G + K G + K 0 σ 2 0. Hence for all u C m and all ω R, we have from equation (34) that U 1 (jω)u 2 U 2 (jω)u 2 (G+K)u 2 σ 2 u 2 0, (35) and so U 1 (jω)u U 2 (jω)u. (36) Since V ( s) X 1 ( s), we deduce that U1 ( s) K( s) V ( s) and so, (37) V 21 ( s)u 1 ( s) + V 22 ( s) (38) for s C + 0. We claim that ker(u 2( jω)) {0} for all ω R. Suppose on the contrary that there exists x 0 such that U 2 ( jω 0 )x 0. Then from (36), we obtain U 1 ( jω 0 )x 0, which violates (38). Concluding, we have that det(u 2 (jω)) 0 for all ω R, and so U 2 (jω) 1 exists for all ω R. From (34), we deduce that U 1 (jω)u 2 (jω) 1 y 2 y 2 ω R, (39)

8 and so U 1 ( s)u2 1 L (jr; C p m ) satisfies U 1 U Step 2: We now construct a Q( s) H, (C c p m ) such that Q(j ) 1 and K( s) (V 11 ( s)q( s) +V 12 ( s))(v 21 ( s)q( s) + V 22 ( s)) 1. Consider MH. We know that X 21 ( s) is strictly proper and both X 22 ( s) and M(s) are proper with a nonzero limit at infinity in C + 0. So there exists a R > 0 such that for every s C 0 +, with s > R, det(u 2( s)) 0. Since det() is holomorphic in C + 0, it follows that if its zeros have an accumulation point in the compact set {s C + 0 s R}, it must lie on the imaginary axis. But det() 0 on the imaginary axis, since det() det() det(m(s)), (40) and neither det(u 2 ) nor det(m) have any zeros on the imaginary axis. Thus det() has only finitely many zeros in C 0 +, and they are all contained in the open right halfplane. So det() S and it follows from Lemma 2.2 that U 1 2 ( s) is an element of MH, c and Q( s) : U 1 ( s)u 1 2 ( s) is a well-defined element of MH,. c We also note that ( M 1 (s)) is invertible as an element of MH, c and Q( s) : U 1 ( s)u2 1 ( s). From Step 1 we see that Q(j ) 1. Now from (31) we obtain and so K( s) X 1 U1 ( s) ( s) U1 ( s) V ( s) K( s) V 11 ( s)u 1 ( s) + V 12 ( s) (41) V 21 ( s)u 1 ( s) + V 22 ( s). (42) Thus K( s) (V 11 ( s)q( s) + V 12 ( s)) (V 11 ( s)q( s)+v 12 ( s)) (V 21 ( s)q( s)+v 22 ( s)) 1, as claimed. Step 3: We show that U 1 ( s), are right coprime over MH c. Since M, N are right coprime, there exist R, S MH c such that RM SN I. Now it is readily verified using V X I that Consider now V 11 U 1 + V 12 U 2 KM N (43) V 21 U 1 + V 22 U 2 M. (44) ( SV 12 + RV 22 )U 2 (SV 11 RV 21 )U 1 R(V 21 U 1 + V 22 U 2 ) S(V 11 U 1 + V 12 U 2 ) RM SN I. Thus U 1 ( s), are right coprime over MH. c Step 4: Q( s) H (C c p m ). The zeros of det(v 22 ( s)), det(m(s)) and det() are contained in some half-plane C + ɛ, where ɛ > 0. Since V22 1 V 21 < 1, there exists a r > 0 such that V22 1 V 21 1 r. It follows from Lemma 2.8 that δ 1 > 0 such that δ 1 < ɛ and for any ζ satisfying 0 < ζ < δ 1, V 1 22 ( ζ )V 21( ζ ) 1 r 2. (45) Similarly it follows Lemma 2.8 that δ 2 > 0 such that δ 2 < ɛ and for any ζ satisfying 0 < ζ < δ 2, Q( ζ j ) 1 + r 4 1 r r. (46) 4 Let δ :min(δ 1, δ 2 ), and fix a ζ satisfying 0 < ζ < δ. Let φ(α, s) det(αv 21 ( s ζ)u 1 ( s ζ) +V 22 ( s ζ)u 2 ( s ζ)), where α 0, φ(0, ) det(v 22 ( ζ)u 1 ( ζ)) and φ(1, ) det(v 21 ( ζ)u 1 ( ζ) + V 22 ( ζ)u 2 ( ζ)) are meromorphic in C + 0 ζ/2. 2. φ(0, ) has a nonzero limit at infinity in C 0 + : det(v 22 ( s)) has a nonzero limit at infinity in C + 0 and det(u 2 ) has a nonzero limit at infinity in C + 0, since det(u 2 ) S. φ(1, ) has a nonzero limit at infinity in C + 0, since V 21 ( s) is strictly proper, U 1 ( s) is proper in C + 0, and the above. 3. φ(α, s) : 0, 1 jr C is a continuous function; and φ(0, jω) det(v 22 ( ζ jω)u 1 ( ζ jω)) det(v 22 ( ζ jω)).det(u 2 ( ζ jω)), φ(1, jω) det(v 21 ( ζ jω)u 1 ( ζ jω) +V 22 ( ζ jω)u 2 ( ζ jω)). 4. We have φ(α, jω) 0 since αv22 1 ( ζ j ) V 21 ( ζ j ) U 1 ( ζ j ) U 1 2 ( ζ j )) < 1 and det(u 2 ( ζ jω)) φ(α, ) 0, since V 21 ( s) is strictly proper, U 1 ( s) is proper in C + 0, and det(v 22( s)). det() has a nonzero limit at infinity in C + 0. Thus the assumptions in Lemma A.1.18 (Curtain and Zwart 8, page 570) are satisfied by φ, and hence it follows that the Nyquist indices of φ(0, s) and φ(1, s) are the same. Consequently, the number of zeros are the same (the

9 number of poles is zero, as φ(0, s), φ(1, s) are holomorphic in C + 0 δ/2 ) and so the sum of the number of zeros of s det(v 22 ( ζ s)) in C + 0 and the number of zeros of s det(u 2 ( ζ s)) in C 0 + equals the number of zeros of s det(v 21 ( ζ s)u 1 ( ζ s) +V 22 ( ζ s)u 2 ( ζ s)) (det(m(ζ + s), using 44) in C + 0, i.e., Thus S Ŝ σ l+1. Proof (of Theorem 4.2) Let K( s) Ĝ(s) + F (s), where Ĝ has MacMillan degree l, and F H (C p m ). Thus, G + K G + Ĝ H Γ G + ΓĜ σ l+1 (G). (51) l + d(q) l, (47) where d(q) denotes the number of zeros of s det(u 2 ( ζ s)) in C + 0. Thus Q has no poles in C+ ζ. But ζ > 0 can be chosen arbitrarily small, which implies that Q H c (C p m ). Our main result is the following: Theorem 5.4 Suppose that Σ(A, B, C) is an exponentially stable infinite-dimensional system with B L(C m, Z), C L(Z, C p ) and σ l+1 < σ < σ l. Let X( ) be given by (5). K( s) H,l c (Cp m ) and G + K σ iff K( s) R 1 ( s)r 2 ( s) 1, where R1 ( s) R 2 ( s) X 1 ( s) Q( s) for some Q( s) H c (C p m ) satisfying Q 1. (48) Finally we remark that the results in this paper can be extended to the Pritchard-Salamon class of systems which allows for unbounded input and output operators. (The paper is available at the following website: amols/index.html.) 6 Appendix Proof (of Lemma 4.1) Let ˆΠ be the projection from Z 2 onto span{w 1, w 2,...,w l+1 }; then ˆΠ(S Ŝ) S Ŝ. (49) Consider the following restriction of ˆΠŜ: ˆΠŜ : span{v 1, v 2,..., v l+1 } span{w 1, w 2,..., w l+1 } (50) which has rank l, and hence there exists a z ker(ˆπŝ), z 1, say z l+1 i1 a iv i with l+1 i1 a i 2 1. Then ˆΠSz l+1 a i σ i w i, i1 S Ŝ 2 ˆΠSz ˆΠŜz 2 ˆΠSz 2 l+1 σi 2 a 2 i i1 l+1 σl+1 2 a 2 i σl+1. 2 i1 References 1 Adamjan V.M., Arov D.Z. and Krein M.G. Infinite Hankel block matrices and related extension problems. American Mathematical Society Translations, Vol. 111, pp , Ball J.A. and Helton J.W. A Beurling-Lax theorem for the Lie group U(m, n) which contains most classical interpolation theory. Journal of Operator Theory, Vol. 9, pp , Ball J.A. and Ran A.C.M. Optimal Hankel norm model reductions and Wiener-Hopf factorization I: The canonical case. SIAM Journal on Control and Optimization, Vol. 25, No.2, pp , Curtain R.F. and Ichikawa A. The Nehari problem for infinite- dimensional systems of parabolic type. Integral Equations and Operator Theory, Vol. 26, pp.29-45, Curtain R.F. and Oostveen J.C. The Nehari problem for nonexponentially stable systems. Integral Equations and Operator Theory, Vol. 31, pp , Curtain R.F. and Ran A.C.M. Explicit formulas for Hankel norm approximations of infinite-dimensional systems. Integral Equations and Operator Theory, Vol. 13, pp , Curtain R.F. and Zwart H.J. The Nehari problem for the Pritchard-Salomon class of infinite-dimensional linear systems: a direct approach. Integral Equations and Operator Theory, Vol. 18, , Curtain R.F. and Zwart H.J. An Introduction to Infinite- Dimensional Systems Theory. Springer-Verlag, New York, Glover K. All optimal Hankel-norm approximations of linear multivariable systems and their L error bounds. International Journal of Control, Vol. 39, pp , Glover K., Curtain R.F. and Partington J.R. Realization and approximation of linear infinite-dimensional systems with error bounds. SIAM Journal on Control and Optimization, Vol. 26, pp , 1988.

10 11 Kato T. Perturbation Theory of Linear Operators. Springer-Verlag, New York, Nikol skii N.K. Ha-plitz operators: a survey of some recent results. In S.C. Power, editor, Operators and Function Theory. Reidel, Boston, Ran A.C.M. Hankel norm approximation for infinitedimensional systems and Wiener-Hopf factorization. In R.F. Curtain, editor, Modelling Robustness and Sensitivity Reduction in Control Systems, NATO ASI Series, pp.57-70, Springer-Verlag, Rudin W. Real and Complex Analysis. 3rd edition, McGraw-Hill, Sasane A.J. and Curtain R.F. Inertia theorems for operator Lyapunov equations. Proceedings of the 38th IEEE Conference on Decision and Control, Phoenix, U.S.A., 1999.

semigroups are consistent, we shall simply use the notation T (t). Assume further that U and Y are separable Hilbert spaces (the input and output spac

semigroups are consistent, we shall simply use the notation T (t). Assume further that U and Y are separable Hilbert spaces (the input and output spac CDC-REG467 Optimal Hankel norm approximation for the Pritchard-Salamon class of non-exponentially stable innite-dimensional systems A.J. Sasane and R.F. Curtain Department of Mathematics, University of

More information

Explicit formulae for J pectral factor for well-poed linear ytem Ruth F. Curtain Amol J. Saane Department of Mathematic Department of Mathematic Univerity of Groningen Univerity of Twente P.O. Box 800

More information

Hankel Optimal Model Reduction 1

Hankel Optimal Model Reduction 1 Massachusetts Institute of Technology Department of Electrical Engineering and Computer Science 6.242, Fall 2004: MODEL REDUCTION Hankel Optimal Model Reduction 1 This lecture covers both the theory and

More information

Balanced Truncation 1

Balanced Truncation 1 Massachusetts Institute of Technology Department of Electrical Engineering and Computer Science 6.242, Fall 2004: MODEL REDUCTION Balanced Truncation This lecture introduces balanced truncation for LTI

More information

On the continuity of the J-spectral factorization mapping

On the continuity of the J-spectral factorization mapping On the continuity of the J-spectral factorization mapping Orest V. Iftime University of Groningen Department of Mathematics and Computing Science PO Box 800, 9700 AV Groningen, The Netherlands Email: O.V.Iftime@math.rug.nl

More information

Reduced-order Model Based on H -Balancing for Infinite-Dimensional Systems

Reduced-order Model Based on H -Balancing for Infinite-Dimensional Systems Applied Mathematical Sciences, Vol. 7, 2013, no. 9, 405-418 Reduced-order Model Based on H -Balancing for Infinite-Dimensional Systems Fatmawati Department of Mathematics Faculty of Science and Technology,

More information

Algebraic Properties of Riccati equations. Ruth Curtain University of Groningen, The Netherlands

Algebraic Properties of Riccati equations. Ruth Curtain University of Groningen, The Netherlands Algebraic Properties of Riccati equations Ruth Curtain University of Groningen, The Netherlands Special Recognition Peter Falb Jan Willems Tony Pritchard Hans Zwart Riccati equation Ṗ(t) + A(t) P(t) +

More information

Classes of Linear Operators Vol. I

Classes of Linear Operators Vol. I Classes of Linear Operators Vol. I Israel Gohberg Seymour Goldberg Marinus A. Kaashoek Birkhäuser Verlag Basel Boston Berlin TABLE OF CONTENTS VOLUME I Preface Table of Contents of Volume I Table of Contents

More information

EXPLICIT UPPER BOUNDS FOR THE SPECTRAL DISTANCE OF TWO TRACE CLASS OPERATORS

EXPLICIT UPPER BOUNDS FOR THE SPECTRAL DISTANCE OF TWO TRACE CLASS OPERATORS EXPLICIT UPPER BOUNDS FOR THE SPECTRAL DISTANCE OF TWO TRACE CLASS OPERATORS OSCAR F. BANDTLOW AND AYŞE GÜVEN Abstract. Given two trace class operators A and B on a separable Hilbert space we provide an

More information

Left invertible semigroups on Hilbert spaces.

Left invertible semigroups on Hilbert spaces. Left invertible semigroups on Hilbert spaces. Hans Zwart Department of Applied Mathematics, Faculty of Electrical Engineering, Mathematics and Computer Science, University of Twente, P.O. Box 217, 75 AE

More information

Stability of an abstract wave equation with delay and a Kelvin Voigt damping

Stability of an abstract wave equation with delay and a Kelvin Voigt damping Stability of an abstract wave equation with delay and a Kelvin Voigt damping University of Monastir/UPSAY/LMV-UVSQ Joint work with Serge Nicaise and Cristina Pignotti Outline 1 Problem The idea Stability

More information

NORMS ON SPACE OF MATRICES

NORMS ON SPACE OF MATRICES NORMS ON SPACE OF MATRICES. Operator Norms on Space of linear maps Let A be an n n real matrix and x 0 be a vector in R n. We would like to use the Picard iteration method to solve for the following system

More information

J-SPECTRAL FACTORIZATION

J-SPECTRAL FACTORIZATION J-SPECTRAL FACTORIZATION of Regular Para-Hermitian Transfer Matrices Qing-Chang Zhong zhongqc@ieee.org School of Electronics University of Glamorgan United Kingdom Outline Notations and definitions Regular

More information

The Tuning of Robust Controllers for Stable Systems in the CD-Algebra: The Case of Sinusoidal and Polynomial Signals

The Tuning of Robust Controllers for Stable Systems in the CD-Algebra: The Case of Sinusoidal and Polynomial Signals The Tuning of Robust Controllers for Stable Systems in the CD-Algebra: The Case of Sinusoidal and Polynomial Signals Timo Hämäläinen Seppo Pohjolainen Tampere University of Technology Department of Mathematics

More information

Positive Stabilization of Infinite-Dimensional Linear Systems

Positive Stabilization of Infinite-Dimensional Linear Systems Positive Stabilization of Infinite-Dimensional Linear Systems Joseph Winkin Namur Center of Complex Systems (NaXys) and Department of Mathematics, University of Namur, Belgium Joint work with Bouchra Abouzaid

More information

DISTANCE BETWEEN BEHAVIORS AND RATIONAL REPRESENTATIONS

DISTANCE BETWEEN BEHAVIORS AND RATIONAL REPRESENTATIONS DISTANCE BETWEEN BEHAVIORS AND RATIONAL REPRESENTATIONS H.L. TRENTELMAN AND S.V. GOTTIMUKKALA Abstract. In this paper we study notions of distance between behaviors of linear differential systems. We introduce

More information

SPECTRAL THEOREM FOR SYMMETRIC OPERATORS WITH COMPACT RESOLVENT

SPECTRAL THEOREM FOR SYMMETRIC OPERATORS WITH COMPACT RESOLVENT SPECTRAL THEOREM FOR SYMMETRIC OPERATORS WITH COMPACT RESOLVENT Abstract. These are the letcure notes prepared for the workshop on Functional Analysis and Operator Algebras to be held at NIT-Karnataka,

More information

Linear Passive Stationary Scattering Systems with Pontryagin State Spaces

Linear Passive Stationary Scattering Systems with Pontryagin State Spaces mn header will be provided by the publisher Linear Passive Stationary Scattering Systems with Pontryagin State Spaces D. Z. Arov 1, J. Rovnyak 2, and S. M. Saprikin 1 1 Department of Physics and Mathematics,

More information

3 Stabilization of MIMO Feedback Systems

3 Stabilization of MIMO Feedback Systems 3 Stabilization of MIMO Feedback Systems 3.1 Notation The sets R and S are as before. We will use the notation M (R) to denote the set of matrices with elements in R. The dimensions are not explicitly

More information

A linear algebra proof of the fundamental theorem of algebra

A linear algebra proof of the fundamental theorem of algebra A linear algebra proof of the fundamental theorem of algebra Andrés E. Caicedo May 18, 2010 Abstract We present a recent proof due to Harm Derksen, that any linear operator in a complex finite dimensional

More information

A linear algebra proof of the fundamental theorem of algebra

A linear algebra proof of the fundamental theorem of algebra A linear algebra proof of the fundamental theorem of algebra Andrés E. Caicedo May 18, 2010 Abstract We present a recent proof due to Harm Derksen, that any linear operator in a complex finite dimensional

More information

FEL3210 Multivariable Feedback Control

FEL3210 Multivariable Feedback Control FEL3210 Multivariable Feedback Control Lecture 8: Youla parametrization, LMIs, Model Reduction and Summary [Ch. 11-12] Elling W. Jacobsen, Automatic Control Lab, KTH Lecture 8: Youla, LMIs, Model Reduction

More information

Chapter 2: Linear Independence and Bases

Chapter 2: Linear Independence and Bases MATH20300: Linear Algebra 2 (2016 Chapter 2: Linear Independence and Bases 1 Linear Combinations and Spans Example 11 Consider the vector v (1, 1 R 2 What is the smallest subspace of (the real vector space

More information

Stabilization of Distributed Parameter Systems by State Feedback with Positivity Constraints

Stabilization of Distributed Parameter Systems by State Feedback with Positivity Constraints Stabilization of Distributed Parameter Systems by State Feedback with Positivity Constraints Joseph Winkin Namur Center of Complex Systems (naxys) and Dept. of Mathematics, University of Namur, Belgium

More information

The Dirichlet-to-Neumann operator

The Dirichlet-to-Neumann operator Lecture 8 The Dirichlet-to-Neumann operator The Dirichlet-to-Neumann operator plays an important role in the theory of inverse problems. In fact, from measurements of electrical currents at the surface

More information

E 231, UC Berkeley, Fall 2005, Packard

E 231, UC Berkeley, Fall 2005, Packard Preliminaries 1. Set notation 2. Fields, vector spaces, normed vector spaces, inner product spaces 3. More notation 4. Vectors in R n, C n, norms 5. Matrix Facts (determinants, inversion formulae) 6. Normed

More information

1 Math 241A-B Homework Problem List for F2015 and W2016

1 Math 241A-B Homework Problem List for F2015 and W2016 1 Math 241A-B Homework Problem List for F2015 W2016 1.1 Homework 1. Due Wednesday, October 7, 2015 Notation 1.1 Let U be any set, g be a positive function on U, Y be a normed space. For any f : U Y let

More information

COMPLETELY INVARIANT JULIA SETS OF POLYNOMIAL SEMIGROUPS

COMPLETELY INVARIANT JULIA SETS OF POLYNOMIAL SEMIGROUPS Series Logo Volume 00, Number 00, Xxxx 19xx COMPLETELY INVARIANT JULIA SETS OF POLYNOMIAL SEMIGROUPS RICH STANKEWITZ Abstract. Let G be a semigroup of rational functions of degree at least two, under composition

More information

Robust Control of Time-delay Systems

Robust Control of Time-delay Systems Robust Control of Time-delay Systems Qing-Chang Zhong Distinguished Lecturer, IEEE Power Electronics Society Max McGraw Endowed Chair Professor in Energy and Power Engineering Dept. of Electrical and Computer

More information

Short note on compact operators - Monday 24 th March, Sylvester Eriksson-Bique

Short note on compact operators - Monday 24 th March, Sylvester Eriksson-Bique Short note on compact operators - Monday 24 th March, 2014 Sylvester Eriksson-Bique 1 Introduction In this note I will give a short outline about the structure theory of compact operators. I restrict attention

More information

TOEPLITZ OPERATORS. Toeplitz studied infinite matrices with NW-SE diagonals constant. f e C :

TOEPLITZ OPERATORS. Toeplitz studied infinite matrices with NW-SE diagonals constant. f e C : TOEPLITZ OPERATORS EFTON PARK 1. Introduction to Toeplitz Operators Otto Toeplitz lived from 1881-1940 in Goettingen, and it was pretty rough there, so he eventually went to Palestine and eventually contracted

More information

Exercise Solutions to Functional Analysis

Exercise Solutions to Functional Analysis Exercise Solutions to Functional Analysis Note: References refer to M. Schechter, Principles of Functional Analysis Exersize that. Let φ,..., φ n be an orthonormal set in a Hilbert space H. Show n f n

More information

Problem Set 5 Solutions 1

Problem Set 5 Solutions 1 Massachusetts Institute of Technology Department of Electrical Engineering and Computer Science 6.245: MULTIVARIABLE CONTROL SYSTEMS by A. Megretski Problem Set 5 Solutions The problem set deals with Hankel

More information

5 Compact linear operators

5 Compact linear operators 5 Compact linear operators One of the most important results of Linear Algebra is that for every selfadjoint linear map A on a finite-dimensional space, there exists a basis consisting of eigenvectors.

More information

Throughout these notes we assume V, W are finite dimensional inner product spaces over C.

Throughout these notes we assume V, W are finite dimensional inner product spaces over C. Math 342 - Linear Algebra II Notes Throughout these notes we assume V, W are finite dimensional inner product spaces over C 1 Upper Triangular Representation Proposition: Let T L(V ) There exists an orthonormal

More information

A Case Study for the Delay-type Nehari Problem

A Case Study for the Delay-type Nehari Problem Proceedings of the 44th EEE Conference on Decision and Control, and the European Control Conference 5 Seville, Spain, December -5, 5 MoA A Case Study for the Delay-type Nehari Problem Qing-Chang Zhong

More information

Set, functions and Euclidean space. Seungjin Han

Set, functions and Euclidean space. Seungjin Han Set, functions and Euclidean space Seungjin Han September, 2018 1 Some Basics LOGIC A is necessary for B : If B holds, then A holds. B A A B is the contraposition of B A. A is sufficient for B: If A holds,

More information

Problem set 5 solutions 1

Problem set 5 solutions 1 Massachusetts Institute of Technology Department of Electrical Engineering and Computer Science 6.242, Fall 24: MODEL REDUCTION Problem set 5 solutions Problem 5. For each of the stetements below, state

More information

Componentwise perturbation analysis for matrix inversion or the solution of linear systems leads to the Bauer-Skeel condition number ([2], [13])

Componentwise perturbation analysis for matrix inversion or the solution of linear systems leads to the Bauer-Skeel condition number ([2], [13]) SIAM Review 4():02 2, 999 ILL-CONDITIONED MATRICES ARE COMPONENTWISE NEAR TO SINGULARITY SIEGFRIED M. RUMP Abstract. For a square matrix normed to, the normwise distance to singularity is well known to

More information

A NOTE ON QUASI ISOMETRIES II. S.M. Patel Sardar Patel University, India

A NOTE ON QUASI ISOMETRIES II. S.M. Patel Sardar Patel University, India GLASNIK MATEMATIČKI Vol. 38(58)(2003), 111 120 A NOTE ON QUASI ISOMETRIES II S.M. Patel Sardar Patel University, India Abstract. An operator A on a complex Hilbert space H is called a quasi-isometry if

More information

Analysis Preliminary Exam Workshop: Hilbert Spaces

Analysis Preliminary Exam Workshop: Hilbert Spaces Analysis Preliminary Exam Workshop: Hilbert Spaces 1. Hilbert spaces A Hilbert space H is a complete real or complex inner product space. Consider complex Hilbert spaces for definiteness. If (, ) : H H

More information

only nite eigenvalues. This is an extension of earlier results from [2]. Then we concentrate on the Riccati equation appearing in H 2 and linear quadr

only nite eigenvalues. This is an extension of earlier results from [2]. Then we concentrate on the Riccati equation appearing in H 2 and linear quadr The discrete algebraic Riccati equation and linear matrix inequality nton. Stoorvogel y Department of Mathematics and Computing Science Eindhoven Univ. of Technology P.O. ox 53, 56 M Eindhoven The Netherlands

More information

Spectral theory for compact operators on Banach spaces

Spectral theory for compact operators on Banach spaces 68 Chapter 9 Spectral theory for compact operators on Banach spaces Recall that a subset S of a metric space X is precompact if its closure is compact, or equivalently every sequence contains a Cauchy

More information

08a. Operators on Hilbert spaces. 1. Boundedness, continuity, operator norms

08a. Operators on Hilbert spaces. 1. Boundedness, continuity, operator norms (February 24, 2017) 08a. Operators on Hilbert spaces Paul Garrett garrett@math.umn.edu http://www.math.umn.edu/ garrett/ [This document is http://www.math.umn.edu/ garrett/m/real/notes 2016-17/08a-ops

More information

Determinant lines and determinant line bundles

Determinant lines and determinant line bundles CHAPTER Determinant lines and determinant line bundles This appendix is an exposition of G. Segal s work sketched in [?] on determinant line bundles over the moduli spaces of Riemann surfaces with parametrized

More information

Operators with numerical range in a closed halfplane

Operators with numerical range in a closed halfplane Operators with numerical range in a closed halfplane Wai-Shun Cheung 1 Department of Mathematics, University of Hong Kong, Hong Kong, P. R. China. wshun@graduate.hku.hk Chi-Kwong Li 2 Department of Mathematics,

More information

Remarks on the Spectrum of Bounded and Normal Operators on Hilbert Spaces

Remarks on the Spectrum of Bounded and Normal Operators on Hilbert Spaces An. Şt. Univ. Ovidius Constanţa Vol. 16(2), 2008, 7 14 Remarks on the Spectrum of Bounded and Normal Operators on Hilbert Spaces M. AKKOUCHI Abstract Let H be a complex Hilbert space H. Let T be a bounded

More information

Notes on Complex Analysis

Notes on Complex Analysis Michael Papadimitrakis Notes on Complex Analysis Department of Mathematics University of Crete Contents The complex plane.. The complex plane...................................2 Argument and polar representation.........................

More information

1 Definition and Basic Properties of Compa Operator

1 Definition and Basic Properties of Compa Operator 1 Definition and Basic Properties of Compa Operator 1.1 Let X be a infinite dimensional Banach space. Show that if A C(X ), A does not have bounded inverse. Proof. Denote the unit ball of X by B and the

More information

1 Introduction. or equivalently f(z) =

1 Introduction. or equivalently f(z) = Introduction In this unit on elliptic functions, we ll see how two very natural lines of questions interact. The first, as we have met several times in Berndt s book, involves elliptic integrals. In particular,

More information

AN ELEMENTARY PROOF OF THE SPECTRAL RADIUS FORMULA FOR MATRICES

AN 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 information

Normed Vector Spaces and Double Duals

Normed Vector Spaces and Double Duals Normed Vector Spaces and Double Duals Mathematics 481/525 In this note we look at a number of infinite-dimensional R-vector spaces that arise in analysis, and we consider their dual and double dual spaces

More information

Representation of integrated autoregressive processes in Banach space

Representation of integrated autoregressive processes in Banach space Representation of integrated autoregressive processes in Banach space Won-Ki Seo Department of Economics, University of California, San Diego April 2, 218 Very preliminary and incomplete. Comments welcome.

More information

SPECTRAL THEOREM FOR COMPACT SELF-ADJOINT OPERATORS

SPECTRAL THEOREM FOR COMPACT SELF-ADJOINT OPERATORS SPECTRAL THEOREM FOR COMPACT SELF-ADJOINT OPERATORS G. RAMESH Contents Introduction 1 1. Bounded Operators 1 1.3. Examples 3 2. Compact Operators 5 2.1. Properties 6 3. The Spectral Theorem 9 3.3. Self-adjoint

More information

Lecture 7. Econ August 18

Lecture 7. Econ August 18 Lecture 7 Econ 2001 2015 August 18 Lecture 7 Outline First, the theorem of the maximum, an amazing result about continuity in optimization problems. Then, we start linear algebra, mostly looking at familiar

More information

Boston College, Department of Mathematics, Chestnut Hill, MA , May 25, 2004

Boston College, Department of Mathematics, Chestnut Hill, MA , May 25, 2004 NON-VANISHING OF ALTERNANTS by Avner Ash Boston College, Department of Mathematics, Chestnut Hill, MA 02467-3806, ashav@bcedu May 25, 2004 Abstract Let p be prime, K a field of characteristic 0 Let (x

More information

Math Linear Algebra II. 1. Inner Products and Norms

Math Linear Algebra II. 1. Inner Products and Norms Math 342 - Linear Algebra II Notes 1. Inner Products and Norms One knows from a basic introduction to vectors in R n Math 254 at OSU) that the length of a vector x = x 1 x 2... x n ) T R n, denoted x,

More information

Problem 1A. Calculus. Problem 3A. Real analysis. f(x) = 0 x = 0.

Problem 1A. Calculus. Problem 3A. Real analysis. f(x) = 0 x = 0. Problem A. Calculus Find the length of the spiral given in polar coordinates by r = e θ, < θ 0. Solution: The length is 0 θ= ds where by Pythagoras ds = dr 2 + (rdθ) 2 = dθ 2e θ, so the length is 0 2e

More information

A Simple Derivation of Right Interactor for Tall Transfer Function Matrices and its Application to Inner-Outer Factorization Continuous-Time Case

A Simple Derivation of Right Interactor for Tall Transfer Function Matrices and its Application to Inner-Outer Factorization Continuous-Time Case A Simple Derivation of Right Interactor for Tall Transfer Function Matrices and its Application to Inner-Outer Factorization Continuous-Time Case ATARU KASE Osaka Institute of Technology Department of

More information

Model reduction for linear systems by balancing

Model reduction for linear systems by balancing Model reduction for linear systems by balancing Bart Besselink Jan C. Willems Center for Systems and Control Johann Bernoulli Institute for Mathematics and Computer Science University of Groningen, Groningen,

More information

C.I.BYRNES,D.S.GILLIAM.I.G.LAUK O, V.I. SHUBOV We assume that the input u is given, in feedback form, as the output of a harmonic oscillator with freq

C.I.BYRNES,D.S.GILLIAM.I.G.LAUK O, V.I. SHUBOV We assume that the input u is given, in feedback form, as the output of a harmonic oscillator with freq Journal of Mathematical Systems, Estimation, and Control Vol. 8, No. 2, 1998, pp. 1{12 c 1998 Birkhauser-Boston Harmonic Forcing for Linear Distributed Parameter Systems C.I. Byrnes y D.S. Gilliam y I.G.

More information

Raktim Bhattacharya. . AERO 632: Design of Advance Flight Control System. Norms for Signals and Systems

Raktim Bhattacharya. . AERO 632: Design of Advance Flight Control System. Norms for Signals and Systems . AERO 632: Design of Advance Flight Control System Norms for Signals and. Raktim Bhattacharya Laboratory For Uncertainty Quantification Aerospace Engineering, Texas A&M University. Norms for Signals ...

More information

The following definition is fundamental.

The following definition is fundamental. 1. Some Basics from Linear Algebra With these notes, I will try and clarify certain topics that I only quickly mention in class. First and foremost, I will assume that you are familiar with many basic

More information

Denis ARZELIER arzelier

Denis ARZELIER   arzelier COURSE ON LMI OPTIMIZATION WITH APPLICATIONS IN CONTROL PART II.2 LMIs IN SYSTEMS CONTROL STATE-SPACE METHODS PERFORMANCE ANALYSIS and SYNTHESIS Denis ARZELIER www.laas.fr/ arzelier arzelier@laas.fr 15

More information

9. Banach algebras and C -algebras

9. Banach algebras and C -algebras matkt@imf.au.dk Institut for Matematiske Fag Det Naturvidenskabelige Fakultet Aarhus Universitet September 2005 We read in W. Rudin: Functional Analysis Based on parts of Chapter 10 and parts of Chapter

More information

A GENERALIZATION OF THE YOULA-KUČERA PARAMETRIZATION FOR MIMO STABILIZABLE SYSTEMS. Alban Quadrat

A GENERALIZATION OF THE YOULA-KUČERA PARAMETRIZATION FOR MIMO STABILIZABLE SYSTEMS. Alban Quadrat A GENERALIZATION OF THE YOULA-KUČERA ARAMETRIZATION FOR MIMO STABILIZABLE SYSTEMS Alban Quadrat INRIA Sophia Antipolis, CAFE project, 2004 Route des Lucioles, B 93, 06902 Sophia Antipolis cedex, France.

More information

RIEMANN MAPPING THEOREM

RIEMANN MAPPING THEOREM RIEMANN MAPPING THEOREM VED V. DATAR Recall that two domains are called conformally equivalent if there exists a holomorphic bijection from one to the other. This automatically implies that there is an

More information

Iterative Solution of a Matrix Riccati Equation Arising in Stochastic Control

Iterative Solution of a Matrix Riccati Equation Arising in Stochastic Control Iterative Solution of a Matrix Riccati Equation Arising in Stochastic Control Chun-Hua Guo Dedicated to Peter Lancaster on the occasion of his 70th birthday We consider iterative methods for finding the

More information

Part III. 10 Topological Space Basics. Topological Spaces

Part III. 10 Topological Space Basics. Topological Spaces Part III 10 Topological Space Basics Topological Spaces Using the metric space results above as motivation we will axiomatize the notion of being an open set to more general settings. Definition 10.1.

More information

THE CENTRAL INTERTWINING LIFTING AND STRICT CONTRACTIONS

THE CENTRAL INTERTWINING LIFTING AND STRICT CONTRACTIONS PROCEEDINGS OF THE AMERICAN MATHEMATICAL SOCIETY Volume 124, Number 12, December 1996, Pages 3813 3817 S 0002-9939(96)03540-X THE CENTRAL INTERTWINING LIFTING AND STRICT CONTRACTIONS RADU GADIDOV (Communicated

More information

Estimates for the resolvent and spectral gaps for non self-adjoint operators. University Bordeaux

Estimates for the resolvent and spectral gaps for non self-adjoint operators. University Bordeaux for the resolvent and spectral gaps for non self-adjoint operators 1 / 29 Estimates for the resolvent and spectral gaps for non self-adjoint operators Vesselin Petkov University Bordeaux Mathematics Days

More information

Elementary linear algebra

Elementary linear algebra Chapter 1 Elementary linear algebra 1.1 Vector spaces Vector spaces owe their importance to the fact that so many models arising in the solutions of specific problems turn out to be vector spaces. The

More information

Algebraic Properties of Riccati equations

Algebraic Properties of Riccati equations Algebraic Properties of Riccati equations Ruth Curtain University of Groningen, The Netherlands (collaboration with Amol Sasane, Royal Institute of Technology, Stockholm, Sweden) Summary Summary Algebraic

More information

CDS Solutions to Final Exam

CDS Solutions to Final Exam CDS 22 - Solutions to Final Exam Instructor: Danielle C Tarraf Fall 27 Problem (a) We will compute the H 2 norm of G using state-space methods (see Section 26 in DFT) We begin by finding a minimal state-space

More information

MATH 722, COMPLEX ANALYSIS, SPRING 2009 PART 5

MATH 722, COMPLEX ANALYSIS, SPRING 2009 PART 5 MATH 722, COMPLEX ANALYSIS, SPRING 2009 PART 5.. The Arzela-Ascoli Theorem.. The Riemann mapping theorem Let X be a metric space, and let F be a family of continuous complex-valued functions on X. We have

More information

Zeros and zero dynamics

Zeros and zero dynamics CHAPTER 4 Zeros and zero dynamics 41 Zero dynamics for SISO systems Consider a linear system defined by a strictly proper scalar transfer function that does not have any common zero and pole: g(s) =α p(s)

More information

On compact operators

On compact operators On compact operators Alen Aleanderian * Abstract In this basic note, we consider some basic properties of compact operators. We also consider the spectrum of compact operators on Hilbert spaces. A basic

More information

Trace Class Operators and Lidskii s Theorem

Trace 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 self-contained derivation of the celebrated Lidskii Trace

More information

Algebra C Numerical Linear Algebra Sample Exam Problems

Algebra C Numerical Linear Algebra Sample Exam Problems Algebra C Numerical Linear Algebra Sample Exam Problems Notation. Denote by V a finite-dimensional Hilbert space with inner product (, ) and corresponding norm. The abbreviation SPD is used for symmetric

More information

A Brief Introduction to Functional Analysis

A Brief Introduction to Functional Analysis A Brief Introduction to Functional Analysis Sungwook Lee Department of Mathematics University of Southern Mississippi sunglee@usm.edu July 5, 2007 Definition 1. An algebra A is a vector space over C with

More information

Spectrum of infinite block matrices and pitriangular

Spectrum of infinite block matrices and pitriangular Electronic Journal of Linear Algebra Volume 16 Article 20 2007 Spectrum of infinite block matrices and pitriangular operators Michael I. Gil gilmi@cs.bgu.ac.il Follow this and additional works at: http://repository.uwyo.edu/ela

More information

c 2005 Society for Industrial and Applied Mathematics

c 2005 Society for Industrial and Applied Mathematics SIAM J. CONTROL OPTIM. Vol. 44, No. 3, pp. 991 1018 c 2005 Society for Industrial and Applied Mathematics THE SUBOPTIMAL NEHARI PROBLEM FOR WELL-POSED LINEAR SYSTEMS RUTH F. CURTAIN AND MARK R. OPMEER

More information

Lebesgue Measure on R n

Lebesgue Measure on R n CHAPTER 2 Lebesgue Measure on R n Our goal is to construct a notion of the volume, or Lebesgue measure, of rather general subsets of R n that reduces to the usual volume of elementary geometrical sets

More information

Composition Operators on Hilbert Spaces of Analytic Functions

Composition Operators on Hilbert Spaces of Analytic Functions Composition Operators on Hilbert Spaces of Analytic Functions Carl C. Cowen IUPUI (Indiana University Purdue University Indianapolis) and Purdue University First International Conference on Mathematics

More information

On Transitive and Localizing Operator Algebras

On Transitive and Localizing Operator Algebras International Mathematical Forum, Vol. 6, 2011, no. 60, 2955-2962 On Transitive and Localizing Operator Algebras Ömer Gök and Elif Demir Yıldız Technical University Faculty of Arts and Sciences Mathematics

More information

Perturbation Theory for Self-Adjoint Operators in Krein spaces

Perturbation Theory for Self-Adjoint Operators in Krein spaces Perturbation Theory for Self-Adjoint Operators in Krein spaces Carsten Trunk Institut für Mathematik, Technische Universität Ilmenau, Postfach 10 05 65, 98684 Ilmenau, Germany E-mail: carsten.trunk@tu-ilmenau.de

More information

On the Stabilization of Neutrally Stable Linear Discrete Time Systems

On the Stabilization of Neutrally Stable Linear Discrete Time Systems TWCCC Texas Wisconsin California Control Consortium Technical report number 2017 01 On the Stabilization of Neutrally Stable Linear Discrete Time Systems Travis J. Arnold and James B. Rawlings Department

More information

Chapter 8 Integral Operators

Chapter 8 Integral Operators Chapter 8 Integral Operators In our development of metrics, norms, inner products, and operator theory in Chapters 1 7 we only tangentially considered topics that involved the use of Lebesgue measure,

More information

Real Analysis Prelim Questions Day 1 August 27, 2013

Real Analysis Prelim Questions Day 1 August 27, 2013 Real Analysis Prelim Questions Day 1 August 27, 2013 are 5 questions. TIME LIMIT: 3 hours Instructions: Measure and measurable refer to Lebesgue measure µ n on R n, and M(R n ) is the collection of measurable

More information

Chapter 6: The metric space M(G) and normal families

Chapter 6: The metric space M(G) and normal families Chapter 6: The metric space MG) and normal families Course 414, 003 04 March 9, 004 Remark 6.1 For G C open, we recall the notation MG) for the set algebra) of all meromorphic functions on G. We now consider

More information

b jσ(j), Keywords: Decomposable numerical range, principal character AMS Subject Classification: 15A60

b jσ(j), Keywords: Decomposable numerical range, principal character AMS Subject Classification: 15A60 On the Hu-Hurley-Tam Conjecture Concerning The Generalized Numerical Range Che-Man Cheng Faculty of Science and Technology, University of Macau, Macau. E-mail: fstcmc@umac.mo and Chi-Kwong Li Department

More information

Approximate Low Rank Solution of Generalized Lyapunov Matrix Equations via Proper Orthogonal Decomposition

Approximate Low Rank Solution of Generalized Lyapunov Matrix Equations via Proper Orthogonal Decomposition Applied Mathematical Sciences, Vol. 4, 2010, no. 1, 21-30 Approximate Low Rank Solution of Generalized Lyapunov Matrix Equations via Proper Orthogonal Decomposition Amer Kaabi Department of Basic Science

More information

Problem Set 2: Solutions Math 201A: Fall 2016

Problem Set 2: Solutions Math 201A: Fall 2016 Problem Set 2: s Math 201A: Fall 2016 Problem 1. (a) Prove that a closed subset of a complete metric space is complete. (b) Prove that a closed subset of a compact metric space is compact. (c) Prove that

More information

A Necessary and Sufficient Condition for High-Frequency Robustness of Non-Strictly-Proper Feedback Systems

A Necessary and Sufficient Condition for High-Frequency Robustness of Non-Strictly-Proper Feedback Systems A Necessary and Sufficient Condition for High-Frequency Robustness of Non-Strictly-Proper Feedback Systems Daniel Cobb Department of Electrical Engineering University of Wisconsin Madison WI 53706-1691

More information

Compact operators on Banach spaces

Compact operators on Banach spaces Compact operators on Banach spaces Jordan Bell jordan.bell@gmail.com Department of Mathematics, University of Toronto November 12, 2017 1 Introduction In this note I prove several things about compact

More information

Online Exercises for Linear Algebra XM511

Online Exercises for Linear Algebra XM511 This document lists the online exercises for XM511. The section ( ) numbers refer to the textbook. TYPE I are True/False. Lecture 02 ( 1.1) Online Exercises for Linear Algebra XM511 1) The matrix [3 2

More information

SPECTRAL PROPERTIES AND NODAL SOLUTIONS FOR SECOND-ORDER, m-point, BOUNDARY VALUE PROBLEMS

SPECTRAL PROPERTIES AND NODAL SOLUTIONS FOR SECOND-ORDER, m-point, BOUNDARY VALUE PROBLEMS SPECTRAL PROPERTIES AND NODAL SOLUTIONS FOR SECOND-ORDER, m-point, BOUNDARY VALUE PROBLEMS BRYAN P. RYNNE Abstract. We consider the m-point boundary value problem consisting of the equation u = f(u), on

More information

On the Connection between Balanced Proper Orthogonal Decomposition, Balanced Truncation, and Metric Complexity Theory for Infinite Dimensional Systems

On the Connection between Balanced Proper Orthogonal Decomposition, Balanced Truncation, and Metric Complexity Theory for Infinite Dimensional Systems 1 American Control Conference Marriott Waterfront, Baltimore, MD, USA June 3-July, 1 FrA6.6 On the Connection between Balanced Proper Orthogonal Decomposition, Balanced Truncation, and Metric Complexity

More information

Spectral analysis for rank one perturbations of diagonal operators in non-archimedean Hilbert space

Spectral analysis for rank one perturbations of diagonal operators in non-archimedean Hilbert space Comment.Math.Univ.Carolin. 50,32009 385 400 385 Spectral analysis for rank one perturbations of diagonal operators in non-archimedean Hilbert space Toka Diagana, George D. McNeal Abstract. The paper is

More information

L p Spaces and Convexity

L p Spaces and Convexity L p Spaces and Convexity These notes largely follow the treatments in Royden, Real Analysis, and Rudin, Real & Complex Analysis. 1. Convex functions Let I R be an interval. For I open, we say a function

More information