AN ESTIMATE OF GREEN S FUNCTION OF THE PROBLEM OF BOUNDED SOLUTIONS IN THE CASE OF A TRIANGULAR COEFFICIENT V. G. KURBATOV AND I.V. KURBATOVA arxiv:90.00792v [math.na] 3 Jan 209 Abstract. An estimate of Green s function of the bounded solutions problem for the ordinary differential equation x (t) Bx(t) = f(t) is proposed. It is assumed that the matrix coefficient B is triangular. This estimate is a generalization of the estimate of the matrix exponential proved by Ch. F. Van Loan. Introduction Any square complex matrix A can be represented in the triangular Schur form A = Q BQ, where B is triangular and Q is unitary. Moreover, the Schur decomposition is calculated effectively and constitutes the preliminary stage of many algorithms connected with spectral theory. Thus, in numerical analysis, having a matrix A, one may assume that its triangular form is known. In its turn, any triangular matrix B can be easily represented as the sum B = D N of a diagonal matrix D and a strictly triangular matrix N. In [25] it is established the estimate n e At = e Bt e αt Nt k, t 0, ( ) k! where the matrix A has the size n n and α = max{reλ : λ σ(a)}. For other estimates of e At, see, e.g., [4, 9], [6, ch. 9, ]. In this paper, we establish an estimate similar to ( ) for Green s function of the bounded solutions problem (Theorem 4). Different estimates for Green s function are obtained in [5, 0, 8]. We consider the differential equation x (t) Ax(t) = f(t), t R, ( ) where A is a square complex matrix. The bounded solutions problem is the problem of finding a bounded solution x that corresponds to a bounded free term f. The bounded solutions problem is closely connected with the problem of exponential dichotomy of solutions. For the discussion of the bounded solutions problem from different points of view and related questions, see [, 2, 3, 6, 7,, 3, 4, 22, 23, 24] and the references therein. It is known (Theorem ) that equation ( ) has a unique bounded solution x for any bounded continuous free term f if and only if the spectrum of the Date: January 4, 209. 99 Mathematics Subject Classification. Primary 65F60; Secondary 97N50. Key words and phrases. bounded solutions problem, Green s function, estimate, Schur decomposition.
2 V. G. KURBATOV AND I.V. KURBATOVA coefficient A is disjoint from the imaginary axis. In this case, the solution can be represented in the form x(t) = G(A,s)f(t s)ds. The kernel G is called Green s function. Assuming that A is triangular or a triangular form of A is known, we establish an effective estimate of Green s function (Theorem 4). Estimates of Green s function are important, e.g., for numerical methods [4, 5, 0, 7, 2].. The definition of Green s function For λ C and t R, we consider the functions { exp t (λ) = e λt, if t > 0, 0, if t < 0, { exp t (λ) = 0, if t > 0, e λt, if t < 0, { exp t (λ), if Reλ > 0, g t (λ) = exp t (λ), if Reλ < 0. These functions are undefined for t = 0. The function g t is also undefined for Reλ = 0. For any fixed t 0, all three functions are analytic on their domains. Let A be a complex matrix of the size n n. We consider the differential equation x (t) = Ax(t)f(t), t R. () The bounded solutions problem is the problem of finding bounded solution x : R C n under the assumption that the free term f : R C n is a bounded function. Theorem ([7, Theorem 4., p.8]). Let A be a complex matrix of the size n n. Equation () has a unique bounded on R solution x for any bounded continuous function f if and only if the spectrum σ(a) of A does not intersect the imaginary axis. This solution admits the representation x(t) = G(A,s)f(t s)ds, where G(A,t) = g t (λ)(λ A) dλ, t 0, 2πi Γ the contour Γ encloses σ(a), and is the identity matrix. The function G is called [7] Green s function of the bounded solutions problem for equation ().
AN ESTIMATE OF GREEN S FUNCTION 3 Corollary 2. Let A = Q BQ, where Q is unitary. Then G(A,t) 2 2 = G(B,t) 2 2, t 0, where 2 2 is the matrix norm induced by the norm z 2 = z 2 z n 2 on C n. Proof. It is well known that for any analytic function f f(a) = Q f(b)q. 2. The estimate Proposition 3. For any square matrices A and B of the same size with the spectrum disjoint from the imaginary axis G(A,t) G(B,t) = G(A,t)(A B)G(B,t)dt. Proof. We organize the proof as a sequence of references. By [20, Corollary 47], G(A,t) G(B,t) = g t (λ)(λ A) (A B)(λ B) dλ, 2πi Γ where the contour Γ surrounds σ(a) σ(b). By [20, Theorem 45], g t (λ)( A) (A B)( B) dλ 2πi Γ = g [] (2πi) 2 t (λ,µ)(λ A) (A B)(µ B) dµdλ, Γ 2 Γ where g [] t is the divided difference of the function g t, Γ surrounds σ(a), and Γ 2 surrounds σ(b). We recall that the divided difference [8, 5] of a function f is the function { f(λ) f(µ) f [], if λ µ, λ µ (λ,µ) = (2) f (λ), if λ = µ. For more about g [] t, see, e.g., [7]. By [9, Theorem 23], (2πi) 2 t (λ,µ)(λ A) (A B)(µ B) dµdλ Γ Γ 2 g [] = The main result of this paper is the following theorem. G(A,s)(A B)G(B,t s)ds. Theorem 4. Let a complex triangular matrix B have the size n n and be represented as the sum B = DN of a diagonal matrix D and a strictly triangular matrix N. Let γ > 0 and γ > 0 be chosen so that the strip {z C : γ < Rez < γ } is disjoint from the spectrum σ(b) of the matrix B. Let the norm
4 V. G. KURBATOV AND I.V. KURBATOVA on the space of matrices be induced by a norm on C n and possesses the property D max i d ii for any diagonal matrix D. Then G(B,t) ( η(t)e γ t η( t)e ) ( n γ t N k t k γ t e γ t /2 K k γ t ), (3) πk! where K is the modified Bessel function of the second kind [26], η is the Heaviside function { if t > 0, η(t) = 0 if t < 0 and γ = γ γ. Remark. WiththehelpofCorollary2,inthecaseofthe 2 2 norm,theorem4 canbeappliedtoanymatrixaprovideditstriangularrepresentationa = Q BQ is known. Remark 2. It can be shown that the function t tk ( γte γt/2 K k γt) is a πk! polynomial of degree k with the leading term tk. For example, if B is normal, k! then N = 0 and, thus, ( n N k t k γ t e γ t /2 K k γ t ) ( = N 0 t 0 γ t e γ t /2 K γ t ) =. πk! π0! Remark 3. Another similar estimate of Green s function is proved in [8]: m G(A,t) e γ t G(A,t) e γ t l j ( ) li t j i (2 A ) lj, t > 0, l (j i)! γ li i=0 j ( ) mi t j i (2 A ) mj, t < 0. m (j i)! γ mi i=0 Here m (respectively, l) is the number of eigenvalues of A counted according to their algebraic multiplicities that lie in the open left (right) half-plane. In this estimate the maximal powers of t are m and l whereas the highest power of t in (3) is n. On the other hand, estimate from [8] contains the factors A k instead of N k in (3), which may be essentially smaller; for example, N = 0 if A is normal. Proof of Theorem 4. From Proposition 3 it follows that G(B,t) = G(D N,t) = G(D,t) G(D N,s)NG(D,t s)ds. (4)
AN ESTIMATE OF GREEN S FUNCTION 5 We substitute this representation into itself: G(B,t) = G(D,t) G(D,s )NG(D,t s )ds G(D N,s 2 )NG(D,s s 2 )NG(D,t s )ds 2 ds. Then we substitute (4) again into the obtained formula several times: G(B,t) = G(D,t) G(D,s )NG(D,t s )ds G(D,s 2 )NG(D,s s 2 )NG(D,t s )ds 2 ds G(D,s 3 )NG(D,s 2 s 3 ) NG(D,s s 2 )NG(D,t s )ds 3 ds 2 ds... G(D,s n )NG(D,s n 2 s n )... NG(D,s s 2 )NG(D,t s )ds n... ds. (5) The subsequent terms are zero because of the following reason. Since the matrix D is diagonal, the matrix function G(D, ) is also diagonal. The matrix N is strictly triangular. Therefore the diagonal of NG( )NG( ) closest to the main one is zero, and so on. From representation (5) it follows that G(B,t) G(D,t) N N 2 N n G(D,s ) G(D,t s ) ds G(D,s 2 ) G(D,s s 2 ) G(D,t s ) ds 2 ds... G(D,s n ) G(D,s n 2 s n )... G(D,s s 2 ) G(D,t s ) ds n... ds. We set h(t) = { e γ t for t 0, e γ t for t 0. Since the matrix D is diagonal, G(D,t) h(t).
6 V. G. KURBATOV AND I.V. KURBATOVA Therefore, G(D N,t) h(t) N N 2 N n h(s )h(t s )ds h(s 2 )h(s s 2 )h(t s )ds 2 ds... h(s n )h(s n 2 s n )...h(t s )ds n... ds. We denote the k-fold convolution of h with itself by h k : h k (t) = h(s k )h(s k 2 s k )...h(t s )ds k... ds. With this notation, the previous estimate takes the from n G(B,t) N k h (k) (t). Next we calculate h k (t). We do some calculations with the help of Mathematica [27]. Clearly, the Fourier transform ĥ of h is ĥ(ω) = γ iω γ iω. Therefore the Fourier transform ĥ k of h k is ( ĥ k (ω) = h k (ω) = γ iω ) k = γ iω Further we have k j (γ iω) j (γ iω) = ( j q k j q Consequently, ĥ k (ω) = = ( ) k j ) (γ iω) j (γ iω) k j. γ jq (γ iω) k j q j ( ) k j q q γ k jq (γ iω) j q. ( ) k j (γ iω) j (γ iω) k j ( ) k j k ( ) j q j q γ jq (γ iω) k j q ( ) k j ( ) k j q j q γ k jq (γ iω) j q
AN ESTIMATE OF GREEN S FUNCTION 7 It is well known that the inverse Fourier transform takes the functions ω (γ iω) k j q, ω (γ iω) j q, respectively, to the functions Hence t k j q t η(t) (k j q )! e γ t, t η( t) ( )j q t j q e γt. (j q )! h k (t) = e γ t η(t) e γt η( t) ( k j ) k j We change the order of summation: k p h k (t) = e γ t η(t) e γt η( t) ( ) j q t k j q q γ jq (k j q )! ( ) k j ( ) k j q ( ) j q t j q. j q γ k jq (j q )! p=0 k p p= ( k j )( p p j ( k pq ) γ p )( k p q t k p (k p )! ) γ k p ( ) p t p (p )!. First, we calculate the internal sums (here we essentially use Mathematica ): k h k (t) = e γ t t k p (k p )! η(t) γ p (k )!p!(k p )! e γt η( t) p=0 p= ( ) k p t k p (k p )! γ p (k )!p!(k p )!. Then, by calculating the final sums (here we again essentially use Mathematica ), we arrive at h k (t) = e γ t η(t) tk γte 2 γt K 2 k ( 2 γt) π(k )! e γt η( t)( ) k tk γte ( 2 γt K 2 k γt) 2 π(k )! (6) = h(t) t k ( γ t e γ t /2 K 2 k γ t ) 2. π(k )! Remark 4. When γ, we have e γt 0 for t < 0 and (which is calculated in Mathematica ) t k ( γ t e γ t /2 K k γ t ) =. πk! lim γ
8 V. G. KURBATOV AND I.V. KURBATOVA Therefore estimate (3) turns into the estimate from [25]: n G(B,t) η(t)e γ t N k t k. k! Sometimes N k may decrease faster than N k. In such a case it may be more convenient to use the following variant of Theorem 4. Corollary 5. Let the norm on the space of matrices be induced by the norm z = z z n or the norm z = max i z i on C n. Then under the assumptions of Theorem 4 G(B,t) ( η(t)e γ t η( t)e ) ( n γ t N k t k γ t e γ t /2 K k γ t ), πk! where A is the matrix consisting of absolute values a ij of the entries a ij of a matrix A, and the inequality A B for matrices means the corresponding inequalities for their entries. Remark 5. The idea of using the power N k was used in [2, Theorem.2.2] for the estimating of the general function of a matrix. Proof. Both the norms in question possess the evident properties: A B A B, A B A B, A = A, and A B if A B. Hence, from (5) we have Therefore, G(B,t) G(D,t) G(D,s )NG(D,t s ) ds G(D,s 2 )NG(D,s s 2 )NG(D,t s ) ds 2 ds G(D,s n )NG(D,s n 2 s n )... NG(D,s s 2 )NG(D,t s ) ds n... ds. G(B,t) h(t) (h(s )) N (h(t s ))ds (h(s 2 )) N (h(s s 2 )) N (h(t s ))ds 2 ds (h(s n )) N (h(s n 2 s n ))... N (h(s s 2 )) N (h(t s ))ds n... ds.
AN ESTIMATE OF GREEN S FUNCTION 9 Or G(B,t) h(t) n = N k h (k) (t). h(s )h(t s ) N ds h(s 2 )h(s s 2 )h(t s ) N 2 ds 2 ds h(s n )(h(s n 2 s n )...h(s s 2 ) N n ds n... ds Now the proof follows from estimate (6). References [] R. R. Akhmerov and V. G. Kurbatov, Exponential dichotomy and stability of neutral type equations, J. Differential Equations 76 (988), no., 25. MR 96460 [2] A. G. Baskakov, Estimates for the Green s function and parameters of exponential dichotomy of a hyperbolic operator semigroup and linear relations, Mat. Sb. 206 (205), no. 8, 23 62, (in Russian); English translation in Sb. Math., 206 (205), no. 8, 049 086. MR 3438589 [3] A. A. Boichuk and A. A. Pokutnii, Bounded solutions of linear differential equations in a Banach space, Nonlinear Oscillations 9 (2006), no., 3 4. MR 2369770 [4] A. Ya. Bulgakov, An effectively calculable parameter of stability quality of systems of linear differential equations with constant coefficients, Sibirsk. Mat. Zh. 2 (980), no. 3, 32 4, 235, (in Russian); English translation in Siberian Mathematical Journal, 2(980), no. 3, 339 347. MR 574960 [5], An estimate of the Green matrix and the continuity of the dichotomy parameter, Sibirsk. Mat. Zh. 30 (989), no., 78 82, (in Russian); English translation in Siberian Mathematical Journal, 30(989), no., 39 42. MR 995033 [6] C. Chicone and Y. Latushkin, Evolution semigroups in dynamical systems and differential equations, Mathematical Surveys and Monographs, vol. 70, American Mathematical Society, Providence, RI, 999. MR 707332 [7] Ju. L. Daleckiĭ and M. G. Kreĭn, Stability of solutions of differential equations in Banach space, Translations of Mathematical Monographs, vol. 43, American Mathematical Society, Providence, RI, 974. MR 0352639 [8] A. O. Gel fond, Calculus of finite differences, International Monographs on Advanced Mathematics and Physics, Hindustan Publishing Corp., Delhi, 97, Translation of the third Russian edition. MR 0342890 [9] M. I. Gil, Estimate for the norm of matrix-valued functions, Linear and Multilinear Algebra 35 (993), no., 65 73. MR 30964 [0] S. K. Godunov, The problem of the dichotomy of the spectrum of a matrix, Sibirsk. Mat. Zh. 27 (986), no. 5, 24 37, 204. MR 873707 [], Ordinary differential equations with constant coefficient, Translations of Mathematical Monographs, vol. 69, American Mathematical Society, Providence, RI, 997, Translated from the 994 Russian original. MR 465434 [2] G. H. Golub and Ch. F. Van Loan, Matrix computations, third ed., Johns Hopkins Studies in the Mathematical Sciences, Johns Hopkins University Press, Baltimore, MD, 996. MR 47720
0 V. G. KURBATOV AND I.V. KURBATOVA [3] Ph. Hartman, Ordinary differential equations, second ed., Classics in Applied Mathematics, Society for Industrial and Applied Mathematics (SIAM), Philadelphia, PA, 2002. MR 92904 [4] D. Henry, Geometric theory of semilinear parabolic equations, Lecture Notes in Mathematics, vol. 840, Springer-Verlag, Berlin New York, 98. MR 60244 [5] Ch. Jordan, Calculus of finite differences, third ed., Chelsea Publishing Co., New York, 965. MR 083987 [6] T. Kato, Perturbation theory for linear operators, Classics in Mathematics, Springer Verlag, Berlin, 995, Reprint of the 980 edition. [7] V. G. Kurbatov and I. V. Kurbatova, Computation of Green s function of the bounded solutions problem, Computational Methods in Applied Mathematics (208), no. 4, 673 685. [8], The Gelfand Shilov type estimate for Green s function of the bounded solutions problem, Qualitative Theory of Dynamical Systems 7 (208), no. 3, 69 629. [9], Green s function of the problem of bounded solutions in the case of a block triangular coefficient, April 208, arxiv: 804.0022. [20] V. G. Kurbatov, I. V. Kurbatova, and M. N. Oreshina, Analytic functional calculus for two operators, April 206, arxiv: 604.07393. [2] A. N. Malyshev, Introduction to numerical linear algebra, Nauka, Novosibirsk, 99, (in Russian). [22] J. L. Massera and J. J. Schäffer, Linear differential equations and function spaces, Pure and Applied Mathematics, Vol. 2, Academic Press, New York London, 966. MR 022324 [23] A. A. Pankov, Bounded and almost periodic solutions of nonlinear operator differential equations, Mathematics and its Applications (Soviet Series), vol. 55, Kluwer Academic Publishers, Dordrecht, 990, Translated from the 985 Russian edition. MR 2078 [24] A. V. Pechkurov, Bisectorial operator pencils and the problem of bounded solutions, Izv. Vyssh. Uchebn. Zaved. Mat. (202), no. 3, 3 4, (in Russian); English translation in Russian Math. (Iz. VUZ), 56 (202), no. 3, 26 35. MR 307656 [25] Ch. F. Van Loan, The sensitivity of the matrix exponential, SIAM J. Numer. Anal. 4 (977), no. 6, 97 98. MR 046837 [26] G. N. Watson, A treatise on the theory of Bessel functions, Cambridge Mathematical Library, Cambridge University Press, Cambridge, 995, Reprint of the second (944) edition. MR 3490 [27] S. Wolfram, The Mathematica book, fifth ed., Wolfram Media, New York, 2003. Department of Mathematical Physics, Voronezh State University,, Universitetskaya Square, Voronezh 39408, Russia E-mail address: kv5@inbox.ru Department of Software Development and Information Systems Administration, Voronezh State University,, Universitetskaya Square, Voronezh 39408, Russia E-mail address: la soleil@bk.ru