## Analysis of some Krylov subspace approximations to the matrix exponential operator (1992)

Venue: | SIAM J. Numer. Anal |

Citations: | 84 - 7 self |

### BibTeX

@ARTICLE{Saad92analysisof,

author = {Y. Saad},

title = {Analysis of some Krylov subspace approximations to the matrix exponential operator},

journal = {SIAM J. Numer. Anal},

year = {1992},

volume = {29},

pages = {209--228}

}

### Years of Citing Articles

### OpenURL

### Abstract

Abstract. In this note we present a theoretical analysis of some Krylov subspace approximations to the matrix exponential operation exp(A)v and establish a priori and a posteriori error estimates. Several such approximations are considered. The main idea of these techniques is to approximately project the exponential operator onto a small Krylov subspace and carry out the resulting small exponential matrix computation accurately. This general approach, which has been used with success in several applications, provides a systematic way of defining high order explicit-type schemes for solving systems of ordinary differential equations or time-dependent Partial Differential Equations. 1. Introduction. The

### Citations

1831 | Perturbation Theory for Linear Operators - Kato - 1966 |

1383 | GMRES: a generalized minimal residual algorithm for solving nonsymmetric linear systems
- Saad, Schultz
- 1986
(Show Context)
Citation Context ...nt; (iii) The minimal polynomial of v is of degree equal to m. Proof. That (iii) is equivalent to (ii) is a simple and well-known result on Krylov subspaces and was essentially proved above, see also =-=[11]-=-. In addition, the previous proposition shows that (ii) → (i). It remains to show that (i) → (ii). Taking the derivative of order k (with respect to τ) of the condition in (i) we get, A k e τA v = βVm... |

592 |
The theory of matrices
- Gantmacher
- 1959
(Show Context)
Citation Context ...−1 be the interpolating polynomial of the function f(z), in the Hermite sense, at the roots of the minimal polynomial of A, repeated according to their multiplicities. Then: (19) f(A) = pν−1(A) . See =-=[5]-=- for a proof. Consider now a Hessenberg matrix H. It is known that whenever hj+1,j �= 0, j = 1, 2,...,m − 1, the geometric multiplicity of each eigenvalue is one, i.e., the minimal polynomial of H is ... |

23 |
Unitary quantum time evolution by iterative Lanczos reduction
- Park, Light
- 1986
(Show Context)
Citation Context ...fferential equations (ODE’s) or time-dependent partial differential equations (PDE’s). Recently, the use of Krylov subspace techniques in this context has been actively investigated in the literature =-=[2, 3, 4, 9, 10]-=-. Friesner et al. [2] and Gallopoulos and Saad [3] introduced a few different ways of applying this approximation to the solution of systems of ordinary differential equations. The paper [3] presents ... |

17 |
A method of exponential propagation of large systems of sti nonlinear dierential equations
- Friesner, Tuckerman, et al.
- 1989
(Show Context)
Citation Context ...fferential equations (ODE’s) or time-dependent partial differential equations (PDE’s). Recently, the use of Krylov subspace techniques in this context has been actively investigated in the literature =-=[2, 3, 4, 9, 10]-=-. Friesner et al. [2] and Gallopoulos and Saad [3] introduced a few different ways of applying this approximation to the solution of systems of ordinary differential equations. The paper [3] presents ... |

13 |
Extended numerical computations on the 1/9 conjecture in rational approximation theory, in Rational Approximation and Interpolation
- CARPENTER, RUTTAN, et al.
- 1984
(Show Context)
Citation Context ...lex arithmetic. The rational functions that are used in [3] as well as in the numerical experiments section in this paper are based on the Chebyshev approximation on [0, ∞) of the function e −x , see =-=[1]-=- and the references therein. This approximation is remarkably accurate and, in general, a small degree approximation, as small as p = 14, suffices to obtain a good working accuracy. We also observe th... |

11 |
Applications of the Lanczos algorithm
- NOUR-OMID
- 1989
(Show Context)
Citation Context ...fferential equations (ODE’s) or time-dependent partial differential equations (PDE’s). Recently, the use of Krylov subspace techniques in this context has been actively investigated in the literature =-=[2, 3, 4, 9, 10]-=-. Friesner et al. [2] and Gallopoulos and Saad [3] introduced a few different ways of applying this approximation to the solution of systems of ordinary differential equations. The paper [3] presents ... |

8 |
Y: On the parallel solution of parabolic equations
- Gallopoulos, Saad
(Show Context)
Citation Context ... the problem of evaluating exp(H)b where H is a small, possibly dense, matrix. One way in which the exponential of a matrix is evaluated is by using rational approximation to the exponential function =-=[3, 4]-=-. Using a rational approximation R(z), 5 � .sof the type (p,p), i.e., such that the denominator has the same degree p as the numerator, it is more viable and economical to use the partial fraction exp... |

7 |
New approach to many-state quantum dynamics: The recursiveresidue-generation method, Phys
- Nauts, Wyatt
- 1983
(Show Context)
Citation Context ...as been successfully used in several applications, in spite of the lack of theory to justify it. For example, it is advocated in the paper by Park and Light [10] following the work by Nauts and Wyatt =-=[7, 8]-=-. Also, the idea of exploiting the Lanczos algorithm to evaluate terms of the exponential of Hamiltonian operators seems to have been first used in Chemical Physics, by Nauts and Wyatt [7]. The genera... |

3 |
The asymptotic accuracy of rational best approximation to e z on a disk
- Trefethen
- 1984
(Show Context)
Citation Context ... = xm m! (1 + o(1)) ≤ xmex m! Proof. The integral form of the remainder in the Taylor series for e x is tm(x) = xm � 1 e (m − 1)! 0 (1−τ)x τ m−1 dτ Following the argument in the proof of Theorem 1 in =-=[12]-=-, we notice that as m increases the term inside the integrand becomes an increasingly narrow spike centered around τ = 1. As a result the function e (1−τ)x in that narrow interval can be assimilated t... |

3 |
An Iterative Solution Method for Solving f(A) = b, using Krylov Subspace Information Obtained for the Symmetric Positive Definite Matrix
- Vorst
- 1987
(Show Context)
Citation Context ... which consists of projecting the exponential operator on the Krylov subspace. The evaluation of arbitrary functions of a matrix with Krylov subspaces has also been briefly mentioned by van der Vorst =-=[13]-=-. The purpose of this paper is to explore these techniques a little further both from a practical and a theoretivcal viewpoint. On the practical side we introduce new schemes that can be viewed as sim... |

1 |
Parallel solution of parabolic equations using polynomial approximations to the exponential
- Gallopoulos, Saad
(Show Context)
Citation Context |

1 |
Theory of laser-module interaction: The recursive-residue-generation method
- Nauts, Wyatt
- 1984
(Show Context)
Citation Context ...as been successfully used in several applications, in spite of the lack of theory to justify it. For example, it is advocated in the paper by Park and Light [10] following the work by Nauts and Wyatt =-=[7, 8]-=-. Also, the idea of exploiting the Lanczos algorithm to evaluate terms of the exponential of Hamiltonian operators seems to have been first used in Chemical Physics, by Nauts and Wyatt [7]. The genera... |