On Krylov Subspace Approximations to the Matrix Exponential Operator
Open accessThe DOI matches its Crossref record; 29 of the 30 references it cites were found in a registry or an index, and 1 are not checked yet.
Abstract
Krylov subspace methods for approximating the action of matrix exponentials are analyzed in this paper. We derive error bounds via a functional calculus of Arnoldi and Lanczos methods that reduces the study of Krylov subspace approximations of functions of matrices to that of linear systems of equations. As a side result, we obtain error bounds for Galerkin-type Krylov methods for linear equations, namely, the biconjugate gradient method and the full orthogonalization method. For Krylov approximations to matrix exponentials, we show superlinear error decay from relatively small iteration numbers onwards, depending on the geometry of the numerical range, the spectrum, or the pseudospectrum. The convergence to exp$(\tau A)v$ is faster than that of corresponding Krylov methods for the solution of linear equations $(I-\tau A)x=v$, which usually arise in the numerical solution of stiff ordinary differential equations (ODEs). We therefore propose a new class of time integration methods for large systems of nonlinear differential equations which use Krylov approximations to the exponential function of the Jacobian instead of solving linear or nonlinear systems of equations in every time step.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“Analogous to the eigenvalue problem in DMRG, both Eqs. ( 7) and ( 8) can be integrated using a Lanczos scheme [28] in order to keep the computational cost at O(D 3 ).”
From Unifying time evolution and optimization with matrix product states (2016) · cites this paper as [28]
Citations
Cites
Solving Ordinary Differential Equations II
Hairer, E., & Wanner, G. (1996) · Springer Series in Computational Mathematics · doi:10.1007/978-3-642-05221-7 · Cited by 4,342
Solving Ordinary Differential Equations II
Hairer, E., & Wanner, G. (1991) · Springer Series in Computational Mathematics · doi:10.1007/978-3-662-09947-6 · Cited by 1,456
Theory of Functions of a Complex Variable
Markushevich, A. I., Silverman, R. A., & Gillis, J. (1966) · Physics Today · doi:10.1063/1.3048359 · Cited by 1,211
Unitary quantum time evolution by iterative Lanczos reduction
Park, T. J., & Light, J. C. (1986) · The Journal of Chemical Physics · doi:10.1063/1.451548 · Cited by 888
Uniform Numerical Methods for Problems with Initial and Boundary Layers.
Berger, A. E., Doolan, E. P., Miller, J. J. H., & Schilders, W. H. A. (1982) · Mathematics of Computation · doi:10.2307/2007353 · Cited by 754
Analysis of Some Krylov Subspace Approximations to the Matrix Exponential Operator
Saad, Y. (1992) · SIAM Journal on Numerical Analysis · doi:10.1137/0729014 · Cited by 670
Numerical Ranges of Operators on Normed Spaces and of Elements of Normed Algebras
Bonsall, F. F., & Duncan, J. (1971) · Cambridge University Press eBooks · doi:10.1017/cbo9781107359895 · Cited by 614
Propagation Methods for Quantum Molecular Dynamics
Kosloff, R. (1994) · Annual Review of Physical Chemistry · doi:10.1146/annurev.pc.45.100194.001045 · Cited by 602
Krylov subspace methods for solving large unsymmetric linear systems
Saad, Y. (1981) · Mathematics of Computation · doi:10.1090/s0025-5718-1981-0616364-6 · Cited by 449
Generalized Runge-Kutta Processes for Stable Systems with Large Lipschitz Constants
Lawson, J. D. (1967) · SIAM Journal on Numerical Analysis · doi:10.1137/0704033 · Cited by 328
Research works printed but not linked: 5
- A. Friedli, Verallgemeinerte Runge‐Kutta Verfahren zur Lösung steifer Differentialgleichungen , in Lecture Notes in Mathematics 631, Springer‐Verlag, Berlin, 1978, pp. 35–50. · doi:10.1007/bfb0067462
- Hochbruck M., 1997, SIAM J. Sci. Comput., 18
- Knizhnerman L. A., 1991, Comput. Math. and Math. Phys., 31, 1
- Knizhnerman L. A., 1992, Comput. Math. and Math. Phys., 32, 1199
- D. R. Taylor, Analysis of the look‐ahead Lanczos algorithm , Ph.D. thesis, University of California at Berkeley, Berkeley, CA, 1982.
Cited by
Nineteen Dubious Ways to Compute the Exponential of a Matrix, Twenty-Five Years Later
Moler, C., & Van Loan, C. (2003) · SIAM Review · doi:10.1137/s00361445024180 · Cited by 2,357
Continuous-time Monte Carlo methods for quantum impurity models
Gull, E., Millis, A. J., Lichtenstein, A. I., Rubtsov, A. N., Troyer, M., & Werner, P. (2011) · Reviews of Modern Physics · doi:10.1103/revmodphys.83.349 · Cited by 1,546
“(Hochbruck and Lubich, 1997) it has been shown rigorously that these Krylov space algorithms converge rapidly as a function of Niter, typically reaching convergence for very small iteration numbers p ≪ Ndim, although the number of iterations depends on the time interval τ .”
BackgroundThe Magnus expansion and some of its applications
Blanes, S., Casas, F., Oteo, J. A., & Ros, J. (2009) · Physics Reports · doi:10.1016/j.physrep.2008.11.001 · Cited by 1,157
“More details are contained in the references [103,105,106].”
BackgroundResultHochbruck, M., & Ostermann, A. (2010) · Acta Numerica · doi:10.1017/s0962492910000048 · Cited by 1,061
Self-citationUnifying time evolution and optimization with matrix product states
Haegeman, J., Lubich, C., Oseledets, I., Vandereycken, B., & Verstraete, F. (2016) · Physical Review B · doi:10.1103/physrevb.94.165116 · Cited by 804
Self-citationIserles, A., Munthe-Kaas, H. Z., Nørsett, S. P., & Zanna, A. (2000) · Acta Numerica · doi:10.1017/s0962492900002154 · Cited by 782
“Surprisingly small values of M produce remarkably good and affordable approximants (Hochbruck and Lubich 1997).”
BackgroundMethodExpokit: a software package for computing matrix exponentials
Sidje, R. B. (1998) · ACM Transactions on Mathematical Software · doi:10.1145/285861.285868 · Cited by 781
“, Lawson [1967], Hochbruck and Lubich [1996], Hochbruck et al. [1996]. A direct way to define the matrix exponential exp(A) is undoubtedly through the exponential power series expansion exp(A) = ∑∞ k=0 A /k! whose convergence is guaranteed for any square matrix A.”
BackgroundTime-evolution methods for matrix-product states
Paeckel, S., Köhler, T., Swoboda, A., Manmana, S. R., Schollwöck, U., & Hubig, C. (2019) · Annals of Physics · doi:10.1016/j.aop.2019.167998 · Cited by 768
“Errors Elaborate bounds are available to estimate the error incurred during the Krylov approximation[60].”
BackgroundInfluentialThe Magnus expansion and some of its applicationsPreprint
Blanes, S., Casas, F., Oteo, J. Á., & Ros, J. (2008) · Cited by 676
Propagators for the time-dependent Kohn–Sham equations
Castro, A., Marques, M. A. L., & Rubio, A. (2004) · The Journal of Chemical Physics · doi:10.1063/1.1774980 · Cited by 599
Exponential Integrators for Large Systems of Differential Equations
Hochbruck, M., Lubich, C., & Selhofer, H. (1998) · SIAM Journal on Scientific Computing · doi:10.1137/s1064827595295337 · Cited by 509
“Here the theory in [5, 13] and [31] tells us that the number of necessary matrix–vector multiplications with the Jacobian in Krylov iterations for exponential methods and the number of function evaluations in Runge–Kutta–Chebyshev methods needed to attain stability are both of the magnitude of √ h‖A‖.”
BackgroundInfluentialSelf-citationComputational Methods for Linear Matrix Equations
Simoncini, V. (2016) · SIAM Review · doi:10.1137/130912839 · Cited by 487
“, [20],[115],[125],[77],[156],[137],[116],[108],[82],[126],[194] and their references.”
Recent computational developments in Krylov subspace methods for linear systems
Simoncini, V., & Szyld, D. B. (2007) · Numerical Linear Algebra with Applications · doi:10.1002/nla.499 · Cited by 375
“2; for KKT systems or saddle point problems [33], [100]; or for integral operators [203]; considerations for parallel computers [82], [104], [178]; methods for singular matrices [134], [198], [262]; or for regularization of ill-posed problems [58], [179]; multiple right-hand side and block methods [172]; nonlinear equ…”
BackgroundA New Iterative Method for Solving Large-Scale Lyapunov Matrix Equations
Simoncini, V. (2007) · SIAM Journal on Scientific Computing · doi:10.1137/06066120x · Cited by 334
“In particular, assuming for the sake of the presentation t = 1, if the numerical range of A is contained in the complex disk |z + ρ| ≤ ρ for some ρ > 0, then a bound of the form ‖ exp(A)v − Vk exp(Hk)e1‖ = O(( eρ k )), holds, for k ≥ 2ρ; see [21].”
BackgroundExpansion Dynamics of Interacting Bosons in Homogeneous Lattices in One and Two Dimensions
Ronzheimer, J. P., Schreiber, M., Braun, S., Hodgman, S. S., Langer, S., McCulloch, I. P., Heidrich-Meisner, F., Bloch, I., & Schneider, U. (2013) · Physical Review Letters · doi:10.1103/physrevlett.110.205301 · Cited by 300
Du, Q., Ju, L., Li, X., & Qiao, Z. (2021) · SIAM Review · doi:10.1137/19m1243750 · Cited by 294
A New Class of Time Discretization Schemes for the Solution of Nonlinear PDEs
Beylkin, G., Keiser, J. M., & Vozovoi, L. P. (1998) · Journal of Computational Physics · doi:10.1006/jcph.1998.6093 · Cited by 287
“[11] and references therein) due to the ubiquitous nature of this operator in physics and mathematics.”
BackgroundGeneralized integrating factor methods for stiff PDEs
Krogstad, S. (2005) · Journal of Computational Physics · doi:10.1016/j.jcp.2004.08.006 · Cited by 233
Ultrafast and reversible control of the exchange interaction in Mott insulators
Mentink, J. H., Balzer, K., & Eckstein, M. (2015) · Nature Communications · doi:10.1038/ncomms7708 · Cited by 241
Exponential Rosenbrock-Type Methods
Hochbruck, M., Ostermann, A., & Schweitzer, J. (2009) · SIAM Journal on Numerical Analysis · doi:10.1137/080717717 · Cited by 241
“Since exprb43 uses significantly larger time steps, we know from the convergence analysis of Krylov subspace methods [7, 11] that this requires more Krylov steps.”
BackgroundMethodSelf-citationA Gautschi-type method for oscillatory second-order differential equations
Hochbruck, M., & Lubich, C. (1999) · Numerische Mathematik · doi:10.1007/s002110050456 · Cited by 239
“Otherwise (or possibly in combination with a partial eigendecomposition), such matrix-function vector products can be computed with Krylov subspace methods [2,7].”
BackgroundSelf-citationNovel Numerical Methods for Solving the Time-Space Fractional Diffusion Equation in Two Dimensions
Yang, Q., Turner, I., Liu, F., & Ilić, M. (2011) · SIAM Journal on Scientific Computing · doi:10.1137/100800634 · Cited by 254
“This approximation has been considered by van der Vorst [50], Saad [39], Druskin and Knizhnerman [7], Hochbruck and Lubich [20], Sidje [45], van den Eshof et al. [49], Eiermann and Ernst [8], Lopez and Simoncini [31], Ilić, Turner, and Anh [23], and Ilić, Turner, and Simpson [24] as well as many other researchers du…”
MethodProjective Methods for Stiff Differential Equations: Problems with Gaps in Their Eigenvalue Spectrum
Gear, C. W., & Kevrekidis, I. G. (2003) · SIAM Journal on Scientific Computing · doi:10.1137/s1064827501388157 · Cited by 236
“They often form the subspace from the Krylov sequence (see [7], [8], and [3]), or they may use Chebyshev polynomials to get a lower-dimensional approximation to the solution operator in the fast subspace [10].”
Preconditioning Lanczos Approximations to the Matrix Exponential
van den Eshof, J., & Hochbruck, M. (2006) · SIAM Journal on Scientific Computing · doi:10.1137/040605461 · Cited by 217
“For matrices A that stem from a self-adjoint elliptic partial differential equation, it has been shown [7, 8, 13] that Krylov approximations to y(τ) always yield superlinear error reduction with the superlinear decay starting after the number of steps exceeds ‖A‖.”
BackgroundMethodSelf-citationNiesen, J., & Wright, W. M. (2012) · ACM Transactions on Mathematical Software · doi:10.1145/2168773.2168781 · Cited by 200
“Hochbruck and Lubich (1997) show that for particular classes of matrices, the convergence of the action of the matrix exponential on a vector is faster than that for the solution of the corresponding linear system.”
Predicting Parking Lot Occupancy in Vehicular Ad Hoc Networks
Caliskan, M., Barthels, A., Scheuermann, B., & Mauve, M. (2007) · 2007 IEEE 65th Vehicular Technology Conference - VTC2007-Spring · doi:10.1109/vetecs.2007.69 · Cited by 187
“Scaling and squaring with rational Padé approximation in combination with Krylov subspace approximations [10] seems very promising for the specific structure of our tri-diagonal Q -matrix.”
Exponential Runge–Kutta methods for parabolic problems
Hochbruck, M., & Ostermann, A. (2005) · Applied Numerical Mathematics · doi:10.1016/j.apnum.2004.08.005 · Cited by 194
“For more general boundary conditions, Krylov subspace methods, e.g. the Lanczos method, the Arnoldi method, or the Chebyshev method, can be used, see [4] [8].”
MethodSelf-citationMany-Body Localization Dynamics from Gauge Invariance
Brenes, M., Dalmonte, M., Heyl, M., & Scardicchio, A. (2018) · Physical Review Letters · doi:10.1103/physrevlett.120.030601 · Cited by 184
Time Integration of Tensor Trains
Lubich, C., Oseledets, I. V., & Vandereycken, B. (2015) · SIAM Journal on Numerical Analysis · doi:10.1137/140976546 · Cited by 183
“…case of an autonomous linear ODE . the above differential equations are constant-coefficient linear differential equations for K <i and S i , respectively, which can be solved efficiently with a few iterations of a Krylov subspace method for computing the action of the operator exponential [8, 26, 9].”
Self-citationRational Krylov approximation of matrix functions: Numerical methods and optimal pole selection
Güttel, S. (2013) · GAMM-Mitteilungen · doi:10.1002/gamm.201310002 · Cited by 172
Graph spectral image smoothing using the heat kernel
Zhang, F., & Hancock, E. R. (2008) · Pattern Recognition · doi:10.1016/j.patcog.2008.05.007 · Cited by 176
“The exact solution of the algorithm can be efficiently computed without iterations by using the Krylov subspace projection technique [22,23].”
MethodTime integration of tensor trainsPreprint
Lubich, C., Oseledets, I. V., & Vandereycken, B. (2016) · Cited by 171
Self-citationRD-Rational Approximations of the Matrix Exponential
Moret, I., & Novati, P. (2004) · BIT Numerical Mathematics · doi:10.1023/b:bitn.0000046805.27551.3b · Cited by 167
“The corresponding approximations, which are often viewed as projection methods, are discussed, among the others, in the papers [22], [7], [14], [34], [16].”
BackgroundTokman, M. (2006) · Journal of Computational Physics · doi:10.1016/j.jcp.2005.08.032 · Cited by 159
Error Estimates and Evaluation of Matrix Functions via the Faber Transform
Beckermann, B., & Reichel, L. (2009) · SIAM Journal on Numerical Analysis · doi:10.1137/080741744 · Cited by 159
“These formulas show the bound (4.3) for |f m | to be asymptotically sharp as m → ∞ up to a constant independent of m. Hochbruck and Lubich [51, Theorems 2, 4, 5, and 6] apply the bound (3.12) to the exponential function when W(A) ⊂ E for (i) E = [−4c, 0] an interval on the negative real axis and ψ(w) = c(w + 1/w − 2),…”
BackgroundResultInfluentialTime evolution of Matrix Product States
García-Ripoll, J. J. (2006) · New Journal of Physics · doi:10.1088/1367-2630/8/12/305 · Cited by 152
“|0i and |1i are the eigenstates of sz.By restricting ourselves to “small” problems (L≤ 20), we can compareall algorithmswith accurate solutionsbased on exact diagonalizationsand the Lanczos algorithm [29, 30]. Notice that we measure the error in the full wavefunction, ε:= kψD(T) − U(T)ψ(0)k2 and not on the expectation…”
MethodEXPINT---A MATLAB package for exponential integrators
Berland, H., Skaflestad, B., & Wright, W. M. (2007) · ACM Transactions on Mathematical Software · doi:10.1145/1206040.1206044 · Cited by 152
“What we need, is just their action to a given state vector v. Krylov subspace approximations to the exponential and some related functions, have been studied by many authors, see for example [12, 14 , 29].”
BackgroundA Restarted Krylov Subspace Method for the Evaluation of Matrix Functions
Eiermann, M., & Ernst, O. G. (2006) · SIAM Journal on Numerical Analysis · doi:10.1137/050633846 · Cited by 139
“1) directly is based on a Krylov subspace of A with initial vector b [8, 9, 28, 14, 17, 4, 19].”
BackgroundMethodInfluentialComputing Semiclassical Quantum Dynamics with Hagedorn Wavepackets
Faou, E., Grădinaru, V. C., & Lubich, C. (2009) · SIAM Journal on Scientific Computing · doi:10.1137/080729724 · Cited by 138
“The computation of the matrix exponential times a vector can then be done efficiently using just a few Lanczos iterations with F [12].”
MethodSelf-citationZhang, D.-W., Chen, Y.-L., Zhang, G.-Q., Lang, L.-J., Li, Z., & Zhu, S.-L. (2020) · Physical review. B./Physical review. B · doi:10.1103/physrevb.101.235150 · Cited by 137
Implementation of exponential Rosenbrock-type integrators
Caliari, M., & Ostermann, A. (2008) · Applied Numerical Mathematics · doi:10.1016/j.apnum.2008.03.021 · Cited by 121
“The attractive computational features of the method are clear: there is no Krylov subspace to store and the complexity of the two-term recurrence (4.1a) is linear in m, whereas the long-term recurrence in the standard Krylov method is quadratic in m for non-symmetric matrices hJ , see, e.g., [3,7, 10 ].”
MethodA review of exponential integrators for first order semi-linear problemsPreprint
Minchev, B. V., & Wright, W. M. (2005) · CERN Document Server (European Organization for Nuclear Research) · Cited by 121
Out-of-equilibrium dynamics with matrix product states
Wall, M. L., & Carr, L. D. (2012) · New Journal of Physics · doi:10.1088/1367-2630/14/12/125015 · Cited by 146
Polynomially scaling spin dynamics simulation algorithm based on adaptive state-space restriction
Kuprov, I., Wagner-Rundell, N., & Hore, P. J. (2007) · Journal of Magnetic Resonance · doi:10.1016/j.jmr.2007.09.014 · Cited by 118
“The alternative option of using the Krylov space technique [42] to compute expði^̂ LDtÞq̂ðtÞ directly from ^̂ L and q̂ðtÞ requires less memory but is much slower, because it has to be applied at every step.”
MethodPARAEXP: A Parallel Integrator for Linear Initial-Value Problems
Gander, M. J., & Güttel, S. (2013) · SIAM Journal on Scientific Computing · doi:10.1137/110856137 · Cited by 129
ROWMAP—a ROW-code with Krylov techniques for large stiff ODEs
Weiner, R., Schmitt, B. A., & Podhaisky, H. (1997) · Applied Numerical Mathematics · doi:10.1016/s0168-9274(97)00067-6 · Cited by 128
Langer, S., Heidrich‐Meisner, F., Gemmer, J., McCulloch, I. P., & Schollwöck, U. (2009) · Physical Review B · doi:10.1103/physrevb.79.214409 · Cited by 122
Orecchia, L., Sachdeva, S., & Vishnoi, N. K. (2012) · ACM Symposium on Theory of Computing (STOC) · doi:10.1145/2213977.2214080 · Cited by 115
“Finally, we note that Theorem 5 is also implied by an independent earlier work of Hochbruck and Lubich (Theorem 2 in [12]) of which we were unaware when we first wrote this paper.”
ResultPueyo, A. G., Marques, M. A. L., Rubio, Á., & Castro, A. (2018) · Journal of Chemical Theory and Computation · doi:10.1021/acs.jctc.8b00197 · Cited by 119
“) 6 evaluated adiabatically with the timedependent density. Remembering that the density is evaluated from the KS orbitals, we can write the energy as a functional of these E[ϕ] = TS[ϕ]+V[ϕ]+EHxc[ϕ]. (28) Representing the orbitals by the new variables (q,p), we define a Hamiltonian function H(q,p) = TS(q,p)+V(q,p)+EHxc…”
High-order commutator-free exponential time-propagation of driven quantum systems
Alvermann, A., & Fehske, H. (2011) · Journal of Computational Physics · doi:10.1016/j.jcp.2011.04.006 · Cited by 127
“Calculation of exponentials Two powerful approaches for the computation of matrix expon entials, particularly ofe−iM with sparse hermitian matricesM, are the Krylov [19, 20] and the Chebyshev technique [21].”
Lanczos-type solvers for nonsymmetric linear systems of equations
Gutknecht, M. H. (1997) · Acta Numerica · doi:10.1017/s0962492900002737 · Cited by 107
“Further cases where the BlO and BIORES algorithms simplify We have mentioned before that in the symmetric case the BlO algorithm simplifies: the left and the right Lanczos vectors coincide, and therefore only one matrix-vector product is needed per step. In fact, this simplification applies in a somewhat more general…”
BackgroundMethodSpin-charge separation in two-component Bose gases
Kleine, A., Kollath, C., McCulloch, I. P., Giamarchi, T., & Schollwöck, U. (2008) · Physical Review A · doi:10.1103/physreva.77.013607 · Cited by 105
Solution of Large Scale Evolutionary Problems Using Rational Krylov Subspaces with Optimized Shifts
Druskin, V. L., Knizhnerman, L., & Zaslavsky, M. (2009) · SIAM Journal on Scientific Computing · doi:10.1137/080742403 · Cited by 105
Lord, G. J., & Tambue, A. (2012) · IMA Journal of Numerical Analysis · doi:10.1093/imanum/drr059 · Cited by 116
High order splitting methods for analytic semigroups exist
Hansen, E., & Ostermann, A. (2009) · BIT Numerical Mathematics · doi:10.1007/s10543-009-0236-x · Cited by 103
“If further efficiency is required, a possible alte rnative to the expmroutine would be to use Krylov type methods instead; see for e xample [9].”
MethodAdiabatic Quantum Algorithm for Search Engine Ranking
Garnerone, S., Zanardi, P., & Lidar, D. A. (2012) · Physical Review Letters · doi:10.1103/physrevlett.108.230506 · Cited by 102
Discretized Dynamical Low-Rank Approximation in the Presence of Small Singular Values
Kieri, E., Lubich, C., & Walach, H. (2016) · SIAM Journal on Numerical Analysis · doi:10.1137/15m1026791 · Cited by 132
“, using a step of a Runge–Kutta method or, when F is independent of t and linear in Y , by Krylov methods for computing the action of a matrix exponential [9, 21].”
MethodSelf-citationFaster Algorithms via Approximation Theory
Sachdeva, S., & Vishnoi, N. K. (2014) · Foundations and Trends® in Theoretical Computer Science · doi:10.1561/0400000065 · Cited by 159
Schmelzer, T., & Trefethen, L. N. (2007) · Oxford University Research Archive (ORA) (University of Oxford) · Cited by 96
Chanda, T., Sierant, P., & Zakrzewski, J. (2020) · Physical review. B./Physical review. B · doi:10.1103/physrevb.101.035148 · Cited by 103
Confinement and Lack of Thermalization after Quenches in the Bosonic Schwinger Model
Chanda, T., Zakrzewski, J., Lewenstein, M., & Tagliacozzo, L. (2020) · Physical Review Letters · doi:10.1103/physrevlett.124.180602 · Cited by 114
Comparative performance of exponential, implicit, and explicit integrators for stiff systems of ODEs
Loffeld, J. J., & Tokman, M. (2012) · Journal of Computational and Applied Mathematics · doi:10.1016/j.cam.2012.09.038 · Cited by 92
Implementation of a restarted Krylov subspace method for the evaluation of matrix functions
Afanasjew, M., Eiermann, M., Ernst, O. G., & Güttel, S. (2008) · Linear Algebra and its Applications · doi:10.1016/j.laa.2008.06.029 · Cited by 90
“Here Krylov subspace approximations of (1) have become popular (cf. [4,9,14]) and, with regard to solving initial value problems, have had a large impact on so-called exponential integrators, in which evaluations of the exponential function applied to the Jacobian are incorporated directly into…”
BackgroundFourth-order compact schemes for the numerical simulation of coupled Burgers’ equation
Bhatt, H., & Khaliq, A. Q. M. (2015) · Computer Physics Communications · doi:10.1016/j.cpc.2015.11.007 · Cited by 90
Sheen, D., Sloan, I. H., & Thomée, V. (1999) · Mathematics of Computation · doi:10.1090/s0025-5718-99-01098-4 · Cited by 90
“, [11, Chapter 9], [5], and [6].”
Stripes, Antiferromagnetism, and the Pseudogap in the Doped Hubbard Model at Finite Temperature
Wietek, A., He, Y.-Y., White, S. R., Georges, A., & Stoudenmire, E. M. (2021) · Physical Review X · doi:10.1103/physrevx.11.031007 · Cited by 68
“Therefore, we use a Lanczos method [114, 115] to compute the exact reference state |ψED〉.”
BackgroundMethodEfficient and Stable Arnoldi Restarts for Matrix Functions Based on Quadrature
Frommer, A., Güttel, S., & Schweitzer, M. (2014) · SIAM Journal on Matrix Analysis and Applications · doi:10.1137/13093491x · Cited by 89
“By far the most important class of iterative methods for this purpose are Krylov subspace methods, see [12, 20, 28, 31, 38].”
MethodA class of explicit multistep exponential integrators for semilinear problems
Calvo, M. P., & Palencia, C. (2005) · Numerische Mathematik · doi:10.1007/s00211-005-0627-0 · Cited by 89
“behave like e, we can efficiently evaluate (4), by using a Krylov approach [8, 9], at least in case A is a sparse matrix.”
MethodResidual, Restarting, and Richardson Iteration for the Matrix Exponential
Botchev, M. A., Grimm, V., & Hochbruck, M. (2013) · SIAM Journal on Scientific Computing · doi:10.1137/110820191 · Cited by 99
“Except the classical Arnoldi/Lanczos method [8, 23, 4, 14], we have been tested the SaI method of Van den Eshof and Hochbruck [27].”
InfluentialSelf-citationImaginary-Time Matrix Product State Impurity Solver for Dynamical Mean-Field Theory
Wolf, F. A., Go, A., McCulloch, I. P., Millis, A. J., & Schollwöck, U. (2015) · Physical Review X · doi:10.1103/physrevx.5.041032 · Cited by 115
“using a Krylov algorithm [51], which represents the timeevolution operator in a local Krylov space and is able to treat Hamiltonians with long-ranged interactions.”
MethodApproximating the exponential from a Lie algebra to a Lie group
Celledoni, E., & Iserles, A. (2000) · Mathematics of Computation · doi:10.1090/s0025-5718-00-01223-0 · Cited by 85
Numerical assessment for accuracy and GPU acceleration of TD-DMRG time evolution schemes
Li, W., Ren, J., & Shuai, Z. (2020) · The Journal of Chemical Physics · doi:10.1063/1.5135363 · Cited by 76
“32,67 In our calculation, the dimension of the Krylov subspace is adaptive and the Lanczos iteration continues until | Ψ ( t + τ ) ⟩ converges.”
BackgroundExponential Integrators for Quantum-Classical Molecular Dynamics
Hochbruck, M., & Lubich, C. (1999) · BIT Numerical Mathematics · doi:10.1023/a:1022335122807 · Cited by 83
Self-citationTwo-photon finite-pulse model for resonant transitions in attosecond experiments
Jiménez-Galán, Á., Martín, F., & Argenti, L. (2016) · Physical Review A · doi:10.1103/physreva.93.023429 · Cited by 88
Analysis of Projection Methods for Rational Function Approximation to the Matrix Exponential
Lopez, L., & Simoncini, V. (2006) · SIAM Journal on Numerical Analysis · doi:10.1137/05062590 · Cited by 86
Dynamical phase transitions in the two-dimensional transverse-field Ising model
Hashizume, T., McCulloch, I. P., & Halimeh, J. C. (2022) · Physical Review Research · doi:10.1103/physrevresearch.4.013250 · Cited by 88
A numerical study of large sparse matrix exponentials arising in Markov chains
Sidje, R. B., & Stewart, W. J. (1999) · Computational Statistics & Data Analysis · doi:10.1016/s0167-9473(98)00062-0 · Cited by 90
DECAY BOUNDS AND ( ) ALGORITHMS FOR APPROXIMATING FUNCTIONS OF SPARSE MATRICES
Benzi, M., & Razouk, N. (2007) · Scuola Normale Superiore di Pisa · Cited by 124
Calculation of Pseudospectra by the Arnoldi Iteration
Toh, K.-C., & Trefethen, L. N. (1996) · SIAM Journal on Scientific Computing · doi:10.1137/0917002 · Cited by 76
Interpolating discrete advection–diffusion propagators at Leja sequences
Caliari, M., Vianello, M., & Bergamaschi, L. (2004) · Journal of Computational and Applied Mathematics · doi:10.1016/j.cam.2003.11.015 · Cited by 88
Symplectic exponential Runge–Kutta methods for solving nonlinear Hamiltonian systems
Mei, L., & Wu, X. (2017) · Journal of Computational Physics · doi:10.1016/j.jcp.2017.03.018 · Cited by 74
“Results for Problem 4: the numerical wave forms of ERK4 on different time intervals: the left t ∈ [0,1] and the right t ∈ [9,10].”
BackgroundMethodInfluentialRational Krylov Methods for Operator Functions
Güttel, S. (2010) · MIMS EPrints (University of Southampton) · Cited by 73
Efficient Split-Lanczos propagator for strong-field ionization of atoms
Jiang, W., & Tian, X. (2017) · Optics Express · doi:10.1364/oe.25.026832 · Cited by 72
EXPONENTIAL PROPAGATORS (INTEGRATORS) FOR THE TIME-DEPENDENT SCHRÖDINGER EQUATION
Bandrauk, A. D., & Lu, H. (2013) · Journal of Theoretical and Computational Chemistry · doi:10.1142/s0219633613400014 · Cited by 75
Zhu, L., Ju, L., & Zhao, W. (2015) · Journal of Scientific Computing · doi:10.1007/s10915-015-0117-1 · Cited by 72
“For those problems, the classic ETD discretizations and Krylov subspace approximations for computing exponential matrix functions could be employed (see [12]).”
MethodPeng, L.-Y., & Starace, A. F. (2006) · The Journal of Chemical Physics · doi:10.1063/1.2358351 · Cited by 85
$\mathcal{H}$ -Matrix approximation for the operator exponential with applications
Gavrilyuk, I. P., Hackbusch, W., & Khoromskij, B. N. (2002) · Numerische Mathematik · doi:10.1007/s002110100360 · Cited by 70
“The most commonly used algorithms are based on Krylov subspace methods [22,18].”
MethodManipulating magnetism by ultrafast control of the exchange interaction
Mentink, J. H. (2017) · Journal of Physics Condensed Matter · doi:10.1088/1361-648x/aa8abf · Cited by 77
“For larger systems, a more efficient scheme is required and the Krylov technique was used [52].”
MethodAcceleration Techniques for Approximating the Matrix Exponential Operator
Popolizio, M., & Simoncini, V. (2008) · SIAM Journal on Matrix Analysis and Applications · doi:10.1137/060672856 · Cited by 73
“Krylov subspaces have been extensively explored to this purpose, due to their favorable computational and approximation properties, see, e.g., [14], [15], [16], [20], [24], [25], [41].”
BackgroundExponential Rosenbrock methods of order five — construction, analysis and numerical comparisons
Luân, V. T., & Ostermann, A. (2013) · Journal of Computational and Applied Mathematics · doi:10.1016/j.cam.2013.04.041 · Cited by 67
Three‐dimensional Model of the Structure and Evolution of Coronal Mass Ejections
Tokman, M., & Bellan, P. M. (2002) · The Astrophysical Journal · doi:10.1086/338699 · Cited by 67
“We instead brieÑy describe the main ideas behind the construction of an exponential propagation scheme and refer the interested reader to references (Hochbruck & Lubich 1997 ; Hochbruck et al. 1998 ; Gallopoulos & Saad 1992) for a more exhaustive treatment of the subject.”
BackgroundMethodKuprov, I. (2008) · Journal of Magnetic Resonance · doi:10.1016/j.jmr.2008.08.008 · Cited by 78
“The initial density matrix q0 (shown as a vector with blue dots denoting non-zeros) is propagated forward in time for k Larmor steps using Krylov technique [26,31] to avoid the computation of Liouvillian exponential.”
MethodJung, G., Hanke, M., & Schmid, F. (2018) · Soft Matter · doi:10.1039/c8sm01817k · Cited by 66
Muñoz‐Jaramillo, A., Nandy, D., & Martens, P. C. H. (2010) · The Astrophysical Journal Letters · doi:10.1088/2041-8205/727/1/l23 · Cited by 65
“However, although this issue has been common knowledge for more than a decade, it’s only because of current improvements in computational techniques (Hochbruck & Lubich 1997; Hochbruck, Lubich & Selhofer; Muñoz-Jaramillo, Nandy & Martens 2009; MNM09 from here on), that this question can be finally…”
BackgroundAn efficient algorithm for time propagation as applied to linearized augmented plane wave method
Dewhurst, J. K., Krieger, K., Sharma, S., & Gross, E. K. U. (2016) · Computer Physics Communications · doi:10.1016/j.cpc.2016.09.001 · Cited by 72
Inexact Uniformization Method for Computing Transient Distributions of Markov Chains
Sidje, R. B., Burrage, K., & MacNamara, S. (2007) · SIAM Journal on Scientific Computing · doi:10.1137/060662629 · Cited by 68
“• Only when t = 100 does the Krylov method become the fastest due to its inherent higher convergence rate [8].”
BackgroundFrommer, A., & Simoncini, V. (2008) · Mathematics in industry · doi:10.1007/978-3-540-78841-6_13 · Cited by 81
“Similar results also hold in the case when A is skew-symmetric, or when A is non-symmetric, under certain hypotheses on the location of its spectrum, see [18], [34].”
Efficient computation of the exponential operator for large, sparse, symmetric matrices
Bergamaschi, L., & Vianello, M. (2000) · Numerical Linear Algebra with Applications · doi:10.1002/(sici)1099-1506(200001/02)7:1<27::aid-nla185>3.0.co;2-4 · Cited by 73
Matrix functions in network analysis
Michele Benzi, A., & Boito, P. (2020) · GAMM-Mitteilungen · doi:10.1002/gamm.202000012 · Cited by 75
“See also [99] and [148] for more results on approximation of eAv.”
ResultA new class of exponential propagation iterative methods of Runge–Kutta type (EPIRK)
Tokman, M. (2011) · Journal of Computational Physics · doi:10.1016/j.jcp.2011.08.023 · Cited by 75
“In [29] Hochbruck and Lubich derived a bound for the error of the Krylov approximation of ψ(A)b.”
Background
Metadata
- Authors
- Hochbruck, M., & Lubich, C.
- Affiliations
- University of Tübingen (DE)
- Venue
- SIAM Journal on Numerical Analysis
- Published
- 1997-10
- Volume
- 34
- Issue
- 5
- Pages
- 1911-1925
- ISSN
- 0036-1429 (ISSN-L)
- 1095-7170
- OpenAlex
- W1984347942
- Cited by
- 795
- References
- 30
- Citations by year
- 2026: 27 · 2025: 23 · 2024: 19 · 2023: 23 · 2022: 23 · 2021: 37 · 2020: 42 · 2019: 41 · 2018: 46 · 2017: 41 · 2016: 42 · 2015: 32 · 2014: 44 · 2013: 44 · 2012: 30
- FWCI
- 14.18
- Type
- journal article
- Language
- en
- Open access
- green
- doi.org
- Licence
- other-oa
- Locations
- SIAM Journal on Numerical Analysis (publishedVersion)
- Repository KITopen (Karlsruhe Institute of Technology) (publishedVersion)
- KITopen (other-oa)
- Topics
- Physical Sciences › Computer Science › Computational Theory and Mathematics › Matrix Theory and Algorithms
- Physical Sciences › Mathematics › Numerical Analysis › Numerical Solvers for Ordinary Differential Equations
- Physical Sciences › Engineering › Computational Mechanics › Advanced Numerical Methods in Computational Mathematics
- Keywords
- matrix exponential, Krylov subspace methods, Arnoldi method, Lanczos method, functional calculus, error bounds, biconjugate gradient method, numerical range, pseudospectra, superlinear convergence, stiff ordinary differential equations, time integration methods, nonlinear differential equations, Jacobian matrix
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.