Novel Mathematics Inspired by Industrial

Challenges Mathematics in Industry 38 Michael

Günther (Editor) install download

https://ebookmeta.com/product/novel-mathematics-inspired-by-
industrial-challenges-mathematics-in-industry-38-michael-gunther-
editor/

Download more ebook from https://ebookmeta.com

We believe these products will be a great fit for you. Click
the link to download now, or visit ebookmeta.com
to discover even more!

Advanced Engineering Mathematics with MATLAB®,(Advances

in Applied Mathematics), 5th Edition Dean G. Duffy

https://ebookmeta.com/product/advanced-engineering-mathematics-
with-matlabadvances-in-applied-mathematics-5th-edition-dean-g-
duffy/

Specialist Mathematics 20th Edition Michael Evans

https://ebookmeta.com/product/specialist-mathematics-20th-
edition-michael-evans/

Engineering Mathematics by Example Robert Sobot

https://ebookmeta.com/product/engineering-mathematics-by-example-
robert-sobot/

Love That Sass 1st Edition C D Gorri

https://ebookmeta.com/product/love-that-sass-1st-edition-c-d-
gorri/
Developmental Mathematics: Prealgebra, Beginning
Algebra, & Intermediate Algebra 2nd Edition Julie
Miller

https://ebookmeta.com/product/developmental-mathematics-
prealgebra-beginning-algebra-intermediate-algebra-2nd-edition-
julie-miller-2/

Imagined Audiences: How Journalists Perceive and Pursue

the Public 1st Edition Jacob L. Nelson

https://ebookmeta.com/product/imagined-audiences-how-journalists-
perceive-and-pursue-the-public-1st-edition-jacob-l-nelson/

Christmas in Eden 1st Edition Margot Scott Natalie

Knight

https://ebookmeta.com/product/christmas-in-eden-1st-edition-
margot-scott-natalie-knight-2/

C# Concurrency (MEAP V03) Nir Dobovizki

https://ebookmeta.com/product/c-concurrency-meap-v03-nir-
dobovizki/

The Fabric of Religious Life in Medieval Ashkenaz 1000

1300 Creating Sacred Communities 1st Edition Jeffrey R
Woolf

https://ebookmeta.com/product/the-fabric-of-religious-life-in-
medieval-ashkenaz-1000-1300-creating-sacred-communities-1st-
edition-jeffrey-r-woolf/
Up and to the Right: The Story of John W. Dobson and
Formula Growth Second Edition Craig Toomey

https://ebookmeta.com/product/up-and-to-the-right-the-story-of-
john-w-dobson-and-formula-growth-second-edition-craig-toomey/
Mathematics in Industry 38
The European Consortium for Mathematics in Industry

Michael Günther
Wil Schilders Editors

Novel
Mathematics
Inspired
by Industrial
Challenges
Mathematics in Industry

The European Consortium for Mathematics in Industry

Volume 38

Managing Editor
Michael Günther, University of Wuppertal, Wuppertal, Germany

Series Editors
Luis L. Bonilla, University Carlos III Madrid, Escuela, Leganes, Spain
Otmar Scherzer, University of Vienna, Vienna, Austria
Wil Schilders, Eindhoven University of Technology, Eindhoven, The Netherlands
The ECMI subseries of the Mathematics in Industry series is a project of The
European Consortium for Mathematics in Industry. Mathematics in Industry
focuses on the research and educational aspects of mathematics used in industry and
other business enterprises. Books for Mathematics in Industry are in the following
categories: research monographs, problem-oriented multi-author collections,
textbooks with a problem-oriented approach, conference proceedings. Relevance to
the actual practical use of mathematics in industry is the distinguishing feature of
the books in the Mathematics in Industry series.

More information about this subseries at https://link.springer.com/bookseries/4651

Michael Günther • Wil Schilders
Editors

Novel Mathematics Inspired

by Industrial Challenges
Editors
Michael Günther Wil Schilders
Applied Math. & Numerical Analysis Mathematics
Bergische Universität Wuppertal TU Eindhoven
Wuppertal, Nordrhein-Westfalen Eindhoven, Noord-Brabant
Germany The Netherlands

ISSN 1612-3956 ISSN 2198-3283 (electronic)

Mathematics in Industry
The European Consortium for Mathematics in Industry
ISBN 978-3-030-96172-5 ISBN 978-3-030-96173-2 (eBook)
https://doi.org/10.1007/978-3-030-96173-2

Mathematics Subject Classification (2020): 00A69, 68U99, 68T09

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland
AG 2022
This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher,
whether the whole or part of the material is concerned, specifically the rights of translation, reprinting,
reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical
way, and transmission or information storage and retrieval, electronic adaptation, computer software,
or by similar or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the
relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, expressed or implied, with respect to the material contained
herein or for any errors or omissions that may have been made. The publisher remains neutral with
regard to jurisdictional claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
To our loved ones
Preface

Mathematics is essential for innovations in industry and science. In many countries,

books with success stories of mathematics have been published 1, reports by inde-
pendent accountants have shown the high economic value of mathematics234, and
the Study Groups of Mathematics with Industry are spread all over the world 5. Con-
ferences like the biennial ECMI conference, the biennial SIAM CSE conference and
the ICIAM conference organized every 4 years contain a large variety of applications
of mathematics to challenges from industry and other sciences.
Despite all of these success stories, it is still claimed that these successes build on
existing mathematics, that no new mathematics is generated and that mathematics for
industry is not challenging at all. Some even feel that new mathematics is only created
by brilliant theoretical mathematicians. A prominent and well-known example is
formed by the number-theoretic results obtained by the famous mathematician Hardy
in the 1940s, that are now the basis of all encryption algorithms used for financial
transactions. Hardy himself would never have dreamed about this, in fact he would
have considered it a nightmare that his methods are used for something practical.
With this book, we wish to demonstrate that mathematics for industry is challeng-
ing and extremely rewarding, leading to new mathematical methods and sometimes
even to entirely new fields within mathematics. A nice example from our own ex-
perience is the solution of indefinite linear systems which originated from working
on electronic circuit simulation. Solving the large linear systems associated with
electronic circuits often led to problems with pivoting. Then the idea came up to
use the fact that there are two variables in the problem: currents and voltages. This
inspired us to re-order the matrix in terms of 2x2 blocks, coupling voltages to cur-

1 https://www.eu-maths-in.eu/wp-content/uploads/2017/01/2011_FLMI-EU_IndustrialMaths-
SuccessStories.pdf
2 https://www.platformwiskunde.nl/wp-content/uploads/2016/10/Deloitte-rapport-20140115-Ma-
thematical-sciences.pdf
3 https://www.eu-maths-in.eu/wp-content/uploads/2016/02/2015-France-SocioEconomic_Im-
pact_of_Mathematics.pdf
4 http://www.eu-maths-in.eu/wp-content/uploads/2019/04/MathematicsImpactStudy_Spain.pdf
5 https://ecmiindmath.org/study-groups/

vii
viii Preface

rents. The success was immediate, no pivoting was required anymore. The method
was generalised, so that it is applicable to all kinds of indefinite systems, and not
only those coming from electronic circuits. It led to very nice new research results,
and new methods for indefinite linear systems 67. Another example is the develop-
ment of methods within the field of model order reduction. This field has benefited
much from demands of and developments in the electronics industry. Methods like
PRIMA 8 and SPRIM 9 originated here, as did many other developments, but all
methods are generally applicable. In recent years we observe a much wider variety
of applications of model order reduction.
This book presents methods that fall into the category sketched in the foregoing
paragraph. The starting point is always an industrial challenge. The chapters describe
how the authors addressed the challenge and developed new methods that were
initially specific for the application, but later formulated for general application. The
book is split into two parts, one on engineering applications and one on stochastics
and finance.
All chapters contained in this book clearly show that industrial challenges do lead
to the development of new mathematical methods, or even completely new fields of
mathematics, needed to address these challenges. The starting point maybe an appli-
cation of existing mathematical methods, but when it is found that more is needed, or
different methods, then the interaction between application and mathematics starts.
Mathematicians can then on the one hand develop new mathematical techniques, on
the other hand solve the challenges. This is extremely rewarding, it often leads to
nice journal papers on the theoretical results, which subsequently are the starting
point of a lot of further research inside the mathematics area. It also leads to papers
in applied journals.
Concluding, we may say that “mathematics for industry” or, even broader, “applied
mathematics”, is much more than just applying existing mathematical methods to
industrial problems. In many cases, the application of existing methods does not lead
to the desired solution, and hence adaptations of existing methods or even entirely
new methods need to be developed in order to effectively address the industrial
challenge. In some cases, this has led to entirely new fields within mathematics. The
interplay between mathematics and industry is, hence, beneficial for both. Industry
benefits by having their problems solved and mathematics benefits because new
methods are developed that are versatile in nature.

6 H.S. Dollar, N.I.M. Gould, W.H.A. Schilders, A.J. Wathen: On iterative methods and implicit-
factorization preconditioners for regularized saddle-point systems, SIAM J. Matr. Anal. Appl. (27)
170–189 (2006)
7 W.H.A. Schilders: Solution of indefinite linear systems using an LQ decomposition for the linear
constraints, Linear Algebra and its Applications 431:30-4 381–395 (2009)
8 A. Odabasioglu, M. Celik and L.T. Pileggi: PRIMA: passive reduced-order interconnect macro-
modeling algorithm, IEEE Trans. Comp. Aid. Dsg. Int. Circ. Syst. 17:8 645–654 (1998)
9 R.W. Freund: Structure-Preserving Model Order Reduction of RCL Circuit Equations, in: W.H.A.
Schilders, H.A. van der Vorst and J. Rommes (eds): Model Order Reduction: Theory, Research
Aspects and Applications, Springer Verlag, Heidelberg, 51–75 (2008)
Preface ix

The book contains two parts on applications in Computational Science and Engi-
neering and Data Analysis and Finance. It should be remarked that all authors have
been asked to start and end their chapter with a brief description of why their chapter
fits into this volume: explaining which industrial challenges have been instrumental
for their inspiration, and which methods have been developed as a result.

Wuppertal and Eindhoven, Michael Günther

Spring 2021 Wil Schilders
Acknowledgements

We are grateful to Dr. Jörg Mittelsten Scheid for his generous grant to the Bergische
Universität Wuppertal, which had made possible the Mittelsten Scheid Visiting
Professorship of Prof.dr. W.H.A. Schilders in the winter semester 2020/2021 at the
Faculty of Mathematics and Natural Sciences. In the course of this semester, the
work on this book could be advanced a decisive piece, which would not have been
possible so easily without Prof. Schilders’ stay in Wuppertal.
We are grateful to Dr. Harshit Bansal for his valuable help in a number of LaTex
and layout problems.

xi
Contents

Part I Computational Science and Engineering

Multirate Schemes — An Answer of Numerical Analysis to a Demand

from Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Andreas Bartel and Michael Günther
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2 Strategies for multirate and convergence . . . . . . . . . . . . . . . . . . . . . . 7
2.1 Combining extra- and interpolation for multirate properly 7
2.2 Linear multistep methods . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.3 Runge-Kutta schemes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.4 Overview on multirate strategies . . . . . . . . . . . . . . . . . . . . . 13
3 Dynamic iteration and multiphysics . . . . . . . . . . . . . . . . . . . . . . . . . . 14
4 Applications in circuit simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
4.1 Partitioned network modeling . . . . . . . . . . . . . . . . . . . . . . . 17
4.2 Multirate schemes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
4.3 Thermal-electric coupling — silicon on insulator . . . . . . . 18
5 Molecular dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
6 Conclusion and outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

Electronic Circuit Simulation and the Development of New

Krylov-Subspace Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
Roland W. Freund
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
1.1 The Central Numerical Task in Circuit Simulation . . . . . . 30
1.2 Large-Scale Matrix Computations and Krylov-Subspace
Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
1.3 The Special Case of Circuit Interconnect Analysis . . . . . . 33
1.4 Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2 From AWE to the PVL Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

xiii
xiv Contents

2.1 Elmore Delay and AWE . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

2.2 PVL Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.3 An Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3 Krylov Subspaces with Multiple Starting Vectors . . . . . . . . . . . . . . . 41
3.1 Block Krylov Subspaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.2 Block Lanczos Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4 A New Approch: the Band Lanczos Method . . . . . . . . . . . . . . . . . . . 43
4.1 Defining Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.2 Reduced-Order Models and Matrix Padé Approximants . . 45
4.3 An Actual Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
5 Structure Preservation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
6 Band Arnoldi Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
7 Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

Modular Time Integration of Coupled Problems in System Dynamics . . . . 57

Martin Arnold
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
2 Model based simulation of pantograph-catenary interaction . . . . . . 58
Engineering application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
Mathematical aspects of modular simulation . . . . . . . . . . . . . . . . . . . . . . . . . 63
3 Modular time integration: The ODE case . . . . . . . . . . . . . . . . . . . . . 63
4 Modular time integration: The DAE case . . . . . . . . . . . . . . . . . . . . . . 66
Preconditioning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
Related work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

Differential-Algebraic Equations and Beyond: From Smooth to

Nonsmooth Constrained Dynamical Systems . . . . . . . . . . . . . . . . . . . . . . . . . 73
Jan Kleinert and Bernd Simeon
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
2 Differential-algebraic equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
2.1 How the topic of DAEs emerged . . . . . . . . . . . . . . . . . . . . . 74
2.2 Electrical circuits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
2.3 Constrained mechanical systems . . . . . . . . . . . . . . . . . . . . . 80
3 Major results and numerical methods . . . . . . . . . . . . . . . . . . . . . . . . . 84
3.1 Perturbation index and implicit Runge-Kutta methods . . . 85
3.2 DAEs and differential geometry . . . . . . . . . . . . . . . . . . . . . . 87
3.3 Singularly perturbed problems and regularization . . . . . . . 89
3.4 General fully implicit DAEs . . . . . . . . . . . . . . . . . . . . . . . . . 91
3.5 Constrained Hamiltonian systems . . . . . . . . . . . . . . . . . . . . 91
4 Beyond classical DAEs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
4.1 Navier-Stokes incompressible . . . . . . . . . . . . . . . . . . . . . . . 93
4.2 Stochastic DAEs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
Contents xv

5 Nonsmooth dynamical systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95

5.1 A short zoology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
5.2 Nonsmooth mechanical systems with impacts . . . . . . . . . . 105
5.3 Numerical solution strategies . . . . . . . . . . . . . . . . . . . . . . . . 110
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122

Fast Numerical Methods to Compute Periodic Solutions of

Electromagnetic Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
Alfredo Bermúdez, Dolores Gómez, Marta Piñeiro and Pilar Salgado
1 Starting Point in Electrical Engineering . . . . . . . . . . . . . . . . . . . . . . . 134
1.1 Methodology Motivation from a Toy Model . . . . . . . . . . . 135
2 Statement of the Problem. Mathematical Modelling . . . . . . . . . . . . . 137
3 Existing Mathematics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
4 A Novel and Efficient Methodology to Solve the Problem . . . . . . . . 143
4.1 Reduced Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
4.2 Approximating the Initial Currents in Rotor Bars . . . . . . . 145
4.3 Numerical Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
5 Current State of Art . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
Challenges in the Simulation of Radio Frequency Circuits . . . . . . . . . . . . . 155
Kai Bittner, Hans Georg Brachtendorf, and Roland Pulch
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
2 Network equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
3 Simulation of Radio Frequency Circuits . . . . . . . . . . . . . . . . . . . . . . 163
4 The Embedding Technique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174

An Integrated Data-Driven Computational Pipeline with Model Order

Reduction for Industrial and Applied Mathematics. . . . . . . . . . . . . . . . . . . . 179
Marco Tezzele, Nicola Demo, Andrea Mola, and Gianluigi Rozza
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
2 From digital twin to real-time analysis . . . . . . . . . . . . . . . . . . . . . . . . 181
3 Advanced geometrical parametrization with automatic CAD files
interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
4 Parameter space dimensionality reduction . . . . . . . . . . . . . . . . . . . . . 187
5 Data driven model order reduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
5.1 Dynamic mode decomposition . . . . . . . . . . . . . . . . . . . . . . . 189
5.2 Proper orthogonal decomposition with interpolation . . . . 191
6 Simulation-based design optimization framework . . . . . . . . . . . . . . 192
xvi Contents

7 Conclusions and perspectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194

Acknowledgment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
Competing Interests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
Ethics approval and consent to participate . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
Availability of data and materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
Funding acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196

From Rotating Fluid Masses and Ziegler’s Paradox to Pontryagin- and

Krein Spaces and Bifurcation Theory. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
Oleg N. Kirillov and Ferdinand Verhulst
1 Historical background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
1.1 Stability of Kelvin’s gyrostat and spinning artillery
shells filled with liquid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
1.2 Secular instability of the Maclaurin spheroids by
viscous and radiative losses . . . . . . . . . . . . . . . . . . . . . . . . . 207
1.3 Brouwer’s rotating vessel . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
2 Ziegler’s paradox . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
3 Bottema’s analysis of Ziegler’s paradox . . . . . . . . . . . . . . . . . . . . . . . 220
4 An umbrella without dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
5 Hopf bifurcation near 1:1 resonance and structural stability . . . . . . 227
6 Abscissa minimization, robust stability and heavy damping . . . . . . 231
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235

Part II Data Analysis and Finance

Topological Data Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247

Jean-Daniel Boissonnat and Frédéric Chazal and Bertrand Michel
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
2 The need of new mathematical and algorithmic tools . . . . . . . . . . . . 249
3 The emergence of geometric inference and persistent homology . . 250
3.1 Distance-based geometric inference . . . . . . . . . . . . . . . . . . 251
3.2 Persistent homology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
4 New research directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262
A brief glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 264
Prediction Models with Functional Data for Variables Related with
Energy Production . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271
Manuel Febrero–Bande, Wenceslao González–Manteiga and Manuel
Oviedo de la Fuente
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272
Contents xvii

2 Functional Data Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275

2.1 Application to Iberian Market Energy . . . . . . . . . . . . . . . . . 278
3 Variable selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283
3.1 The selection algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285
3.2 Numerical results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 288
4 Real data application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 290
4.1 Energy Market Demand . . . . . . . . . . . . . . . . . . . . . . . . . . . . 290
4.2 Energy Price . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292
5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 294

Quantization Methods for Stochastic Differential Equations . . . . . . . . . . . . 299

J. Kienitz, T.A. McWalter, R. Rudd, and E. Platen
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 300
1.1 Finance and Stochastic Differential Equations . . . . . . . . . . 300
1.2 Quantization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 302
1.3 Outline of the Paper . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303
2 Vector Quantization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 304
2.1 Optimal Quantization Grids . . . . . . . . . . . . . . . . . . . . . . . . . 305
2.2 Numerical Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 307
3 Recursive Marginal Quantization . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311
3.1 Numerical Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313
3.2 The Zero Boundary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 315
3.3 Higher-order Updates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 316
4 Recursive Marginal Quantization for Stochastic Volatility Models . 317
5 Numerical Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319
5.1 Numerical Convergence Results . . . . . . . . . . . . . . . . . . . . . 319
5.2 An Example of a Local Volatility Model . . . . . . . . . . . . . . 321
5.3 Stochastic Volatility Models . . . . . . . . . . . . . . . . . . . . . . . . . 322
5.4 Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 324
6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 326
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 327

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 331
List of Contributors

Martin Arnold
Martin Luther University Halle-Wittenberg, Institute of Mathematics, 06099 Halle
(Saale), Germany, e-mail: [email protected]
Andreas Bartel
Bergische Universität Wuppertal, School of Mathematics and Natural Sciences,
IMACM, Gaußstraße 20, D-42119 Wuppertal, e-mail: bartel@math.
uni-wuppertal.de
A. Bermúdez
Dpto. de Matemática Aplicada & Instituto de Matemáticas (IMAT) &
Instituto Tecnológico de Matemática Industrial (ITMATI), Universidade de
Santiago de Compostela, ES-15782 Santiago de Compostela, Spain, e-mail:
[email protected]
Kai Bittner
University of Applied Sciences of Upper Austria, Softwarepark 11, Hagenberg im
Mühlkreis, 4232, Austria, e-mail: [email protected]
Jean-Daniel Boissonnat
Université Côte d’Azur Inria, e-mail: [email protected]
Hans Georg Brachtendorf
University of Applied Sciences of Upper Austria, Softwarepark 11, Hagenberg im
Mühlkreis, 4232, Austria, e-mail: [email protected]
Frédéric Chazal
Inria Saclay, e-mail: [email protected]
Nicola Demo
Mathematics Area, mathLab, SISSA, International School of Advanced Studies, via
Bonomea 265, I-34136 Trieste, Italy, e-mail: [email protected]
Manuel Febrero-Bande

xix
xx List of Contributors

Department of Statistics, Mathematical Analysis and Optimization, Universidade

de Santiago de Compostela, e-mail: [email protected]
Roland W. Freund
Department of Mathematics, University of California at Davis, One Shields Avenue,
Davis, California 95616, USA, e-mail: [email protected]
D. Gómez
Dpto. de Matemática Aplicada & Instituto de Matemáticas (IMAT) &
Instituto Tecnológico de Matemática Industrial (ITMATI), Universidade de
Santiago de Compostela, ES-15782 Santiago de Compostela, Spain, e-mail:
[email protected]
Wenceslao González-Manteiga
Department of Statistics, Mathematical Analysis and Optimization, Universidade
de Santiago de Compostela, e-mail: [email protected]
Michael Günther
Bergische Universität Wuppertal, School of Mathematics and Natural Sciences,
IMACM, Gaußstraße 20, D-42119 Wuppertal, e-mail: guenther@math.
uni-wuppertal.de
J. Kienitz
Fachbereich Mathematik und Naturwissenschaften, Bergische Universität Wuppertal
/ The African Institute for Financial Markets and Risk Management (AIFMRM),
University of Cape Town / Quaternion Risk Management GmbH, e-mail:
[email protected]
Oleg N. Kirillov
Northumbria University, NE1 8ST Newcastle upon Tyne, UK, e-mail:
[email protected]
Jan Kleinert
Hochschule Bonn-Rhein-Sieg, Grantham-Allee 20, D-53757 Sankt Augustin,
German Aerospace Center, Simulation and Software Technology, Linder Höhe,
D-51147 Cologne, e-mail: [email protected],jan.kleinert@
dlr.de
T.A. McWalter
The African Institute for Financial Markets and Risk Management (AIFMRM),
University of Cape Town / Faculty of Science, Department of Statistics, University
of Johannesburg, e-mail:
Bertrand Michel
Ecole Centrale de Nantes, e-mail: [email protected]
Andrea Mola
Mathematics Area, mathLab, SISSA, International School of Advanced Studies, via
Bonomea 265, I-34136 Trieste, Italy, e-mail: [email protected]
Manuel Oviedo de la Fuente
Technological Institute for Industrial Mathematics and Department of Statistics,
List of Contributors xxi

Mathematical Analysis and Optimization, Universidade de Santiago de Compostela,

e-mail: [email protected]
M. Piñeiro
Dpto. de Matemática Aplicada, Universidade de Santiago de Compostela, ES-15782
Santiago de Compostela, Spain, e-mail: [email protected]
E. Platen
Finance Discipline Group and School of Mathematical and Physical Sciences,
University of Technology Sydney / The African Institute for Financial Markets and
Risk Management (AIFMRM), University of Cape Town, e-mail:
Roland Pulch
Institute of Mathematics and Computer Science, University of Greifswald,
Walther-Rathenau-Str. 47, Greifswald, 17489, Germany, e-mail: roland.pulch@
uni-greifswald.de
Gianluigi Rozza
Mathematics Area, mathLab, SISSA, International School of Advanced Studies, via
Bonomea 265, I-34136 Trieste, Italy, e-mail: [email protected]
R. Rudd
The African Institute for Financial Markets and Risk Management (AIFMRM),
University of Cape Town, e-mail:
P. Salgado
Dpto. de Matemática Aplicada & Instituto de Matemáticas (IMAT) &
Instituto Tecnológico de Matemática Industrial (ITMATI), Universidade de
Santiago de Compostela, ES-15782 Santiago de Compostela, Spain, e-mail:
[email protected]
Bernd Simeon
Felix-Klein-Zentrum, TU Kaiserslautern, D-67663 Kaiserslautern, Germany,
e-mail: [email protected]
Marco Tezzele
Mathematics Area, mathLab, SISSA, International School of Advanced Studies, via
Bonomea 265, I-34136 Trieste, Italy, e-mail: [email protected]
Ferdinand Verhulst
Mathematisch Instituut, PO Box 80.010, 3508TA Utrecht, Netherlands, e-mail:
[email protected]
 Part I
Computational Science and Engineering
2

Computational Science and Engineering has become a very flourishing area of

research. Amongst the conferences organised by SIAM, the CSE series attracts the
highest number of participants nowadays. Also the attendence of ECMI and ICIAM
conferences is growing rapidly in the past decade. Engineers are about the top users
of mathematical software, and hence of mathematical methods. In turn, they provide
many challenges and also many opportunities for mathematicians to adapt their
methods or construct even entirely new ones. Clearly, this is very advantageous for
the development of mathematical methods and theories.
Let us, for example, consider ordinary differential equations (ODE). In the last few
decades, the theory of numerical methods for general (non-stiff and stiff) ordinary
differential equations has reached a certain maturity, and excellent general-purpose
codes, mainly based on Runge–Kutta methods or linear multistep methods, have
become available. The motivation for developing structure preserving algorithms
for special classes of problems came independently from such different areas of
research as astronomy, molecular dynamics, mechanics, theoretical physics, and
numerical analysis as well as from other areas of both applied and pure mathematics.
It turned out that the preservation of geometric properties of the flow not only
produces an improved qualitative behaviour, but also allows for a more accurate
long-time integration than with general-purpose methods. Such methods belong to
the more general class of mimetic methods. An important shift of view-point came
about by ceasing to concentrate on the numerical approximation of a single solution
trajectory and instead to consider a numerical method as a discrete dynamical system
which approximates the flow of the differential equation – and so the geometry
of phase space comes back again through the window. This view allows a clear
understanding of the preservation of invariants and of methods on manifolds, of
symmetry and reversibility of methods, and of the symplecticity of methods and
various generalizations. It has brought an enormously rich additional theory into the
area of mathematical methods for ODEs.
The chapters in this first part on "Engineering applications" describe similar
developments, inspired by different branches of industry. By sheer coincidence,
most of the chapters in this first part of the book originate from the electronics
industry. On the one hand, this is owing to the connections of the two editors, who
have been very active in this area. On the other hand, the electronics industry has
been, and still is, extremely important for the progress of mankind in recent decades.
Hence, it is not surprising that many researchers have felt attracted to this branch of
industry.
The first chapter is by Bartel and Günther. They observed the need, in the elec-
tronics industry, for so-called multirate methods. It turns out that electronic circuits
often have different parts of the circuit being latent, and other parts being active. In
principle, using the same time steps for both parts would be a waste of computational
effort. The latent part could do with much larger time steps, until this part of the cir-
cuit becomes active again. This observation, as well as the desire to be able to carry
out such time stepping procedure, has led to multirate methods. A lot of research
has been carried out in this area, and the methods developed are clearly not only
applicable to electronic circuits. The chapter contains a very detailed and concise
 3

description of the developments, as well as applications other than those from the
electronics industry.
The chapter by Roland Freund on model order reduction does not need much
further introduction, we discussed MOR already in the previous section. Freund
has been at the basis of the enormous developments in the area of MOR, with the
seminal paper on PVL 10, written together with Peter Feldmann. It established a
connection of methods for model order reduction, then still in their infancy, and
numerical linear algebra. As argued already in the previous section, from then on
MOR benefited much from the electronics industry, as the continuous miniaturisation
of semiconductor devices led to many challenges. Even now, challenges from the
electronics industry are an inspiration for new developments in the MOR area.
Martin Arnold concentrates on stability problems frequently encountered when
solving systems of differential-algebraic equations. Well-known methods such as
multi-rate methods or waveform relaxation may suffer from so-called exponential
instabilities. Arnold developed an entirely new framework for the analysis of such
instability as well as for the convergence behaviour, which led to new methods that
do not suffer from the aforementioned problems. As a result, a totally new theory of
methods for coupled problems within system dynamics has been developed.
The next chapter in this part is by a team of Spanish researchers around Alfredo
Bermudez, pioneer in the area of mathematical methods for industrial problems, that
concentrates on electromagnetic models. The problem at hand is to calculate the
steady-state solution of an electromagnetic problem. This can often take many days,
due to the fact that the transient are dying out only very slowly. Engineers had the
desire to be able to reach the steady state much faster, and to this end new numerical
methods were developed that can handle this task. The chapter sketches the historical
development, as well as the final result of the search for such novel methods. In many
areas nowadays this kind of challenge is encountered, and the ideas underlying the
methods in this chapter can immediately be used to achieve the desired speed-up of
the simulations. A related method can be found in [10].
In the chapter by Bittner, Brachtendorf and Pulch, the emphasis is on so-called
MPDEs, in full multirate partial differential equations. These MPDE are of hyper-
bolic type, which have a physically meaningful solution only along one characteristic
curve. The problem is analysed in-depth, and it is shown that the theory is applica-
ble also to other fields where such MPDE may occur. The chapter contains several
examples that illustrate the challenge very nicely.
The lust but one final chapter in this part on engineering applications is by
a group of Italian researchers around Gianluigi Rozza, well-known for the many
contributions in the area of model order reduction and reduced basis methods.
The chapter discusses an integrated data-driven computational pipeline, containing
several model order reduction techniques, for addressing industrial challenges. It
contains many different ingredients, amongst which dimension reduction of the
parameter space. The framework developed has been put in a general context, and
hence the methodologies are usable in many different application fields.

10 https://pure.tue.nl/ws/files/97050457/publicsummary_giovannideluca.pdf
 The final chapter by Oleg N. Kirillov and Ferdinand Verhulst is an example
how the the interplay between applications and progress in mathematics may take
place. Starting point is a curious phenomenon observed in the natural sciences,
destabilization by dissipation — usually it is the other way round. The solution of
this problem, which involved the eigenvalue calculus of matrices, finally was one
source for the mathematical theory of structural stability of matrices. An important
consequence of these results is that in a large number of problem-fields one can now
predict and characterise precisely this type of instability.
Multirate Schemes
—
An Answer of Numerical Analysis to a Demand
from Applications

Andreas Bartel and Michael Günther

Abstract In science and engineering, simulation tasks often involve numerical time
integration of differential equations. Usually, these systems contain different time
constants of the involved components and/or right-hand side. This multirate behavior
may be caused by coupling subsystems in multiphysics problems acting on different
time scales. Such a behavior does already occur if one deals with just single-physics
problems: for example, the activity level of components in electrical networks may
strongly vary depending on the according functional purpose, physics or time; an-
other example is given in lattice QCD, where the equations of motion may depend
on weak and strong forces, which demand to sample these forces with different fre-
quencies to gain the same rate of approximation.
To be efficient or to speed up simulation of highly complex coupled systems is neces-
sary for many design and optimization work flows. To this end, numerical integration
schemes have to be adapted to exploit this multirate behavior. One idea proposed
by Rice in 1960 are multirate schemes, which use different step sizes adapted to
the various activity levels. In the last 50 years, the methodology of numerical time
integration schemes has been advanced in a constant interplay between the demands
defined by the need of exploiting multirate behavior in different fields of applications
and the development of tailored multirate schemes to answer these demands.

1 Introduction

In many technical applications, ranging from electric circuits to multibody systems

and in particular for multiphysics simulation, the governing set of differential equa-
tions is characterized by a multirate behavior in time domain: that is to say, some
parts of the right-hand side follow a fast dynamics, whereas the other parts are char-

Andreas Bartel · Michael Günther

Bergische Universität Wuppertal, School of Mathematics and Natural Sciences, IMACM, Gauß-
straße 20, D-42119 Wuppertal, e-mail: {bartel,guenther}@math.uni-wuppertal.de

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 5

M. Günther, W. Schilders (eds.), Novel Mathematics Inspired by Industrial Challenges,
Mathematics in Industry 38, https://doi.org/10.1007/978-3-030-96173-2_1
6 Andreas Bartel and Michael Günther

acterized by a comparatively slow dynamics. Often, the slower parts represents even
a much larger part/partition of the overall system in question. A traditional, standard
time integration scheme has to resolve the fastest dynamics (according to the given
tolerance). Thus the time stepping is adapted to these fast components, and yields
rather small time steps for the slower components, which results an an oversampling
revealing only superfluous detail. Now, to be efficient, a numerical (time) integration
scheme needs to exploit this multirate potential.
For the moment, let us consider the case of an initial-value problem of ordinary
differential equations ẇ = h(a, b)

ẇ = h(t, w), w(t0 ) = w0 (1)

with h : R × Rn → Rn assumed to be Lipschitz continuous in w. We denote the

unique solution of (1) at time point t by w(t; w0 ). Furthermore, we assume that in
h we have some multirate potential, i.e., some coordinates or summands of h are
slower than some remaining ones.
On the one hand, the interpretation of fast and slow coordinates in h leads to a
partitioning of the unknowns w> = (y> S , yF ) in slow yS (t) ∈ R and fast components
> m

yS (t) ∈ Rn−m . In this way, the ODE (1) is transferred to a partitioned system:

ẏS = fS (t, yS , yF ), yS (t0 ) = yS, 0 ,

(2)
ẏF = fF (t, yS , yF ), yF (t0 ) = yF, 0 ,

with corresponding slow right-hand side fS : R × Rm × Rn−m → Rm and fast right-

hand side fF : R × Rm × Rn−m → Rn−m . This is termed component-wise partition-
ing. On the other hand, one can split the right-hand side:

ẇ = hs (t, w) + h f (t, w), w(t0 ) = y0 (3)

into slow and fast terms. This induces an additive splitting of the unknown w =
ws + w f into slow and fast varying parts ws and w f . This is referred to as right-hand
side partitioning. Note that both formulations are equivalent in the sense that each
component-wise partitioned system can be rewritten as a right-hand side partitioned
system and vice versa.
Multiorder as multirate. There are many ways to exploit multirate behavior in
such systems. One idea is to use multi-order methods. Here a single method with
a single step size is employed for the whole system. To adapt to the activity level,
the order of the method is modified accordingly. This class comprises, for example,
the schemes MURX [17] and MUR8 [16] by Engstler and Lubich. The first method
(MURX) is based on Richardson extrapolation of the explicit Euler scheme. Thereby
the computation of the extrapolation tableau is stopped if a component is accurate
enough. The latter MUR8 uses low-order methods embedded in a high-order method,
where the update of the slow components is deactivated after a first few function
evaluations.
Multirate Schemes 7

Multiple step sizes as multirate. To our knowledge, the first method based on
exploiting multirate behavior by adapting step sizes to the activity level of compo-
nents was derived by Rice [33] in 1960 for missile simulations. This method is based
on Runge-Kutta schemes and employs a so-called compound step. Later, Gear and
Wells [21] proposed an alternative approach based on linear multistep methods as
well as extrapolation and interpolation.
The work at hand focuses on the multiple step size approaches. In Section 2, we
discuss the ideas of Rice and Gear/Wells. The class of extra- and interpolation cou-
pling is strongly linked to waveform relaxation/dynamic iteration schemes [3, 12, 29],
which we treat in Section 3. In fact, dynamic iteration enables the general applica-
tion to multiphysics systems. Subsequently, we treat some applications. Section 4
covers electric circuits simulation, where the compound-step approach has been suc-
cessfully applied to develop multirate strategies for single-physics problems. Then
molecular dynamics is discussed in Section 5, in which the multirate potential has
to be exploited, without violating the preservation of geometric properties. For this
case, operator splitting has turned out to be the right framework for developing mul-
tirate schemes, as properties of the base scheme are easily preserved. Finally, we
draw a conclusion and give an outlook on open problems.

2 Strategies for multirate and convergence

Before discussing in detail the methods by Rice [33] as well as by Gear and Wells
[21], we first focus on the common idea behind these approaches: the combination
of basic numerical integration schemes with extra- and interpolation techniques to
solve (2) with a small step size h for the fast variable z and large step size H = M · h
for the slow y. Here, the multirate factor is a fixed number M ∈ N.

2.1 Combining extra- and interpolation for multirate properly

Depending on the sequence of the computation of the unknowns yS and yF , one

distinguishes the following three versions of extra-/and interpolation techniques for
a macro step from t to t + H [11]:
1) fully-decoupled approach: fast and slow variables are integrated in parallel using
in both cases extrapolated waveforms; computations are based on information
from the initial data of the current macro step which is given at t;
2) slowest-first approach [21]: in the first step, the slow variables are integrated,
using an extrapolated waveform of yF based on the information available at
t; this is used to evaluate the coupling variable yF in the slow system on the
current macro step [t, t + H]. In the second step, M micro steps are performed to
integrate the fast variables yF from t to t + H, using an interpolated waveform
8 Andreas Bartel and Michael Günther

of the coupling variable yS . Thereby the interpolation is based on information

from the current macro step [t,t + H].
3) fastest-first approach [21]: now, in the first step, M micro steps are performed to
integrate the fast variables on the current macro step [t, t + H]. To this end, the
coupling variable yS are extrapolated based on the information available at t. In
the second step, one macro step is performed to integrate the slow variables yS
from t to t + H. Thereby, the interpolated waveform of yF is computed based on
the information from the current macro step [t,t + H] and this is used to evaluate
the coupling variable yF .
We have the following main result for the decoupled case:

Theorem 2.1 We consider (2) with both right-hand sides fS and fF Lipschitz con-
tinuous in both variables yS and yF . Furthermore, we consider an arbitrary macro
step from t → t + Mh and respective initial values yS (t) = yS,t , yF (t) = yF,t . Let
M ∈ N be a fixed multirate factor be given and two integration schemes of order p
be applied: a first scheme for one macro step of size H = M · h for the slow yS and a
second scheme is applied for M steps of size h for fast yF . If we use an integration
scheme of order p and combine this with an extrapolation procedure of order p − 1,
the overall scheme has order p.

Proof For the coupled system (2) with initial data yS (t) = yS,t , yF (t) = yF,t , we refer
to the unique solution by (yS (t; yS,t , yF,t )>, yF (t; yS,t , yF,t )> ). Now, we replace the
coupled system (2) by a modified system which decouples both parts:

ẏS = fS (t, yS , yeF ) =: feS (t, yS ), yS (t) = yS,t ,

(4)
ẏF = fF (t, yeS , yF ) =: feF (t, yF ), yF (t) = yF,t ,

where yeS and yeF are extrapolations of order p − 1, i.e.,

yS (t)− yeS (t) = O(H p ) and yF (t)− yeF (t) = O(H p ) for any t ∈ [t, t +Mh]. (5)

We denote the unique solution of (4) by (b yS (t; yS,t , yF,t )>, ybF (t; yS,t , yF,t )> ).
Next, we solve the decoupled system (4) with two numerical integration schemes
of order p, with M step sizes h applied to yF and one step size H = M · h applied to yS .
The numerical solution obtained at t ∗ = t + Mh is denoted by (yS, H (t ∗ ), yF, H (t ∗ ))> .
For the difference between the numerical multirate approximation and the exact
solution at t ∗ , the triangle inequality yields

kyS, H (t ∗ ) − yS (t ∗ )k kyS, H (t ∗ ) − ybS (t ∗ )k yS (t ∗ ) − yS (t ∗ )k

     
kb
≤ + . (6)
kyF, H (t ∗ ) − yF (t ∗ )k kyF, H (t ∗ ) − ybF (t ∗ )k kbyF (t ∗ ) − yF (t ∗ )k

The first term on the right-hand side represents the error of the applied integration
schemes. Employing for both coordinates an integration scheme of order p, with one
macro step of size H = M · h for the slow yS and M steps of size h for fast yF , we
have
Multirate Schemes 9

kyS, H (t ∗ ) − ybS (t ∗ )k
   
cS
≤ cF H p+1 (7)
kyF, H (t ∗ ) − ybF (t ∗ )k M p+1

with respective leading error coefficients cS and cF .

For the second term on the right-hand side (6), we get from the Lipschitz continuity
of fS , fF with corresponding constants (Li, j for fi w.r.t. y j )
 Z t ∗
yS (t ∗ ) − yS (t ∗ )k



kb k fS τ, ybS (τ), yeF (τ)
− fS τ, yS (τ), yF (τ)
k
≤ dτ
kbyF (t ∗ ) − yF (t ∗ )k t k fF τ, yeS (τ), ybF (τ) − fF τ, yS (τ), yF (τ) k
Z t ∗ 
LS,S kb
yS (τ) − yS (τ)k + LS,F keyF (τ) − yF (τ)k
≤ dτ
t LF,S keyS (τ) − yS (τ)k + LF,F kbyF (τ) − yF (τ)k

actually a decoupled estimate. Using that yeS and yeF are approximations of order p − 1
(5) and respective Lipschitz constants LS , LF of the corresponding extrapolation
operators, we find
∗
 
 ∗ ∗
 LS,F ·LF p+1 Rt
kb
yS (t ) − yS (t )k + LS,S kb yS (τ) − yS (τ)kdτ 
 H
t
≤
   
 t∗ 
LF,S · LS · H p+1 + LF,F kb
R
yF (t ∗ ) − yF (t ∗ )k yF (τ) − yF (τ)kdτ
 
kb
t

Applying Gronwall’s lemma, it follows:

∗
yS (t ∗ ) − yS (t ∗ )k LS,F LF eLS,S (t −t) H p+1
   
kb
≤ ∗ .
yF (t ∗ ) − yF (t ∗ )k
kb LF,S LS eLF,F (t −t) H p+1

Finally combining this with the integration error (7) into the split error (6), we obtain
that the multirate scheme has consistency order p for the compound step from t to
t + Mh (on the macro step level). 

Remark 2.1 a) One can show that the order p of the underlying numerical integration
scheme is preserved for the slowest- and fastest-first approach, too, provided that the
order of the used extra-/interpolation schemes is of order p − 1 at least [11]. These
methods are referred to as extrapolation/interpolation-based multirate scheme.
b) Notice for a working multirate scheme, we still have to define the ex-
trapolation/interpolation routines. Furthermore, arbitrary high orders of the extra-
/interpolation are not possible in the one-step-method context.

Corollary 2.1 Under the assumptions of Thm. 2.1, we have an overall multirate
scheme of convergence order p if
a) we use one-step integration schemes.
b) we use multistep schemes, where both schemes are 0-stable.

Summing up, the art of defining multirate (multistep) schemes lies in implicitly
defining the extrapolation and interpolation procedures used within the scheme to
be of high enough order.
10 Andreas Bartel and Michael Günther

Corollary 2.2 We consider the following initial values problem (IVP)

ẏS = fS (t, yS , yF , zS , zF ), yS (t0 ) = yS,0 , ẏF = fF (t, yS , yF , zS , zF ), yF (t0 ) = yF,0 ,

0 = gS (t, yS , yF , zS , zF ) 0 = gF (t, yS , yF , zS , zF )

of coupled semi-explicit DAEs with slow subsystem (yS , zS ) and fast subsystem
(yF , zF ). Futhermore, we assume that the overall system is index-1 and both subsys-
tems are index-1 with Lipschitz continuous fλ , gλ (λ ∈ {S, F}) with uniform Lipschitz
constants on any macro step [t,t + Mh] in [t0 , T ]. Applying a multirate method, which
is for coupled ODEs of order p (Thm. 2.1) and the algebraic variables zS , zF are
always consistently computed (i.e., implicit), then the method has still order p.
Proof The proof is a direct consequence of the recursion estimate Lemma 3.1 in [2]
for dynamic iteration schemes of coupled index-1 DAE systems, if only one iteration
is considered and the initial iteration error is given by the extrapolation/interpolation
error. 

2.2 Linear multistep methods

How to define the extrapolated and interpolated approximations? Linear multistep

schemes are based on polynomial interpolation using information of possibly several
previous steps. Here, these methods are advantageous over one-step schemes. This
can be seen as follows: let us assume, we use a linear K-step scheme for the slow
part, and a k-step scheme for the fast variables. Accordingly, the extrapolation can be
based on K macro step approximations for the slow variable and k micro step values
for the fast variables. This yields a global error of order min{K, k} (Thm. 2.1) if the
respective schemes are at least of order K and k.
Based on preliminary work by Gear [20], Orailoglu [31] and Wells [49], the
first comprehensive study and still fundamental work on multirate linear multistep
methods was published in 1984 by Gear and Wells [21]. It covers efficiency consid-
erations, convergence and error analysis, absolute stability and numerical test results
for an Adams-type based algorithm. In addition, the concepts of slowest-first and
fastest-first methods were first introduced and discussed. At a first glance, the fastest
first strategy seems to be advantageous. The extrapolation error in the computation
of the fast part is acceptable, as one is extrapolating over an interval of size H,
which is tailored to the activity level of the slow part. In the slowest first strategy,
we are extrapolating the fast variables over many micro step h to compute the slow
variables, which may only be tolerable if the coupling of the fast into the slow part,
measured by k∂ fS /∂ yF k, is small. As a rule of thumb, this quantity will be usually
small, as otherwise a high level of activity would be transferred from the fast into
the slow part, and the slow part would not be slow anymore. Certainly, this is only a
rule of thumb, since there exist — although academic — counterexamples [21].
However, considering an adaptive step sizes selection based on error control, the
slowest first strategy can become advantageous. For instance, if a secondly performed
Multirate Schemes 11

macro step [t, t + H] (slow variables) fails, the macro step size H has to be decreased
and the computation has to be repeated. If the approximations of the micro steps (fast
variable) are not stored (for instance, due to memory reasons) or if the computation
of the fast variables is based on extrapolated values on the failed macro step size
Hold , these values have to be recomputed. In contrast, the slowest first strategy does
not introduce any problems: if an integration of the fast part fails, one only has to
repeat it with a smaller step size. The necessary information to interpolate the slow
part does not change. See [20, 21] for further details.

Remark 2.2 Some systems do not allow for a static partitioning into fast and slow
subsystems (e.g. the inverter chain benchmark [23]). In these settings, slow compo-
nents can become fast (wake up) and vice versa. Thus, the partitions in fast and slow
need to be adapted over time. This causes a rather large step size modification, which
is more problematic for multistep methods, see e.g. [26], than for one-step schemes.

Remark 2.3 Other linear-multistep approaches have been discussed, for instance, by
Verhoeven et al. [48] and [32] (both BDF-based), or by Sandu and Constantinescu
(Adams-based) [34]. Stabilty issues have been addressed first by Skelboe [40] and
by Skelboe and Andersen [41].

2.3 Runge-Kutta schemes

In one-step schemes, only information of the last step is available. This limits the
approximation order to 1, as only constant or linear interpolation is available in the
multirate scheme with extra- or interpolation (cf. Thm. 2.1). Of course, the use of
approximations at previous steps (in the extra- and interpolation) would turn the
one-step scheme into a multistep scheme.
About twenty years before Gear’s paper on multirate multistep schemes, Rice
proposed a solution to overcome this problem for one-step schemes [33]: split Runge-
Kutta schemes. Firstly, these schemes perform one large Runge-Kutta compound
step with macro step size H for the joint system, but use only the result for the slow
variables as approximates. Secondly, to get the approximations for the fast part in
all micro steps of the macro step, the fast part is then integrated using interpolated
information of the slow variables based on the Runge-Kutta increments of the macro
step. In 2008, this approach has been applied by Verhoeven to BDF schemes [48],
see also Section 4.
We give some details for the split Runge-Kutta schemes [33] on system (2). The
numerical approximation yS, H (t ? ) is given by one explicit Runge-Kutta step from t
to t ? = t + H:
12 Andreas Bartel and Michael Günther
s
yS, H (t ? ) = yS,t + ∑ bi ki ,
i=1
 i−1 i−1 i−1 
ki = H fS t + ∑ ai, j H, yS,t + ∑ ai, j k j , yF,t + ∑ ai, j l j , (i = 1, . . . , s),
j=1 j=1 j=1
 i−1 i−1 i−1 
li = H fF t + ∑ ai, j H, yS,t + ∑ ai, j k j , yF,t + ∑ ai, j l j , (i = 1, . . . , s),
j=1 j=1 j=1

using internal stages li for the active part, which will not be used later on. This is
referred to as compound step and it employs the coefficients bi , ai, j with s stages.
Secondly, for the M micro-steps, another Runge-Kutta scheme with coefficients
bi , aei, j and sestages is used to compute yF, H (t i + lh) for yF at micro grid points t + λ h
e
for λ = 1, . . . , M:

se
yF, H t + (λ + 1)h = yF, H (t + λ h) + ∑ ebi kiλ , (λ = 1, . . . , M − 1)
i=1
 i−1  i−1  i−1 
kiλ = h fF t + λ h + ∑ aei, j h, ybS t + λ h + ∑ aei j h , yF, H (t + λ h) + ∑ aei, j kλj ,
j=1 j=1 j=1

(i = 1, . . . , se). Hereby, the values ybS (t + λ h + ∑i−1 ei j ) of the slow components at

j=1 a
i−1
grid points t +λ h+ ∑ j=1 aei j h are approximated by a dense output of the Runge-Kutta
approximation of the compound step, i.e.,
s
ybS (t + θ H) := yS,t + ∑ bi (θ )ki ,
i=1

such that it holds

y(t + θ H) − y(t + θ H)k ≤ cS H p−1 .

max kb
0≤θ ≤1

for all 0 ≤ θ ≤ 1 and some constant cS > 0.

Remark 2.4 Note that the increments li according to yF are only used within the
computation of ki . Consequently, for an explicit Runge-Kutta scheme, ls does not
have to be computed.

Remark 2.5 Other multirate one-step schemes have been developed and analyzed,
for example, by Savcenco et al. [35] and by Günther and Rentrop [23], both based on
Rosenbrock-Wanner methods (ROW), by Sandu and Günther [25] based on GARK
methods, and by Striebel et al [42] based on ROW methods for index-1 DAE systems,
and others.
Multirate Schemes 13

2.4 Overview on multirate strategies

To conclude this section, we state two main multirate strategies which exist for
component-wise splitting:
• Extra-/Interpolation based multirate schemes: here, one computes the split vari-
ables one after the other. We have slowest-first and fastest-first, where the cou-
pling variables are extrapolated or interpolated based on previously computed
approximations. One may use any numerical integration scheme as basis scheme
of such multirate schemes; the order of the scheme p is preserved provided that
the extra-/interpolation is at least of order p − 1.
• Compound-step based multirate methods: here, slow and fast variables are jointly
computed using one macro step (compound step); then, the fast approximation
is disregarded and replaced by M micro steps using the fast dynamics and dense
output of the slow variables (compound-fast approach).
As the dense output for the coupling term yS (t) is available on the whole macro
step, one may discard the idea of using M micro steps of the same step size h
and use the different step sizes according to the step size prediction of the numerical
integration scheme. The latter approach, combined with the compound step, is called
mixed multirate [4].
The time stepping approach introduced by Savcenco, Hundsdorfer and Ver-
wer [35] is a generalization of the compound-step approach. First, an approximation
for all components after one macro step is computed. For those components not
accurate enough the computation is redone with smaller steps. The refinement is
recursively continued until the error estimator is below a given tolerance for all
components.
One criticism of this compound step approach is the use of the macro step size H
also for the fast component yF inside the computation of the new approximate for the
slow variables yS . One may overcome this problem by combining the macro step for
the slow part with the first micro step of the active part (generalized compound-step
approach). As micro and macro step are interwoven in this case, additional coupling
conditions have to be fulfilled for the coefficients of one-step methods to preserve
the order of the method for the slow components. This approach has been introduced
by Kværnø and Rentrop in [28] for Runge-Kutta schemes. Corresponding further
methods are based on the W-method [9] or on generalized additive Runge-Kutta
schemes [25].
If right-hand side splitting is concerned, operator splitting might be the method
of choice. For the split system (3), the idea reads as follows: suppose that the slow
dynamics hs is characterized by an expensive evaluation, whereas the fast dynamics
h f can be cheaply evaluated. In this case, one may develop an operator splitting
approach where the slow dynamics is solved only once on a macro step, and the fast
systems M-times during one macro step. See Section 5 for some more details, in
particular, in the context of geometric integration.
14 Andreas Bartel and Michael Günther

3 Dynamic iteration and multiphysics

Due to downscaling in electric devices and due to higher accuracy requests to

numerical simulation, more and more coupled problems need to be studied and
thus simulated for industrial applications. This gives naturally rise to multiphysics
problems.
In modular time integration of such multiphysics problems, different subsystems
are modeled and simulated by different simulation packages, see e.g. [3, 12, 50]. This
allows to exploit the multirate potential, which is caused by different time scales of
the subsystems, efficiently within a waveform relaxation approach. We describe this
in the following.
To start with, we consider again the component-wise partitioned ODE (2), which
comprises a slow subsystem in yS and fast subsystem in yF . Furthermore, we assume
that the solution or an approximation yS, H , yF, H is available on [t0 , t]. The coupled
initial value problem (2) can be solved iteratively on a time window [t, t + H]
employing old iterates (i−1) and current iterates (i) and respective splitting functions
FS , FF to encode usage of old and current iterates inside the coupled system. This
reads:
(i+1) (i+1) (i) (i+1) (i) (i+1) (0)
ẏS = FS (t, yS , yS , yF , yF ), yS (t) = yS (t),
(8)
(i+1) (i+1) (i) (i+1) (i) (i+1) (0)
ẏF = FF (t, yS , yS , yF , yF ), yF (t) = yF (t),
where the initial waveforms y(0) , z(0) are given by extrapolating from the previous
time window:
!
  (0)   
yS, H yS yS, H
→ (0) := Φ .
yF, H [t−H,t] y [t,t+H] yF, H [t−H,t]
F

Here, Φ denotes an extrapolation operator and let LΦ be the respective Lipschitz

constant. Furthermore the splitting functions shall fulfill Lipschitz conditions with
respect to all components and are consistent, and the consistency with IVP (2):

FS (t, yS , yS , yF , yF ) = fS (t, yS , yF ) and FF (t, yS , yS , yF , yF ) = fF (t, yS , yF ).

Depending on the choice of the splitting functions, one may define different instants
(i+1) (i) (i+1) (i)
of waveform relaxation schemes: Y = (t, yS , yS , yF , yF )
• Picard iteration:
(i) (i) (i) (i)
FS (Y ) = fS (t, yS , yF ), FF (Y ) = fF (t, yS , yF ).

• Jacobi-type iteration:
(i+1) (i) (i) (i+1)
FS (Y ) = fS (t, yS , yF ), FF (Y ) = fF (t, yS , yF ). (9)

• Gauß-Seidel type iteration (slowest-first):

Multirate Schemes 15

(i+1) (i) (i+1) (i+1)

FS (Y ) = fS (t, yS , yF ), FF (Y ) = fF (t, yS , yF ). (10)

• Gauß-Seidel type iteration (fastest-first):

(i+1) (i+1) (i) (i+1)
FS (Y ) = fS (t, yS , yF ), FF (Y ) = fF (t, yS , yF ). (11)

The iterates converge monotonically to the exact solution provided that the macro
step size H is small enough.
(i+1) (i+1)
If (9) is used, then both new iterates yS and yF in (8) can be computed in
parallel, with extrapolation. If the exact integration of (10) or (11) is replaced by a
(i+1) (i+1)
numerical integration, the computation of the new iterates yS , yF is equivalent
to applying an extra-/interpolation based multirate scheme (to slow yS and fast
yF component): (10) represents slowest-first and (11) fastest-first setting And vice
versa, extra-/interpolation based multirate is equivalent to stopping the iteration of
the waveform-relaxation after the first step. See also Section 4 for further discussion.
IVPs of coupled DAE systems (cf. Cor. 2.2)

ẏS = fS (t, yS , yF , zS , zF ), yS (t0 ) = yS,0 , ẏF = fF (t, yS , yF , zS , zF ), yF (t0 ) = yF,0 ,

0 = gS (t, yS , yF , zS , zF ) 0 = gF (t, yS , yF , zS , zF ) (12)

arise, for example, in circuit simulation or in electro-thermal coupling (see Section 4).
A corresponding dynamic iteration scheme needs splitting functions FL , GL , FA and
GA , which fulfill Lipschitz conditions with respect to all arguments and are consistent,
i.e.,

F? (t, yS , yS , yF , yF , zS , zS , zF , zF ) = f? (t, yS , yF , zS , zF ),
G? (t, yS , yS , yF , yF , zS , zS , zF , zF ) = g? (t, yS , yF , zS , zF ) with ? ∈ {S, F}.

Then, the dynamic iteration scheme reads: for ? ∈ {S, F}

(i+1) (i+1) (i) (i+1) (i) (i+1) (i) (i+1) (i)
y? = F? (t, yS , yS , yF , yF , zS , zS , zF , zF ),
(13)
(i+1) (i) (i+1) (i) (i+1) (i) (i+1) (i)
0 = G? (t, yS , yS , yF , yF , zS , zS , zF , zF ).

(0) (0) (0) (0)

Again for the actual time window [t, t + H], initial waveforms yL , yA , zL , zA
are given by extrapolation of the waveform approximates of the last window [t −
H, t]. In contrast to coupled ODE systems, monotone convergence can no longer be
guaranteed by choosing the window step size H small enough. In addition, one has
to fulfill two additional contractivity conditions, see e.g. [3]). For the Gauss-Seidel
type approaches, this reads:
• Convergence within one window:

α < 1, (14)
16 Andreas Bartel and Michael Günther

whereas the contractivity constant α is given by the sum of the Schur complent-
like quantity
 ∂ GS ∂ GS
−1  ∂ G ∂ GS

S
(i+1) (i+1) (i) (i)
zS zF zS zF
max  ·  ∂G
   
 ∂G ∂ GF ∂ GF

τ, with F F
t ≤t+τH ≤t+H (i+1) (i+1) (i) (i)
zS zF zS zF

and a term of order O(H), which tends to zero for the limit H → 0.
• Stable error propagation from window to window:

LΦ α k < 1 (15)

with k iterations in the current window and Lipschitz constant LΦ of the extrap-
olation operator. Hence, depending on the LΦ , more than one iteration may be
necessary, though convergence within one window is given.
If we replace the exact solution of (13) by a numerical integration and stop after the
first iteration, again an extra-/interpolation based multirate scheme is defined, now
for coupled DAE systems (12). Note that stability can only be guaranteed if (15)
holds; here, additional iterations of the multirate scheme can be necessary to obtain
a convergent scheme.

Remark 3.1 Note that in the case of DAE-ODE coupling, i.e., when the second
subsystem to be solved in a Gauss-Seidel type approach is defined by an ODE, then
α = 0 (14), since both splitting functions GS and GF do not depend on old iterates
of the algebraic variables. In this case, one can show that the dynamic iteration is
convergent of first order in the macro step size H (cf. [5]). In Section 4.3, a electro-
thermal problems, which yields a DAE-ODE coupling is discussed. Depending on
the fine structure of the coupling, up to second order convergence is possible, see [5].

Multirate schemes on basis of dynamic iteration schemes have been successfully

applied to multiphysics problems. See, for example, field-circuit coupling [36, 37],
electro-thermal coupling refined network modelling [6].

4 Applications in circuit simulation

Large integrated electrical networks are usually modeled via differential algebraic
equations (e.g., see [22]): for a charge-flux oriented formulation, we have compactly

d
F (t, x, w(x)) = 0, (16)
dt
in terms of node potential and currents (through voltage defining elements) x, charges
d
and fluxes w = q(x) as well as time t. In (16), the matrix dw F · dx
d
w is generally not
Multirate Schemes 17

regular (i.e., in this case, (16) represents a DAE). In fact, these networks are built up by
numerous coupled subcircuits of different functionality. As the subcircuits constitute
different functional units, the overall system often shows multirate behavior To
exploit this multirate potential, such a subcircuit partitioning has to be taken into
account.

4.1 Partitioned network modeling

The subcircuits are modeled independently and composed to one macro system by
connecting respective terminals, i.e., each pair of connected (terminal) nodes merge
to a one node of the coupled system. This can be modeled by inserting virtual voltage
sources for each pair of connected (terminal) nodes, where virtual means zero applied
voltage. This approach preserves the macro circuit’s block structure and produces
additional variables: the branch currents u through the coupling voltage sources.
These currents are implicitly determined by the property, that the node potentials of
each pair of connected boundary nodes have to coincide. Now, we assume given a
macro circuit composed of λ = 1, . . . , r subcircuits each of type (16) and coupled
via the algebraic constraints of virtual voltages sources. This reads:

d
Fλ (t, x, w (x ), u) = 0, (λ = 0, . . . , r) (17a)
dt λ λ
G (x1 , . . . , xr ) = 0, (17b)

with xλ internal node potentials and voltages, internal fluxes and charges wλ =
qλ (xλ ) of subsystem λ . Several index-1 conditions can be formulated for system
(17) and its subsystems: [10]:
(C1) The overall system (17) has index 1 (with respect to x1 , . . . , xr , u).
(C2) All systems (17a) define index-1 systems with respect to x¸ (and u given as
input).
(C3) For any λ ∈ {1, . . . , r}, the λ -th system (17a) with coupling condition (17b)
has index-1 with respect to xλ ¸ and u given all other xi (i.e.,i 6= λ ) as input).
In analogy to the procedure described in [18], topological conditions to guarantee
the index conditions (C1)-(C3) can be derived. Given the above index-1 conditions,
the overall model (17) can be transformed into algebraically coupled semi-explicit
systems 
ẏλ (t) = fλ (t, zλ , u)
λ = 1, . . . , r
0 = hλ (t, yλ , zλ , u) (18)
0 = g(z1 , . . . , zr )
with algebraic variables zλ as node potentials and inner currents of system λ as
well as differential variables yλ defining the charges and fluxes (of system λ ). In
particular, setting (18) is the basis for developing more efficient schemes.
18 Andreas Bartel and Michael Günther

Remark 4.1 Notice we have two sets of coupled DAEs: one abstractly coupled
DAEs (12), one with dedicated coupling equation (18), which represents networks,
which are coupled via virtual voltage sources. In fact, for the latter case, the coupling
condition (18-3) is linear in all arguments.

4.2 Multirate schemes

To get a mixed multirate scheme for coupled index-1 DAEs of type (18) (semi-
explicit), we first regard the case r = 2 with F (fast) and S (slow) scale. It is natural
to assume the coupling variable u (defined by 0 = g) behaves slowly like yS and zS .
This amounts to the following structure:

ẏS = fS (t, zS , u), ẏF = fF (t, zF , u),

0 = hS (t, yS , zS , u), 0 = hF (t, yF , zF , u), (19)
0 = g(zS , zF ).

With the index assumption (C3), the slow subsystem [ẏS = fS , 0 = hS , 0 = g] is index-
1 for algebraic variables zS and u; in addition, the fast subsystem [ẏA = fA , 0 = hA ]
is of index-1 with respect to zF if (C2) holds.
Now, we can apply multirate schemes to solve system (19). We present two
options:
• Compound-step approach: Verhoeven et al. have developed a series of multirate
schemes [45] for coupled network equations: slowest-first [46] and compound-
step approach [48], equipped with step size prediction for both macro and micro
step size and error control [48]. His approach does not demand a given sub-
circuit partitioning, but allows for automatic partitioning of the overall network
equations [47].
• Mixed multirate: Striebel et al. [42] have developed a mixed multirate scheme on
the basis of Rosenbrock-Wanner methods, which uses a generalized compound
step to (jointly) compute macro step for the slow part and the first micro step
of the fast part. The remaining micro steps of the active part are done by mixed
multirate. The scheme allows for more than two time scales in a hierarchical
setting by nesting compound steps and later micro-steps in a way that at each
time merely a two-level multirate scheme is engaged [43].
Remark 4.2 In circuit simulation, the slow variable is usually referred to as latent
and the fast variable is referred to as active.

4.3 Thermal-electric coupling — silicon on insulator

The transistor technology, silicon on insulator (SOI), introduces a thin insulator

layer within the device. This layer can made of an oxide, which has the purpose
Multirate Schemes 19

to electrically decouple the substrate and the channel area, see Fig. 1 for a sketch
of such a device. Furthermore, it confines the channel to a rather small area and
thus lowers parasitic (electrical) capacitance. In the end, the power consumption can
be reduced and frequencies enhanced. However, since the electric insulator is often
also a thermal insulator, the thermal coupling gains importance and needs to be
investigated, see, e.g., [44]. Thus the SOI technology also drives the development of
thermal-electric simulation.

gate
source drain
n channel
n+ n+
p
insulator

substrate

Fig. 1: Sketch of an SOI device.

Such a thermal-electric system is a multiphyics problem. On larger spatial units,

one often obtains a multirate system: On the one hand, the electric subsystem switches
on a very fast scale (e.g. on GHz scale). On the other hand, the temperature is
dissipated and conducted slowly; e.g. the heat diffusivity in silicon is at 300K about
87 10−6 m2 /s. This yields a multirate setting and demands for an according simulation
strategy.
In [7, 8], an according modeling and a simulation strategy was addressed. To
this end, an accompanying (thermal) network (AN) for an electric network was pro-
posed. The AN allows the connection of simple, thermal elements. These elements
are spatially one dimensional and thus enable heat conduction along dedicated 1d
structures.

Coupled thermal-electric model. The standard electric network model is a

DAE of type (16), see e.g. [24] for details. To enable a thermal-electric coupling, we
introduce a set of parameters p (which are going to be temperature dependent). This
slightly generalizes the network equations (16) to the following initial value problem
(for given p):

d
F (t, x, w(x); p) = 0, with consistent initial value x(t0 ) = x0 . (20)
dt
Thermally, the AN model comprises a set of heat conduction equations for
distributed temperatures: T : [t0 ,te ] × [0, 1] → Rk ; all 1d elements shall have a
normalized spatial dimension [0, 1]. Furthermore, we have lumped temperatures
θ : [t0 ,te ] → Rm , which denote the temperatures at the coupling nodes of the AN.
20 Andreas Bartel and Michael Günther

This can be cast into the following type of equations (cf. [8])

∂t T = ∂rr T − (T − Tenv ) + P1 ,
(21)
∂t θ = g (∂r T (0,t), ∂r T (1,t)) − (θ − Tenv ) + P2

with source terms P1 , P2 (for the power dissipation) and material coefficients nor-
malized to one, for simplicity of notation. In addition, we have boundary conditions
(or coupling conditions) for the AN:

T (0,t) = M0 θ , T (1,t) = M1 θ

with suitable matrices M0 , M1 identifying coupling of the 1d-elements. Finally, we

need to supply suitable initial conditions: θ (t0 ) = θ0 ≥ Tenv , T (t0 , x) = T0 (x) ≥ Tenv ,
where Tenv is a given ambient temperature.
Now, the coupling conditions are as follows. On the one hand, the source terms
Pi indicate the dissipated powers, which are a function of the network variables x:

Pi = Pi (x) ≥ 0 (for i = 1, 2).

On the other hand, the electric network does depend on the temperatures T, θ , via
the network parameters:
p = p(T, θ ).
Later on, the accompanying thermal network was generalized to patches, where
two spatial dimensional are involved, see [1, 14].

Dedicated multirate strategy for simulation. First, we consider a method of

lines approach, where we apply a suitable spatial discretization to the thermal part.
Due to the structure of the AN equation (21), we can obtain an ODE model for the
thermal part, and thus, a coupled DAE-ODE system.
For the time integration, we apply a dynamic iteration, cf. Section 3, with win-
dowing. Due to the DAE-ODE coupling, the contraction factor α = 0 (14) and thus
we have linear convergence of the dynamic iteration scheme [6].
Practically, we first solve the electric subsystem. To exploit the multirate setting
with fast electric signals and slow thermal adjustments, we pursue an energy cou-
pling, which is based on averaging, cf. [6, 15]. That is, instead of coupling via the
fast changing instantaneous power dissipation, one can compute the total dissipated
power E H of electric subsystem during a certain time interval [t, t + H]. This is an
energy, and may be computed by simply adding according energy variables E to
the network variables and appending the network equations (20) with corresponding
differential equations
Ė = P(x).
Then, the total amount of the dissipated power is added to the thermal network as
averaged quantity: Pi = E H /H. This way, the total amount of the dissipated (electric)
energy is added to the thermal system. The addition of the actual power as source
term to the heat conduction might happen with a little time shift on the scale of
Multirate Schemes 21

the fast system. Of course, this is for a slow heat system of minor importance. The
reverse coupling needs to adjust the temperature dependent network parameters, i.e.,
an updated device temperature needs to be assigned. To this end, the heat equation
is solved. Since we expect on the H−scale only minor temperature changes, we
can even skip any iterations of the dynamic scheme and compute the preceding
communication step (e.g. [t + h, t + 2H]), i.e., solve again first the electric subsystem
with the new temperatures. This is referred to as multirate co-simulation, since both
subsystems may be solved on their time scale and no overhead of iterations occurs.
This multirate co-simulation strategy was successfully applied to solve a ring
oscillator circuit, see [7]. In fact, it is reasonable to exchange data after one step of
the slow subsystem, i.e., the communication step size can be chosen as the inherent
step size of the solve subsystem. As a further enhancement, one can apply a dedicated
midpoint rule to solve the slow (heat) subsystem and gain formally a second order
method, see [6].

5 Molecular dynamics

In molecular dynamics, one has to solve initial-value problems of the type

ẇ = h(w), w(t0 ) = w0 (22a)

with
     
q −1 0 −1 q
w := , h(w) := J ∇H(q, p), J := , w0 := 0 (22b)
p 1 0 p0

with positions q : [t0 , T ] → Rn , momenta p : [t0 , T ] → Rn and Hamiltonian H :

Rn × Rn → R. The separable Hamiltonian H(q, p) = T (p) +V (q) is composed of
a kinetic energy T and a potential energy V .
The standard numerical integration scheme applied to Hamiltonian systems is the
Störmer-Verlet (or leap-frog) scheme, which we refer to as ψhH0 for step size h0 . In
fact, we can describe ψhH0 for (22a-22b) via operator splitting based on explicit Euler
steps ϕh for the components q and p:

ψhH0 (q0 , p0 ) := ϕhV0 /2 ◦ ϕhT0 ◦ ϕhV0 /2 (q0 , p0 ). (23)

Here, ϕhT advances the kinetic energy H = T with an explicit Euler step of size h and
ϕhV analogously the potential energy H = V . In fact, this numerical approximation
(23) solves exactly the shadow Hamiltonian H, b which differs from the original
Hamiltonian H by the following expression:
2
b − H = − h0 ([V, [V, T ]] + 2[T, [V, T ]]) + O(h40 ).
H
24
22 Andreas Bartel and Michael Günther

If the forces derived from the Hamiltonian are hierarchical and the larger forces
turn out to be cheap to compute, the integration of molecular dynamics can be
accelerated by means of multirate schemes. For example in lattice quantum chro-
modynamics [27] holds: the gauge force is cheap to evaluated and largest in size the
fermionic force is expensive but smaller in size. Thus, let us assume that the potential
energy consists of two parts V1 and V2 , which differ as follows:
• V1 : the contribution
f1 (q) := J −1 ∇V1 (q),
to the force is strong and with a fast dynamics, but its evaluation is cheap;
• V2 : the contribution
f2 (q) := J −1 ∇V2 (q),
to the force is weak with a slow dynamics, but its evaluation is expensive.
Consequently, one may split the ODE (22) with respect to the right-hand side:

hs (w,t) := f2 (q), h f (w,t) := f1 (q) + J −1 ∇T (p),

where we have assumed that the dynamics related to the kinetic energy is fast, but
cheap to be evaluated. Thus, we have the structure of (3).
Naturally, one idea to exploit this multirate behavior is an according evaluation:
within one macro step H evaluate the slow, expensive part hs (only) once; the
fast, cheap part h f several times. However, the multirate approach has to preserve the
symplectic and time-reversible structure of the Hamiltonian flow. Hence, an operator
splitting approach appears to be favorable, as the composition of symplectic and time
reversible schemes is again symplectic and time reversible.
For the right-hand side split ODE (3), a multirate method ψ ehH (q0 , p0 ) [38] with
0
macro step size h0 and m1 mirco steps of step size h1 = h0 /m1 is obtained from
 m 1
ψehH0 (q0 , p0 ) := ϕhV2/2 ◦ ψhT +V1 ◦ ϕhV2/2 (q0 , p0 ),
0 1 0

where we used the Störmer-Verlet scheme ψhH of the single-rate case with H :=
T +V1 and step size h1 . This multirate scheme conserves now the following shadow
Hamiltonian H,
b which differs from the original Hamiltonian H by

2 
b − H = − h0 ([V2 , [V2 , T ]] − 2[V1 , [V2 , T ]] − 2[T, [V2 , T ]] +
H
24 
1
([V1 , [V1 , T ]] − 2[T, [V1 , T ]]) + O(h40 ).
m21

Remark 5.1 Let us assume that the dynamics level of V1 over V2 is proportional to m1 ,
then all commutators involving one or two instances of V1 are properly scaled with
1/m21 besides the commutator [V1 , [V2 , T ]]. However, if the Störmer-Verlet scheme for
m1 inner micro-steps is replaced by a force-gradient scheme [30], the commutator
Multirate Schemes 23

[V1 , [V2 , T ]] is eliminated, and the shadow Hamiltonian for this multirate version [39]
is given by
 2  
h0 1 1
H+ [V1 , [T,V1 ]] + [T, [T,V1 ]] + O(h40 ).
m1 96 48

This multirate approach can be applied to Hamiltonians with a further splitting

of the potential energy in a hierarchical manner as follows: let us assume that the
Hamiltonian is given by
N
H(q, p) = T (p) + ∑ Vl (q).
l=1

If the energies are ordered such that the computational cost are increasing, while
at the same time the strength of the associated forces is decreasing, a multirate
integration based on the Störmer-Verlet scheme can be defined [38]: using macro
step size h0 and micro step size h1 = h0 /M proceeds as follows:
 m 1
ϕh0 (q0 , p0 ) = ϕhH2/2 ϕhH11 ϕhH2/2 ,
0 0

with H1 (q, p) = T (p)+ ∑N−1

l=1 Vl (q) and H2 (q, p) = VN (q). This scheme can be nested,
by introducing a next finer step size h2 = h1 /m2 and further splitting H1 into

H1 (q, p) = H11 (q, p) + H12 (q, p)

H1
with H11 (q, p) = T (p) + ∑N−2
l=1 Vl (q) and H12 (q, ) = VN−1 (q) in order to replace ϕh1
above by   m2
ϕhH11 = ϕhH12/2 ϕhH211 ϕhH12/2 .
1 1

This procedure can be applied recursively to obtain N different step size ratios at the
end, corresponding to the activity levels of the N potential energies Vl (q).

6 Conclusion and outlook

In many applications, the governing differential equations are characterized by

strongly varying time scales. Numerical time integration of these type of models
demands integration schemes able to exploit this behavior, not only for efficiency
reason, but quite often to enable numerical simulation at all.
In this paper, we have discussed two basic classes of multirate schemes based on
extra-/interpolation and compound-step approaches: the pioneering work by Rice for
missile applications (compound step approach for one-step methods) and the seminal
work by Gear and Wells driven by electrical circuit application (extra-/interpolation
24 Andreas Bartel and Michael Günther

for multi-step methods). For the latter, we characterized the overall convergence
properties for coupled ODE systems.
Another trend in applications — refined modeling yielding coupled systems con-
sisting of multiphysical subsystems — initialized further development in dynamic
iteration approaches. They deliver another class of (full) multirate schemes by stop-
ping the iteration after the first sweep. We discussed the basics of this approach
and verified the efficiency of such multirate schemes by inspecting the coupled
thermal-electric problem, where we gave a tailored numerical algorithm.
If the multirate behavior is not given by varying solution components, but by
different characteristics of the right-hand side, an operator splitting approach is the
method of choice to derive multirate schemes, especially if preservation of properties
such as in geometric integration is mandatory. Here, molecular dynamics serves as
an example of applications in this field.
Summing up, the need for the simulation of complex, time consuming applications
inspired the numerical analysis to develop a new class of methods, multirate methods.
And this class is still growing. In turn, the enhancement of the algorithm’s efficiency
enabled an increase in complexity of applications.
The increasing interest in parallel-in-time schemes in the last years offers a new
way to speed up simulation time by the help of multirate schemes. Here the com-
bination of parallel-in-time schemes with dynamic iteration and multirate schemes
seems to be promising, see e.g. [19]. This will be an interesting topic for research
on multirate schemes in the coming years.

References

[1] Alì, G., Bartel, A., Culpo, M., De Falco, C.: Analysis of a PDE thermal element
model for electrothermal circuit simulation. In: J. Roos and L.R.J. Costa (eds.),
Scientific Computing in Electrical Engineering (SCEE 2008). Springer, 2010,
273–280.
[2] Alì, G., Bartel, B., Günther, M., Romano, V., Schöps, S.: Simulation of coupled
PDAEs: Dynamic iteration and multirate simulation. In: M. Günther (ed.):
Coupled Multiscale Simulation and Optimization in Nanoelectronics, Springer
Berlin, 2015, 103–156.
[3] Arnold, M., Günther, M.: Preconditioned Dynamic Iteration for Coupled
Differential-Algebraic Systems. BIT 41:1 (2001), 1–25.
[4] Bartel, A.: Multirate ROW Methods of Mixed Type for Circuit Simulation. In:
van Rienen, U., Günther, M., Hecht, D. (eds.): Scientific Computing in Elec-
trical Engineering. Lecture Notes in Computational Science and Engineering
18, Springer-Verlag, Berlin, 2001, 241–249.
[5] Bartel, A., Brunk, M, Schöps, S.: On the convergence rate of dynamic iteration
for coupled problems with multiple subsystems, J. Comp. Appl. Math. 262
(2014), 14–24.
Multirate Schemes 25

[6] Bartel, A.: Partial Differential-Algebraic Models in Chip-Design — Ther-

mal and Semiconductor Problems. PhD-Thesis TU Munich, 2004. Fortschritt-
Berichte, VDI Verlag, Düsseldorf.
[7] Bartel, A., Feldmann, U.: Modeling and Simulation for Thermal-Electric Cou-
pling in an SOI-Circuit. In: A.M. Anile, G. Alì, G. Mascali (eds.): Scien-
tifc Computing in Electrical Engineering (SCEE-2004 Proceedings), Springer,
Berlin, 2006, 27–32.
[8] Bartel, A., Günther, M.: From SOI to abstract electric-thermal-1D multiscale
modeling for first order thermal effects. Math. Comp. Modell. Dyn. Syst 9:1
(2003), 25–44.
[9] Bartel, A., Günther, M.: A multirate W-method for electrical networks in
state–space formulation. J. of Comput. Appl. Math. 147 (2002), 411–425.
[10] Bartel, A., Günther, M.: PDAEs in Refined ElectricalNetwork Modeling. SIAM
Review 60:1 (2018), 59–91.
[11] Bartel, A. and Günther, M.: Inter/extrapolation-based multirate schemes –
a dynamic-iteration perspective. Submitted for publication (ArXiV math.NA
2001.02310).
[12] Burrage, K.: Parallel and sequential methods for ordinary differential equations.
Oxford Science Publications, 1995.
[13] Ciuprina, G., Ioan D. (eds.): Scientific Computing in Electrical Engineering.
Mathematics in Industry. Springer, Berlin, 2007.
[14] Culpo, M.: Numerical Algorithms for System Level Electro-Thermal Simula-
tion. PhD-Thesis at BU Wuppertal, 2009. http://elpub.bib.uni-wuppertal.de
[15] Deml, Ch., Türkes, P.: Fast Simulation Technique for Power Electronic Circuits
with Widely Different Time Constants. IEEE Trans. on Industry Applications
35:3 (1999), 657–662.
[16] Engstler Ch., Lubich, Ch.: MUR8: a multirate extension of the eighth-order
Dormand-Prince method, Appl. Num. Math. 25:2–3 (1997), 185–192.
[17] Engstler Ch., Lubich Ch.: Multirate extrapolation methods for differential
equations with different time scales, Computing 85:2 (1997), 173–185.
[18] Estévez Schwarz, D., Tischendorf, C.: Structural Analysis for Electric Circuits
and Consequences for MNA, Int. J. of Circuit Theory and Applications 28
(2000), 131–162.
[19] Gander, M.J., Kulchytska-Ruchka, I., Niyonzima, I., Schöps, S.: A New Parareal
Algorithm for Problems with Discontinuous Sources. SIAM J. on Scientific
Computing 41:2 (2019), B375–B395.
[20] Gear, C.W.: Automatic multirate methods for ordinary differential equations.
In: Proceedings IFIP, North-Holland, Amsterdam 1980, 712-722.
[21] Gear, C.W., Wells, D.: Multirate linear multistep methods. BIT 24 (1984),
484–502.
[22] Günther, M., Feldmann, U.: CAD-based electric-circuit modeling in industry I:
mathematical structure and index of network equations Surveys on Mathematics
for Industry 8:2 (1999), 97-130.
[23] Günther M., Rentrop P.: Multirate ROW methods and latency of electric cir-
cuits, Appl. Num. Math. 13 (1992), 83–102.
26 Andreas Bartel and Michael Günther

[24] Günther, M., Rentrop, P.: Numerical Simulation of Electrical Circuits, GAMM-
Mitteilungen 23:1 (2000).
[25] Günther M., Sandu, A.: Multirate generalized additive Runge Kutta methods
Numerische Mathematik 133 (2016), 497–524.
[26] Hairer, E., Wanner, G.: Solving Ordinary Differential Equations II, Springer,
Berlin, 2002.
[27] Knechtli, F., Günther, M., Peardon. M.: Lattice Quantum Chromodynamics
Practical Essentials, Springer Briefs in Physics, 2016.
[28] Kværnø, A., Rentrop, P.: Low order multirate Runge-Kutta methods in electric
circuit simulation, Preprint No. 99/1, IWRMM, Universität Karlsruhe (TH),
1999.
[29] Lelarasmee, E. Ruehli, A.E, Sangiovanni-Vincentelli, A.L.: The Waveform Re-
laxation Method for Time-Domain Analysis of Large Scale Integrated Circuits,
IEEE Transactions on Computer-Aided Design of Integrated Circuits and Sys-
tems 1:3 (1982), 131–145.
[30] Omelyan, I.P., Mryglod, I.M., Folk, R.: Symplectic analytically integrable de-
composition algorithms: classification, derivation, and application to molecu-
lar dynamics, quantum and celestial mechanics, Comput. Phys. Commun. 151
(2003), 272–314.
[31] Orailoglu, A.: A multirate ordinary differential equation integrator. Master The-
sis UIUCDCS-R-79-959, University of Illinois at Urbana-Champaign, Dept. of
Computer Science, 1979.
[32] Rodríguez-Gómez, G., González-Casanova, P., Martínez-Carballido, J.: Com-
puting general companion matrices and stability regions of multirate methods,
Int. J.for Numerical Methods in Engineering 61 (2004), 255–273.
[33] Rice, J.R.: Split Runge-Kutta method for simultaneous equations, J. Res. Nat.
Bur. Standards 64B (1960), 151–170.
[34] Sandu, A., Constantinescu, E.M.: Multirate Explicit Adams Methods for Time
Integration of Conservation Laws. J. of Scientific Computing 38:2 (2009),
229–249.
[35] Savcenco, V., Hundsdorfer. W., Verwer, J.G.: A multirate time stepping strategy
for stiff ordinary differential equations, BIT 47 (2007), 137–155.
[36] Schöps, S.: Multiscale Modeling and Multirate Time-Integration of Field/Cir-
cuit Coupled Problems of Elektrotechnik. (PhD Thesis, Univ. Wuppertal). VDI
Verlag, Düsseldorf, 2011.
[37] Schöps, S., Bartel, A., De Gersem, H.: Multirate Time Integration of Field/Cir-
cuit Coupled Problems by Schur Complements. In: Michielsen, B., Poirier, J.-R.
(eds.): Scientific Computing in Electrical Engineering SCEE 2010, Springer
Berlin Heidelberg, 2012, 243–251.
[38] Sexton, J.C., Weingarten, D.H.: Hamiltonian evolution for the hybrid Monte
Carlo algorithm. Nucl. Phys. B 380:3 (1992), 665–677.
[39] Shcherbakov, D., Ehrhardt, M., Günther M., Peardon, M.: Force-gradient nested
multirate methods for Hamiltonian systems. Comput. Phys. Commun. 187
(2015), 91–97.
Multirate Schemes 27

[40] Skelboe, S.: Stability properties of backward differentiation multirate formulas,

Appl. Num. Math. 5 (1989), 151–160.
[41] Skelboe S., Andersen, U.: Stability properties of backward Euler multirate
formulas, SIAM J. Sci. Stat. Comput. 10 (1989), 1000–1009.
[42] Striebel, M., Bartel A., Günther, M.: A multirate ROW-scheme for index-1
network equations, Appl. Num. Math. 59:3–4 (2009), 800–814.
[43] Striebel, M., Günther, M.: Hierarchical mixed multirating in circuit simulation.
In: [13], 221–228.
[44] Su, L.T., Chung, J.E., Antoniadis, D.A., Goodson K.E., Flik, M.I.: Measure-
ment and modeling of self-heating in SOI nMOSFET’s, IEEE Trans. on Elec-
tron Devices 41:1, 69–75, 1994.
[45] Verhoeven, A., El Guennouni, A., ter Maten, E.J.W., Mattheij, R.M.M.: A gen-
eral compound multirate method for circuit simulation problems. In: Scientific
Computing in Electrical Engineering, Springer, 2006, 143–149.
[46] Verhoeven, A., ter Maten, E.J.W., Mattheij, R.M.M. Tasic, B.: Stability analysis
of the BDF slowest-first multirate methods, Int. J. of Comput. Math. 84:6
(2007), 895–923.
[47] Verhoeven, A., Tasic, B., Beelen, T.G.J., ter Maten, E.J.W. Mattheij, R.M.M.:
Automatic partitioning for multirate methods. In: [13], 229–236.
[48] Verhoeven, A., Tasic, B., Beelen, T.G.J., ter Maten, E.J.W., Mattheij, R.M.M.:
BDF compound-fast multirate transient analysis with adaptive stepsize control,
J. Numer. Anal. Ind. Appl. Math 3 (2008), 275–297.
[49] Wells, D.R.: Multirate linear multistep methods for the solution of systems
of ordinary differential equationsequation integrator, PhD Thesis UIUCDCS-
R-82-1093, Unviersity of Illinois at Urbana-Champaign, Dept. of Computer
Science, 1982.
[50] White J.K., Sangiovanni-Vincentelli A.: Waveform Relaxation. In: Relaxation
Techniques for the Simulation of VLSI Circuits. Kluwer International Series
in Engineering and Computer Science 20, Springer, Boston, MA, 1987.
Other documents randomly have
different content
Cotton (Sir Robert), an eminent book collector, i. 438.
Cotton (Roger), a minor poet, of the age of Shakspeare, i. 680.
Country inns, picture of, i. 216-218.
Country life, manners and customs during the age of
Shakspeare, i. 68-122.
Description of its holidays and festivals, amusements, 123-
313.
Superstitions, 314-400.
Literature but little cultivated, 430, 431.
Country squires, rank of, in Shakspeare's age, i. 68.
Description of their mansion houses, 72, 73.
And halls, 74, 77-79.
Distinctions observed at their tables, 74, 75.
Their diet, 75, 76.
But little skilled in literature, 430, 431.
Portrait of a country squire in the reign of Queen Anne, 88.
note [86:B].
Courtiers of Elizabeth, sometimes wrote lyrics, for music, i. 731.
Instances of her rough treatment of them, ii. 150, 151.
Courting chair of Shakspeare, notice of, i. 61.
Courtship, how anciently conducted, i. 220.
Cox (Captain), an eminent book collector, i. 434.
List of romances in his library, 518, 519.
Remarks on it by Mr. Dibdin, 520.
Crab-tree, Shakspeare's, still remaining at Bidford, i. 49.
 Roasted crabs and ale a favourite mess, 105, 106.
Credulity of the age of Shakspeare, instances of, i. 314-400. ii.
154.
Criticism, state of, in the age of Elizabeth and James I., i. 456.
Severity of controversial criticism, 457.
Lampooning critics, 459.
Notice of the critical labours of Gascoigne, 461.
Of James I. ibid. 462, 463.
Of Webbe, 463, 464.
Of Spenser, 464.
Of Fraunce, 464.
Of Hake, ibid. 465.
Of Puttenham, 465, 466.
Of Sir John Harrington, 466.
Of Sir Philip Sidney, 467.
Of Meres, 468.
Of Campion, ibid.
and of Bolton, 470.
Crocodiles, legendary tales concerning, noticed, i. 389.
Cromek (Mr.), accounts by, of the fairy superstitions in Scotland,
ii. 325, 326.
Cross-bow, chiefly used for killing game, ii. 182.
Culrose (Elizabeth), a minor poetess of the age of Shakspeare, i.
680.
Curiosity of the age of Shakspeare, illustrations of, ii. 155.
Cutwode (T.), a minor poet of the age of Shakspeare, i. 680.
Cymbeline, probable date of, ii. 466.
Beauty of its fable, ibid.
Remarks on the character of Imogen, 467.
And of Cloten, 468.

Passages of this drama, illustrated in the present work.

Act ii. scene 2., ii. 115. 117.

scene 4., ii. 113.

Act iii. scene 2., i. 297.

scene 4., ii. 91.
Act iv. scene 1., i. 243.
scene 2., i. 214. 395.
Act v. scene 3., i. 308.

scene 5., i. 397.

Czartoryska (Princess), the purchaser of Shakspeare's chair, i.

22, 23.

D
"Damon and Pythias," illustration of, i. 106.
Dancing, a favourite amusement in the age of Shakspeare, ii.
174.
Notice of different kinds of dances, The Brawl, 175.
 The Pavin, ibid. 176.
Canary Dance, 177.
Corantoes, ibid. 178.
Dancing Horse, in the time of Shakspeare, notice of, ii. 186.
Danes, massacre of, i. 149, 150.
Danger, supposed omens of, i. 351-354.
Daniel (Samuel), critical notice of his "Defence of Ryme," i. 169,
470.
And of his poems, 611.
Causes of the unpopularity of his poem on the "Civil Wars
between the Houses of York and Lancaster," ibid.
General observations on his style and versification, 612.
Notice of his sonnets, ii. 55.
Was the prototype of Shakspeare's amatory verse, 57, 58.
Daniel's History of England, character of, i. 176, 477.
Darwin's (Dr.), poetical description of the night-mare, i. 348.
note.
Davenant (Sir William), anecdote of his attachment to
Shakspeare, ii. 589.
Davidstone (John), a minor poet of the age of Elizabeth, i. 680.
Davies (Sir John), notice of, i. 613.
Critical merits of his poem, entitled "Nosce Teipsum," ibid.
Davies (John), a minor poet of the age of Shakspeare, list of the
pieces of, i. 680. and note [680:B].
Davison (Francis and Walter), minor poets in the time of
Shakspeare, i. 680, 681.
Critical notice of their "Poetical Rapsodie," i. 728-730.
Davors (John), critical remarks on the poems of, i. 614.
Days (particular), superstitious notions concerning, i. 323.
St. Valentine's-Day, 324.
Midsummer-Eve, 329.
Michaelmas-Day, 334.
All-Hallow-Eve, 341.
Dead, bodies, frequently rifled of their hair, ii. 92, 93.
Death, account of supposed omens of, i. 351-362.
Delineation of, ii. 455, 456.
Decker (Thomas), character of as a miscellaneous writer, i. 486.
Notice of his "Gul's Horn Booke," 487.
Of his "Belman in London," ibid.
Of his "Lanthern and Candlelight," ibid.
His quarrel with Ben Jonson, ibid.
Probable time of his death, 488.
Estimate of his merits, as a dramatic poet, ii. 566, 567.
Extract from his "Gul's Horn Book," on the fashions of that
age, ii. 102.

Passages of his Plays, which are illustrated or explained.

The Honest Whore, i. 75.
More Dissemblers besides Women, ii. 147.
 Seven Deadly Sinnes of London, i. 251.
Villanies Discovered by Lantorne and Candle-light, i. 273.
396.

Dedications of plays, customary reward for, ii. 225.

Dee (Dr. John), an eminent book-collector, i. 434.
And magician, ii. 510.
Account of his singular character, 510-513.
Catalogue of his library, 511, 512. notes.
Deer-stealing, Shakspeare punished for, i. 404, 407, 408.
De la Casa (John), the "Galatea" of, translated into English, i.
453.
Delone (Thomas), a minor poet of the age of Shakspeare, i.
681.
Notice of his "Ballads," ibid. note [681:A].
Demoniacal voices and shrieks, superstitious notions
concerning, i. 355.
The presence of demons supposed to be indicated by lights
burning blue, 358.
Dennys, or Davors, (John), "Treatyse on Fishing," notice of, i.
291.
Beautiful quotation from, 292, 293.
His book translated into prose by Markham, 293, 294.
Derricke (John), a minor poet of the age of Shakspeare, i. 681.
Descriptions, exquisite, in Shakspeare's "Venus and Adonis," ii.
21-26, 27.
Desdemona, beautiful ditty quoted by, i. 592.
Remarks on her character, ii. 531.
Desserts, where taken, in the age of Shakspeare, ii. 144.
Devil, supposed compact with, of witches, account of, ii. 483-
485.
Dibdin's (Rev. T. F.), "Bibliomania," notice of, i. 432.
His character of "Stubbes's Anatomie of Abuses," 502.
Account of Dr. Dee's library, ii. 511, 512. notes.
Dicer's Oaths, falsehood of, illustrated, ii. 171, 172.
Dictionaries, list of, in use in Shakspeare's time, i. 25. note.
Cooper's Latin and English Dictionary used by him, 26.
Diet of country squires in the age of Shakspeare, i. 75, 76.
Of country gentlemen, 79, 80.
Of farmers or yeomen, on ordinary occasions, 103-108.
On festivals, 109.
Of the sovereigns and higher classes during the age of
Shakspeare, ii. 120-129.
Digby (Sir Kenelm), marvellous properties ascribed to his
sympathetic powder, i. 375, 376.
Dinner, hour of, in the time of Shakspeare, ii. 125.
Account of the dinners of the higher classes, 126-129.
Hands, why always washed before dinner, 145.
Dionysius's angelic hierarchy, account of, i. 335.
Distaff's (Saint) Day, festival of, i. 135.
Verses on, ibid. 136.
Diversions, in the age of Shakspeare, enumeration of, i. 246,
247.
Account of the itinerant stage, 247-252.
Cotswold games, 252-254.
Hawking, 255.
Hunting, 272.
Fowling, 287.
Bird-batting, 289.
Fishing, 289.
Horse-racing, 297.
The Quintaine, 300.
Wild-goose chace, 304.
Hurling, 305.
Shovel-board, 306.
Shove-groat, 307, 308.
Juvenile sports, 308.
Barley breake, 309.
Whipping a top, 312.
Diversions of the metropolis and court, ii. 168.
Card-playing, 169.
Tables and dice, 171.
Dancing, 172.
Bull-baiting and bear-baiting, 176.
 Archery, 178.
Frequenting of Paul's Walk, 182.
Sagacious horses, 186.
Masques and Pageants, 187.
Royal Progresses, 193.
The stage, 201-226.
Dives, or evil genii of the Persians, ii. 303.
Dogberry, origin of the character of, ii. 589.
Donne (Dr.), critical notice of the poems of, i. 615.
Doublets, fashion of, in the age of Shakspeare, ii. 104, 105.
Douce (Mr.), beautiful version of a Christmas carol by, i. 200.
On the source of Shakspeare's Merchant of Venice, ii. 385,
386.
His vindication of Shakspeare's love of music, against Mr.
Steevens's flippant censures, 390.
Conjectures on the probable date of Shakspeare's Tempest,
504.
His "Illustrations of Shakspeare" cited, passim.
Dowricke (Anne), a minor poetess of the age of Shakspeare, i.
681.
Dragon, introduction of, into the May-games, i. 166.
Drake (Sir Francis), costly new year's gift of, to Queen Elizabeth,
ii. 99. note.
Tobacco first introduced into England by him, 135.
Drake (Lady), beautiful sonnet to, i. 621.
Drama, patronized by Elizabeth and her ministers, ii. 202. 205.
By private individuals, whose names they bore, 205.
And by James I., 206.
Dramatic Poets, remuneration of, in the time of Shakspeare, ii.
224, 225.
Dramatic Poetry, sketch of, from the birth of Shakspeare to the
period of his commencing a writer for the stage, i. 227.
Mysteries, moralities, and interludes, the first performances,
ibid.
Ferrex and Porrex, the first regular tragedy, ibid.
Gammar Gurton's Needle, the first regular comedy, ibid.
Dramatic Histories, 228.
Composite drama of Tarleton, 229.
Account of eminent dramatic poets during this period, 230-
251.
Conjectures as to the extent of Shakspeare's obligation to
his predecessors, 253-255.
Brief view of dramatic poetry, and its principal cultivators,
during Shakspeare's connection with the stage, ii. 556.
Account of the dramatic works of Fletcher, 557.
Massinger, 561.
Ford, 563.
Webster, 564.
Middleton, 565.
Decker, 566.
 Marston, 567.
Heywood, 568.
Chapman, 569.
Rowley, 570.
Other minor dramatic poets, 570, 571.
Ben Jonson, 572-580.
Drant (Thomas), a minor poet of the age of Shakspeare, i. 681.
Drayton (Michael), notice of, i. 615.
Critical remarks on his historical poetry, 615, 616.
On his topographical, epistolary, and pastoral poems,
616, 617.
And on his miscellaneous poetry, 617.
Poetical description by him of the dress, &c. of young
women, i. 83, 84.
Of Robin Hood, 159.
Of Tom the Piper, 164.
Sheep-shearing, 182.
Of the carbuncle, 397.
Encomium on Lilly's Euphues, 442.
Commendatory verses by, on Shakspeare's Rape of Lucrece,
ii. 39.
His tragedies, totally lost, 571.
Character of his Sonnets, ii. 56.
Dreams, considered as prognostics of good or evil, i. 354, 355.
Dress of country gentlemen, in Shakspeare's time, i. 82, 83.
 Of farmers or yeomen, 110.
Wedding dress of a rustic, 229.
Proper for anglers, 293. note.
Of the inhabitants of London, during the age of Shakspeare,
ii. 87-89.
Of Queen Elizabeth, 89, 91.
Of the ladies of that time, 91, 92. 100.
Of the gentlemen, 87, 88, 89. 101-109.
Of the citizen, 110, 111.
Of servants, 138.
Drinking of healths, origin of, i. 127, 128.
Drummond (William), biographical notice of, i. 617.
His merits as a poet, considered, 618.
Drunkenness, propensity of the English to, in the age of
Shakspeare, ii. 128, 129.
Dryden's testimony to the priority of Shakspeare's Pericles,
considered, ii. 280, 281.
Duelling, prevalence of, in the age of Shakspeare, ii. 158.
Dunlop (Mr.), opinion of on the source of Shakspeare's Romeo
and Juliet, ii. 360-362.
And of Measure for Measure, 453.
Durham, Easter gambols at, i. 148. note.
Dyer's "Fleece," illustration of, i. 183.
Dying, form of prayers for, i. 233.
 Superstitious notions concerning the last moments of
persons dying, i. 390, 391.

E
Earle (Bishop), character of his "Microcosmography," i. 511.
His portrait of an upstart country squire or knight, i. 84.
Of a country fellow, or clown, 120-122.
Earthquake of 1580, alluded to by Shakspeare, i. 52.
Account of, ibid. 53.
Easter-tide, festival of, i. 146.
Early rising on Easter Sunday, ibid.
Amusements, ibid.
Handball, 147, 148.
Presenting of eggs, 148.
Edgar, remarks on the assumed madness of, i. 588.
Contrast between his insanity and the madness of Lear, ii.
462. 464.
Education, state of, during Shakspeare's youth, i. 25-28.
Edwardes (C.), a minor poet of the age of Shakspeare, i. 681.
Edward (Richard), specimen of the poetical talents of, i. 713,
714.
Character of his dramatic compositions, ii. 231, 232.
Eggs, custom of giving, at Easter, i. 148.
Elderton (William), a minor poet of the age of Shakspeare, i.
681.
Elizabeth (Queen), school books commanded by, to be used, i.
26.
Visit of, to the Earl of Leicester, at Kenelworth Castle, 37, 38,
39. ii. 191-199.
Account of presents made to her on New-Year's Day, i. 125,
126.
Magnificent reception of her, at Norwich, 192. note.
Her wisdom in establishing the Flemings in this country, 192.
note.
A keen huntress, 285, 286.
Touched persons for the evil, 371.
Cultivated bibliography, 428.
The ladies of her court skilled in Greek equally with herself,
429.
Classical literature encouraged at her court, ibid. 431, 432.
Notice of her Prayer-book, 432.
Influence of her example, 433.
Notice of her works, 451.
Deeply skilled in Italian literature, ibid.
Notice of her poetical pieces, 704. note.
Proof that Shakspeare's Sonnets were not, and could not be
addressed to her, ii. 61, 73. note.
Instances of her vanity and love of dress, 90, 91.
Description of her dress, 89, 90.
Amount of her wardrobe, 91, 92.
 Silk stockings first worn by her, 98.
Costly New-Year's gifts made to her, 99.
Furniture of her palaces, 111, 112.
Description of the mode in which her table was served, 122,
123.
Her character as a sovereign, 145, 146.
Her industry, 146.
Instances of her vanity and coquetry, 147.
Affectation of youth, 148.
Artfulness, 149.
Extreme jealousy, 150.
Ill treatment of her courtiers, 150, 151.
Excelled in dancing, 172.
Delighted with bear-baiting, 176.
Account of her progresses, 193-199.
Passionately fond of dramatic performances, 202, 205.
Ordered Shakspeare's "Merry Wives of Windsor," 435.
And bestowed many marks of her favour upon him, 590.
Elfland or Fairy Land, description of, ii. 318, 319.
Elves or fairies of the Scandinavians, ii. 308.
Account of the Bright Elves, or benevolent fairies, 308, 309.
Of the Swart Elves, or malignant fairies, 309, 310.
And of the Scottish Elves, 314-336.
Elviden (Edmond), a minor poet of the age of Shakspeare, i.
681.
"England's Helicon," a collection of poems, critical notice of, i.
721-723.
English Language but little cultivated prior to the time of
Ascham, i. 439.
Improved by the labours of Wilson, 440.
Corrupted by Lilly, in the reign of Elizabeth, 441.
And by the interlarding of Latin quotations in that of
James I., 442.
This affectation satyrised by Sir Philip Sidney, 444,
445.
And by Shakspeare, 445, 446.
The English language improved by Sir Walter Raleigh and his
contemporaries, 446, 447.
Remarks on the prose writers of the reign of James I., 447,
448.
Notice of Mulcaster's labours for improving it, 455.
And of Bullokar's, ibid. 456.
English Mercury, the first newspaper ever published, i. 508.
Specimen of, ibid.
English nation, character of, ii. 154.
"Epicedium," a funeral song on the death of Lady Branch, ii. 38.
note.
Extract from, in commendation of Shakspeare's Rape of
Lucrece, 39. note.
Epilogue, concluded with prayer in the time of Shakspeare, ii.
222, 223.
Epitaph on Shakspeare, in Stratford church, ii. 619.
Epitaphs by Shakspeare:—a satirical one on Mr. Combe, ii. 605.
On Sir Thomas Stanley, 607.
And on Elias James, 607. note.
Erskine (Mr.) exquisite poetical allusions of, to fairy mythology, ii.
327, 328, 336.
Espousals, ceremony of, i. 220-223.
Essays, critical account of the writers of, in the age of Elizabeth,
i. 511-517.
Evans (Lewes and William), minor poets of the age of
Shakspeare, i. 682.
Evergreens, why carried at funerals, i. 239.
Evil spirits, supposed to be driven away by the sound of the
passing-bell, i. 232, 233.

F
Facetiæ, notice of writers of, during the age of Shakspeare, i.
515-517.
"Faerie Queene" of Spenser, critical remarks on, i. 646-649.
Fairefax (Edward), biographical notice of, i. 619.
Examination of his version of Tasso, ibid.
His original poetry lost, 620.
Fairies, superstitious traditions concerning, i. 320.
Their supposed influence on All-Hallow-Eve, 333.
Supposed to haunt fountains and wells, 392.
Critical account of the fairy mythology of Shakspeare, ii. 302.
Oriental fairies, 302, 303.
The knowledge of the oriental fairy mythology introduced
from the Italians, 303.
Origin of the Gothic system of fairy mythology, 304.
Known in England in the eleventh century, 306.
Scandinavian system of fairy mythology, 308-312.
Scandinavian system current in England in the thirteenth
century, 313.
Scottish elves, ibid. 314.
Their dress and weapons, 315.
Lowland fairies, 316.
Allusions to fairy superstitions by Chaucer, 313. 317.
Description of Elf or Fairy-land, 318, 319.
Allusions to it by various poets, 319-321.
Fairy processions at Roodsmass, 322.
Fairies in Scotland supposed to appear most commonly by
moonlight, 323.
Their supposed influence on pregnant women, 324.
Children said to be stolen and changed by them, 325,
326.
Expedients for recovering them, 326, 327.
Their speech, food, and work, 328, 329.
 Account of the malignant fairy called the Wee Brown Man of
the Muirs, 329, 330.
Traditions relative to the benevolent sprite, Brownie, 330-
336.
The fairy mythology of Shakspeare, though partly founded
on Scottish tradition, yet, from its novelty and poetic beauty,
meriting the title of the English System, 337, 338.
Critical illustrations of his allusions to fairies and Fairy-
land, 337-353.
Scandinavia the parent of our popular fairy mythology, which
has undergone various modifications, 353-355.
Fairs, how celebrated antiently, i. 214-216.
Falconer, an important officer in the households of the great, i.
265, 266.
His qualifications, 266.
Falconry, when introduced into England, i. 255.
Universal among the nobility and gentry, ibid. 256.
Notices of books on, 257. note.
Falconry an expensive diversion, 257-259.
Prohibited to the clergy, 259. note.
Remarks on this sport, 260-262.
Poetical description of it by Massinger, 262, 263.
A favourite diversion of the ladies, 265.
Falcons, different sorts of, i. 263, 264.
Account of their training, 266-271.
Falstaff, analysis of the character of, as introduced in
Shakspeare's plays of Henry IV., Parts I. and II., ii. 381-384.
And in the Merry Wives of Windsor, 436.
Fans, structure and fashion of, in the age of Shakspeare, ii. 98,
99.
Fare of country squires in the age of Shakspeare, i. 73, 76.
Of country gentlemen, 79, 80.
And of the sovereign and higher classes, ii. 120-129.
Farmers, character of, in the time of Edward VI., i. 100, 101.
In Queen Elizabeth's time, 98.
Description of their houses or cottages, 99, 100.
Their furniture and household accommodations, 101.
103.
Their ordinary diet, 103-108.
Diet on festivals, 109.
Dress, 110.
Qualifications of a good farmer's wife, 111, 112.
Occupations, &c. of their servants, 113.
Manners, &c. of Scottish farmers during the same period,
117, 118.
Progress of extravagance among this class of persons, 119.
Farmer (Dr.), conclusion of, as to the result of Shakspeare's
school education, i. 29, 30.
His conclusion controverted, 30, 31.
 His opinion as to the extent of Shakspeare's knowledge of
French and Italian literature considered, 54-56, 57.
Faulconbridge, analysis of the character of, ii. 120.
Feasts (ordinary), curious directions for, i. 80. note.
Felton's portrait of Shakspeare, authenticity of, ii. 623.
Fenner (Dudley), a minor poet of the age of Shakspeare, i. 682.
Fenton's (Geffray), account of his "Certain Tragicall Discourses,"
a popular collection of Italian novels, i. 542.
Fern-seed, supposed to be visible on Midsummer-Eve, i. 329.
"Ferrex and Porrex," the first regular tragedy ever performed in
England, i. 227.
Ferrers (George), a minor poet of the age of Shakspeare, i. 682.
Ferriar (Dr.), theory of apparitions of, ii. 406.
Application of it to the character of Hamlet, 407.
His opinion of the merits of Massinger as a dramatic poet
controverted, 562.
Festivals, account of those observed in Shakspeare's time, i.
123.
New-Year's Day, 123-126.
Twelfth Day, 127-134.
St. Distaff's Day, 135.
Plough Monday, 136-138.
Candlemas Day, 138-140.
Shrove Tide, 141-145.
Easter Tide, 146-148.
 Hock Day, 149-151.
May Day, 152-174.
Whitsuntide, 175-180.
Sheep-shearing, 181-185.
Harvest-home, 185-190.
Martinmas, 192.
Christmas, 193-208.
Wakes or fairs, 209-249.
Weddings, 219-229.
Christenings, 230, 231.
Burials, 232-245.
Fete, magnificent, at Kenelworth Castle, given to Queen
Elizabeth, i. 37-39.
Fetherstone (Christopher), a minor poet of the age of
Shakspeare, i. 682.
Fires kindled on Midsummer-Eve, of Pagan origin, i. 328, 329;
and on All-Hallow-Eve, 341.
Fire Spirits, machinery of, introduced in the Tempest, ii. 521,
522.
Fishing, pursued with avidity, in the 16th century, i. 289.
Account of books on this sport, 290, 291.
Poetical description of, 292, 293.
Qualifications requisite for, 294-297.
Fitzgeffrey (Charles), Biographical notice of, i. 620.
Specimen of his poetical talents, 621.
Fitzherbert (Sir Anthony), notice of his agricultural treatises, i.
115. note.
His precepts to a good housewife, 116, 117. notes.
Fleming (Abraham), a miscellaneous writer, account of, i. 504.
Character of his style, 505.
Poems of, 682.
Fletcher (Robert), a minor poet of the age of Shakspeare, i.
682.
Fletcher (Giles), critical remarks on the poetry of, i. 621, 622.
Fletcher (Phineas), notice of, i. 622.
Critical observations on his "Purple Island," 623.;
and on his "Piscatory Eclogues," ib.
Fletcher (John), the chief author of the plays extant under his
name, ii. 557.
How far he was assisted by Beaumont, 558.
Critical estimate of his character as a dramatic poet, 558-
560.
His feeble attempts to emulate Shakspeare, 560, 561.
His Faithful Shepherdess (act v. sc. 1.) illustrated, i. 130.
See also Beaumont, in this index.
Floralia (Roman), perpetuated in May-Day, i. 152.
Florio (John), pedantry of, satyrised by Shakspeare, i. 415.
Appointed reader of the Italian language to the Queen of
James I., 451.
Flowers, antiently scattered on streams at sheep-shearing time,
i. 185.
Garlands of flowers carried at funerals, and buried with the
deceased, 240-242.
Graves in Wales still decorated with flowers, 242-244.
Allusions to this custom by Shakspeare, 243.
Fools of Shakspeare's plays, &c. remarks on, i. 587. ii. 550.
Description of their apparel and condition, ii. 141, 142.
Apes or monkies kept as companions for them, 145, 146.
Ford, merits of, as a dramatic poet, considered, ii. 563, 564.
Forks, when introduced into England, ii. 126.
Fortescue's (Thomas), "Forest of Historyes," a popular collection
of novels, notice of, i. 543.
"Fortune my Foe," a popular song, quoted by Shakspeare, i.
477.
Fountains and wells, why superstitiously visited, i. 391.
Supposed to be the haunts of fairies and spirits, 392.
Pilgrimages made to them, 393.
Fowling, how pursued in the sixteenth century, i. 287-289.
Fox's "Acts and Monuments," character of, i. 482.
Fraunce (Abraham), notice of his "Arcadian Rhetoricke," i. 464.
List of his poetical works, 682, 683.
Freeman (Thomas), a minor poet of the age of Shakspeare, i.
683.
French Language, Shakspeare's knowledge of, when acquired, i.
53, 54.
Proofs that he had some acquaintance with it, 55, 56.
List of French grammars which he might have read, 57.
"Friar of Orders Grey," a beautiful ballad, notice of, i. 579, 580.
Quoted by Shakspeare, 589, 590.
Friend, absence from, exquisitely pourtrayed by Shakspeare, ii.
78.
Friendship, beautiful delineation of, ii. 389.
Fulbeck's account of Roman factions, i. 476.
Fulbroke Park, the scene of Shakspeare's deer-stealing, i. 402,
403.
Fuller (Thomas), character of Shakspeare, i. 29.;
and of Dr. Dee, and his assistant Kelly, ii. 512, 513.
Fullwell (Ulpian), a minor poet of the age of Shakspeare, i. 683.
Funeral ceremonies described, i. 232-237.
Entertainments given on those occasions, 238.
Furniture, splendid, of Queen Elizabeth's palaces, ii. 111, 112.
Of the inhabitants of London, 112-120.
Of the halls of country gentlemen, i. 77-79.
Fuseli's picture of the night-mare, description of, i. 348. note
[348:B].

G
Gale (Dunstan), a minor poet of the age of Shakspeare, i. 683.
Gamage (William), a minor poet of the age of Shakspeare, i.
684, and note [684:A].
Games (Cotswold), account of, i. 252-254.
Gaming, prevalence of, in the age of Shakspeare, ii. 157, 158.
"Gammer Gurton's Needle," illustration of, i. 106.
The earliest comedy ever written or performed in England, ii.
227.
Critical remarks on, 233.
Garlands, anciently used at funerals, and buried with the
deceased, i. 240-242.
Garnier's Henriade probably seen by Shakspeare, i. 54, 55.
Garter (Barnard), a minor poet of the age of Shakspeare, i. 684.
Garter (Thomas), a dramatic poet in the reign of Elizabeth,
character of, ii. 235.
Gascoigne (George), notice of the "Posies" of, i. 461.
Biographical sketch of, 623, 624.
Remarks on his poetry, 624, 625.
Character of, as a dramatic poet, ii. 233, 234.
Gastrell (Rev. Francis), purchases Shakspeare's house at
Stratford, ii. 584. note.
Cuts down his mulberry tree, ibid.
And destroys the house itself, 585. note.
Gay's Trivia, quotation from, on the influence of particular days,
i. 323. note.
Poetical description of spells, 332.
Genius of Shakspeare's drama considered, ii. 536-541.
Gentlemen, different sorts of, in the age of Shakspeare, i. 69.
Their virtues and vices, ibid. 70.
Description of the mansion houses of country gentlemen,
72-74.
Their usual fare, 79, 80-82.
Employments and dress of their daughters, 83, 84.
Character of country gentlemen towards the commencement
of the 17th century, 84, 85.
When they began to desert their halls for the metropolis, 85.
Portraits of, in the close of the 17th, and at the beginning of
the 18th century, 86, 87. notes.
Dress of gentlemen in the metropolis, ii. 87, 88, 89. 101-109.
Gerbelius (Nicholas), rapturous declamation of, on the
restoration of some Greek authors, i. 435.
Gerguntum, a fabulous Briton, notice of, i. 192. note.
Germans, fairy mythology of, ii. 312.
Gesta Romanorum, a popular romance in Shakspeare's time, i.
534.
Different translations of the continental Gesta, ibid. 535.
Critical account of the English Gesta, 535, 536. ii. 386.
Notice of its different editions, i. 537, 538.
Long continuance of its popularity, 538.
Ghosts, superstitious notions concerning, prevalent in the age of
Shakspeare, i. 318, 319.
 Remarks on the supposed agency of ghosts, as received at
that time, ii. 399-405.
Considerations on the introduction of the ghost in Hamlet,
and its strict consonance to the popular superstitions shewn,
411-417.
Its superiority over all other ghostly representations, ancient
or modern, 417, 418.
Gifford (Humphrey), a minor poet of the age of Shakspeare, i.
684.
Gifford (Mr.), conjecture of, on the date of Shakspeare's Henry
VIII. ii. 442, 443.
Observations on the excellent plan of his notes on Massinger,
561. note.
His estimate of the merits of Ben Jonson, as a dramatic
poet, 575, 576.
Vindicates Jonson from the cavils of Mr. Malone, 578. note.
Gilchrist (Mr.) on the character of Puttenham's "Arte of English
Poesie," i. 466.
Gleek, a fashionable game at cards, notice of, ii. 170.
Glen Banchar, anecdote of a peasant of, i. 233, 234.
Globe Theatre, license to Shakspeare for, ii. 207, 208.
Account of it, 208, 209.
Description of its interior, 210-214.
Gloves, costly, presented to Elizabeth, ii. 99.
Goblins and spectres, superstitious notions concerning, i. 316,
317.
Machinery of goblins or spirits of earth, introduced into the
Tempest, ii. 523, 524.
Goder Norner, or beneficent elves of the Goths, notice of, ii.
308.
Godwin (Mr.), remarks of, on Shakspeare's Troilus and Cressida,
ii. 440, 441.
His estimate of the merits of Ben Jonson, as a dramatic
poet, 574-579.
Golding (Arthur), a minor poet of the age of Shakspeare, i. 684.
Googe (Barnaby), description of Midsummer-Eve superstitions, i.
328.
Notice of his poetical works, 684.
Gorboduc, critical remarks on Sackville's tragedy of, ii. 230, 231.
Gordon (Patrick), a minor poet of the age of Shakspeare, i. 684.
"Gorgeous Gallery of Gallant Inventions," a collection of poems,
critical account of, i. 715-717.
Gorges (Sir Arthur), a minor poet of the age of Shakspeare, i.
684, 685. and notes.
Gossipping, prevalence of, in the age of Shakspeare, ii. 159,
160.
Gosson (Stephen), a Puritanical wit, in Shakspeare's time,
account of, i. 500, 501.
Notice of his "Speculum humanum," 685. and note [685:C].
Gowns, materials and fashions of, in the age of Shakspeare, ii.
97, 98.
Grammars and dictionaries, list of, in use in Shakspeare's time, i.
25. note.
Henry VII.'s grammar learned by Shakspeare, 26.
The English grammar but little cultivated, previous to the
time of Ascham, 439.
Improved by him, ibid.;
and by Wilson, 440.
Notice of eminent Latin grammarians, 454, 455.
English grammar of Ben Jonson, 456.
Grange (John), a minor poet of the age of Shakspeare, i. 685.
Grant (Edward), an eminent Latin philologer, notice of, i. 454.
Graves, why planted with flowers, i. 242-244. and note.
Allusions to this custom by Shakspeare, 243.
Grave-digger in Hamlet, songs mis-quoted by, probably by
design, i. 591.
Greek literature, cultivated and encouraged at the court of
Queen Elizabeth, i. 429-431, 432.
Promoted essentially by the labours of Sir Thomas Smith, Sir
Henry Savile, and Dr. Boys, 453, 454.
List of Greek authors, translated into English in the time of
Shakspeare, 483.
Greene (Thomas), the barrister, an intimate friend of
Shakspeare's, ii. 600.
Greene (Thomas), the player, notice of, i. 417.
Character of, ibid.
Whether a townsman and relation of Shakspeare, 420.
Greene (Thomas), a minor poet of the age of Shakspeare, i.
685.
Greene (Robert), a miscellaneous writer in the time of
Shakspeare, biographical account of, i. 486.
Studies and dissipations of his early years, 486, 487.
His marriage, 487.
Pleasing sketch of his domestic life, 488.
Returns to the dissipations of the metropolis, 489.
Affectionate demeanour of his wife, 490.
His beautiful address, "By a Mother to her Infant," 492, 493.
Becomes a writer for bread, 494.
Character of Greene as a prose writer, 494.
List of his principal pieces, 495.
Poetical extract from his "Never Too Late," 496.
Extract entitled "The Farewell of a Friend," 497.
His death, ibid.
Miserable state of his latter days, 498.
Satirical sonnet addressed to him, 499.
Critical notice of his poetry, 627.
List of his dramatic productions, with remarks, ii. 249-251.
"Green Sleeves," a popular song, quoted by Shakspeare, i. 477.