Professional Documents
Culture Documents
PDF Proof Complexity Encyclopedia of Mathematics and Its Applications 1St Edition Jan Krajicek Ebook Full Chapter
PDF Proof Complexity Encyclopedia of Mathematics and Its Applications 1St Edition Jan Krajicek Ebook Full Chapter
https://textbookfull.com/product/mathematics-of-the-bond-market-
a-levy-processes-approach-encyclopedia-of-mathematics-and-its-
applications-1st-edition-michal-barski/
https://textbookfull.com/product/quasi-hopf-algebras-a-
categorical-approach-encyclopedia-of-mathematics-and-its-
applications-1st-edition-daniel-bulacu/
https://textbookfull.com/product/discrete-mathematics-and-its-
applications-8th-edition-rosen/
https://textbookfull.com/product/crossing-numbers-of-graphs-
discrete-mathematics-and-its-applications-1st-edition-marcus-
schaefer/
R Programming and Its Applications in Financial
Mathematics 1st Edition Shuichi Ohsaki
https://textbookfull.com/product/r-programming-and-its-
applications-in-financial-mathematics-1st-edition-shuichi-ohsaki/
https://textbookfull.com/product/combinatorics-and-number-theory-
of-counting-sequences-discrete-mathematics-and-its-
applications-1st-edition-istvan-mezo/
https://textbookfull.com/product/discrete-mathematics-via-logic-
and-proof-calvin-jongsma/
https://textbookfull.com/product/systems-of-nonlinear-partial-
differential-equations-applications-to-biology-and-engineering-
mathematics-and-its-applications-a-w-leung/
https://textbookfull.com/product/proof-technology-in-mathematics-
research-and-teaching-gila-hanna/
PROOF COMPLEXITY
This series is devoted to significant topics or themes that have wide application in
mathematics or mathematical science and for which a detailed development of the
abstract theory is less important than a thorough and concrete exploration of the
implications and applications.
All the titles listed below can be obtained from good booksellers or from Cambridge
University Press. For a complete series listing visit
www.cambridge.org/mathematics.
122 S. Khrushchev Orthogonal Polynomials and Continued Fractions
123 H. Nagamochi and T. Ibaraki Algorithmic Aspects of Graph Connectivity
124 F. W. King Hilbert Transforms I
125 F. W. King Hilbert Transforms II
126 O. Calin and D.-C. Chang Sub-Riemannian Geometry
127 M. Grabisch et al. Aggregation Functions
128 L. W. Beineke and R. J. Wilson (eds.) with J. L. Gross and T. W. Tucker Topics in Topological
Graph Theory
129 J. Berstel, D. Perrin and C. Reutenauer Codes and Automata
130 T. G. Faticoni Modules over Endomorphism Rings
131 H. Morimoto Stochastic Control and Mathematical Modeling
132 G. Schmidt Relational Mathematics
133 P. Kornerup and D. W. Matula Finite Precision Number Systems and Arithmetic
134 Y. Crama and P. L. Hammer (eds.) Boolean Models and Methods in Mathematics, Computer Science,
and Engineering
135 V. Berthé and M. Rigo (eds.) Combinatorics, Automata and Number Theory
136 A. Kristály, V. D. Rădulescu and C. Varga Variational Principles in Mathematical Physics, Geometry,
and Economics
137 J. Berstel and C. Reutenauer Noncommutative Rational Series with Applications
138 B. Courcelle and J. Engelfriet Graph Structure and Monadic Second-Order Logic
139 M. Fiedler Matrices and Graphs in Geometry
140 N. Vakil Real Analysis through Modern Infinitesimals
141 R. B. Paris Hadamard Expansions and Hyperasymptotic Evaluation
142 Y. Crama and P. L. Hammer Boolean Functions
143 A. Arapostathis, V. S. Borkar and M. K. Ghosh Ergodic Control of Diffusion Processes
144 N. Caspard, B. Leclerc and B. Monjardet Finite Ordered Sets
145 D. Z. Arov and H. Dym Bitangential Direct and Inverse Problems for Systems of Integral and
Differential Equations
146 G. Dassios Ellipsoidal Harmonics
147 L. W. Beineke and R. J. Wilson (eds.) with O. R. Oellermann Topics in Structural Graph Theory
148 L. Berlyand, A. G. Kolpakov and A. Novikov Introduction to the Network Approximation Method for
Materials Modeling
149 M. Baake and U. Grimm Aperiodic Order I: A Mathematical Invitation
150 J. Borwein et al. Lattice Sums Then and Now
151 R. Schneider Convex Bodies: The BrunnMinkowski Theory (Second Edition)
152 G. Da Prato and J. Zabczyk Stochastic Equations in Infinite Dimensions (Second Edition)
153 D. Hofmann, G. J. Seal and W. Tholen (eds.) Monoidal Topology
154 M. Cabrera Garcı́a and Á. Rodrı́guez Palacios Non-Associative Normed Algebras I: The Vidav–Palmer
and Gelfand–Naimark Theorems
155 C. F. Dunkl and Y. Xu Orthogonal Polynomials of Several Variables (Second Edition)
156 L. W. Beineke and R. J. Wilson (eds.) with B. Toft Topics in Chromatic Graph Theory
157 T. Mora Solving Polynomial Equation Systems III: Algebraic Solving
158 T. Mora Solving Polynomial Equation Systems IV: Buchberger Theory and Beyond
159 V. Berthé and M. Rigo (eds.) Combinatorics, Words and Symbolic Dynamics
160 B. Rubin Introduction to Radon Transforms: With Elements of Fractional Calculus and Harmonic Analysis
161 M. Ghergu and S. D. Taliaferro Isolated Singularities in Partial Differential Inequalities
162 G. Molica Bisci, V. D. Radulescu and R. Servadei Variational Methods for Nonlocal Fractional Problems
163 S. Wagon The Banach–Tarski Paradox (Second Edition)
164 K. Broughan Equivalents of the Riemann Hypothesis I: Arithmetic Equivalents
165 K. Broughan Equivalents of the Riemann Hypothesis II: Analytic Equivalents
166 M. Baake and U. Grimm (eds.) Aperiodic Order II: Crystallography and Almost Periodicity
167 M. Cabrera Garcı́a and Á. Rodrı́guez Palacios Non-Associative Normed Algebras II: Representation
Theory and the Zel’manov Approach
168 A. Yu. Khrennikov, S. V. Kozyrev and W. A. Zúñiga-Galindo Ultrametric Pseudodifferential Equations
and Applications
169 S. R. Finch Mathematical Constants II
170 J. Krajı́ček Proof Complexity
171 D. Bulacu, S. Caenepeel, F. Panaite and F. Van Oystaeyen Quasi-Hopf Algebras
E N C Y C L O P E D I A O F M AT H E M AT I C S A N D I T S A P P L I C AT I O N S
Proof Complexity
J A N K R A J Í Č E K
Charles Univesity, Prague
University Printing House, Cambridge CB2 8BS, United Kingdom
One Liberty Plaza, 20th Floor, New York, NY 10006, USA
477 Williamstown Road, Port Melbourne, VIC 3207, Australia
314–321, 3rd Floor, Plot 3, Splendor Forum, Jasola District Centre, New Delhi – 110025, India
79 Anson Road, #06–04/06, Singapore 079906
www.cambridge.org
Information on this title: www.cambridge.org/9781108416849
DOI: 10.1017/9781108242066
© Jan Krajı́ček 2019
This publication is in copyright. Subject to statutory exception
and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without the written
permission of Cambridge University Press.
First published 2019
Printed and bound in Great Britain by Clays Ltd., Elcograf S.P.A.
A catalogue record for this publication is available from the British Library.
Library of Congress Cataloging-in-Publication Data
Names: Krajı́ček, Jan, author.
Title: Proof complexity / Jan Krajı́ček.
Description: Cambridge ; New York, NY : Cambridge University Press, 2019. |
Series: Encyclopedia of mathematics and its applications ; 170 | Includes
bibliographical references and index.
Identifiers: LCCN 2018042527 | ISBN 9781108416849 (hardback : alk. paper)
Subjects: LCSH: Proof theory. | Computational complexity.
Classification: LCC QA9.54 .K72 2019 | DDC 511.3/6–dc23
LC record available at https://lccn.loc.gov/2018042527
ISBN 978-1-108-41684-9 Hardback
Cambridge University Press has no responsibility for the persistence or accuracy
of URLs for external or third-party internet websites referred to in this publication
and does not guarantee that any content on such websites is, or will remain,
accurate or appropriate.
Let no one ignorant of logic enter.
Introduction 1
4.3 Skolemization of G 87
4.4 Bibliographical and Other Remarks 91
5 Resolution 93
5.1 The Resolution Rule and Its Completeness 93
5.2 R∗ and the DPLL Procedure 95
5.3 Regular Resolution and Read-Once Branching Programs 97
5.4 Width and Size 98
5.5 Height and Space 102
5.6 Interpolation 106
5.7 DNF-Resolution 108
5.8 Extended Resolution 111
5.9 Bibliographical and Other Remarks 112
6 Algebraic and Geometric Proof Systems 115
6.1 Equational Calculus 116
6.2 Algebraic Geometry and the Nullstellensatz 117
6.3 Integer Linear Programming and the Cutting Plane Method 121
6.4 Semi-Algebraic Geometry and the Positivstellensatz 124
6.5 Between the Discrete and the Continuum 128
6.6 Beyond Propositional Logic 130
6.7 Bibliographical and Other Remarks 131
7 Further Proof Systems 134
7.1 Combined Proof Systems 134
7.2 Circuit and WPHP Frege Systems 138
7.3 Implicit Proof Systems 141
7.4 Auxiliary Logic Proof Systems 145
7.5 Auxiliary Algebraic Proof Systems 149
7.6 Outside the Box 154
7.7 Bibliographical and Other Remarks 157
Bibliography 481
Special Symbols 506
Index 510
Preface
The concept of propositional formulas and their logical validity developed over
almost two and a half millennia and reached its current mathematical definition at
the advent of mathematical logic in the second half of the nineteenth century. It is
remarkably simple. We have propositional atoms and a few logical connectives to
combine them into formulas. We identify truth assignments with functions assign-
ing the values (true) and ⊥ (false) to atoms and use simple algebraic rules to
quickly compute the truth value of any formula under a particular truth assignment.
Logically valid formulas, tautologies, are then those formulas receiving the value
under any truth assignments. The definition of logical validity offers a mechanical
way to decide whether a formula is a tautology: compute its truth value under all
truth assignments to its atoms. The definition is amenable to countless variations,
leading to formulations of various non-classical logics. In some of these alternative
definitions a truth assignment may be a more complicated notion than in others but
the idea of determining the truth value bottom up by a few rules is common to all.
The preoccupation of logic over the ages with logically sound modes of reasoning
has led to various lists of logically valid inference schemes and eventually to the
concept of a logical calculus. Aristotle and Frege are the main figures at the advent
of logic and of mathematical logic, respectively. In such calculi one starts with initial
tautologies, axioms, of a small number of transparent forms and uses a finite number
of inference rules to derive increasingly more complicated tautologies. Typically one
can simulate the mechanical process of going through all truth assignments by such
a derivation.
A property of these specific derivations simulating computations is that they do
not involve formulas more complicated than the one being proved. But one may
prove a tautology entirely differently, deriving first intermediate formulas (lemmas),
seemingly having nothing to do with the target formula, and eventually combining
them into a valid derivation. It is a fundamental problem of logic to understand what
a general derivation looks like and, in particular, when it can be significantly shorter
than a trivial mechanical derivation and how can we find it.
xiv Preface
When the truth values are represented by the bits 1 and 0, truth assignments
can be identified with binary strings, truth functions with the properties of such
strings and propositional formulas become algorithms for computing them. Thus,
at once we find propositional logic at the heart of the theory of computing and the
above questions about the efficiency of proofs become fundamental problems of
computational complexity theory such as the P vs. NP problem or the NP vs. coNP
problem.
The mathematical area studying the complexity of propositional proofs, proof
complexity, provides at present one of the richest approaches to these problems. It
is proof complexity which, together with its logical foundations, I emphasize in this
book. Proof complexity has, in my view, at least as good a chance of achieving
a significant breakthrough on the above-mentioned problems as any other approach
contemplated at present. And, contrary to some other, even partial results, often lower
bounds for specific proof systems, can be genuinely interpreted as saying something
interesting and relevant about these problems.
We should also keep an open mind about how the eventual answers may turn out.
Mathematics can be seen as a special kind of language developed for the world of
virtual objects like points and lines, numbers and equations, sets, spaces, algebras
and other occupants of this mathematical world and for our thoughts about them.
This language is quite powerful and is akin in many respects to natural languages.
And just as a talented poet can describe in few lines or even in few words a thought
or a situation for which you or I may need volumes, miraculously short and eloquent
proofs which nobody would have dreamt of may exist. I think that at present we have
only a tentative understanding of the power of the language of mathematics.
Acknowledgements
I am grateful to Neil Thapen for a number of discussions about the book and about
various topics covered in it, and for extensive comments on the draft. I am indebted
to Pavel Hrubeš, Moritz Müller, Igor C. Oliveira, Ján Pich and Pavel Pudlák for
offering comments or suggestions concerning significant parts of the draft.
I thank Albert Atserias, Olaf Beyersdorff, Sam Buss, Steve Cook, Gilda Ferreira,
Michal Garlı́k, Azza Gaysin, Emil Jeřábek, Leszek Kolodziejczyk, Tomáš Krňák,
Jakob Nordström, George Osipov, Theodoros Papamakarios, Zenon Sadowski and
Iddo Tzameret for pointing out minor errors in the draft or for answering my ques-
tions and providing bibliographical information.
Introduction
the spectrum problem in the model theory of Scholz [462] and Asser [25] (in the
early 1950s), which, as we now know, asks for a characterization of NE sets (i.e.
non-deterministic time 2O(n) sets) and, in particular, whether the class NE is closed
under complementation. They also include problems involving the rudimentary sets
of Smullyan [476] and Bennett [65] (in the early 1960s). The first problem explicitly
mentioning time complexity was perhaps the problem posed by Gödel to von Neu-
mann in a 1956 letter [201]; there he asks about the time complexity of a problem
we now know to be NP-complete (see the introduction to the volume of Clote and
Krajı́ček [144] for the letter and its translation, and Buss [111] for the completeness
result). Sipser [468] describes these developments clearly and succinctly.
In the light of this history it seems plausible that the P vs. NP problem has signifi-
cant logical facets connected to the foundations of mathematics. It can even be seen
as a miniaturization of the Entscheidungsproblem: replace first-order formulas by
propositional formulas and ask for a feasible algorithm rather than just any algorithm.
We do not know how hard the problem is. It may require just one trick (as some
famous problems of the past did), one dramatically new idea (as was forcing for
the independence of the continuum hypothesis) or a development of a whole part
of mathematics (as was needed for Fermat’s last theorem). In either case it seems
sensible to try to develop an understanding of the logical side of the problem.
This thinking is, however, foreign to the canons of contemporary computational
complexity theory. That field developed as a part of combinatorics and some early
successes stimulated a lot of faith in that approach. However, although many inter-
esting developments have taken place, progress on combinatorial insight into the
original fundamental problems remains tentative. Despite this, complexity theorists
sometimes go to great lengths to talk about logic without ever mentioning the word.
Although it is remarkable that one can discuss provability or unprovability without
using words like “a theory” or “a formula,” it is hard to imagine that doing so is
faithful to the topic at hand.
There are several research areas connecting mathematical logic and computational
complexity theory; some of these connections are tighter than others. We shall con-
centrate on propositional proof complexity. Proof complexity (the adjective proposi-
tional is often left out, as it is tacitly understood) is now a fairly rich subject drawing
its methods from many fields (logic, combinatorics, algebra, computer science, . . . ).
The first results appeared in the 1960s, with a significant acceleration from the late
1980s to the early 2000s. For some years now, however, progress has lain more in
sharpening and generalizing earlier advances and recasting them in a new formalism
and in improving the technical tools at hand, but not in fact that much (at least not
at the level of the progress in the 1990s) in developing completely new ideas. It
thus seems a good time to write an up-to-date presentation of the field to enable
newcomers to learn the subject and to allow active researchers to step back and
contemplate a larger picture.
There are two approaches to proof complexity, which are complementary, in a
sense, to each other. The first views problems, especially length-of-proof problems,
Introduction 3
as purely combinatorial questions about the existence and properties of various com-
binatorial configurations. Prominent early examples of this approach are Tseitin’s
[492] 1968 lower bound for regular resolution or Haken’s [216] 1985 lower bound
for general resolution proofs of the pigeonhole principle. This approach, when suc-
cessful, typically gives qualitatively very detailed results. It works only for weak
proof systems however.
The second approach views problems through the eyes of mathematical logic.
Representative examples of this view are Cook’s [149] 1975 and Paris and Wilkie’s
[389] 1985 translations of bounded arithmetic proofs into (sequences of) proposi-
tional proofs, Ajtai’s [5] 1988 lower bound, obtained by forcing, for pigeonhole-
principle proofs in constant-depth Frege systems or my 1991 idea of the feasible
interpolation method formulated in [276] (1994, preprint circulating in 1991). This
approach typically yields a general concept, and its applications in specific cases
require combinatorics as well.
The two approaches are not disjoint and can be fruitfully combined. Nevertheless,
there has been much emphasis in recent years on a purely combinatorial analysis
of proof systems related to specific SAT-solving or optimization algorithms. Only
a small number of researchers have involved themselves in mathematical logic or
investigations aimed at explaining some general phenomena. Of course, the original
fundamental – logic-motivated – problems are dutifully mentioned when the goals
in this area are discussed in talks or papers, but in actual research programs these
problems are rarely studied. It seems that not only are the original problems aban-
doned but some of the proof complexity theory developed around them is becoming
exotic to many researchers. I suppose this is due in part to the demise of logic from a
standard computer science curriculum. Being able to formulate a theorem or a proof
does not provide sufficient education in logic, just as being able to draw a graph
is not a sufficient education in combinatorics. Without at least an elementary logic
background the only part of proof complexity that remains accessible is the algorithm
analysis aspect. Of course, it is a perfectly sound research topic but it is not the whole
of proof complexity; it is not even close to half of it.
I think that logic will play eventually a greater role also in proof complexity
research motivated by applications. I am not competent to judge how significant a
contribution current proof complexity makes to the analysis and practice of real-
life algorithms. Such contributions often contain ingenious technical innovations
and ought to shed light on the performance of the respective algorithms. But SAT
solving (or the design of other algorithms) is mostly an engineering activity, with
little theory, as we mathematicians understand it, involved. Even the basic issue
of how to compare the efficiency of two algorithms is in reality approached in an
ad hoc, heuristic, mathematically non-principled way. Therefore I think that proof
complexity could contribute most by introducing new original theoretical concepts
and ideas that could influence how the whole field of SAT solving is approached in
future. A significant contribution may come even from technically simple ideas, for
example, how to use stronger proof systems in proof search or how to combine the
4 Introduction
present-day optimization algorithms with a little logic. This could be, in my opinion,
quite analogous to the fact that fairly general ideas of Turing and others influenced
the thinking in computer science significantly. But from where else could genuinely
new concepts and ideas, and not just technical innovations, come other than from
considering the very foundational problems of the field?
In the book I stress my view of proof complexity as a whole entity rather than as
a collection of various topics held together loosely by a few notions. The frame that
supports it is logic. This is not a dogma but represents the current state of affairs. For
this reason I choose as the motto at the start of the book a paraphrase of the famous
sign on the gate to Plato’s Academy, in order to underline the point, not to distress
the reader.
Basic concepts, classical results, current work and possible future directions are
presented. I have not attempted to compile a compendium of all known results,
and have concentrated more on ideas than on the technically strongest statements.
These ideas might be enlightening definitions and concepts, stimulating problems
or original proof methods. I have given preference to statements that attempt some
generality over statements that are strong in specific situations only.
The intended readers include researchers and doctoral students in mathematics
(mathematical logic and discrete mathematics, in particular) and in theoretical com-
puter science (in computational complexity theory, in particular). I do not necessarily
present the material in the most elementary way. But I try to present it honestly,
meaning that I attempt to expose the fundamental ideas underlying the topics and do
not hide difficult or technically delicate things under the carpet. The book is meant to
be self-contained, in a reasonable sense of the word, as a whole; it is not a collection
of self-contained chapters and there is a lot of cross-referencing.
address all proof systems. It is here where there is, I think, a chance to develop some
deeper understanding of the fundamental problems. I present there topics for which
some general theory has emerged in the past and discuss directions that seem to me
to be promising for future research. In the last chapter we step back a little, look at
what has been achieved and offer a few thoughts on the nature of proof complexity.
Each part is divided into chapters, which further divide into sections with, I hope,
informative titles. This gives a structure to the material that should be apparent at
first glance. There are various novel results, topics, constructions and proofs.
In the main narrative, typically I attribute to sources only the main results, con-
cepts, problems and ideas discussed. The qualification main is subjective, of course,
and does not necessarily refer only to the technically hardest topics but sometimes
also to simple observations that have proved to be very useful: I rather agree with
the classical dictum that the really important ideas are also simple. Each chapter
ends with a section on bibliographical and other remarks, where I attempt to give
full bibliographical information for all the results covered, point to other relevant
literature and discuss at least some topics that are related but were omitted.
Occasionally I use the adjectives interesting and important and similar qualifiers.
This should be understood as meaning interesting (or important) for me. In fact, it
is a dangerous illusion, to which some people succumb, that these words have an
objective meaning in mathematical literature.
makes the structure of a long text more comprehensible. All lemmas, theorems and
corollaries are syntactically distinguished and numbered in the conventional way.
All unfamiliar symbols are properly defined in the text and the special-symbol
index near the end of the book lists for each such symbol the section where it appears
first.
Background
Although I review some elements of propositional and first-order logic in the first two
sections of Chapter 1, this is mostly to set the scene and to introduce some formal-
Introduction 7
ism and notation. It is helpful if the reader has a basic knowledge of mathematical
logic at the level of an introductory course in first-order logic and model theory.
This should include the completeness and compactness theorems and Herbrand’s
theorem (although I do present a proof of this theorem). I do not assume any specific
knowledge of bounded arithmetic or the model theory of arithmetic.
From computational complexity theory I assume a knowledge of the basic con-
cepts of Turing machines and their time complexity and of the definitions of standard
complexity classes (although some of this is briefly recalled in Section 1.3). I do not
assume a particular knowledge of circuit complexity (it is explained when needed)
but it could make some ideas and arguments more transparent.
Part I
Basic Concepts
Part I is devoted to the introduction of basic concepts and problems and to a number
of examples of propositional proof systems. Chapter 1 also recalls some very basic
notions of propositional and first-order logic (Sections 1.1 and 1.2) and of computa-
tional complexity theory (Sections 1.3 and 1.4). The former is aimed at those readers
with a primarily computer science (CS) background without any logic education
while the latter is aimed at readers who have avoided CS until now. These sections
should give the reader enough intuition to follow the presentation in the later chapters
but they cannot replace a proper education in these two subjects.
Section 1.5 introduces the key concepts of proof complexity and three fundamental
problems. Chapters 2–5 present the main examples of logical proof systems and
explain their basic properties, Chapter 6 introduces several examples of algebraic and
geometric proof systems and Chapter 7 collects together a number of proof systems
that do not fit into the previous categories.
1
Concepts and Problems
• α is a constant or an atom;
• there is a formula β such that α = (¬β);
• there are formulas β, γ such that α = (β ∨ γ );
• there are formulas β, γ such that α = (β ∧ γ ).
Another random document with
no related content on Scribd:
« Tant qu’elle vivra, le Maure, et Sforce et les couleuvres des
Visconti seront redoutés, des neiges hyperboréennes aux rivages de
la mer Rouge, de l’Inde aux monts qui donnent passage à la mer.
Elle morte, eux et le royaume d’Insubrie tomberont en esclavage, au
grand dommage de toute l’Italie. Sans elle la suprême prudence
paraîtra aventureuse.
« Il en existera encore d’autres, portant le même nom, et qui
naîtront bien des années avant elle. L’une d’elles ornera ses beaux
cheveux de la splendide couronne de Pannonie. Une autre, après
avoir délaissé les biens terrestres, sera placée au nombre des
saintes sur la terre d’Ausonie et se verra rendre un culte et élever
des autels.
« Je me tairai sur les autres, car, comme j’ai dit, il serait trop long
de parler de toutes, bien que chacune pût faire l’objet d’un chant
héroïque et éclatant. Je passerai sous silence les Blanche, les
Lucrèce, les Constance et les autres, mères ou réparatrices de tant
d’illustres maisons qui régneront en Italie.
« Plus que toutes celles qui ont jamais existé, ta maison sera
célèbre par ses femmes, et je ne sais si elle ne le sera pas plus par
les qualités des filles, que par la haute chasteté des épouses. Sache
également à ce sujet que Merlin m’a éclairée sur ce point, pensant
que j’aurais peut-être à te le répéter. J’ai donc un vif désir de t’en
entretenir.
« Et je te parlerai d’abord de Ricciarda, modèle de courage et de
chasteté. Jeune encore, elle restera veuve et en proie aux coups de
la fortune, ce qui arrive souvent aux meilleurs. Elle verra ses fils
dépouillés du royaume paternel, errer en exil sur la terre étrangère,
laissant leurs jeunes enfants aux mains de leurs ennemis. Mais elle
finira par être amplement dédommagée de ses malheurs.
« Je ne puis me taire sur l’illustre reine de l’antique maison
d’Aragon dont je ne vois pas l’égale, pour la chasteté et la sagesse,
dans l’histoire grecque ou latine. Je n’en connais pas non plus à qui
la fortune se soit montrée plus amie, puisqu’elle sera choisie par la
Bonté divine pour être la mère de cette belle race : Alphonse,
Hippolyte et Isabelle.
« Ce sera la sage Éléonore qui viendra se greffer sur ton arbre
fortuné. Que te dirai-je de sa seconde belle-fille qui doit lui succéder
peu après, Lucrèce Borgia, dont la beauté, la vertu, le renom de
chasteté [66] et la fortune, croîtront d’heure en heure, comme la
jeune plante dans un terrain fertile ?
« Comme l’étain est à l’argent, le cuivre à l’or, le pavot des
champs à la rose, le saule pâle au laurier toujours vert, le verre peint
à la pierre précieuse, ainsi, comparées à celle que j’honore avant
qu’elle soit née, seront les plus estimées pour leur sagesse et leurs
autres vertus.
« Et par-dessus tous les grands éloges qui lui seront donnés
pendant sa vie et après sa mort, on la louera d’avoir inculqué de
nobles sentiments à Hercule et à ses autres fils, qui, par la suite,
s’illustreront sous la toge et dans les armes, car le parfum qu’on
verse dans un vase neuf ne s’en va point si facilement, qu’il soit bon
ou mauvais.
« Je ne veux pas non plus passer sous silence Renée de France,
belle-fille de la précédente, et fille de Louis XII et de l’éternelle gloire
de la Bretagne. Je vois réunies dans Renée toutes les vertus qu’ait
jamais possédées une femme, depuis que le feu échauffe, que l’eau
mouille et que le ciel tourne autour de la terre.
« J’en aurais long à te dire sur Alde de Saxe, la comtesse de
Selano, Blanche-Marie de Catalogne, la fille du roi de Sicile, la belle
Lippa de Bologne et autres. Mais si j’entreprenais de te dire les
grandes louanges qu’elles mériteront toutes, j’entrerais dans une
mer qui n’a pas de rivages. — »
Après qu’elle lui eut fait connaître, à son vif contentement, la plus
grande partie de sa postérité, elle lui répéta à plusieurs reprises
comment Roger avait été attiré dans le palais enchanté. Arrivée près
de la demeure du méchant vieillard, Mélisse s’arrêta et ne jugea pas
à propos d’aller plus loin, de peur d’être vue par Atlante.
Et elle renouvela à la jeune fille les conseils qu’elle lui avait déjà
mille fois donnés, puis elle la laissa seule. Celle-ci ne chevaucha pas
plus de deux milles, dans un étroit sentier, sans voir quelqu’un qui
ressemblait à son Roger. Deux géants, à l’aspect féroce, le serraient
de près pour lui donner la mort.
Dès que la dame voit dans un tel péril celui qui a toutes les
apparences de Roger, elle change en doute la foi qu’elle avait dans
les avis de Mélisse, et elle oublie toutes ses belles résolutions. Elle
croit que Mélisse hait Roger pour quelque nouvelle injure ou pour
des motifs qu’elle ignore, et qu’elle a ourdi cette trame inusitée pour
le faire périr de la main de celle qui l’aime.
Elle se disait : « — N’est-ce pas là Roger, que je vois toujours
avec le cœur, et qu’aujourd’hui je vois avec mes yeux ? Et si
maintenant je ne le vois pas ou si je ne le reconnais pas, comment le
verrai-je, comment le reconnaîtrai-je jamais ? Pourquoi veux-je en
croire plutôt à autrui qu’à mes propres yeux ? A défaut de mes yeux,
mon cœur me dit s’il est loin ou près. — »
Pendant qu’elle se parle ainsi, elle croit entendre la voix de
Roger qui appelle à son secours. Elle le voit en même temps
éperonner son cheval rapide et lui retenir le mors, tandis que ses
deux féroces ennemis le suivent et le chassent à toute bride. La
dame s’empresse de les suivre et arrive avec eux dans la demeure
enchantée.
Elle n’en a pas plus tôt franchi les portes, qu’elle tombe dans
l’erreur commune. Elle cherche en vain Roger de tous côtés, en
haut, en bas, au dedans et au dehors. Elle ne s’arrête ni jour ni nuit ;
et l’enchantement était si fort, et l’enchanteur avait été si habile,
qu’elle voit sans cesse Roger et lui parle sans qu’elle le reconnaisse,
ou sans que Roger la reconnaisse elle-même.
Mais laissons Bradamante, et n’ayez pas de regret de la savoir
en proie à cet enchantement. Quand il sera temps qu’elle en sorte,
je l’en ferai sortir, et Roger aussi. De même que le changement de
nourriture ranime l’appétit, ainsi il me semble que mon histoire
risquera d’autant moins d’ennuyer qui l’entendra, qu’elle sera plus
variée.
Il faut aussi que je me serve de beaucoup de fils pour tisser la
grande toile à laquelle je travaille. Qu’il ne vous déplaise donc pas
d’écouter comment, sortie de ses tentes, l’armée des Maures a pris
les armes pour défiler devant le roi Agramant, lequel, fortement
menacé par les lis d’or, l’a rassemblée pour une nouvelle revue, afin
de savoir combien elle compte de combattants.
Outre que bon nombre de cavaliers et de fantassins avaient
disparu, beaucoup de chefs manquaient, et des meilleurs, parmi les
troupes d’Espagne, de Libye et d’Éthiopie. Les divers corps de
nations erraient sans direction propre. Afin de leur donner un chef, et
de remettre de l’ordre dans chacun d’eux, tout le camp était
rassemblé pour la revue.
Pour remplacer les pertes subies dans les batailles et les conflits
sanglants, le roi d’Espagne et le roi d’Afrique avaient envoyé des
ordres chacun dans leur pays, pour en faire venir de nombreux
renforts, et ils les avaient distribués sous les différents chefs. Avec
votre agrément, seigneur, je remettrai à l’autre chant l’exposé de
cette revue.
CHANT XIV.