Eigenfunctions of the Laplacian of a

Riemannian Manifold Steven Zelditch

Visit to download the full and correct content document:
https://textbookfull.com/product/eigenfunctions-of-the-laplacian-of-a-riemannian-manif
old-steven-zelditch/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...

Introduction to Riemannian manifolds Corrected version

of second edition Lee

https://textbookfull.com/product/introduction-to-riemannian-
manifolds-corrected-version-of-second-edition-lee/

Oracle of Wynter 02 0 Magic Menopause 1st Edition Lisa

Manifold

https://textbookfull.com/product/oracle-of-wynter-02-0-magic-
menopause-1st-edition-lisa-manifold/

Macdonald Polynomials: Commuting Family of q-Difference

Operators and Their Joint Eigenfunctions Masatoshi
Noumi

https://textbookfull.com/product/macdonald-polynomials-commuting-
family-of-q-difference-operators-and-their-joint-eigenfunctions-
masatoshi-noumi/

Principles of Macroeconomics 2nd Edition Steven A.

Greenlaw: University Of Mary Washington

https://textbookfull.com/product/principles-of-
macroeconomics-2nd-edition-steven-a-greenlaw-university-of-mary-
washington/
Bourdieu's Theory of the State: A Critical Introduction
1st Edition Steven Loyal (Auth.)

https://textbookfull.com/product/bourdieus-theory-of-the-state-a-
critical-introduction-1st-edition-steven-loyal-auth/

Currencies of the Indian Ocean World Steven Serels

https://textbookfull.com/product/currencies-of-the-indian-ocean-
world-steven-serels/

A New Development at the Intersection of Nuclear

Structure and Reaction Theory Steven Karataglidis

https://textbookfull.com/product/a-new-development-at-the-
intersection-of-nuclear-structure-and-reaction-theory-steven-
karataglidis/

A History of Capital Punishment in the Australian

Colonies, 1788 to 1900 Steven Anderson

https://textbookfull.com/product/a-history-of-capital-punishment-
in-the-australian-colonies-1788-to-1900-steven-anderson/

Learn Linux in a Month of Lunches 1st Edition Steven

Ovadia

https://textbookfull.com/product/learn-linux-in-a-month-of-
lunches-1st-edition-steven-ovadia/
 Eigenfunctions of the Laplacian of Riemannian
manifolds
Updated: August 15, 2017

Steve Zelditch

Department of Mathematics, Northwestern University, 2033 Sheri-

dan Road, Evanston, IL 60208
E-mail address: zelditch@math.university.edu
Key words and phrases. Laplacian, eigenfunction, nodal set, Lp norms, Weyl law,
quantum limits

Research partially supported by NSF grant and DMS-1541126 and by the Stefan
Bergman trust.
 Contents

Preface xi
0.1. Organization xii
0.2. Topics which are not covered xiii
0.3. Topics which are double covered xiv
0.4. Notation xiv
Acknowledgments xiv

Chapter 1. Introduction 1
1.1. What are eigenfunctions and why are they useful 1
1.2. Notation for eigenvalues 3
1.3. Weyl’s law for (−∆)-eigenvalues 3
1.4. Quantum Mechanics 4
1.5. Dynamics of the geodesic or billiard flow 6
1.6. Intensity plots and excursion sets 7
1.7. Nodal sets and critical point sets 8
1.8. Local versus global analysis of eigenfunctions 9
1.9. High frequency limits, oscillation and concentration 10
1.10. Spectral projections 11
1.11. Lp norms 12
1.12. Matrix elements and Wigner distributions 12
1.13. Egorov’s theorem 13
1.14. Eherenfest time 14
1.15. Weak* limit problem 14
1.16. Ergodic versus completely integrable geodesic flow 16
1.17. Ergodic eigenfunctions 17
1.18. Quantum unique ergodicity (QUE) 18
1.19. Completely integrable eigenfunctions 18
1.20. Heisenberg uncertainty principle 19
1.21. Sequences of eigenfunctions and length scales 19
1.22. Localization of eigenfunctions on closed geodesics 20
1.23. Some remarks on the contents and on other texts 21
1.24. References 22

Bibliography 23

Chapter 2. Geometric preliminaries 27

2.1. Symplectic linear algebra and geometry 27
2.2. Symplectic manifolds and cotangent bundles 29
2.3. Lagrangian submanifolds 30
2.4. Jacobi fields and Poincaré map 31
v
vi CONTENTS

2.5. Pseudo-differential operators 32

2.6. Symbols 33
2.7. Quantization of symbols 34
2.8. Action of a pseudo-differential operator on a rapidly oscillating
exponential 35

Bibliography 37

Chapter 3. Main results 39

3.1. Universal Lp bounds 39
3.2. Self-focal points and extremal Lp bounds for high p 40
3.3. Low Lp norms and concentration of eigenfunctions around geodesics 41
3.4. Kakeya-Nikodym maximal function and extremal Lp bounds for small
p 42
3.5. Concentration of joint eigenfunctions of quantum integrable ∆ around
closed geodesics 43
3.6. Quantum ergodic restriction theorems for Cauchy data 46
3.7. Quantum ergodic restriction theorems for Dirichlet data 48
3.8. Counting nodal domains and nodal intersections with curves 51
3.9. Intersections of nodal lines and general curves on negatively curved
surfaces 54
3.10. Complex zeros of eigenfunctions 55

Bibliography 59

Chapter 4. Model spaces of constant curvature 61

4.1. Euclidean space 61
4.2. Euclidean wave kernels 65
4.3. Flat torus Tn 73
4.4. Spheres S n 74
4.5. Hyperbolic space and non-Euclidean plane waves 80
4.6. Dynamics and group theory of G = P SL(2, R) 81
4.7. The Hyperbolic Laplacian 82
4.8. Wave kernel and Poisson kernel on Hyperbolic space Hn 83
4.9. Poisson kernel 86
4.10. Spherical functions on H2 87
4.11. The non-Euclidean Fourier transform 87
4.12. Hyperbolic cylinders 87
4.13. Irreducible representations of G 88
4.14. Compact hyperbolic quotients XΓ = Γ\H2 88
4.15. Representation theory of G and spectral theory of ∆ on compact
quotients 89
4.16. Appendix on the Fourier transform 89

Bibliography 93

Chapter 5. Local structure of eigenfunctions 95

5.1. Local versus global eigenfunctions 95
5.2. Small balls and local dilation 96
5.3. Local elliptic estimates of eigenfunctions 98
 CONTENTS vii

5.4. λ-Poisson operators 102

5.5. Bernstein estimates 105
5.6. Frequency function and doubling index 105
5.7. Carleman estimates 108
5.8. Norm square of the Cauchy data 110
5.9. Hyperbolic aspects 113

Bibliography 117

Chapter 6. Hadamard parametrices on Riemannian manifolds 119

6.1. Hadamard parametrix 119
6.2. Hadamard-Riesz parametrix 121
6.3. The Hadamard-Feynman fundamental solution and Hadamard’s
parametrix 122
6.4. Sketch of proof of Hadamard’s construction 123
6.5. Convergence in the real analytic case 126
6.6. Away from CR 127
6.7. Hadamard parametrix on a manifold without conjugate points 127
6.8. Dimension 3 128
6.9. Appendix on Homogeneous distributions 132

Bibliography 135

Chapter 7. Lagrangian distributions and Fourier integral operators 137

7.1. Introduction 137
7.2. Homogeneous Fourier integral operators 139
7.3. Semi-classical Fourier integral operators 148
7.4. Principal symbol, testing and matrix elements 153
7.5. Composition of half-densities on canonical relations in cotangent
bundles 159

Bibliography 163

Chapter 8. Small time wave group and Weyl asymptotics 165

8.1. Hörmander parametrix 165
8.2. Wave group and spectral projections 166
8.3. Small-time asymptotics for microlocal wave operators 167
8.4. Weyl law and local Weyl law 169
8.5. Fourier Tauberian approach 172
8.6. Tauberian Lemmas 176

Bibliography 179

Chapter 9. Matrix elements 181

9.1. Invariance properties 182
9.2. Proof of Egorov’s theorem 182
9.3. Weak* limit problem 184
9.4. Matrix elements of spherical harmonics 185
9.5. Quantum ergodicity and mixing of eigenfunctions 186
9.6. Hassell’s scarring result for stadia 194
9.7. Appendix on Duhamel’s formula 198
viii CONTENTS

Bibliography 201

Chapter 10. Lp norms 203

10.1. Discrete Restriction theorems 205
10.2. Random spherical harmonics and extremal spherical harmonics 206
10.3. Sketch of proof of the Sogge Lp estimates 207
10.4. Maximal eigenfunction growth 209
10.5. Geometry of loops and return maps. 216
10.6. Proof of Theorem 10.21. Step 1: Safarov’s pre-trace formula 222
10.7. Proof of Theorem 10.29. Step 2: Estimates of remainders at L-points 228
10.8. Completion of the proof of Proposition 10.30 and Theorem 10.29:
study of R̃j1 229
10.9. Infinitely many twisted self-focal points 233
10.10. Dynamics of the first return map at a self-focal point 234
10.11. Proof of Proposition 10.20 235
10.12. Uniformly bounded orthonormal basis 237
10.13. Appendix: Integrated Weyl laws in the real domain 238

Bibliography 241

Chapter 11. Quantum Integrable systems 245

11.1. Classical integrable systems 245
11.2. Normal forms of integrable Hamiltonians near non-degenerate
singular orbits 248
11.3. Joint eigenfunctions 249
11.4. Quantum toral integrable systems 250
11.5. Lagrangian torus fibration and classical moment map 252
11.6. Lp norms of Quantum integrable eigenfunctions 252
11.7. Sketch of proof of Theorem 11.8 253
11.8. Mass concentration of special eigenfunctions on hyperbolic orbits in
the quantum integrable case 255
11.9. Details on Mh 256
11.10. Concentration of quantum integrable eigenfunctions on submanifolds 257

Bibliography 261

Chapter 12. Restriction theorems 263

12.1. Null restrictions, degenerate restrictions and ‘goodness’ 264
12.2. L2 upper bounds on Dirichlet or Neumann data of eigenfunctions 266
12.3. Cauchy data of Dirichlet eigenfunctions for manifolds with boundary 267
12.4. Restriction bounds for Neumann eigenfunctions 268
12.5. Periods and Fourier coefficients of eigenfunctions on a closed geodesic 268
12.6. Kuznecov sum formula: Proofs of Theorems 12.8 and 12.10 270
12.7. Restricted Weyl laws 271
12.8. Relating matrix elements of restrictions to global matrix elements 273
12.9. Geodesic geometry of hypersurfaces 274
12.10. Tangential cutoffs 276
12.11. Canonical relation of γH 276
∗
12.12. The canonical relation of γH OpH (a)γH 277
∗
12.13. The canonical relation Γ ◦ CH ◦ Γ 279
 CONTENTS ix

12.14. The pullback ΓH := ∆∗t Γ∗ ◦ CH ◦ Γ 280

12.15. The pushforward πt∗ ∆∗t Γ∗ ◦ CH ◦ Γ 280
12.16. The symbol of U (t1 )∗ (γH
∗
OpH (a)γH )≥ε U (t2 ) 282
12.17. Proof of the restricted local Weyl law: Proposition 12.14 283
12.18. Asymptotic completeness and orthogonality of Cauchy data 284
12.19. Expansions in Cauchy data of eigenfunctions 286
12.20. Bochner-Riesz means for Cauchy data 288
12.21. Quantum ergodic restriction theorems 289
12.22. Rellich approach to QER: Proof of Theorem 12.33 292
12.23. Proof of Theorem 12.36 and Corollary 12.37 294
12.24. Quantum ergodic restriction (QER) theorems for Dirichlet data 296
12.25. Time averaging 298
12.26. Completion of the proofs of Theorems 12.39 and 12.40 300

Bibliography 303

Chapter 13. Nodal sets: Real domain 307

13.1. Fundamental existence theorem for nodal sets 308
13.2. Curvature of nodal lines and level lines 309
13.3. Sub-level sets of eigenfunctions 310
13.4. Nodal sets of real homogeneous polynomials 312
13.5. Rectifiability of the nodal set 313
13.6. Doubling estimates 315
13.7. Lower bounds for Hm−1 (Nλ ) for C ∞ metrics 317
13.8. Counting nodal domains 323

Bibliography 335

Chapter 14. Eigenfunctions in the complex domain 339

14.1. Grauert tubes and complex geodesic flow 340
14.2. Analytic continuation of the exponential map 341
14.3. Maximal Grauert tubes 341
14.4. Model examples 342
14.5. Analytic continuation of eigenfunctions 343
14.6. Maximal holomorphic extension 344
14.7. Husimi functions 345
14.8. Poisson wave operator and Szegő projector on Grauert tubes 346
14.9. Poisson operator and analytic Continuation of eigenfunctions 346
14.10. Analytic continuation of the Poisson wave group 346
14.11. Complexified spectral projections 347
14.12. Poisson operator as a complex Fourier integral operator 348
14.13. Complexified Poisson kernel as a complex Fourier integral operator 349
14.14. Analytic continuation of the Poisson wave kernel 349
14.15. Hörmander parametrix for the Poisson wave kernel 349
14.16. Subordination to the heat kernel 350
14.17. Fourier integral distributions with complex phase 350
14.18. Analytic continuation of the Hadamard parametrix 351
14.19. Analytic continuation of the Hörmander parametrix 351
14.20. ∆g , 2g and characteristics 351
14.21. Characteristic variety and characteristic conoid 352
x CONTENTS

14.22. Hadamard parametrix for the Poisson wave kernel 353

14.23. Hadamard parametrix as an oscillatory integral with complex phase 353
14.24. Tempered spectral projector and Poisson semi-group as complex
Fourier integral operators 356
14.25. Complexified wave group and Szegő kernels 357
14.26. Growth of complexified eigenfunctions 358
14.27. Siciak extremal functions: Proof of Theorem 14.14 (1) 360
14.28. Pointwise phase space Weyl laws on Grauert tubes 362
14.29. Proof of Corollary 14.16 365
14.30. Complex nodal sets and sequences of logarithms 365
14.31. Real zeros and complex analysis 367
14.32. Background on hypersurfaces and geodesics 368
14.33. Proof of the Donnelly-Fefferman lower bound (A. Brudnyi) 374
14.34. Properties of eigenfunctions in good balls 375
14.35. Background on good-ness 376
14.36. A. Brudnyi’s proof of Proposition 14.38 376
14.37. Equidistribution of complex nodal sets of real ergodic eigenfunctions 379
14.38. Sketch of the proof 380
14.39. Growth properties of complexified eigenfunctions 381
14.40. Proof of Lemma 14.48 383
14.41. Proof of Lemma 14.47 384
14.42. Intersections of nodal sets and analytic curves on real analytic
surfaces 384
14.43. Counting nodal lines which touch the boundary in analytic plane
domains 385
14.44. Application to Pleijel’s conjecture 391
14.45. Equidistribution of intersections of nodal lines and geodesics on
surfaces 391
Bibliography 395

Index 399
 Preface

These lecture notes are an expanded version of the author’s CBMS ten Lectures
at the University of Kentucky in June 20-24, 2011. The lectures were devoted to
eigenfunctions of the Laplacian and of Schrödinger operators, in particular to their
Lp -norms and nodal sets. The lecture notes have undergone extensive revisions
in the intervening years, due in part to progress in the field and also to the new
publications on related topics, which made some of the original lecture notes obso-
lete. In particular, the new book [So2] of Chris Sogge and the author’s 2013 Park
City Lecture notes [Ze7] are also devoted to eigenfunctions and includes extensive
background on pseudo-differential operators and harmonic analysis. (References
for the preface can be found at the end of §1.) The book of Maciej Zworski [Zw]
contains a systematic introduction to semi-classical Fourier integral operators and
includes applications to quantum ergodicity of eigenfunctions. The recent book
[GS] of V. Guillemin and S. Sternberg also gives background on the global theory
of Fourier integral operators and in particular on their symbols. Fanghua Lin and
Qing Han also have a book in progress on eigenfunctions from viewpoint of local
elliptic equations. For this reason, we do not feel it is useful in these lecture notes
to provide any systematic background on these techniques, although their proper-
ties will be used freely. We do include some background on symplectic geometry,
pseudo-differential and Fourier integral operators to establish notation and links to
other references. But overall we assume that the reader is willing to consult these
other references for the basic techniques.
The purpose of these lecture notes is to convey inter-related themes and results,
and so we rarely give detailed proofs. Rather we aim to outline key ideas and how
they are related to other results. The lectures concentrate on the following themes:

• Local versus Global analysis of eigenfunctions. The Local analysis of

eigenfunctions belongs to the theory of elliptic equations, and pertains to
local solutions of the eigenvalue problem (∆ + λ)ϕ = 0 on small balls of
radius √Cλ . The global analysis belongs to hyperbolic equations, i.e., stud-
√ √
ies the eigenfunctions through the wave equation cos t −∆ϕ = cos t λϕ
and their relations to geodesics as λ → ∞. One of the aims of these
lectures is to survey both local and global methods, and to discuss how
they interact. For instance, the main existence theorem that there exists
A
a zero of ϕλ in each ball B(p, √λg ) whose radius is a certain number Cg
of wavelengths is a local result and global methods are not particularly
useful in proving it. On the other hand, the basic sup-norm estimates of
eigenfunctions are most easily proved using the wave equation. It often
seems that researchers on eigenfunctions split into two disjoint groups,
exclusively using local or global methods. It is likely that many problems
xi
xii PREFACE

require both types of methods. In §5.3 we review the elliptic methods

that have been applied to eigenfunctions by Donnelly-Fefferman, F. H.
Lin, Nazarov-Sodin, Colding-Minicozzi, and many others.
• Quantum analogues of classical dynamical methods for ergodic or com-
pletely integrable systems. For instance, Birkhoff normal forms are local
normal forms on both the classical and quantum level around invariant
sets such as closed geodesics, which are useful in study concentration on
submanifolds.
• Lp bounds on eigenfunctions and their source in the global dynamics of
the geodesic flow.
• Restriction theorems for eigenfunctions under dynamical assumptions mainly
in the ergodic setting.
• Nodal geometry in the complex domain. Considerable space is devoted
to analytic continuation of eigenfunctions of Laplacians of real analytic
Riemannian manifolds to the complexification of the manifold. The ra-
tionale for analytic continuation is that the nodal sets are better behaved
and easier to study in the complex domain than the real domain. From
the viewpoint of quantum mechanics, both the real and complex domains
are equally good representations.

0.1. Organization
Let us go over the sequence of events in these lectures and explain what is and
what is not contained in them and what is the logic of the presentation.
We introduce the subject of eigenfunctions in terms of vibrating membranes and
quantum energy eigenstates. The rich phenomenology of examples developed over
the last two hundred years is rapidly surveyed. In Chapter 3 we give an overview
of the principal new results that will be discussed in detail. The model surfaces of
constant curvature are introduced in §4. Harmonic analysis begins with the Eu-
clidean eigenfunctions eihx,ki on Rn or T n , yet they have very unusual properties
compared to eigenfunctions on other Riemannian manifolds. The eigenfunctions
of S 2 illustrate virtually the entire range of behavior of eigenfunctions of any Rie-
mannian metric with regard to size and concentration. On the other hand, they
are restrictions of harmonic polynomials on R3 and their nodal sets are potentially
tamer than for a general C ∞ metric. Eigenfunctions of hyperbolic surfaces H2 /Γ
come next. They are the material of quantum chaos and are the subject of in-
tense investigation over the last 30 years. In §5-5.3 the local elliptic analysis of
eigenfunctions is surveyed. This leads §6 on the wave equation on a Riemannian
manifold and the Hadamard-Riesz construction of parametrices. This construction
parallels the Minakshisundaraman-Pleijel parametrix construction for heat kernels.
In some ways, the original presentations of Hadamard and Riesz remain the best
expositions, in particular in their presentations of the convergence of the parametrix
construction in the real analytic case. It was a precursor to the Fourier integral op-
erator theory, which is rapidly reviewed in §2.1, §2.5, §7. As mentioned above, this
material is contained in many other references and is principally used to establish
notation. In §8.2 classical results on the pointwise and local Weyl laws are reviewed,
and the results presented give the universal sup-norm estimates on eigenfunctions
and their gradients. The author is not aware of a proof of such estimates using
 0.2. TOPICS WHICH ARE NOT COVERED xiii

elliptic estimates. Geometric analysts who are more familiar with elliptic estimates
might want to compare their methods to the small time wave equation methods
used in the proofs. In §9, the asymptotics and limits of matrix elements hAϕj , ϕj i of
pseudo-differential operators with respect to eigenfunctions are introduced. Matrix
elements are the fundamental quantities in quantum mechanics. They are quadratic
in the eigenfunctions and thus are related to energy estimates. There exist some
results on multilinear eigenfunction estimates but they are not covered in these lec-
tures. In §9.5 the basic facts about quantum ergodic systems are reviewed. At this
point in the lecture notes, the global long-time dynamics of the geodesic flow takes
over as the dominant player. In §11 some parallel results for quantum integrable
systems are presented. At this time there exist only a few results on quantizations
of mixed systems, and despite the great interest in mixed systems we do not present
these results but only record the existence of several articles devoted to them. Lp
norms of eigenfunctions are studied in §10. Sogge’s books [So1, So2] also concern
Lp norms but the material presented here contains both less and more on them.
Less, because the universal Sogge estimates are not presented, and more because
the more advanced results due to Sogge and the author are given in some detail. In
§11.6, Lp norms of eigenfunctions in the quantum integrable case are reviewed. One
of the motivations to include this material is the belief that such QCI eigenfunc-
tions are extremals for Lp norms and restrictions of eigenfunctions. Although it is
very relevant the restriction theorems of Burq-Gérard-Tzvetkov are not discussed
here. Rather we turn to quantum ergodic restriction theorems in §12.21. They
have proved useful in the study of nodal sets and that is the main topic for the rest
of the lectures. Nodal sets in the real domain are discussed in §13, in particular
bounds on hypersurface volumes and counting nodal domains. Starting in §14, the
analytic continuation of eigenfunctions to Grauert tubes and their complex zeros
are studied. Complex nodal sets and their intersections with complexified geodesics
are studied in §14.30. Use of the complexified wave kernel gives a simplified proof
of the Donnelly-Fefferman upper bound on the hypersurface measure of nodal sets.
The lower bound seems to be disconnected from global methods. In §14.33, Alex
Brudnyi has contributed a simplified proof of the Donnelly-Fefferman lower bound.
In §14.37, the author’s results on equidistribution of complexified nodal sets in the
ergodic case are presented. There are parallel results in the completely integrable
case which are still in progress. Other results in this section are those of John Toth
and the author giving upper bounds on numbers of intersection points of nodal
lines with curves in dimension two.

0.2. Topics which are not covered

There are many important topics on eigenfunctions which are not discussed
in these lecture notes, due to time and length constraints. A more comprehensive
treatment of eigenfunctions would include the following topics:
• Arithmetic quantum chaos. These lecture notes are devoted to PDE meth-
ods and therefore we do not get into the special methods available for
Hecke-Maass forms on arithmetic quotients. The sharpest results on Lp
norms or nodal sets of eigenfunctions are for these special joint eigen-
functions of ∆ and of Hecke operators. One might compare their special
xiv PREFACE

properties to those of joint eigenfunctions of a quantum integrable sys-

tem although they are much more complicated and the dynamics is in the
opposite chaotic regime.
• Entropy of quantum limits. The breakthrough results of Anantharaman
and the subsequent work of Anantharaman-Nonnenmacher and Rivière
are very relevant to the theme of these lectures.
• General Lp bounds on restrictions of eigenfunctions, multilinear estimates
and Kakeya-Nikodym bounds.
• Gaussian random spherical harmonics and more general random linear
combinations of eigenfunctions.
• Spectral and scattering theory for non-compact complete Riemannian
manifolds.

0.3. Topics which are double covered

It is impossible to avoid overlaps with the author’s prior expository articles,
such as the article on local and global analysis of eigenfunctions [Ze3], on nodal
sets [Ze6] or Park City Lecture notes [Ze7] and other expository articles on eigen-
functions and nodal sets.
Another double-coverage is with regard to cited references. Each chapter has
a bibliography of the references cited in it. Many references are cited in several
chapters. Although this results in duplicated references it seems preferable to only
listing hundreds of references at the end of the lecture notes.

0.4. Notation
Notation regarding eigenvalue parameters is given in §1.2 and notation for
geometric and dynamical objects is given in §2.

Acknowledgments
Thanks go to Peter Hislop and Peter Perry for organizing the CBMS lecture
series at the University of Kentucky. Both author and reader will thank Alex
Brudnyi for giving alternative arguments to the Donnelly-Fefferman lower bound
in §14.33. Thanks also to Hans Christianson, J. Jung, C.D. Sogge, J.A. Toth for
collaboration and for an infinite number of discussions on the topics discussed here.
My main thanks go to Robert Chang for reading the text and for suggesting many
corrections. Robert also did almost all the technical support in producing the book.
 CHAPTER 1

Introduction

In this chapter, we introduce the main objects and themes of this monograph.
In particular we introduce the quantum mechanical interpretation of eigenfunctions
and their time evolution. At the end we outline the topics emphasized in later
chapters.

1.1. What are eigenfunctions and why are they useful

Eigenfunctions of the Laplacian first arose in the study of vibrating plates and
membranes. The equations of motion of a vibrating membrane Ω are given by the
mixed initial value and Dirichlet problem for the profile u(t, x) on R × Ω:

∂2


 − ∆ u(t, x) = 0;
∂t2
(1.1) ∂ϕ
u(0, x) = ϕ0 (x), ∂t u(0, x) = 0;
u(t, x) = 0, x ∈ ∂Ω.


If ϕλ is a Dirichlet eigenfunction, i.e., a solution of the Helmholtz equation

(1.2) (∆ + λ2 )ϕλ = 0 and ϕλ |∂Ω = 0,

then one obtains a periodic solution of the wave equation on R × Ω:

(1.3) uλ (t, x) = (cos tλ)ϕλ (x).

Thus, ϕλ represents the profile of a periodic vibration, i.e., a mode of vibration. In

our notation, −λ2 is the eigenvalue or energy and λ denotes the frequency.
When the domain Ω is compact, the Laplacian has a discrete spectrum with
finite multiplicities. We write

(1.4) λ20 = 0 < λ21 ≤ λ22 ↑ ∞

for the ordered sequence of eigenvalues, repeated according to multiplicity. The

2
corresponding set {ϕj } of eigenfunctions
R form an orthonormal basis of L (Ω) with
respect to the inner product hu, vi = Ω uv̄ dV , where dV is the volume density:
Z
(1.5) (∆ + λ2j )ϕj =0 and hϕj , ϕk i := ϕj ϕk dV = δjk
M

1
2 1. INTRODUCTION

The nodal set (or zero set) of ϕλ gives the positions at which the vibrating
membrane is still. The nodal patterns have been studied since the time of Chladni
(ca. 1800).

A natural generalization is to consider eigenfunctions of the Laplacian on Rie-

mannian manifolds (M, g), with or without boundary. The Laplacian of (M, g) is
given locally by
n  
1 X ∂ √ ∂
(1.6) ∆g := √ g ij g ,
g i,j=1 ∂xi ∂xj

∂
replacing the Euclidean Laplacian ∆ above. Here, gij = g( ∂x , ∂ ), [g ij ] is the
i ∂xj
√ p
inverse matrix to [gij ] and g := det[gij ]. Since g is usually understood, in
subsequent chapters we suppress the dependency of the metric by writing ∆g = ∆.
It follows that on a Riemannian manifold (M, g), the eigenvalue problem (1.2)
has the form
(1.7) (∆g + λ2 )ϕλ = 0.
If M has a non-empty boundary ∂M then we impose the standard Dirichlet or
Neumann boundary conditions. If M is compact case, there exists an orthonormal
basis {ϕj }j≥0 of L2 (M ) of eigenfunctions,
Z
(1.8) ∆g ϕj = −λ2j ϕj and hϕj , ϕk iL2 (M ) := ϕj ϕk dVg = δjk
M

and as above the eigenvalues

(1.9) 0 = λ20 ≤ λ21 ≤ λ22 ↑ ∞
are repeated according to multiplicity. As in the case of a vibrating membrane, the
eigenfunctions ϕλ represent modes of vibration of M .
 1.3. WEYL’S LAW FOR (−∆)-EIGENVALUES 3

An equivalent definition of the Laplacian is that it is the operator corresponding

to the quadratic form
Z
(1.10) D(f ) = kdf k2 dVg = hdf, df i
M
in the sense that
(1.11) D(f ) = −h∆f, f i.

1.2. Notation for eigenvalues

We often parametrize spectral quantities in terms of the frequencies
√ λj , which
are eigenvalues of the first order elliptic pseudo-differential operator −∆, rather
than by the eigenvalues λ2j . We warn the reader that many others denote the
p
(−∆)-eigenvalues by λj and the frequencies by λj .
Regarding eigenfunctions, we write ϕj when the eigenfunction is part of an
orthonormal basis as in (1.5) and ϕλ to denote any eigenfunction of eigenvalue λ2
with kϕλ kL2 = 1. The notation is ambiguous since the eigenvalue λ2 may not be
simple, i.e., the eigenspace may have dimension greater than one, but it is a useful
notation when we only care about the dependence on the eigenvalue.
There are two reasons to emphasize frequencies over eigenvalues. One is to sim-
plify the notation by getting rid of square roots. The other is to relate frequencies
to Planck’s constant. (Planck’s constant is also written as ~ = 2π h
. We use the two
notations interchangeably.)
(1.12) hj = λ−1
j ,
which is conceptually important because the high frequency asymptotics of eigen-
values and eigenfunctions is equivalent to the semiclassical asymptotics h → 0. We
also use the notation ϕh for ϕλ where it is understood that h = λ−1 as in (1.12).
We sometimes denote an orthonormal basis by ϕhj . Thus we write the Helmholtz
equation in semiclassical notation as
(1.13) ∆ϕh = −h−2 ϕh ⇐⇒ (h2 ∆ − 1)ϕh = 0.
As the semiclassical notation suggests, a compelling motivation to study eigenfunc-
tions comes from their role in quantum mechanics.

1.3. Weyl’s law for (−∆)-eigenvalues

When M is a compact manifold, Weyl’s law counts the number of eigenvalues
of ∆g . Let
(1.14) N (λ) := {j : λj ≤ λ}.
Weyl’s law states the following:
Theorem 1.1. If (M, g) is a compact Riemannian manifold of dimension m,
then
(1.15) N (λ) = Cm Vol(M, g)λm + O(λm−1 ),
where Cm = Vol(B1 ) is a dimensional constant, the volume of the unit ball in Rm .
Here, we say R(λ) = O(λr ) if there exists a constant C independent of λ so
that R(λ) ≤ Cλr as λ → ∞. We also write R(λ) = o(λr ) if R(λ) ≤ ελr as λ → ∞
for any ε > 0.
4 1. INTRODUCTION

1.4. Quantum Mechanics

Much of quantum mechanics is concerned with the eigenvalue problem for the
Schrödinger equation
h2
 
(1.16) − ∆ + V ψ = Eψ.
2
Here, V stands for multiplication by the potential V ∈ C ∞ (M ), and (as above) h
is Planck’s constant, a very small parameter. When V = 0, E = 1 and h = λ−1 ,
(1.16) specializes to (1.2). Thus we think of the limit as λ → ∞ in (1.2) as the
semiclassical limit h → 0.
The Schrödinger eigenvalue problem in quantum mechanics resolves a puzzle
about the stability of atoms. Before quantum mechanics, a hydrogen atom was
roughly pictured as a 2-body planetary system, i.e., as an electron orbiting the
nucleus centered at the origin 0 ∈ R3 according to Kepler’s laws. The orbits are
projections to configuration space R3 of the phase space orbits of the classical
Hamiltonian flow defined by Hamilton’s equations
dxj ∂H

 dt = ∂ξ ,


j
(1.17)
dξj ∂H
=− ,



dt ∂xj
where the Hamiltonian
1 2 1
(1.18) H(x, ξ) = |ξ| − : T ∗ R3 → R
2 |x|
1
is the total Newtonian kinetic plus Coulomb potential energy function V (x) = − |x|
on phase space, the cotangent bundle T ∗ R3 of the configuration space R3 . We
denote the Hamiltonian (geodesic) flow by
(1.19) Gt (x, ξ) = exp tXH ,
where XH is the Hamilton vector field and exp tX is the general notation for the
flow of X.
But this model cannot be right: the electron would radiate energy and spiral
into the nucleus.

The Bohr model (1913) of “old quantum theory” proposed that the electron
can only occupy special stable orbits defined by Bohr-Sommerfeld “quantization
conditions.”
 1.4. QUANTUM MECHANICS 5

However this theory is too specialized. It relies on the special structure of the
orbits of the Coulomb problem, in particular the (hidden) symmetry that makes
all of the orbits periodic. It does not extend in any clear way to more compli-
cated atoms such as Helium or even to the hydrogen atom in an electric or mag-
netic field. In the article Quantisierung als Eigenwertproblem, Annalen der Physik
(1926), Schrödinger [Sch1] proposed to model the electron by a wave function
ψ(x) ∈ L2 (R3 ) with the states of energy Ej (~) solving the eigenvalue problem (as
in (1.16))
~
 2 
(1.20) Ĥψ~,j := − ∆ + V ψ~,j = Ej (~)ψ~,j ,
2
P ∂2
for the Schrödinger operator Ĥ, where ∆ = j ∂x 2 is the Laplacian and V is the
j

potential, a multiplication operator on L2 (R3 ). Here {ψ~,j } denote an orthonormal

basis of eigenfunctions with eigenvalues Ej (~) in non-decreasing order.
Classically, the particle evolves according to Hamilton’s equations (1.17) with
H(x, ξ) = 21 |ξ|2 + V (x). The Hamiltonian is constant along Hamilton orbits and
therefore the orbits lie on level sets {H = E} of H. The projection {x : V (x) ≤ E}
of these level sets to the configuration space Rn is known as the allowed region; a
classical particle cannot enter the forbidden region, which is the set {x : V (x) > E}.
Quantum mechanics thus replaces the classical mechanics of Hamilton’s equa-
tions with linear algebra (an eigenvalue problem). The time evolution of an energy
state is given by
t ~2 tEj (~)
(1.21) U~ (t)ψ~,j = e−i ~ (− 2 ∆+V )
ψ~,j = e−i ~ ψ~,j .
Throughout it is assumed that eigenfunctions are L2 normalized,
Z
(1.22) |ψ~,j |2 dV = 1,

so that the state ψ~,j defines a probability amplitude, i.e., its modulus square is a
probability measure with
(1.23) |ψ~,j (x)|2 dx = the probability density of finding the particle at x .
This probability density is not concentrated in the classically allowed region {V ≤
E}, i.e., a quantum particle has a positive probability of going into the forbidden
region {x : V (x) > E}. The only observable quantities are the matrix elements
Z
(1.24) hAψ~,j , ψ~,j i = ψ~,j Aψ~,j (x) dV
6 1. INTRODUCTION

of observables (A is a self adjoint operator). Under the time evolution (1.21),

tEj (~)
the factors of e−i ~ cancel and so the particle evolves as if “stationary,” i.e.,
observations of the particle are independent of the time t.
Quantum mechanics resolves the puzzle of how the electron can be moving
and stationary at the same time. But it also replaces the geometric (classical
mechanical) Bohr model of classical orbits with eigenfunctions (1.20), which are not
geometric objects and which are difficult to visualize. They are very complicated
functions on high dimensional spaces. How does one reconcile the classical picture
of orbits with the quantum picture of eigenfunctions, as stationary energy states of
atoms? In the semiclassical limit h → 0, the quantum physics should tend to the
classical physics, and the eigenfunctions should be related to the classical orbits.
The Bohr model proposed a close relation between the quantum mechanics of
a hydrogen atom and the classical mechanics of the corresponding classical Hamil-
tonian H(x, ξ) = 12 |ξ|2 + V (x). Can we use classical mechanics to analyze shapes
and sizes of quantum eigenstates?

1.5. Dynamics of the geodesic or billiard flow

The (homogeneous) geodesic flow
(1.25) Gt : T ∗ M \0 → T ∗ M \0
on the punctured cotangent bundle T ∗ M \0 = {(x, ξ) ∈ T ∗ M : ξ 6= 0} is the
Hamiltonian flow of the metric norm function
v
u n
uX
(1.26) H(x, ξ) = t g ij (x)ξi ξj .
i,j=1

It is free particle motion (with V = 0) on M . When ∂M 6= 0 the geodesic reflects off

the boundary by Snell’s law of equal angles. This flow is called the broken geodesic
flow or billiard flow.
The Bohr correspondence principle suggests that as λj → ∞ the asymptotics of
eigenfunctions and eigenvalues should be related to dynamics of the geodesic flow.
The relations between eigenfunctions and the Hamiltonian flow are best established
in two extreme cases: (i) where the Hamiltonian flow is completely integrable on an
energy surface, or (ii) where it is ergodic. The hydrogen atom is completely inte-
grable and that is why the special eigenfunctions which are joint eigenfunctions of
the Schrödinger operator, the total angular momentum and the z-component of the
angular momentum, can be completely understood. These are the eigenfunctions
whose images are graphed here. Integrable systems are rare but important in that
many of the known results are obtained by perturbing
the integrable case. The quantum integrable case is discussed in some detail in
§11.
Ergodic (or more chaotically mixing) dynamical systems are more difficult than
integrable systems because they are not explicitly solvable. However, as with the
law of large numbers or central limit theorem in probability theory, chaos induces a
kind of symmetry or uniformity which makes it possible to prove results by indirect
calculations and results.
The extremes are illustrated below in the case of (a) billiards on rotationally
invariant annulus, (b) chaotic billiards on a cardioid.
 1.6. INTENSITY PLOTS AND EXCURSION SETS 7

A typical trajectory in the case of ergodic billiards is uniformly distributed,

while all trajectories are quasi-periodic in the integrable case.

1.6. Intensity plots and excursion sets

There are several ways to ‘picture’ an eigenfunction and the probability density
(1.23) that it defines. One vivid kind of picture of a hydrogen atom is an intensity
plot which darkens in the regions where |ϕj (x)|2 is large (most probable locations).

The most probable locations are defined by the excursion sets

(1.27) Ωj,~,E = {x : |ψ~,j (x)|2 ≥ E}.

It is particularly interesting to understand the high excursion levels, where E '

A~−r . One would ideally like to know how the excursion sets are distributed in the
semiclassical limit ~ → 0. How many connected components does it have and what
are there shapes and locations? What is the distribution function

(1.28) µψ~,j [E, ∞) := Vol{x : |ψ~,j (x)|2 ≥ E},

where Vol denotes the volume measure of E (corresponding to the metric underlying
the Laplacian ∆). One could also use (1.23) as the measure to determine the relative
proportion of the L2 mass of the eigenfunction which is concentrated near its top
values.
One may also imagine graphing (1.23) over the high-dimensional configuration
space and asking for the prominent features of the graph. In the case of a high
frequency spherical harmonic on S 2 one may obtain the graph:
8 1. INTRODUCTION

One observes that there are many local maxima near the peak values of the
eigenfunction (or its square (1.23)). They appear to be rather uniformly distributed.
Can one at least prove that the number of critical points tends to infinity with
the eigenvalue (or equivalently as ~ → 0)? This is known to be false for some
eigenfunctions of general Riemannian manifolds. How does the distribution or
number of critical points reflect the underlying Hamiltonian dynamics?
These questions are almost completely open and (as in the images) are most
accessible for quantum integrable systems. The high excursion sets are the most
important sets, but also rather intractable since they involve the distribution func-
tion of the eigenfunction. The only case known to the author where the distribution
function has been discussed is in the case of toric eigenfunctions on Kähler mani-
folds [STZ]. It is likely that analogous results can be proved for joint eigenfunctions
of real integrable systems such as surfaces of revolution or (as in the images) the
hydrogen atom eigenfunctions.

1.7. Nodal sets and critical point sets

At the opposite are plots of the nodal hypersurfaces: the zero set
(1.29) Nϕλ = {x ∈ M : ϕλ (x) = 0}.

These are the points where the probability (density) of the particle’s position
vanishes. Here, we consider eigenfunctions of the Laplacian ∆g of a compact Rie-
mannian manifold rather than a general Schrödinger operator.
The nodal domains of ϕλ are the connected components Ωj of M \Nϕλ =
SN (ϕλ )
j=1 Ωj . We write
(1.30) N (ϕλ ) := the number of nodal domains of ϕλ .
In [Br], J. Brüning (and Yau, unpublished) showed that H1 (Zϕλ ) ≥ Cg λ in
the dim M = 2 case, i.e., the length of a nodal line is bounded below by a constant
 1.8. LOCAL VERSUS GLOBAL ANALYSIS OF EIGENFUNCTIONS 9

multiple of the frequency for some constant Cg > 0. A. Logunov has recently proved
the analogous result in all dimensions [Lo].
According to Courant’s nodal domain theorem [C], there exists a universal
upper bound for N (ϕj ):
(1.31) N (ϕj ) ≤ j.
In terms of order of magnitude, this bound is often obtained: when M is the unit
sphere S 2 and ϕ is a random spherical harmonics, then N (ϕλ ) ∼ cλ2 holds almost
surely for some constant c > 0 thanks to Nazarov-Sodin [NS] . However, it is known
that the bound is not always sharp in terms of order of magnitude. In the chapter
on nodal sets ♣ref♣, we will review results of H. Lewy and others that construct
sequences of eigenfunctions with a uniform bound on the number of nodal domains.
On the other hand, it is very plausible that every compact Riemannian manifold
possesses a sequence of eigenfunctions for which the number of nodal domains
tends to infinity. In the same chapter, we prove this to be true for almost the entire
sequence of eigenfunctions of a non-positively curved surface with concave boundary
(for Dirichlet or Neumann boundary conditions) and for negatively curved surfaces
possessing an anti-holomorphic isometric involution with dividing fixed point set.
Closely related to nodal sets are the other level sets
(1.32) Nϕaj = {x ∈ M : ϕj (x) = a}
and the sublevel sets
(1.33) {x ∈ M : |ϕj (x)| ≤ a}.
The zero level is distinguished since the symmetry ϕj → −ϕj in the equation pre-
serves the nodal set. A fundamental existence result states that there exists a
constant A > 0 so that every ball of (M, g) contains a nodal point of any eigen-
function ϕλ if its radius is greater than Aλ.
Of equal interest is the critical point set
(1.34) Cϕj = {x ∈ M : ∇ϕj (x) = 0}.
The critical point set can be a hypersurface in M . In counting problems it is better
to consider the set
(1.35) Vϕj = {ϕj (x) : ∇ϕj (x) = 0}
of critical values. At this time of writing, there exist (to the author’s knowledge)
no rigorous upper bounds on the number of critical values except in separation-of-
variables situations.

1.8. Local versus global analysis of eigenfunctions

As will be discussed in detail in §5.3, the local study of eigenfunctions uses
analysis on small balls of radii Cλ . One does necessarily assume that the eigen-
functions are global, i.e., that they are eigenfunctions on a global closed manifold
without boundary, or that they satisfy Dirichlet or Neumann boundary conditions
on a manifold with boundary.
Global harmonic analysis concerns the properties of global eigenfunctions. A
key property is that they are eigenfunctions of the evolution operator
√
−∆
(1.36) U (t) = eit
10 1. INTRODUCTION

or propagator

it
(1.37) Uh (t) = e ~ Ĥ

for semiclassical Schrödinger operators.

The goal is then to relate the behavior of eigenfunctions in the semiclassical
limit λj → ∞ or ~ → 0 to properties of the geodesic flow, or more generally the
Hamiltonian flow of 21 |ξ|2 + V (x) on a fixed energy surface.

1.9. High frequency limits, oscillation and concentration

The emphasis of these lectures is on high frequency limits λj → ∞ of zeros sets,
norms and mass distribution of sequences of eigenfunctions. For general Schrödinger
operators, one studies the semiclassical limit ~ → 0.
In analogy with polynomials, the degree of a polynomial, resp. the frequency λ
of an eigenfunction, is a measure of its “complexity” and the high frequency limit
is the large complexity limit. A sequence of eigenfunctions of increasing frequency
oscillates more and more rapidly and the problem is to find its “limit shape”. In
the graph below of (sin kx)2 , the square of the eigenfunction tends in the weak*
sense of measures to its mean value. Thus the oscillations smear out to an average
value. The eigenfunction sequence itself always tends to zero weakly in L2 .

But it is also possible that a sequence of squares will concentrate on a low-

dimensional subset, as in this picture of a sequence of Gaussians tending to the
delta function at 0.
 1.10. SPECTRAL PROJECTIONS 11

In the Riemannian case, there exist sequences of squares of eigenfunctions called

Gaussian beams which put the two types together: they oscillate more and more
rapidly along a geodesic γ and have Gaussian decay in the transverse direction, so
that in the limit they tend to a delta function along γ.

1.10. Spectral projections

We now specialize the eigenvalue problem to the setting of Laplacians on com-
pact Riemannian manifold (M, g), i.e. we set V = 0. The quantum Hamiltonian is
then the Laplacian (1.6).
We denote by Π[0,λ] = Π[0,h−1/2 ] the spectral projections kernel for the interval
[0, λ]:
X
(1.38) Π[0,λ] (x, y) = Πλ (x, y) := ϕj (x)ϕj (y).
j : λj ≤λ

It is the Schwartz kernel of the orthogonal projection onto the span of the eigenfunc-
tions with frequencies ≤ λ, and is independent of the choice of orthonormal basis.
Sometimes we wish to consider shorter spectral intervals, and then subscript the
projection by the relevant interval. An important case is the spectral projections
for a short interval:
X
(1.39) Π[λ,λ+1] (x, y) := ϕj (x)ϕj (y).
j : λ≤λj ≤λ+1

We mainly consider compact Riemannian manifolds in this monograph, al-

though many of the same problems and techniques are valid on non-compact com-
plete Riemannian manifolds. Our main concern is to relate the behavior of eigenval-
ues/eigenfunctions to the dynamics of the geodesic flow, and the setting of compact
Riemannian manifolds is sufficiently rich to illustrate the possible relations. In the
non-compact setting, the spectrum is continuous (possibly with embedded eigen-
values) and there is a basis of generalized eigenfunctions. We will briefly consider
examples such as the hyperbolic plane and hyperbolic cylinder.
In the non-compact case, it is also natural to study resonances instead of eigen-
values and resonance states instead of eigenfunctions, but the theory is quite dif-
ferent and is not discussed here. See [DZw] for a comprehensive exposition. It is
12 1. INTRODUCTION

also natural to study the scattering phase shifts (eigenvalues) and eigenfunctions
of the scattering operator S(h) (see [GHZ] for references).
Remark 1.2. We set the potential V equal to zero for the sake of brevity, but
almost everything we do generalizes, often in subtle ways, to Schrödinger operators.
The dynamics of geodesic flows is sufficiently rich to exhibit the relations between
classical and quantum mechanics. More general Schrödinger operators −~2 ∆g + V
(or magnetic Schrödinger operators) give rise to significant additional issues that
in some cases have barely been explored, such as the behavior of nodal sets in the
forbidden region.
Most of the problems stated above are not only unsolved but appear to be
completely intractable. They are most accessible when the Schrödinger operator
is completely integrable on the quantum level in the sense of §1.19, although the
problems remain unsolved even in this case. There are many phenomena which
show up clearly in numerical plots yet which are far beyond mathematical analysis.
Therefore we need to simplify the problems to the point where rigorous results are
possible.

1.11. Lp norms
As mentioned above, excursion sets (1.27) are difficult to study in all but the
simplest cases (such as the standard spheres or surfaces of revolution). A somewhat
more accessible and natural mathematical problem is to study the Lp norms of
eigenfunctions (to the pth power),
Z ∞ Z
(1.40) tp dµϕj (t) = |ϕj |2p dV
0
as a function of the eigenvalue Ej (~). Here, µϕj is the distribution function of |ϕj |2
(1.28). Different powers measure different aspects of the intensity plot. Since ϕj is
L2 -normalized (1.22), high Lp norms, (e.g., the sup norm kϕj k∞ = supx |ϕj (x)|) is
large when there exist a few very high peaks and are not so large when there exist
many relatively shallow peaks. Lower Lp norms are large when the set of rather
large values has a large measure. A random spherical harmonic spreads its mass
rather evenly around the sphere, and thus has relatively small Lp norms for high p.
General upper bounds on Lp norms of eigenfunctions will be discussed in §10.
In §10.4 we discuss the case Riemannian manifolds possessing sequences of eigen-
functions achieving the maximal allowed growth for large p. The case of small p
is not understood in general. Lp norms of quantum integrable eigenfunctions are
discussed in §11.6.

1.12. Matrix elements and Wigner distributions

Lp norms are ‘non-linear’ measures of the size of the eigenfunction. The most
linear measures are the matrix elements (1.24). For instance, if A is multiplication
by the characteristic function 1E of a Borel set E ⊂ M , one is measuring the
L2 -mass
Z
(1.41) |ϕj |2 dV
E
of the eigenfunction in the set E, e.g., to determine where it concentrates most.
In the case where the boundary ∂E (closure of E minus its interior) has measure
 1.13. EGOROV’S THEOREM 13

zero, there are many results on the semiclassical limits of the L2 mass. Virtually
nothing is known for more general E such as Cantor sets of positive measure. It is
possible to allow E to depend on ~ and to let it shrink at a specified rate as ~ → 0.
Such “small-scale mass” is closely related to Lp norms.
The diagonal matrix elements
(1.42) ρj (A) := hAϕj , ϕj i
of an observable A (i.e. a bounded operator on L2 (M )) are interpreted in quantum
mechanics as the expected value of the observable A in the energy state ϕj . The
off-diagonal matrix elements
(1.43) ρjk (A) = hAϕi , ϕj i, (j 6= k)
are interpreted as transition amplitudes. Here, and below, an amplitude is a com-
plex number whose modulus square is a probability.
There is a special class of observables A for which it is possible to study semiclas-
sical limits of matrix elements (1.42), namely (various kinds of) pseudo-differential
operators Op~ (a) = a(x, ~D). Such operators are understood as ‘quantizations’
of classical observables, namely functions a(x, ξ) on phase space. For instance,
one may let a(x, ξ) = 1E (x, ξ) where E ⊂ T ∗ M is a nice Borel set. Then (1.42)
measures the phase space mass of the eigenfunction in E, i.e., the probability
(1.44) h1E ϕj , ϕj i
that its (position, momentum) = (x, ξ) belong to E. Just as ϕj determines the
probability measure (1.23) on configuration space, so it also induces probability
measures on phase space.
If we fix the quantization a → Op~ (a), then the matrix elements can be rep-
resented by Wigner distributions. In the diagonal case, we define Wk ∈ D0 (T ∗ M )
by
Z
(1.45) a dWk := hOp~ (a)ϕk , ϕk i.
T ∗M
Here, we are using semiclassical pseudo-differential operators (see [DSj, Zw]). If
we use homogeneous pseudo-differential operators, the Wigner distributions may
be defined as distributions on the unit co-sphere bundle S ∗ M .
The basic compactness theorem regarding the sequence of probability measures
(1.23) or their microlocal lifts is simply the compactness of probability measures
in the weak* topology. As the name suggests, weak* convergence is a very weak
type of convergence and it is difficult to determine many concrete properties of
eigenfunctions even from knowledge of the limit measures.

1.13. Egorov’s theorem

Egorov’s theorem is the precise statement of the correspondence between the
Heisenberg time evolution Ut AUt∗ of an observable A and the time evolution of the
classical observable (its symbol) σA ◦Gt , where Gt is the corresponding Hamiltonian
(geodesic) flow. It states that if A ∈ Ψ0 (M ) (i.e., A is a pseudo-differential operator
of order zero), then
(1.46) U~t Oph (a)U~−t − Oph (a ◦ Gt ) ∈ Ψ−1
h (M ),
i.e., the difference is a pseudo-differential operator of order −1. In semiclassical
notation, order −1 means of order O(~).
14 1. INTRODUCTION

1.14. Eherenfest time

The aim in quantum chaos is to obtain information about the high energy
asymptotics as λj → ∞ of eigenvalues and eigenfunctions by connecting infor-
mation about U t and Gt . The connection often comes from Egorov’s theorem
(Theorem 1.46). But to use the hypothesis that Gt is ergodic or chaotic, one needs
to exploit the connection as ~ → 0 and t → ∞. The difficulty in quantum chaos is
that the approximation of U t by Gt is only a good one for t less than the Eherenfest
time
log |~|
(1.47) TE = ,
λmax
where λmax is the so-called maximal Lyapunov exponent.
Roughly speaking, the idea is that the evolution of a well constructed “coher-
ent” quantum state or particle is a moving lump that “tracks along” the trajectory
of a classical particle up to time TE and then slowly falls apart and stops acting
like a classical particle. Numerical studies of long time dynamics of wave pack-
ets are given in works of E.J. Heller [He1, He2] and rigorous treatments are in
Bouzouina-Robert [BoR], Combescure-Robert [CR] and Schubert [S].
The basic result expressed in semiclassical notation is that there exists Γ > 0
such that
(1.48) kU~t Op(a)U~−t − Op(a ◦ Gt )k ≤ C~etΓ .
The exponential growth rate in t has long been known to be the essential stumbling
block to precise localization in the spectrum. Thus, one only expects good joint
asymptotics as ~ → 0, t → ∞ for t ≤ TE . As a result, one can only exploit the
approximation of U t by Gt for the relatively short time TE .

1.15. Weak* limit problem

There are two (equivalent) ways to state the weak* limit problem: (i) in terms
of quantum statistical mechanical states on the algebra of observables, or (ii) in
terms of Wigner distributions (or microlocal lifts, microlocal defect measures, etc.).
The first is more abstract or at least less PDE oriented but is useful in not requiring
any choice of quantization of classical observables. The second is more concrete.
The diagonal matrix elements define linear functionals (1.42) on Ψ0 . We observe
that ρj (I) = 1, that ρj (A) ≥ 0 if A ≥ 0 and that
(1.49) ρk (U t AU −t ) = ρk (A).
Indeed, if A ≥ 0 then A = B ∗ B for some B ∈ Ψ0 and we can move B ∗ to the right
side. Similarly (1.49) is proved by moving Ut to the right side and using the fact
that the eigenvalues of Ut are of modulus one. In quantum statistical mechanics,
these properties are summarized by saying that ρj is an invariant state on the
algebra Ψ0 , or more precisely, on its closure in the operator norm. An invariant
state is the analogue in quantum statistical mechanics of an invariant probability
measure.
We denote by MI the convex set of invariant probability measures for the
geodesic flow. Further, we say that a measure is time-reversal invariant if it is
invariant under the anti-symplectic involution (x, ξ) → (x, −ξ) on T ∗ M . We denote
the time-reversal invariant elements of MI by M+ I .
 1.15. WEAK* LIMIT PROBLEM 15

Proposition 1.3. Any weak limit of the sequence {ρj } on Ψ0 is a time-reversal

invariant, Gt invariant probability measure on S ∗ M , i.e. is an element of M+
I .

Proof. For any compact operator K, hKϕj , ϕj i → 0. Hence, any limit of

hAϕj , ϕj i is equally a limit of h(A + K)ϕj , ϕj i. By the norm estimate, the limit is
bounded by inf K kA + Kk (the infimum taken over compact operators). Hence any
weak limit is bounded by a constant times the sup norm kσA kL∞ of the symbol
σA of A and is therefore continuous on C(S ∗ M ). It is a positive functional since
each ρj is and hence any limit is a probability measure. By Egorov’s theorem and
the invariance of ρj , any limit of ρj (A) is a limit of ρj (Op(σA ◦ Gt )) and hence
the limit measure is invariant. It is also time-reversal when the eigenfunctions are
real-valued, i.e., complex conjugation invariant. 

The Wigner distributions (also called microlocal lifts) dWk defined by (1.45) of
course depend on the choice of Op(a). If a is chosen to be homogeneous of degree
0 on T ∗ M − {0} (the zero section) then one can arrange that dwk ∈ D0 (S ∗ M ). In
the semiclassical setting one deals with non-homogeneous symbols. Eigenfunctions
localize on the ‘energy surface’ {H = 1}, i.e., on the unit co-sphere bundle Sg∗ M in
the case of the Laplacian, and the corresponding microlocal lifts concentrate there.
Problem 1.4. Determine the set Q of ‘quantum limits,’ i.e., weak* limit points
of the sequence {dWk }.
The set Q is independent of the definition of quantization a → Op(a). The
simplest examples are the exponentials on a flat torus Rm /Zm . By definition of
pseudo-differential operator, Ae2πihk,xi = a(x, k)e2πihk,xi where a(x, k) is the com-
plete symbol. Thus,
 
k
Z Z
(1.50) hAe2πihk,xi , e2πihk,xi i = a(x, k)dx ∼ σA x, dx.
Rn /Zn Rn /Zn |k|
k
A subsequence e2πihkj ,xi of eigenfunctions has a weak limit if and only if |kjj | tends
to Ra limit vector ξ0 in the unit sphere in Rn . In this case, the associated weak* limit
is Rn /Zn σA (x, ξ0 )dx, i.e., the delta-function on the invariant torus Tξ0 ⊂ S ∗ M for
Gt , defined by the constant momentum condition ξ = ξ0 . The eigenfunctions are
said to localize on this invariant torus. Given ξ0 , we can always define a sequence
k
kj so that |kjj | → ξ0 , and thus, every invariant torus measure arises as a quantum
limit.
In general, there are many possible limit measures. The most important are:
(1) Normalized Liouville measure µL . In fact, the functional ω of integration
against normalized Liouville measure is also a state on Ψ0 for the reason
explained above. A subsequence {ϕjk } of eigenfunctions is considered
diffuse if ρjk → ω.
(2) A periodic orbit measure µγ defined by µγ (A) = L1γ γ σA ds where Lγ
R

is the length of γ. A sequence of eigenfunctions for which ρkj → µγ

obviously concentrates (or strongly ‘scars’) on the closed geodesic.
(3) A finite sum of periodic orbit measures.
(4) A delta-function along an invariant Lagrangian manifold Λ ⊂ S ∗ M . The
associated eigenfunctions are viewed as localizing along Λ.
(5) A more general measure which is singular with respect to dµ.
16 1. INTRODUCTION

All of these possibilities arise as (M, g) varies among Riemannian manifolds.

Indeed, the standard sphere provides an extreme example (see [JZ])
Theorem 1.5. For the standard round sphere S n , Q = M+
I .
In the case where ρkj → ω, the corresponding eigenfunctions become uniformly
distributed on the energy surface S ∗ M . By testing against multiplication operators,
one gets
1 Vol(E)
Z
(1.51) |ϕkj (x)|2 d Vol →
Vol(M ) E Vol(M )
for any measurable set E whose boundary has measure zero. In the interpretation
of |ϕkj (x)|2 d Vol as the probability density of finding a particle of energy λ2kj at
x, this says that the sequence of probabilities tends to uniform measure. However,
ρkj → ω is much stronger since it says that the eigenfunctions become uniformly
distributed on S ∗ M and not just on the configuration space M . For instance, on
the flat torus Rn /Zn , the standard exponentials e2πihk,xi satisfy |e2πihk,xi |2 = 1,
and are thus uniformly distributed in configuration space. On the other hand, as
seen above, in phase space they localize on invariant Lagrange tori in S ∗ M .
The flat torus is a model of a completely integrable system, on both the classical
and quantum levels. On the other hand, if the geodesic flow is ergodic one would
expect the eigenfunctions to be diffuse in phase space. The statement that the all
eigenfunctions are diffuse, i.e., Q = {ω}, is known as quantum unique ergodicity.
It will be discussed in §1.18.
Off-diagonal matrix elements (1.43) are also important as transition amplitudes
between states. They no longer define states since ρjk (I) = 0, are no longer posi-
tive, and are no longer invariant. Indeed, ρj,k (Ut AUt∗ ) = eit(λj −λk ) ρjk (A), so they
are eigenvectors of the automorphism αt (A) = Ut AUt∗ . A sequence of such matrix
elements cannot have a weak limit unless the spectral gap λj − λk tends to a limit
τ ∈ R. In this case, by the same discussion as above, any weak limit of the func-
tionals ρjk will be a time-reversal invariant eigenmeasure of the geodesic flow which
transforms by eiτ t under the action of Gt . Examples of such eigenmeasures are or-
RL
bital Fourier coefficients L1γ 0 γ e−iτ t σA (Gt (x, ξ))dt along a periodic orbit. Here
τ ∈ L2πγ Z. We denote by Qτ such eigenmeasures of the geodesic flow. Problem 1.4
has the following extension to off-diagonal elements:
Problem 1.6. Determine the set Qτ of ‘quantum limits’, i.e., weak* limit
points of the sequence {ρjk } on the classical phase space T ∗ M .
As will be discussed in ♣§9.5.8♣, the asymptotics of off-diagonal elements de-
pends on the weak mixing properties of the geodesic flow and not just its ergodicity.

1.16. Ergodic versus completely integrable geodesic flow

In line with §1.5, one of the principal cases where one can largely control the
weak* limits of the Wigner distributions and the complex nodal currents is that of
Riemannian manifolds (M, g) with ergodic geodesic flow. When M has a boundary
∂M then the geodesic flow is replace by the billiard flow. In the images below,
the left one shows a typical trajectory of the ergodic billiards in a stadium, and
the right one gives intensity plots of the ergodic Dirichlet eigenfunctions, as well
as several modes of ‘bouncing ball type,’ corresponding to vertical bouncing ball
orbits in the middle rectangle.
 1.17. ERGODIC EIGENFUNCTIONS 17

1.17. Ergodic eigenfunctions

A subsequence {ϕjk } of eigenfunctions is called quantum ergodicquantum er-
godic if the only weak* limit of the sequence of ρjk is dµL or equivalently the
Liouville state ω.
One can quantize characteristic functions 1E of open sets in S ∗ M whose bound-
aries have measure zero. Then
(1.52)
hOp(1E )ϕj , ϕj i = the amplitude that the particle in energy state λ2j lies in E.
For an ergodic sequence of eigenfunctions,
µL (E)
(1.53) hOp(1E )ϕjk , ϕjk i → ,
µL (S ∗ M )
so that the particle becomes diffuse, i.e. uniformly distributed on S ∗ M . This is
the quantum analogue of the property of uniform distribution of typical geodesics
of ergodic geodesic flows (Birkhoff’s ergodic theorem).
Another random document with
no related content on Scribd:
dress.
“Do you remember,” she said, “how the blinds used to flap at
Bourton?”
“They did,” he said; and he remembered breakfasting alone, very
awkwardly, with her father; who had died; and he had not written to
Clarissa. But he had never got on well with old Parry, that querulous,
weak-kneed old man, Clarissa’s father, Justin Parry.
“I often wish I’d got on better with your father,” he said.
“But he never liked any one who—our friends,” said Clarissa; and
could have bitten her tongue for thus reminding Peter that he had
wanted to marry her.
Of course I did, thought Peter; it almost broke my heart too, he
thought; and was overcome with his own grief, which rose like a
moon looked at from a terrace, ghastly beautiful with light from the
sunken day. I was more unhappy than I’ve ever been since, he
thought. And as if in truth he were sitting there on the terrace he
edged a little towards Clarissa; put his hand out; raised it; let it fall.
There above them it hung, that moon. She too seemed to be sitting
with him on the terrace, in the moonlight.
“Herbert has it now,” she said. “I never go there now,” she said.
Then, just as happens on a terrace in the moonlight, when one
person begins to feel ashamed that he is already bored, and yet as
the other sits silent, very quiet, sadly looking at the moon, does not
like to speak, moves his foot, clears his throat, notices some iron
scroll on a table leg, stirs a leaf, but says nothing—so Peter Walsh
did now. For why go back like this to the past? he thought. Why
make him think of it again? Why make him suffer, when she had
tortured him so infernally? Why?
“Do you remember the lake?” she said, in an abrupt voice, under the
pressure of an emotion which caught her heart, made the muscles of
her throat stiff, and contracted her lips in a spasm as she said “lake.”
For she was a child, throwing bread to the ducks, between her
parents, and at the same time a grown woman coming to her parents
who stood by the lake, holding her life in her arms which, as she
neared them, grew larger and larger in her arms, until it became a
whole life, a complete life, which she put down by them and said,
“This is what I have made of it! This!” And what had she made of it?
What, indeed? sitting there sewing this morning with Peter.
She looked at Peter Walsh; her look, passing through all that time
and that emotion, reached him doubtfully; settled on him tearfully;
and rose and fluttered away, as a bird touches a branch and rises
and flutters away. Quite simply she wiped her eyes.
“Yes,” said Peter. “Yes, yes, yes,” he said, as if she drew up to the
surface something which positively hurt him as it rose. Stop! Stop! he
wanted to cry. For he was not old; his life was not over; not by any
means. He was only just past fifty. Shall I tell her, he thought, or not?
He would like to make a clean breast of it all. But she is too cold, he
thought; sewing, with her scissors; Daisy would look ordinary beside
Clarissa. And she would think me a failure, which I am in their sense,
he thought; in the Dalloways’ sense. Oh yes, he had no doubt about
that; he was a failure, compared with all this—the inlaid table, the
mounted paper-knife, the dolphin and the candlesticks, the chair-
covers and the old valuable English tinted prints—he was a failure! I
detest the smugness of the whole affair he thought; Richard’s doing,
not Clarissa’s; save that she married him. (Here Lucy came into the
room, carrying silver, more silver, but charming, slender, graceful she
looked, he thought, as she stooped to put it down.) And this has
been going on all the time! he thought; week after week; Clarissa’s
life; while I—he thought; and at once everything seemed to radiate
from him; journeys; rides; quarrels; adventures; bridge parties; love
affairs; work; work, work! and he took out his knife quite openly—his
old horn-handled knife which Clarissa could swear he had had these
thirty years—and clenched his fist upon it.
What an extraordinary habit that was, Clarissa thought; always
playing with a knife. Always making one feel, too, frivolous; empty-
minded; a mere silly chatterbox, as he used. But I too, she thought,
and, taking up her needle, summoned, like a Queen whose guards
have fallen asleep and left her unprotected (she had been quite
taken aback by this visit—it had upset her) so that any one can stroll
in and have a look at her where she lies with the brambles curving
over her, summoned to her help the things she did; the things she
liked; her husband; Elizabeth; her self, in short, which Peter hardly
knew now, all to come about her and beat off the enemy.
“Well, and what’s happened to you?” she said. So before a battle
begins, the horses paw the ground; toss their heads; the light shines
on their flanks; their necks curve. So Peter Walsh and Clarissa,
sitting side by side on the blue sofa, challenged each other. His
powers chafed and tossed in him. He assembled from different
quarters all sorts of things; praise; his career at Oxford; his marriage,
which she knew nothing whatever about; how he had loved; and
altogether done his job.
“Millions of things!” he exclaimed, and, urged by the assembly of
powers which were now charging this way and that and giving him
the feeling at once frightening and extremely exhilarating of being
rushed through the air on the shoulders of people he could no longer
see, he raised his hands to his forehead.
Clarissa sat very upright; drew in her breath.
“I am in love,” he said, not to her however, but to some one raised up
in the dark so that you could not touch her but must lay your garland
down on the grass in the dark.
“In love,” he repeated, now speaking rather dryly to Clarissa
Dalloway; “in love with a girl in India.” He had deposited his garland.
Clarissa could make what she would of it.
“In love!” she said. That he at his age should be sucked under in his
little bow-tie by that monster! And there’s no flesh on his neck; his
hands are red; and he’s six months older than I am! her eye flashed
back to her; but in her heart she felt, all the same, he is in love. He
has that, she felt; he is in love.
But the indomitable egotism which for ever rides down the hosts
opposed to it, the river which says on, on, on; even though, it admits,
there may be no goal for us whatever, still on, on; this indomitable
egotism charged her cheeks with colour; made her look very young;
very pink; very bright-eyed as she sat with her dress upon her knee,
and her needle held to the end of green silk, trembling a little. He
was in love! Not with her. With some younger woman, of course.
“And who is she?” she asked.
Now this statue must be brought from its height and set down
between them.
“A married woman, unfortunately,” he said; “the wife of a Major in the
Indian Army.”
And with a curious ironical sweetness he smiled as he placed her in
this ridiculous way before Clarissa.
(All the same, he is in love, thought Clarissa.)
“She has,” he continued, very reasonably, “two small children; a boy
and a girl; and I have come over to see my lawyers about the
divorce.”
There they are! he thought. Do what you like with them, Clarissa!
There they are! And second by second it seemed to him that the wife
of the Major in the Indian Army (his Daisy) and her two small children
became more and more lovely as Clarissa looked at them; as if he
had set light to a grey pellet on a plate and there had risen up a
lovely tree in the brisk sea-salted air of their intimacy (for in some
ways no one understood him, felt with him, as Clarissa did)—their
exquisite intimacy.
She flattered him; she fooled him, thought Clarissa; shaping the
woman, the wife of the Major in the Indian Army, with three strokes
of a knife. What a waste! What a folly! All his life long Peter had been
fooled like that; first getting sent down from Oxford; next marrying
the girl on the boat going out to India; now the wife of a Major in the
Indian Army—thank Heaven she had refused to marry him! Still, he
was in love; her old friend, her dear Peter, he was in love.
“But what are you going to do?” she asked him. Oh the lawyers and
solicitors, Messrs. Hooper and Grateley of Lincoln’s Inn, they were
going to do it, he said. And he actually pared his nails with his
pocket-knife.
For Heaven’s sake, leave your knife alone! she cried to herself in
irrepressible irritation; it was his silly unconventionality, his
weakness; his lack of the ghost of a notion what any one else was
feeling that annoyed her, had always annoyed her; and now at his
age, how silly!
I know all that, Peter thought; I know what I’m up against, he
thought, running his finger along the blade of his knife, Clarissa and
Dalloway and all the rest of them; but I’ll show Clarissa—and then to
his utter surprise, suddenly thrown by those uncontrollable forces
thrown through the air, he burst into tears; wept; wept without the
least shame, sitting on the sofa, the tears running down his cheeks.
And Clarissa had leant forward, taken his hand, drawn him to her,
kissed him,—actually had felt his face on hers before she could
down the brandishing of silver flashing—plumes like pampas grass
in a tropic gale in her breast, which, subsiding, left her holding his
hand, patting his knee and, feeling as she sat back extraordinarily at
her ease with him and light-hearted, all in a clap it came over her, If I
had married him, this gaiety would have been mine all day!
It was all over for her. The sheet was stretched and the bed narrow.
She had gone up into the tower alone and left them blackberrying in
the sun. The door had shut, and there among the dust of fallen
plaster and the litter of birds’ nests how distant the view had looked,
and the sounds came thin and chill (once on Leith Hill, she
remembered), and Richard, Richard! she cried, as a sleeper in the
night starts and stretches a hand in the dark for help. Lunching with
Lady Bruton, it came back to her. He has left me; I am alone for ever,
she thought, folding her hands upon her knee.
Peter Walsh had got up and crossed to the window and stood with
his back to her, flicking a bandanna handkerchief from side to side.
Masterly and dry and desolate he looked, his thin shoulder-blades
lifting his coat slightly; blowing his nose violently. Take me with you,
Clarissa thought impulsively, as if he were starting directly upon
some great voyage; and then, next moment, it was as if the five acts
of a play that had been very exciting and moving were now over and
she had lived a lifetime in them and had run away, had lived with
Peter, and it was now over.
Now it was time to move, and, as a woman gathers her things
together, her cloak, her gloves, her opera-glasses, and gets up to go
out of the theatre into the street, she rose from the sofa and went to
Peter.
And it was awfully strange, he thought, how she still had the power,
as she came tinkling, rustling, still had the power as she came
across the room, to make the moon, which he detested, rise at
Bourton on the terrace in the summer sky.
“Tell me,” he said, seizing her by the shoulders. “Are you happy,
Clarissa? Does Richard—”
The door opened.
“Here is my Elizabeth,” said Clarissa, emotionally, histrionically,
perhaps.
“How d’y do?” said Elizabeth coming forward.
The sound of Big Ben striking the half-hour struck out between them
with extraordinary vigour, as if a young man, strong, indifferent,
inconsiderate, were swinging dumb-bells this way and that.
“Hullo, Elizabeth!” cried Peter, stuffing his handkerchief into his
pocket, going quickly to her, saying “Good-bye, Clarissa” without
looking at her, leaving the room quickly, and running downstairs and
opening the hall door.
“Peter! Peter!” cried Clarissa, following him out on to the landing. “My
party to-night! Remember my party to-night!” she cried, having to
raise her voice against the roar of the open air, and, overwhelmed by
the traffic and the sound of all the clocks striking, her voice crying
“Remember my party to-night!” sounded frail and thin and very far
away as Peter Walsh shut the door.
Remember my party, remember my party, said Peter Walsh as he
stepped down the street, speaking to himself rhythmically, in time
with the flow of the sound, the direct downright sound of Big Ben
striking the half-hour. (The leaden circles dissolved in the air.) Oh
these parties, he thought; Clarissa’s parties. Why does she give
these parties, he thought. Not that he blamed her or this effigy of a
man in a tail-coat with a carnation in his buttonhole coming towards
him. Only one person in the world could be as he was, in love. And
there he was, this fortunate man, himself, reflected in the plate-glass
window of a motor-car manufacturer in Victoria Street. All India lay
behind him; plains, mountains; epidemics of cholera; a district twice
as big as Ireland; decisions he had come to alone—he, Peter Walsh;
who was now really for the first time in his life, in love. Clarissa had
grown hard, he thought; and a trifle sentimental into the bargain, he
suspected, looking at the great motor-cars capable of doing—how
many miles on how many gallons? For he had a turn for mechanics;
had invented a plough in his district, had ordered wheel-barrows
from England, but the coolies wouldn’t use them, all of which
Clarissa knew nothing whatever about.
The way she said “Here is my Elizabeth!”—that annoyed him. Why
not “Here’s Elizabeth” simply? It was insincere. And Elizabeth didn’t
like it either. (Still the last tremors of the great booming voice shook
the air round him; the half-hour; still early; only half-past eleven still.)
For he understood young people; he liked them. There was always
something cold in Clarissa, he thought. She had always, even as a
girl, a sort of timidity, which in middle age becomes conventionality,
and then it’s all up, it’s all up, he thought, looking rather drearily into
the glassy depths, and wondering whether by calling at that hour he
had annoyed her; overcome with shame suddenly at having been a
fool; wept; been emotional; told her everything, as usual, as usual.
As a cloud crosses the sun, silence falls on London; and falls on the
mind. Effort ceases. Time flaps on the mast. There we stop; there we
stand. Rigid, the skeleton of habit alone upholds the human frame.
Where there is nothing, Peter Walsh said to himself; feeling hollowed
out, utterly empty within. Clarissa refused me, he thought. He stood
there thinking, Clarissa refused me.
Ah, said St. Margaret’s, like a hostess who comes into her drawing-
room on the very stroke of the hour and finds her guests there
already. I am not late. No, it is precisely half-past eleven, she says.
Yet, though she is perfectly right, her voice, being the voice of the
hostess, is reluctant to inflict its individuality. Some grief for the past
holds it back; some concern for the present. It is half-past eleven,
she says, and the sound of St. Margaret’s glides into the recesses of
the heart and buries itself in ring after ring of sound, like something
alive which wants to confide itself, to disperse itself, to be, with a
tremor of delight, at rest—like Clarissa herself, thought Peter Walsh,
coming down the stairs on the stroke of the hour in white. It is
Clarissa herself, he thought, with a deep emotion, and an
extraordinarily clear, yet puzzling, recollection of her, as if this bell
had come into the room years ago, where they sat at some moment
of great intimacy, and had gone from one to the other and had left,
like a bee with honey, laden with the moment. But what room? What
moment? And why had he been so profoundly happy when the clock
was striking? Then, as the sound of St. Margaret’s languished, he
thought, She has been ill, and the sound expressed languor and
suffering. It was her heart, he remembered; and the sudden
loudness of the final stroke tolled for death that surprised in the midst
of life, Clarissa falling where she stood, in her drawing-room. No! No!
he cried. She is not dead! I am not old, he cried, and marched up
Whitehall, as if there rolled down to him, vigorous, unending, his
future.
He was not old, or set, or dried in the least. As for caring what they
said of him—the Dalloways, the Whitbreads, and their set, he cared
not a straw—not a straw (though it was true he would have, some
time or other, to see whether Richard couldn’t help him to some job).
Striding, staring, he glared at the statue of the Duke of Cambridge.
He had been sent down from Oxford—true. He had been a Socialist,
in some sense a failure—true. Still the future of civilisation lies, he
thought, in the hands of young men like that; of young men such as
he was, thirty years ago; with their love of abstract principles; getting
books sent out to them all the way from London to a peak in the
Himalayas; reading science; reading philosophy. The future lies in
the hands of young men like that, he thought.
A patter like the patter of leaves in a wood came from behind, and
with it a rustling, regular thudding sound, which as it overtook him
drummed his thoughts, strict in step, up Whitehall, without his doing.
Boys in uniform, carrying guns, marched with their eyes ahead of
them, marched, their arms stiff, and on their faces an expression like
the letters of a legend written round the base of a statue praising
duty, gratitude, fidelity, love of England.
It is, thought Peter Walsh, beginning to keep step with them, a very
fine training. But they did not look robust. They were weedy for the
most part, boys of sixteen, who might, to-morrow, stand behind
bowls of rice, cakes of soap on counters. Now they wore on them
unmixed with sensual pleasure or daily preoccupations the solemnity
of the wreath which they had fetched from Finsbury Pavement to the
empty tomb. They had taken their vow. The traffic respected it; vans
were stopped.
I can’t keep up with them, Peter Walsh thought, as they marched up
Whitehall, and sure enough, on they marched, past him, past every
one, in their steady way, as if one will worked legs and arms
uniformly, and life, with its varieties, its irreticences, had been laid
under a pavement of monuments and wreaths and drugged into a
stiff yet staring corpse by discipline. One had to respect it; one might
laugh; but one had to respect it, he thought. There they go, thought
Peter Walsh, pausing at the edge of the pavement; and all the
exalted statues, Nelson, Gordon, Havelock, the black, the
spectacular images of great soldiers stood looking ahead of them, as
if they too had made the same renunciation (Peter Walsh felt he too
had made it, the great renunciation), trampled under the same
temptations, and achieved at length a marble stare. But the stare
Peter Walsh did not want for himself in the least; though he could
respect it in others. He could respect it in boys. They don’t know the
troubles of the flesh yet, he thought, as the marching boys
disappeared in the direction of the Strand—all that I’ve been through,
he thought, crossing the road, and standing under Gordon’s statue,
Gordon whom as a boy he had worshipped; Gordon standing lonely
with one leg raised and his arms crossed,—poor Gordon, he
thought.
And just because nobody yet knew he was in London, except
Clarissa, and the earth, after the voyage, still seemed an island to
him, the strangeness of standing alone, alive, unknown, at half-past
eleven in Trafalgar Square overcame him. What is it? Where am I?
And why, after all, does one do it? he thought, the divorce seeming
all moonshine. And down his mind went flat as a marsh, and three
great emotions bowled over him; understanding; a vast philanthropy;
and finally, as if the result of the others, an irrepressible, exquisite
delight; as if inside his brain by another hand strings were pulled,
shutters moved, and he, having nothing to do with it, yet stood at the
opening of endless avenues, down which if he chose he might
wander. He had not felt so young for years.
He had escaped! was utterly free—as happens in the downfall of
habit when the mind, like an unguarded flame, bows and bends and
seems about to blow from its holding. I haven’t felt so young for
years! thought Peter, escaping (only of course for an hour or so)
from being precisely what he was, and feeling like a child who runs
out of doors, and sees, as he runs, his old nurse waving at the wrong
window. But she’s extraordinarily attractive, he thought, as, walking
across Trafalgar Square in the direction of the Haymarket, came a
young woman who, as she passed Gordon’s statue, seemed, Peter
Walsh thought (susceptible as he was), to shed veil after veil, until
she became the very woman he had always had in mind; young, but
stately; merry, but discreet; black, but enchanting.
Straightening himself and stealthily fingering his pocket-knife he
started after her to follow this woman, this excitement, which seemed
even with its back turned to shed on him a light which connected
them, which singled him out, as if the random uproar of the traffic
had whispered through hollowed hands his name, not Peter, but his
private name which he called himself in his own thoughts. “You,” she
said, only “you,” saying it with her white gloves and her shoulders.
Then the thin long cloak which the wind stirred as she walked past
Dent’s shop in Cockspur Street blew out with an enveloping
kindness, a mournful tenderness, as of arms that would open and
take the tired—
But she’s not married; she’s young; quite young, thought Peter, the
red carnation he had seen her wear as she came across Trafalgar
Square burning again in his eyes and making her lips red. But she
waited at the kerbstone. There was a dignity about her. She was not
worldly, like Clarissa; not rich, like Clarissa. Was she, he wondered
as she moved, respectable? Witty, with a lizard’s flickering tongue,
he thought (for one must invent, must allow oneself a little diversion),
a cool waiting wit, a darting wit; not noisy.
She moved; she crossed; he followed her. To embarrass her was the
last thing he wished. Still if she stopped he would say “Come and
have an ice,” he would say, and she would answer, perfectly simply,
“Oh yes.”
But other people got between them in the street, obstructing him,
blotting her out. He pursued; she changed. There was colour in her
cheeks; mockery in her eyes; he was an adventurer, reckless, he
thought, swift, daring, indeed (landed as he was last night from India)
a romantic buccaneer, careless of all these damned proprieties,
yellow dressing-gowns, pipes, fishing-rods, in the shop windows; and
respectability and evening parties and spruce old men wearing white
slips beneath their waistcoats. He was a buccaneer. On and on she
went, across Piccadilly, and up Regent Street, ahead of him, her
cloak, her gloves, her shoulders combining with the fringes and the
laces and the feather boas in the windows to make the spirit of finery
and whimsy which dwindled out of the shops on to the pavement, as
the light of a lamp goes wavering at night over hedges in the
darkness.
Laughing and delightful, she had crossed Oxford Street and Great
Portland Street and turned down one of the little streets, and now,
and now, the great moment was approaching, for now she
slackened, opened her bag, and with one look in his direction, but
not at him, one look that bade farewell, summed up the whole
situation and dismissed it triumphantly, for ever, had fitted her key,
opened the door, and gone! Clarissa’s voice saying, Remember my
party, Remember my party, sang in his ears. The house was one of
those flat red houses with hanging flower-baskets of vague
impropriety. It was over.
Well, I’ve had my fun; I’ve had it, he thought, looking up at the
swinging baskets of pale geraniums. And it was smashed to atoms—
his fun, for it was half made up, as he knew very well; invented, this
escapade with the girl; made up, as one makes up the better part of
life, he thought—making oneself up; making her up; creating an
exquisite amusement, and something more. But odd it was, and
quite true; all this one could never share—it smashed to atoms.
He turned; went up the street, thinking to find somewhere to sit, till it
was time for Lincoln’s Inn—for Messrs. Hooper and Grateley. Where
should he go? No matter. Up the street, then, towards Regent’s
Park. His boots on the pavement struck out “no matter”; for it was
early, still very early.
It was a splendid morning too. Like the pulse of a perfect heart, life
struck straight through the streets. There was no fumbling—no
hesitation. Sweeping and swerving, accurately, punctually,
noiselessly, there, precisely at the right instant, the motor-car
stopped at the door. The girl, silk-stockinged, feathered, evanescent,
but not to him particularly attractive (for he had had his fling),
alighted. Admirable butlers, tawny chow dogs, halls laid in black and
white lozenges with white blinds blowing, Peter saw through the
opened door and approved of. A splendid achievement in its own
way, after all, London; the season; civilisation. Coming as he did
from a respectable Anglo-Indian family which for at least three
generations had administered the affairs of a continent (it’s strange,
he thought, what a sentiment I have about that, disliking India, and
empire, and army as he did), there were moments when civilisation,
even of this sort, seemed dear to him as a personal possession;
moments of pride in England; in butlers; chow dogs; girls in their
security. Ridiculous enough, still there it is, he thought. And the
doctors and men of business and capable women all going about
their business, punctual, alert, robust, seemed to him wholly
admirable, good fellows, to whom one would entrust one’s life,
companions in the art of living, who would see one through. What
with one thing and another, the show was really very tolerable; and
he would sit down in the shade and smoke.
There was Regent’s Park. Yes. As a child he had walked in Regent’s
Park—odd, he thought, how the thought of childhood keeps coming
back to me—the result of seeing Clarissa, perhaps; for women live
much more in the past than we do, he thought. They attach
themselves to places; and their fathers—a woman’s always proud of
her father. Bourton was a nice place, a very nice place, but I could
never get on with the old man, he thought. There was quite a scene
one night—an argument about something or other, what, he could
not remember. Politics presumably.
Yes, he remembered Regent’s Park; the long straight walk; the little
house where one bought air-balls to the left; an absurd statue with
an inscription somewhere or other. He looked for an empty seat. He
did not want to be bothered (feeling a little drowsy as he did) by
people asking him the time. An elderly grey nurse, with a baby
asleep in its perambulator—that was the best he could do for
himself; sit down at the far end of the seat by that nurse.
She’s a queer-looking girl, he thought, suddenly remembering
Elizabeth as she came into the room and stood by her mother.
Grown big; quite grown-up, not exactly pretty; handsome rather; and
she can’t be more than eighteen. Probably she doesn’t get on with
Clarissa. “There’s my Elizabeth”—that sort of thing—why not “Here’s
Elizabeth” simply?—trying to make out, like most mothers, that
things are what they’re not. She trusts to her charm too much, he
thought. She overdoes it.
The rich benignant cigar smoke eddied coolly down his throat; he
puffed it out again in rings which breasted the air bravely for a
moment; blue, circular—I shall try and get a word alone with
Elizabeth to-night, he thought—then began to wobble into hour-glass
shapes and taper away; odd shapes they take, he thought. Suddenly
he closed his eyes, raised his hand with an effort, and threw away
the heavy end of his cigar. A great brush swept smooth across his
mind, sweeping across it moving branches, children’s voices, the
shuffle of feet, and people passing, and humming traffic, rising and
falling traffic. Down, down he sank into the plumes and feathers of
sleep, sank, and was muffled over.
The grey nurse resumed her knitting as Peter Walsh, on the hot seat
beside her, began snoring. In her grey dress, moving her hands
indefatigably yet quietly, she seemed like the champion of the rights
of sleepers, like one of those spectral presences which rise in twilight
in woods made of sky and branches. The solitary traveller, haunter of
lanes, disturber of ferns, and devastator of great hemlock plants,
looking up, suddenly sees the giant figure at the end of the ride.
By conviction an atheist perhaps, he is taken by surprise with
moments of extraordinary exaltation. Nothing exists outside us
except a state of mind, he thinks; a desire for solace, for relief, for
something outside these miserable pigmies, these feeble, these ugly,
these craven men and women. But if he can conceive of her, then in
some sort she exists, he thinks, and advancing down the path with
his eyes upon sky and branches he rapidly endows them with
womanhood; sees with amazement how grave they become; how
majestically, as the breeze stirs them, they dispense with a dark
flutter of the leaves charity, comprehension, absolution, and then,
flinging themselves suddenly aloft, confound the piety of their aspect
with a wild carouse.
Such are the visions which proffer great cornucopias full of fruit to
the solitary traveller, or murmur in his ear like sirens lolloping away
on the green sea waves, or are dashed in his face like bunches of
roses, or rise to the surface like pale faces which fishermen flounder
through floods to embrace.
Such are the visions which ceaselessly float up, pace beside, put
their faces in front of, the actual thing; often overpowering the
solitary traveller and taking away from him the sense of the earth,
the wish to return, and giving him for substitute a general peace, as if
(so he thinks as he advances down the forest ride) all this fever of
living were simplicity itself; and myriads of things merged in one
thing; and this figure, made of sky and branches as it is, had risen
from the troubled sea (he is elderly, past fifty now) as a shape might
be sucked up out of the waves to shower down from her magnificent
hands compassion, comprehension, absolution. So, he thinks, may I
never go back to the lamplight; to the sitting-room; never finish my
book; never knock out my pipe; never ring for Mrs. Turner to clear
away; rather let me walk straight on to this great figure, who will, with
a toss of her head, mount me on her streamers and let me blow to
nothingness with the rest.
Such are the visions. The solitary traveller is soon beyond the wood;
and there, coming to the door with shaded eyes, possibly to look for
his return, with hands raised, with white apron blowing, is an elderly
woman who seems (so powerful is this infirmity) to seek, over a
desert, a lost son; to search for a rider destroyed; to be the figure of
the mother whose sons have been killed in the battles of the world.
So, as the solitary traveller advances down the village street where
the women stand knitting and the men dig in the garden, the evening
seems ominous; the figures still; as if some august fate, known to
them, awaited without fear, were about to sweep them into complete
annihilation.
Indoors among ordinary things, the cupboard, the table, the window-
sill with its geraniums, suddenly the outline of the landlady, bending
to remove the cloth, becomes soft with light, an adorable emblem
which only the recollection of cold human contacts forbids us to
embrace. She takes the marmalade; she shuts it in the cupboard.
“There is nothing more to-night, sir?”
But to whom does the solitary traveller make reply?

So the elderly nurse knitted over the sleeping baby in Regent’s Park.
So Peter Walsh snored.
He woke with extreme suddenness, saying to himself, “The death of
the soul.”
“Lord, Lord!” he said to himself out loud, stretching and opening his
eyes. “The death of the soul.” The words attached themselves to
some scene, to some room, to some past he had been dreaming of.
It became clearer; the scene, the room, the past he had been
dreaming of.
It was at Bourton that summer, early in the ’nineties, when he was so
passionately in love with Clarissa. There were a great many people
there, laughing and talking, sitting round a table after tea and the
room was bathed in yellow light and full of cigarette smoke. They
were talking about a man who had married his housemaid, one of
the neighbouring squires, he had forgotten his name. He had
married his housemaid, and she had been brought to Bourton to call
—an awful visit it had been. She was absurdly over-dressed, “like a
cockatoo,” Clarissa had said, imitating her, and she never stopped
talking. On and on she went, on and on. Clarissa imitated her. Then
somebody said—Sally Seton it was—did it make any real difference
to one’s feelings to know that before they’d married she had had a
baby? (In those days, in mixed company, it was a bold thing to say.)
He could see Clarissa now, turning bright pink; somehow
contracting; and saying, “Oh, I shall never be able to speak to her
again!” Whereupon the whole party sitting round the tea-table
seemed to wobble. It was very uncomfortable.
He hadn’t blamed her for minding the fact, since in those days a girl
brought up as she was, knew nothing, but it was her manner that
annoyed him; timid; hard; something arrogant; unimaginative;
prudish. “The death of the soul.” He had said that instinctively,
ticketing the moment as he used to do—the death of her soul.
Every one wobbled; every one seemed to bow, as she spoke, and
then to stand up different. He could see Sally Seton, like a child who
has been in mischief, leaning forward, rather flushed, wanting to talk,
but afraid, and Clarissa did frighten people. (She was Clarissa’s
greatest friend, always about the place, totally unlike her, an
attractive creature, handsome, dark, with the reputation in those
days of great daring and he used to give her cigars, which she
smoked in her bedroom. She had either been engaged to somebody
or quarrelled with her family and old Parry disliked them both equally,
which was a great bond.) Then Clarissa, still with an air of being
offended with them all, got up, made some excuse, and went off,
alone. As she opened the door, in came that great shaggy dog which
ran after sheep. She flung herself upon him, went into raptures. It
was as if she said to Peter—it was all aimed at him, he knew—“I
know you thought me absurd about that woman just now; but see
how extraordinarily sympathetic I am; see how I love my Rob!”
They had always this queer power of communicating without words.
She knew directly he criticised her. Then she would do something
quite obvious to defend herself, like this fuss with the dog—but it
never took him in, he always saw through Clarissa. Not that he said
anything, of course; just sat looking glum. It was the way their
quarrels often began.
She shut the door. At once he became extremely depressed. It all
seemed useless—going on being in love; going on quarrelling; going
on making it up, and he wandered off alone, among outhouses,
stables, looking at the horses. (The place was quite a humble one;
the Parrys were never very well off; but there were always grooms
and stable-boys about—Clarissa loved riding—and an old coachman
—what was his name?—an old nurse, old Moody, old Goody, some
such name they called her, whom one was taken to visit in a little
room with lots of photographs, lots of bird-cages.)
It was an awful evening! He grew more and more gloomy, not about
that only; about everything. And he couldn’t see her; couldn’t explain
to her; couldn’t have it out. There were always people about—she’d
go on as if nothing had happened. That was the devilish part of her
—this coldness, this woodenness, something very profound in her,
which he had felt again this morning talking to her; an
impenetrability. Yet Heaven knows he loved her. She had some
queer power of fiddling on one’s nerves, turning one’s nerves to
fiddle-strings, yes.
He had gone in to dinner rather late, from some idiotic idea of
making himself felt, and had sat down by old Miss Parry—Aunt
Helena—Mr. Parry’s sister, who was supposed to preside. There she
sat in her white Cashmere shawl, with her head against the window
—a formidable old lady, but kind to him, for he had found her some
rare flower, and she was a great botanist, marching off in thick boots
with a black collecting-box slung between her shoulders. He sat
down beside her, and couldn’t speak. Everything seemed to race
past him; he just sat there, eating. And then half-way through dinner
he made himself look across at Clarissa for the first time. She was
talking to a young man on her right. He had a sudden revelation.
“She will marry that man,” he said to himself. He didn’t even know
his name.
For of course it was that afternoon, that very afternoon, that
Dalloway had come over; and Clarissa called him “Wickham”; that
was the beginning of it all. Somebody had brought him over; and
Clarissa got his name wrong. She introduced him to everybody as
Wickham. At last he said “My name is Dalloway!”—that was his first
view of Richard—a fair young man, rather awkward, sitting on a
deck-chair, and blurting out “My name is Dalloway!” Sally got hold of
it; always after that she called him “My name is Dalloway!”
He was a prey to revelations at that time. This one—that she would
marry Dalloway—was blinding—overwhelming at the moment. There
was a sort of—how could he put it?—a sort of ease in her manner to
him; something maternal; something gentle. They were talking about
politics. All through dinner he tried to hear what they were saying.
Afterwards he could remember standing by old Miss Parry’s chair in
the drawing-room. Clarissa came up, with her perfect manners, like a
real hostess, and wanted to introduce him to some one—spoke as if
they had never met before, which enraged him. Yet even then he
admired her for it. He admired her courage; her social instinct; he
admired her power of carrying things through. “The perfect hostess,”
he said to her, whereupon she winced all over. But he meant her to
feel it. He would have done anything to hurt her after seeing her with
Dalloway. So she left him. And he had a feeling that they were all
gathered together in a conspiracy against him—laughing and talking
—behind his back. There he stood by Miss Parry’s chair as though
he had been cut out of wood, he talking about wild flowers. Never,
never had he suffered so infernally! He must have forgotten even to
pretend to listen; at last he woke up; he saw Miss Parry looking
rather disturbed, rather indignant, with her prominent eyes fixed. He
almost cried out that he couldn’t attend because he was in Hell!
People began going out of the room. He heard them talking about
fetching cloaks; about its being cold on the water, and so on. They
were going boating on the lake by moonlight—one of Sally’s mad
ideas. He could hear her describing the moon. And they all went out.
He was left quite alone.
“Don’t you want to go with them?” said Aunt Helena—old Miss Parry!
—she had guessed. And he turned round and there was Clarissa
again. She had come back to fetch him. He was overcome by her
generosity—her goodness.
“Come along,” she said. “They’re waiting.”
He had never felt so happy in the whole of his life! Without a word
they made it up. They walked down to the lake. He had twenty
minutes of perfect happiness. Her voice, her laugh, her dress
(something floating, white, crimson), her spirit, her adventurousness;
she made them all disembark and explore the island; she startled a
hen; she laughed; she sang. And all the time, he knew perfectly well,
Dalloway was falling in love with her; she was falling in love with
Dalloway; but it didn’t seem to matter. Nothing mattered. They sat on
the ground and talked—he and Clarissa. They went in and out of
each other’s minds without any effort. And then in a second it was
over. He said to himself as they were getting into the boat, “She will
marry that man,” dully, without any resentment; but it was an obvious
thing. Dalloway would marry Clarissa.
Dalloway rowed them in. He said nothing. But somehow as they
watched him start, jumping on to his bicycle to ride twenty miles
through the woods, wobbling off down the drive, waving his hand
and disappearing, he obviously did feel, instinctively, tremendously,
strongly, all that; the night; the romance; Clarissa. He deserved to
have her.
For himself, he was absurd. His demands upon Clarissa (he could
see it now) were absurd. He asked impossible things. He made
terrible scenes. She would have accepted him still, perhaps, if he
had been less absurd. Sally thought so. She wrote him all that
summer long letters; how they had talked of him; how she had
praised him, how Clarissa burst into tears! It was an extraordinary
summer—all letters, scenes, telegrams—arriving at Bourton early in
the morning, hanging about till the servants were up; appalling tête-
à-têtes with old Mr. Parry at breakfast; Aunt Helena formidable but
kind; Sally sweeping him off for talks in the vegetable garden;
Clarissa in bed with headaches.
The final scene, the terrible scene which he believed had mattered
more than anything in the whole of his life (it might be an
exaggeration—but still so it did seem now) happened at three o’clock
in the afternoon of a very hot day. It was a trifle that led up to it—
Sally at lunch saying something about Dalloway, and calling him “My
name is Dalloway”; whereupon Clarissa suddenly stiffened,
coloured, in a way she had, and rapped out sharply, “We’ve had
enough of that feeble joke.” That was all; but for him it was precisely
as if she had said, “I’m only amusing myself with you; I’ve an
understanding with Richard Dalloway.” So he took it. He had not
slept for nights. “It’s got to be finished one way or the other,” he said
to himself. He sent a note to her by Sally asking her to meet him by
the fountain at three. “Something very important has happened,” he
scribbled at the end of it.
The fountain was in the middle of a little shrubbery, far from the
house, with shrubs and trees all round it. There she came, even
before the time, and they stood with the fountain between them, the
spout (it was broken) dribbling water incessantly. How sights fix
themselves upon the mind! For example, the vivid green moss.
She did not move. “Tell me the truth, tell me the truth,” he kept on
saying. He felt as if his forehead would burst. She seemed
contracted, petrified. She did not move. “Tell me the truth,” he
repeated, when suddenly that old man Breitkopf popped his head in
carrying the Times; stared at them; gaped; and went away. They
neither of them moved. “Tell me the truth,” he repeated. He felt that
he was grinding against something physically hard; she was
unyielding. She was like iron, like flint, rigid up the backbone. And
when she said, “It’s no use. It’s no use. This is the end”—after he
had spoken for hours, it seemed, with the tears running down his
cheeks—it was as if she had hit him in the face. She turned, she left
him, went away.