Random matrices high dimensional phenomena 1st Edition Gordon Blower download pdf

vionersaruar 7 views 87 slides Feb 21, 2025
Slide 1
Slide 1 of 87
Slide 1
1
Slide 2
2
Slide 3
3
Slide 4
4
Slide 5
5
Slide 6
6
Slide 7
7
Slide 8
8
Slide 9
9
Slide 10
10
Slide 11
11
Slide 12
12
Slide 13
13
Slide 14
14
Slide 15
15
Slide 16
16
Slide 17
17
Slide 18
18
Slide 19
19
Slide 20
20
Slide 21
21
Slide 22
22
Slide 23
23
Slide 24
24
Slide 25
25
Slide 26
26
Slide 27
27
Slide 28
28
Slide 29
29
Slide 30
30
Slide 31
31
Slide 32
32
Slide 33
33
Slide 34
34
Slide 35
35
Slide 36
36
Slide 37
37
Slide 38
38
Slide 39
39
Slide 40
40
Slide 41
41
Slide 42
42
Slide 43
43
Slide 44
44
Slide 45
45
Slide 46
46
Slide 47
47
Slide 48
48
Slide 49
49
Slide 50
50
Slide 51
51
Slide 52
52
Slide 53
53
Slide 54
54
Slide 55
55
Slide 56
56
Slide 57
57
Slide 58
58
Slide 59
59
Slide 60
60
Slide 61
61
Slide 62
62
Slide 63
63
Slide 64
64
Slide 65
65
Slide 66
66
Slide 67
67
Slide 68
68
Slide 69
69
Slide 70
70
Slide 71
71
Slide 72
72
Slide 73
73
Slide 74
74
Slide 75
75
Slide 76
76
Slide 77
77
Slide 78
78
Slide 79
79
Slide 80
80
Slide 81
81
Slide 82
82
Slide 83
83
Slide 84
84
Slide 85
85
Slide 86
86
Slide 87
87

About This Presentation

Instant Random matrices high dimensional phenomena 1st Edition Gordon Blower download upon payment at https://ebookultra.com/download/random-matrices-high-dimensional-phenomena-1st-edition-gordon-blower. Find more textbooks and ebooks in https://ebookultra.com Download complete PDF chapter.


Slide Content

Visit https://ebookultra.com to download the full version and
explore more ebooks
Random matrices high dimensional phenomena 1st
Edition Gordon Blower
_____ Click the link below to download _____
https://ebookultra.com/download/random-matrices-high-
dimensional-phenomena-1st-edition-gordon-blower/
Explore and download more ebooks at ebookultra.com

Here are some suggested products you might be interested in.
Click the link to download
High Dimensional Covariance Estimation With High
Dimensional Data 1st Edition Mohsen Pourahmadi
https://ebookultra.com/download/high-dimensional-covariance-
estimation-with-high-dimensional-data-1st-edition-mohsen-pourahmadi/
Exploration and Analysis of DNA Microarray and Other High
Dimensional Data 2nd Edition Dhammika Amaratunga
https://ebookultra.com/download/exploration-and-analysis-of-dna-
microarray-and-other-high-dimensional-data-2nd-edition-dhammika-
amaratunga/
Ignition A Guide to Building High Performing Teams Gordon
J. Curphy
https://ebookultra.com/download/ignition-a-guide-to-building-high-
performing-teams-gordon-j-curphy/
Totally positive matrices 1st Edition Pinkus Allan
https://ebookultra.com/download/totally-positive-matrices-1st-edition-
pinkus-allan/

Plumbing Mechanical Services Book 2 Bk 2 5th Edition G. J.
Blower
https://ebookultra.com/download/plumbing-mechanical-services-
book-2-bk-2-5th-edition-g-j-blower/
Low Dimensional Geometry Francis Bonahon
https://ebookultra.com/download/low-dimensional-geometry-francis-
bonahon/
Field Guide to Probability Random Processes and Random
Data Analysis Larry C. Andrews
https://ebookultra.com/download/field-guide-to-probability-random-
processes-and-random-data-analysis-larry-c-andrews/
Random Graph Dynamics 1st Edition Rick Durrett
https://ebookultra.com/download/random-graph-dynamics-1st-edition-
rick-durrett/
Random Vibration 1st Edition Christi Lalanne (Author)
https://ebookultra.com/download/random-vibration-1st-edition-christi-
lalanne-author/

Random matrices high dimensional phenomena 1st
Edition Gordon Blower Digital Instant Download
Author(s): Gordon Blower
ISBN(s): 9780521133128, 0521133122
Edition: 1
File Details: PDF, 2.06 MB
Year: 2009
Language: english

This page intentionally left blank

LONDON MATHEMATICAL SOCIETY LECTURE NOTE SERIES
M anaging Editor: Professor M . Reid, M athematics Institute,
University of Warwick, Coventry CV4 7AL, United K ingdom
The titles b elow are available from b ooksellers, or from Cambridge University Press at
www.cambridge.org/mathematics
247 Analytic number theory, Y. MOTOHASHI (ed)
248 Tame topology and O-minimal structures, L. VAN DEN DRIES
249 The atlas of finite groups - ten years on, R.T. CURTIS & R.A. W ILSON (eds)
250 Characters and blocks of finite groups, G. NAVARRO
251 Gr¨obner bases and applications, B. BUCHBERGER & F. W INKLER (eds)
252 Geometry and cohomology in group theory, P.H. KROPHOLLER, G.A. NIBLO & R.
ST
¨
OHR (eds)
253 The q-Schur algebra, S. DONKIN
254 G alois representations in arithm etic algebraic geom etry, A.J. SCHO LL & R.L. TAYLO R
(eds)
255 Sym m etries and integrability of diff erence equations, P.A. CLARK SO N & F.W .
NIJHOFF (eds)
256 Aspects of Galois theory, H. V
¨
OLKLEIN, J.G. THOMPSON, D. HARBATER & P.
M
¨
ULLER (eds)
257 An introduction to noncommutative differential geometry and its physical applications
(2nd Edition), J. M AD O RE
258 Sets and proofs, S.B. COOPER & J.K. TRUSS (eds)
259 M odels and com putability, S.B. CO O PER & J. TRUSS (eds)
260 Groups St Andrews 1997 in Bath I, C.M. CAMPBELL et al(eds)
261 Groups St Andrews 1997 in Bath II, C.M. CAMPBELL et al(eds)
262 Analysis and logic, C.W . HENSON, J. IOVINO, A.S. KECHRIS & E. ODELL
263 Singularity theory, W. BRUCE & D. MOND (eds)
264 New trends in algebraic geometry, K. HULEK, F. CATANESE, C. PETERS & M. REID
(eds)
265 Elliptic curves in cryptography, I. BLAK E, G . SERO USSI & N. SM ART
267 Surveys in combinatorics, 1999, J.D. LAMB & D.A. PREECE (eds)
268 Spectral asymptotics in the semi-classical limit, M. DIMASSI & J. SJ
¨
OSTRAND
269 Ergodic theory and topological dynamics of group actions on homogeneous spaces, M.B.
BEKKA & M. MAYER
271 Singular perturbations of differential operators, S. ALBEVERIO & P. KURASOV
272 Character theory for the odd order theorem, T. PETERFALVI. Translated by R.
SANDLING
273 Spectral theory and geometry, E.B. DAVIES & Y. SAFAROV (eds)
274 The Mandelbrot set, theme and variations, T. LEI (ed)
275 Descriptive set theory and dynamical systems, M. FOREMAN, A.S. KECHRIS, A.
LOUVEAU & B. WEISS (eds)
276 Singularities of plane curves, E. CASAS-ALVERO
277 Computational and geometric aspects of modern algebra, M. ATKINSON et al(eds)
278 Global attractors in abstract parabolic problems, J.W . CHOLEWA & T. DLOTKO
279 Topics in symbolic dynamics and applications, F. BLANCHARD, A. MAASS & A.
NOGUEIRA (eds)
280 Characters and automorphism groups of compact Riemann surfaces, T. BREUER
281 Explicit birational geometry of 3-folds, A. CORTI & M. REID (eds)
282 Auslander-Buchweitz approximations of equivariant modules, M. HASHIMOTO
283 Nonlinear elasticity, Y.B. FU & R.W . O G D EN (eds)
284 Foundations of computational mathematics, R. DEVORE, A. ISERLES & E. S
¨
ULI (eds)
285 Rational points on curves over finite fields, H. NIEDERREITER & C. XING
286 Clifford algebras and spinors (2nd Edition), P. LOUNESTO
287 Topics on Riemann surfaces and Fuchsian groups, E. BUJALANCE, A.F. COSTA & E.
MART
´
INEZ (eds)
288 Surveys in combinatorics, 2001, J.W .P. HIRSCHFELD (ed)
289 Aspects of Sobolev-type inequalities, L. SALOFF-COSTE
290 Quantum groups and Lie theory, A. PRESSLEY (ed)
291 Tits buildings and the model theory of groups, K. TENT (ed)
292 A quantum groups primer, S. MAJID
293 Second order partial differential equations in Hilbert spaces, G. DA PRATO & J.
ZABCZYK
294 Introduction to operator space theory, G. PISIER
295 G eom etry and integrability, L. M ASO N & Y. NUTK U (eds)
296 Lectures on invariant theory, I. DOLGACHEV
297 The homotopy category of simply connected 4-manifolds, H.-J. BAUES
298 Higher operads, higher categories, T. LEINSTER (ed)
299 Kleinian groups and hyperbolic 3-manifolds, Y. KOMORI, V. MARKOVIC & C. SERIES
(eds)
300 Introduction to M ¨obius differential geometry, U. HERTRICH-JEROMIN
301 Stable modules and the D(2)-problem, F.E.A. JOHNSON
302 Discrete and continuous nonlinear Schr¨ odinger systems, M.J. ABLOW ITZ, B. PRINARI
& A.D. TRUBATCH
303 Number theory and algebraic geometry, M. REID & A. SKOROBOGATOV (eds)
304 Groups St Andrews 2001 in Oxford I, C.M. CAMPBELL, E.F. ROBERTSON & G.C.
SMITH (eds)
305 Groups St Andrews 2001 in Oxford II, C.M. CAMPBELL, E.F. ROBERTSON & G.C.
SMITH (eds)

306 Geometric mechanics and symmetry, J. MONTALDI & T. RATIU (eds)
307 Surveys in combinatorics 2003, C.D. W ENSLEY (ed.)
308 Topology, geometry and quantum field theory, U.L. TILLMANN (ed)
309 Corings and comodules, T. BRZEZINSKI & R. WISBAUER
310 Topics in dynamics and ergodic theory, S. BEZUGLYI & S. KOLYADA (eds)
311 Groups: topological, combinatorial and arithmetic aspects, T.W . M
¨
ULLER (ed)
312 Foundations of computational mathe matics, Minneapolis 2002, F. CUCKER et al(eds)
313 Transcendental aspects of algebraic cycles, S. M
¨
ULLER-STACH & C. PETERS (eds)
314 Spectral generalizations of line graphs, D. CVETKOVI
´
C, P. ROWLINSON & S. SIMI
´
C
315 Structured ring spectra, A. BAKER & B. RICHTER (eds)
316 Linear logic in computer science, T. EHRHARD, P. RUET, J.-Y. GIRARD & P. SCOTT
(eds)
317 Advances in elliptic curve cryptography, I.F. BLAK E, G . SERO USSI & N.P. SM ART
(eds)
318 Perturbation of the boundary in boundary-value problems of partial differential equations,
D. HENRY
319 Double affi ne Hecke algebras, I. CHEREDNIK
320 L-functions and Galois representations, D. BURNS, K. BUZZARD & J. NEKOV
´
A
ˇ
R(eds)
321 Surveys in modern mathematics, V. PRASOLOV & Y. ILYASHENKO (eds)
322 Recent perspectives in random matrix theory and number theory, F. MEZZADRI & N.C.
SNAITH (eds)
323 Poisson geometry, deformation quantisation and group representations, S. GUTT et al
(eds)
324 Singularities and computer algebra, C. LOSSEN & G. PFISTER (eds)
325 Lectures on the Ricci flow, P. TOPPING
326 Modular representations of finite groups of Lie type, J.E. HUMPHREYS
327 Surveys in combinatorics 2005, B.S. W EBB (ed)
328 Fundamentals of hyperbolic manifolds, R. C ANARY, D. EPSTEIN & A. MARDEN (eds)
329 Spaces of Kleinian groups, Y. MINSKY, M. SAKUMA & C. SERIES (eds)
330 Noncommutative localization in algebra and topology, A. RANICKI (ed)
331 Foundations of computational mathemati cs, Santander 2005, L.M PARDO, A. PINKUS,
E. S
¨
ULI & M.J. TODD (eds)
332 Handb ook of tilting theory, L. ANG ELERI H
¨
UGEL, D. HAPPEL & H. KRAUSE (eds)
333 Synthetic differential geometry (2nd Edition), A. KOCK
334 The Navier–Stokes equations, N. RILEY & P. DRAZIN
335 Lectures on the combinatorics of free probability, A. NICA & R. SPEICHER
336 Integral closure of ideals, rings, and modules, I. SWANSON & C. HUNEKE
337 MethodsinBanachspacetheory, J.M.F.CASTILLO & W.B.JOHNSON(eds)
338 Surveys in geometry and number theory, N. YOUNG (ed)
339 Groups St Andrews 2005 I, C.M. CAMPBELL, M.R. QUICK, E.F. ROBERTSON & G.C.
SMITH (eds)
340 Groups St Andrews 2005 II, C.M. CAMPBELL, M.R. QUICK, E.F. ROBERTSON &
G.C. SMITH (eds)
341 Ranks of elliptic curves and random m atrix theory, J.B. CO NREY, D .W . FARM ER, F.
MEZZADRI & N.C. SNAITH (eds)
342 Elliptic cohom ology, H.R. M ILLER & D .C. RAVENEL (eds)
343 Algebraic cycles and motives I, J. NAGEL & C. PETERS (eds)
344 Algebraic cycles and motives II, J. NAGEL & C. PETERS (eds)
345 Algebraic and analytic geometry, A. NEEMAN
346 Surveys in combinatorics 2007, A. HILTON & J. TALBOT (eds)
347 Surveys in contemporary mathematics, N. YOUNG & Y. CHOI (eds)
348 Transcendental dynamics and complex analysis, P.J. RIPPON & G.M. STALLARD (eds)
349 Model theory with applications to algebra and analysis I, Z. CHATZIDAKIS, D.
MACPHERSON, A. PILLAY & A. WILKIE (eds)
350 Model theory with applications to algebra and analysis II, Z. CHATZIDAKIS, D.
MACPHERSON, A. PILLAY & A. WILKIE (eds)
351 Finite von Neumann algebras and masas, A.M. SINCLAIR & R.R. SMITH
352 Number theory and polynomials, J. MCKEE & C. SMYTH (eds)
353 Trends in stochastic analysis, J. BLATH, P. M
¨
ORTERS & M. SCHEUTZOW (eds)
354 Groups and analysis, K. TENT (ed)
355 Non-equilibrium statistical m echanics and turbulence, J. CARDY, G . FALKOVICH & K .
GAWEDZKI
356 Elliptic curves and big G alois representations, D . D ELBO URG O
357 Algebraic theory of differential equations, M.A.H. MACCALLUM & A.V. MIKHAILOV
(eds)
358 Geometric and cohomological methods in group theory, M.R. BRIDSON, P.H.
KROPHOLLER & I.J. LEARY (eds)
359 Moduli spaces and vector bundles, L. BRAMBILA-PAZ, S.B. BRADLOW, O.
GARC
´
IA-PRADA & S. RAMANAN (eds)
360 Zariski geometries, B. ZILBER
361 Words: Notes on verbal width in groups, D. SEGAL
362 Differential tensor algebras and their module categories, R. BAUTISTA, L. SALMER
´
ON
& R. ZUAZUA
363 Foundations of computational mathemati cs, Hong Kong 2008, F. CUCKER, A. PINKUS
&M.J.TODD(eds)
364 Partial diff erential equations and fluid m echanics, J.C. RO BINSO N & J.L. RO D RIG O
(eds)
365 Surveys in combinatorics 2009, S. HUCZYNSKA, J.D. MITCHELL & C.M.
RONEY-DOUGAL (eds)
366 Highly oscillatory problem s, B. ENG Q UIST, A. FO K AS, E. HAIRER & A. ISERLES
(eds)

London Mathematical Society Lecture Note Series: 367
Random Matrices: High Dimensional
Phenomena
GORDON BLOWER
Lancaster University

cambridge university press
Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore,
S˜ao Paulo, Delhi
Cambridge University Press
The Edinburgh Building, Cambridge CB2 8RU, UK
Published in the United States of America
by Cambridge University Press, New York
www.cambridge.org
Information on this title: www.cambridge.org/9780521133128
CffG. Blower 2009
This publication is in copyright. Subject to statutory exception
and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without
the written permission of Cambridge University Press.
First published 2009
Printed in the United Kingdom at the University Press, Cambridge
A catalogue record for this publication is available from the British Library
ISBN 978-0-521-13312-8 Paperback
Cambridge University Press has no responsibility for the persistence or
accuracy of URLs for external or third-party Internet websites referred to
in this publication, and does not guarantee that any content on such
websites is, or will remain, accurate or appropriate.

To the memory of my father
Ronald Frederick Blower

Contents
Introduction page1
1 Metric measure spaces 4
1.1 Weak convergence on compact metric spaces
1.2 Invariant measure on a compact metric group
1.3 Measures on non-compact Polish spaces
1.4 The Brunn–Minkowski inequality
1.5 Gaussian measures
1.6 Surface area measure on the spheres
1.7 Lipschitz functions and the Hausdorff metric
1.8 Characteristic functions and Cauchy transforms
2 Lie groups and matrix ensembles 42
2.1 The classical groups, their eigenvalues and norms
2.2 Determinants and functional calculus
2.3 Linear Lie groups
2.4 Connections and curvature
2.5 Generalized ensembles
2.6 The Weyl integration formula
2.7 Dyson’s circular ensembles
2.8 Circular orthogonal ensemble
2.9 Circular symplectic ensemble
3 Entropy and concentration of measure 84
3.1 Relative entropy
3.2 Concentration of measure
3.3 Transportation
3.4 Transportation inequalities
3.5 Transportation inequalities for uniformly
convex potentials
3.6 Concentration of measure in matrix ensembles
vii

viii Contents
3.7 Concentration for rectangular Gaussian matrices
3.8 Concentration on the sphere
3.9 Concentration for compact Lie groups
4 Free entropy and equilibrium 132
4.1 Logarithmic energy and equilibrium measure
4.2 Energy spaces on the disc
4.3 Free versus classical entropy on the spheres
4.4 Equilibrium measures for potentials on the real line
4.5 Equilibrium densities for convex potentials
4.6 The quartic model with positive leading term
4.7 Quartic models with negative leading term
4.8 Displacement convexity and relative free entropy
4.9 Toeplitz determinants
5 Convergence to equilibrium 177
5.1 Convergence to arclength
5.2 Convergence of ensembles
5.3 Mean field convergence
5.4 Almost sure weak convergence for uniformly
convex potentials
5.5 Convergence for the singular numbers from the
Wishart distribution
6 Gradient flows and functional inequalities 196
6.1 Variation of functionals and gradient flows
6.2 Logarithmic Sobolev inequalities
6.3 Logarithmic Sobolev inequalities for uniformly
convex potentials
6.4 Fisher’s information and Shannon’s entropy
6.5 Free information and entropy
6.6 Free logarithmic Sobolev inequality
6.7 Logarithmic Sobolev and spectral gap inequalities
6.8 Inequalities for Gibbs measures on
Riemannian manifolds
7 Young tableaux 227
7.1 Group representations
7.2 Young diagrams
7.3 The Vershik Ω distribution
7.4 Distribution of the longest increasing subsequence
7.5 Inclusion-exclusion principle

Contents ix
8 Random point fields and random matrices 253
8.1 Determinantal random point fields
8.2 Determinantal random point fields on the real line
8.3 Determinantal random point fields and orthogonal
polynomials
8.4 De Branges’s spaces
8.5 Limits of kernels
9 Integrable operators and differential equations 281
9.1 Integrable operators and Hankel integral operators
9.2 Hankel integral operators that commute with second
order differential operators
9.3 Spectral bulk and the sine kernel
9.4 Soft edges and the Airy kernel
9.5 Hard edges and the Bessel kernel
9.6 The spectra of Hankel operators and rational
approximation
9.7 The Tracy–Widom distribution
10 Fluctuations and the Tracy–Widom distribution 321
10.1 The Costin–Lebowitz central limit theorem
10.2 Discrete Tracy–Widom systems
10.3 The discrete Bessel kernel
10.4 Plancherel measure on the partitions
10.5 Fluctuations of the longest increasing subsequence
10.6 Fluctuations of linear statistics over unitary
ensembles
11 Limit groups and Gaussian measures 352
11.1 Some inductive limit groups
11.2 Hua–Pickrell measure on the infinite unitary group
11.3 Gaussian Hilbert space
11.4 Gaussian measures and fluctuations
12 Hermite polynomials 373
12.1 Tensor products of Hilbert space
12.2 Hermite polynomials and Mehler’s formula
12.3 The Ornstein–Uhlenbeck semigroup
12.4 Hermite polynomials in higher dimensions
13 From the Ornstein–Uhlenbeck process to the
Burgers equation 392
13.1 The Ornstein–Uhlenbeck process

x Contents
13.2 The logarithmic Sobolev inequality for the
Ornstein–Uhlenbeck generator
13.3 The matrix Ornstein–Uhlenbeck process
13.4 Solutions for matrix stochastic differential equations
13.5 The Burgers equation
14 Noncommutative probability spaces 411
14.1 Noncommutative probability spaces
14.2 Tracial probability spaces
14.3 The semicircular distribution
References 424
Index 433

Introduction
The purpose of this book is to introduce readers to certain topics in
random matrix theory that specifically involve the phenomenon of con-
centration of measure in high dimension. Partly this work was moti-
vated by researches in the EC networkPhenomena in High Dimension,
which applied results from functional analysis to problems in statistical
physics. Pisier described this as the transfer of technology, and this book
develops this philosophy by discussing applications to random matrix
theory of:
(i) optimal transportation theory;
(ii) logarithmic Sobolev inequalities;
(iii) exponential concentration inequalities;
(iv) Hankel operators.
Recently some approaches to functional inequalities have emerged that
make a unified treatment possible; in particular, optimal transporta-
tion links together seemingly disparate ideas about convergence to equi-
librium. Furthermore, optimal transportation connects familiar results
from the calculus of variations with the modern theory of diffusions and
gradient flows.
I hope that postgraduate students will find this book useful and, with
them in mind, have selected topics with potential for further develop-
ment. Prerequisites for this book are linear algebra, calculus, complex
analysis, Lebesgue integration, metric spaces and basic Hilbert space
theory. The book does not use stochastic calculus or the theory of inte-
grable systems, so as to widen the possible readership.
In their survey of random matrices and Banach spaces, Davidson
and Szarek present results on Gaussian random matrices and then indi-
cate that some of the results should extend to a wider context by the
1

2 Gordon Blower
theory of concentration of measure [152]. This book follows this pro-
gramme in the context of generalized orthogonal ensembles and com-
pact Lie groups. While the Gaussian unitary ensemble and Wishart
ensembles have special properties, they provide a helpful model for
other cases. The book covers the main examples of the subject, such
as Gaussian random matrices, within the general context of invariant
ensembles.
The coverage of material is deliberately uneven, in that some topics
are treated more thoroughly than others and some results from other ar-
eas of analysis are recalled with minimal discussion. There are detailed
accounts of familiar topics such as the equilibrium measure of the quar-
tic potential, since these illustrate techniques that are useful in many
problems. The book develops classical and free probability in parallel,
in the hope that the analogy makes free probability more accessible.
The presentation is mainly rigorous, although some important proofs
are omitted. In order to understand the standard ensembles of random
matrix theory, the reader must have some knowledge of Lie groups, so
the book contains an abbreviated treatment which covers the main cases
that are required and emphasizes the classical compact linear groups.
Likewise, the presentations of Gaussian measures in
Ornstein–Uhlenbeck process in
but do not give a complete perspective on the theory. Similarly, the
treatment of free probability describes only one aspect of the topic.
Some of the results and proofs are new, although the lack of a specific
reference does not imply originality. In preparing the Sections 2.3, 2.4
and 2.6 on Lie groups, I have used unpublished notes from lectures given
by Brian Steer in Oxford between 1987 and 1991.
sults originally published by the author in [17] and [19], with technical
improvements due to ideas from Bolley’s thesis [29]. The material in
Chapter 6
lecture to postgraduate students attending the North British Func-
tional Analysis Seminar at Lancaster in 2006; likewise, Sections 8.1, 8.2,
and 7.3 are drawn from postgraduate lectures at Lancaster. Conversely,
Sections 7.2, 12.2 and 2.5 are based upon dissertations written by my
former students Katherine Peet, Stefan Olphert and James Groves. Sub-
stantial portions of
McCafferty’s PhD thesis [113], which the author supervised.
In his authoritative guide to lakeland hillwalking [168], Wainwright
offers the general advice that one should keep moving, and he discusses
6 possible ascents of Scafell Pike, the optimal route depending upon the

Introduction 3
starting point, the time available and so on. Similarly, the author pro-
poses 6 routes through the book, in addition to the obvious progression
1.1–14.3 which goes over the local maxima.
(1) Compact groups feature in the first half of the book, especially in
Sections 1.2, 2.3–2.9, 3.8, 3.9, 5.1, 7.4, 10.5.
(2) Generalized orthogonal ensembles feature in the middle of the book,
particularly in 1.5, 2.5, 3.4–3.7, 4.4–4.7, 6.3.
(3) Convergence to equilibrium distributions is the topic in 1.1, 3.4–3.9,
5.2–5.5, 10.6, 11.4.
(4) Free probability features in 4.3, 4.5, 4.8, 6.5, 6.6, 13.5, 14.1–3.
(5) Semicircular and similar special distributions appear in 4.4–4.7, 5.5,
7.3, 13.5, 14.3.
(6) Integrable operators appear in 9.1–9.7 and 11.2.
To summarize the contents of sections or the conclusions of examples,
we sometimes give lists of results or definitions with bullet points. These
should be considered in context, as they generally require elaboration.
There are exercises that the reader should be able to solve in a few hours.
There are also problems, which are generally very difficult and for which
the answer is unknown at the time of writing.
There are many important topics in random matrix theory that this
book does not cover, and for which we refer the reader elsewhere:
(i) the orthogonal polynomial technique and Riemann–Hilbert theory,
as considered by Deift in [56];
(ii) connections with analytic number theory as in [98, 50];
(iii) applications to von Neumann algebras, as developed by Voiculescu
and others [163, 164, 165, 166, 83, 84, 85, 77];
(iv) applications to physics as in [89];
(v) joint distributions of pairs of random matrices as in [76];
(vi) random growth models, and similar applications.
Jessica Churchman first suggested this topic as the subject for a book.
I am most grateful to Graham Jameson, Fran¸cois Bolley, Alex Belton,
Stefan Olphert, Martin Cook and especially James Groves for reading
sections and suggesting improvements. Finally, I express thanks to Roger
Astley of Cambridge University Press for bringing the project to fruition.

1
Metric measure spaces
Abstract
The contents of this chapter are introductory and covered in many stan-
dard books on probability theory, but perhaps not all conveniently in one
place. In Section 1.1 we give a summary of results concerning probability
measures on compact metric spaces. Section 1.2 concerns the existence of
invariant measure on a compact metric group, which we later use to con-
struct random matrix ensembles. In Section 1.3, we resume the general
theory with a discussion of weak convergence of probability measures on
(noncompact) Polish spaces; the results here are technical and may be
omitted on a first reading. Section 1.4 contains the Brunn–Minkowski
inequality, which is our main technical tool for proving isoperimetric
and concentration inequalities in subsequent chapters. The fundamen-
tal example of Gaussian measure and the Gaussian orthogonal ensemble
appear in Section 1.5, then in Section 1.6 Gaussian measure is realised
as the limit of surface area measure on the spheres of high dimension. In
Section 1.7 we state results from the general theory of metric measure
spaces. Some of the proofs are deferred until later chapters, where they
emerge as important special cases of general results. A recurrent theme
of the chapter is weak convergence, as defined in Sections 1.1 and 1.3,
and which is used throughout the book. Section 1.8 shows how weak
convergence gives convergence for characteristic functions, cumulative
distribution functions and Cauchy transforms.
1.1 Weak convergence on compact metric spaces
Definition(Polish spaces). Let (Ω,d ) be a metric space. Then (Ω,d)
is said to be complete if every Cauchy sequence converges; that is,
4

Metric measure spaces 5
whenever a sequence (x
n) in Ω satisfiesd(x n,xm)→0asn, m →∞,
there existsx∈Ω such thatd(x
n,x)→0asn→∞.
A metric space (Ω,d) is said to be separable if there exists a sequence
(x
n)

n=1
in Ω such that for allε>0 and allx∈Ω, there existsx nsuch
thatd(x, x
n)<ε. Such a sequence (x n) is said to be dense.
A complete and separable metric space (Ω,d) is called a Polish space.
A mapϕ:(Ω
1,d1)→(Ω 2,d2) between metric spaces is anisometryif
d
2(ϕ(x),ϕ(y)) =d 1(x, y) for allx, y∈Ω.
LetC
b(Ω;R) be the space of bounded and continuous functionsf:
Ω→Rwith the supremum normΓfΓ
∞=sup{|f(x)|:x∈Ω}.
Definition(Compact metric spaces). A metric space is said to be (se-
quentially) compact if for any sequence (x
n) in Ω there existx∈Ωand
a subsequence (x
nk
) such thatd(x nk
,x)→0asn k→∞. The reader
may be familiar with the equivalent formulation in terms of open covers.
See [150].
Definition(Total boundedness). Let (Ω,d ) be a metric space. Anε-net
is a finite subsetSof Ω such that for allx∈Ω, there existss∈Ω such
thatd(x, s)<ε.If (Ω,d) has anε-net for eachε>0, then (Ω,d)is
totally bounded.
A metric space is compact if and only if it is complete and totally
bounded. See [150].
Proposition 1.1.1Suppose that(K, d)is a compact metric space. Then
C(K;R)is a separable Banach space for the supremum norm.
Proof.Let (x
n) be a dense sequence inKand letf n:K→Rbe
the continuous functionf
n(x)=d(x, x n). Then for any pair of distinct
pointsx, y∈Kthere existsnsuch thatf
n(x)∆=f n(y). Now the algebra
A=

β(0)I+

S:S⊂N
βS

j:j∈S
fj(x):β S∈Qfor allS;β S=0
for all but finitely manyS;Sfinite
Γ
(1.1.1)
that is generated by thef
nand the rationals is countable and dense in
C(K;R) by the Stone–Weierstrass theorem; henceC(K;R) is separable.
See [141]. ff
Definition(Dual space). Let (E, Γ.Γ) be a real Banach space. A
bounded linear functional is a mapϕ:E→Rsuch that:

6 Random Matrices: High Dimensional Phenomena
(i)ϕ(sx+ty)=sϕ(x)+tϕ(y) for allx, y∈Eands, t∈R;
(ii)ΓϕΓ=sup{|ϕ(x)|:x∈E;ΓxΓΛ1}<∞.
LetE

be the space of all bounded linear functionals. LetB={x∈
E:ΓxΓΛ1}; then the product topology on [−1,1]
B
is generated by the
open sets
{(x
b)b∈B:|xbj
−ybj
|<εj;j=1,...,n} (1.1.2)
given byb
j∈B,y bj
∈[−1,1] andε>0forj =1,...,n. Further,
B

={φ∈E

:ΓφΓΛ 1}may be identified with a closed subspace
of [−1,1]
B
via the mapφΘff(φ(x)) x∈B. This is the weak

orσ(E

,E)
topology onB

. See [63, 141].
Theorem 1.1.2(Mazur).LetEbe a separable Banach space. ThenB

is a compact metric space for the weak

topology. Further,Eis linearly
isometric to a closed linear subspace ofC(B

;R).
Proof.By Tychonov’s theorem [141], [−1,1]
B
is a compact topological
space, and hence the closed subspace{(φ(x))
x∈B:φ∈B

}is also
compact; this is known as Alaoglu’s theorem. Now we show thatB

has a metric that gives an equivalent topology; that is, gives the same
collection of open sets.
Let (x
n)

n=1
be a dense sequence inBand let
d(ψ, ϕ)=


n=1
2
−n
|ϕ(xn)−ψ(x n)|(ϕ, ψ∈B

),(1.1.3)
so thatddefines a metric onB

. Now we check thatdinduces a compact
Hausdorff topology onB

, which must coincide with the weak

topology.
Let (ϕ
j) be a sequence inB

. We extract a subsequence (ϕ
j1(k))

k=1
such
thatϕ
j1(k)(x1) converges asj 1(k)→∞; from this we extract a further
subsequence (ϕ
j2(k))

k=1
such thatϕ
j2(k)(x2) converges asj 2(k)→∞;
and so on. Generally we havej
k:N→Nstrictly increasing and
j
k(n)=j k−1(m) for somem≥n. Then we introduce the diagonal
subsequence (ϕ
jk(k)). By Alaoglu’s theorem there existsφ∈B

that
is a weak

cluster point of the diagonal subsequence, and one checks
that
d(φ, ϕ
jk(k))=


n=1
2
−n
|φ(xn)−ϕ
jk(k)(xn)|→0 (1.1.4)
asj
k(k)→∞sinceϕ
jk(k)(xn)→φ(x n)asj k(k)→∞for eachn.

Metric measure spaces 7
Letf∈E. Thenfgives a function
ˆ
f:B

→RbyϕΘffϕ(f)which
is continuous by the definition of the weak

topology. Further, by the
Hahn–Banach Theorem [141] we have
Γ
ˆ

∞=sup{|ϕ(f)|:ϕ∈B

}=ΓfΓ, (1.1.5)
sofΘff
ˆ
fis a linear isometryE→C(B

;R). The range of a linear
isometry on a Banach space is complete and hence closed. ff
Definition(Borel measures). Let (Ω,d ) be a Polish space. Aσ-algebra
Aon Ω is a collection of subsets of Ω such that:
(σ1) Ω∈A;
(σ2) ifA∈A, then Ω\A∈A;
(σ3) if (A
n)

n=1
satisfiesA n∈Afor alln, thenA=


n=1
Anhas
A∈A.
The setsAin aσalgebraAare called events.
The open subsets of Ω generate the Borelσ-algebraB(Ω) andM
b(Ω)
is the space of bounded Borel measuresµ:B(Ω)→Rsuch that
(i)ΓµΓ
var=sup{
Λ
j
|µ(E j)|:E j∈B(Ω) mutually disjoint,j=
1,...,N}<∞;
(ii)µ(


j=1
Ej)=
Λ

j=1
µ(Ej) for all (E j)

j=1
mutually disjointE j∈
B(Ω).
We writeM
+
b
(Ω) for the subspace ofµ∈M b(Ω) such thatµ(E)≥0
for allE∈B(Ω) andProb(Ω) for the subspace{µ∈M
+
b
(Ω) :µ(Ω) =
1}of probability measures. Further, we writeM
1(Ω) ={µ∈M b(Ω) :
ΓµΓ
var≤1}.An event is a Borel-measurable subset of Ω. See [88]. For
any Borel setA,I
Adenotes the indicator function ofAwhich is one on
Aand zero elsewhere, soµ(A)=
Θ

IAdµ.
A probability space (Ω,P) consists of aσalgebraAon Ω, and a
probability measureP:A→R.
Theorem 1.1.3(Riesz representation theorem).Let(Ω,d )be a compact
metric space andϕ:C
b(Ω;R)→Ra bounded linear functional. Then
there exists a uniqueµ∈M
b(Ω)such that
(iii)ϕ(f)=
Θ
f(x)µ(dx)for allf∈C
b(Ω;R).

8 Random Matrices: High Dimensional Phenomena
Conversely, eachµ∈M
b(Ω)defines a bounded linear functionalϕvia
(iii) such thatΓϕΓ=ΓµΓ
var.Further,µis a probability measure if and
only if
(iv)ΓϕΓ=1=ϕ( I).
Proof.See [88]. ff
Definition(Weak convergence). Let (µ
j)

j=1
be a sequence inM b(Ω),
and letµ∈M
b(Ω). If
lim
j→∞
Φ
fdµ
j=
Φ
fdµ (f∈C b(Ω)), (1.1.6)
then (µ
j) converges weakly toµ. The term weak convergence is tradi-
tional in analysis; whereas the term weak

convergence would be more
suggestive, since we have convergence in theσ(M
b(Ω);C b(Ω)) topology.
Proposition 1.1.4LetE=C
b(Ω)and letJ:Prob(Ω) →B


[−1,1]
B
be the mapJ(µ)=(
Θ
fdµ) f∈B.Forasequence( µ j)in Prob(Ω)
andµ∈Prob(Ω),
µ
j→µweakly⇔J(µ j)→J(µ)in[ −1,1]
B
(j→∞).(1.1.7)
Proof.This is immediate from the definitions. ff
Proposition 1.1.5LetKbe a compact metric space. Then Prob(K)
with the weak topology is a compact metric space.
Proof.This follows immediately from Theorems 1.1.2 and Theorem
1.1.3 sinceProb(K) is linearly isometric to a compact subset of
C(K;R)

. ff
Theorem 1.1.2 thus gives a metric for weak convergence on a compact
metric space so thatProbbecomes a compact metric space. The defini-
tion of the metric in Theorem 1.1.2 is rather contrived, so in Section 3.3
we shall introduce a more natural and useful metric for the weak topol-
ogy, called the Wasserstein metric.
Examples.(i) The metric space (M
b(Ω),Γ.Γ var) is nonseparable when
Ω is uncountable. IndeedΓδ
y−δxΓvar= 2 for all distinct pairsx, y∈Ω.
(ii) WhereasB

is compact as a subspace of [−1,1]
B
,J(Prob(Ω))
is not necessarily compact when Ω is noncompact. For example, when
Ω=Nandδ
nis the Dirac unit mass atn∈N,(δ n) does not have any
subsequence that converges to anyµ∈Prob(N).In Proposition 1.2.5

Metric measure spaces 9
we shall introducetightnesscriteria that ensure that measures do not
leak away to infinity in this way.
In applications, one frequently wishes to move measures forward from
one space to another by a continuous function. The following result
defines thepush forwardorinducedmeasureν=ϕ µon a compact
metric space. A more general version appears in Theorem 1.3.5.
Proposition 1.1.6(Induced measure).Letϕ :(Ω
0,d0)→(Ω 1,d1)be
a Borel map between metric spaces where(Ω
1,d1)is compact. Then for
eachµ∈M
b(Ω0)there exists a uniqueν∈M b(Ω1)such that
Φ
Ω1
f(y)ν(dy)=
Φ
Ω0
f(ϕ(x))µ(dx)( f∈C b(Ω1)).(1.1.8)
Proof.Forf∈C
b(Ω1), the functionf◦ϕis also bounded and Borel,
hence integrable with respect toµ. The right-hand side clearly defines a
bounded linear functional onC
b(Ω1), and hence by Theorem 1.1.3 there
exists a unique measureνthat realizes this functional. ff
The following result is very useful when dealing with convergence of
events on probability space. See [73, 88].
Theorem 1.1.7(First Borel–Cantelli lemma).Let(A
n)

n=1
be events
in a probability space(Ω;P),andletCbe the event with elements given
by:ω∈Cif and only ifω∈A
kfor infinitely many values ofk.
If
Λ

n=1
P(A n)<∞, thenP(C)=0.
Proof.We shall begin by checking that
C=


n=1


k=n
Ak. (1.1.9)
By axiom (σ 3),C
n=


k=n
Akis an event for each integern≥1;
consequently,


n=1
Cnis also an event. Ifxbelongs toC, then for each
n, there existsk
n≥nwithx∈A kn
,sox∈C n. Consequentlyxbelongs
to


n=1
Cn. Conversely, ifx∈


n=1
Cn, then for eachn,xbelongs to
C
n; so there existsk n≥nwithx∈A kn
. But thenxbelongs to infinitely
manyA
j, and hencexis an element ofC.
We can estimate the probability ofC
n=


k=n
Akby
P(C
n)≤


k=n
P(A k), (1.1.10)

10 Random Matrices: High Dimensional Phenomena
using the axioms of measure. By hypothesis, the right-hand side is the
tail sum of a convergent series, and hence
Λ

k=n
P(A k)→0asn→∞.
Further,C⊆C
n, so we can form a sandwich
0≤P(C)≤P(C
n)≤


k=n
P(A k)→0(n→∞).(1.1.11)
HenceP(C)=0. ff
Exercise 1.1.8Letµ, ν∈Prob(Ω) be mutually absolutely continuous.
(i) Show that
ρ(µ, ν)=
Φ





1/2

satisfiesρ(µ, ν)≤1.
(ii) Now letδ(µ, ν)=−logρ(µ, ν). Show that:
(a)δ(µ, ν)≥0;
(b)δ(µ, ν) = 0 if and only ifµ=νas measures;
(c)δ(µ, ν)=δ(ν, µ).
(The triangle inequality does not hold forδ.)
1.2 Invariant measure on a compact metric group
•A compact metric group has a unique Haar probability measure.
Definition(Compact metric group). A topological group is a topological
spaceGthat is a group with neutral elementesuch that multiplication
G×G→G:(x, y)Θffxyand inversionG→G:xΘffx
−1
are continuous.
Furthermore, if the topology onGis induced by a metricd, then (G, d)
is a metric group. Finally, if (G, d) is a metric group that is compact as
a metric space, thenGis acompact metric group.
Metric groups can be characterized amongst topological groups by their
neighbourhoods of the identity as in [87, page 49], and we mainly use
metric groups for convenience. Our first application of Theorem 1.2.1 is
to show that a compact metric group has a unique probability measure
that is invariant under left and right translation. The proof given here
is due to von Neumann and Pontrjagin [130]. Compactness is essential
to several stages in the proof; the result is actually valid for compact
Hausdorff topological groups [87, 130].

Metric measure spaces 11
Definition(Haar measure). LetGbe a compact metric group. A Haar
measure is aµ
G∈Prob(G) such that
Φ
G
f(gh)µ G(dh)=
Φ
G
f(hg)µ G(dh)
=
Φ
G
f(h)µ G(dh)(g∈G, f∈C(G)), (1.2.1)
and is also known as an invariant probability measure.
Examples 1.2.1(i) LetGbe a finite group, soG={g
1,...,gn}
where theg
jare distinct. The Haar probability measure onGis
µ
G=n
−1
Λ
n
j=1
δgj
. Thenµ Ghas the special property that for all
f:G→R, the left and right averages with respect toµ
G, namely
1
n
n

j=1
f(gjg)and
1
n
n

j=1
f(ggj)(g∈G) (1.2.2)
are constant.
(ii) The circle groupT={e

:0≤θ<2π}is an abelian compact
metric group under the usual metric and multiplication on complex num-
bers. The normalized arclength measuredθ/2πis invariant under rota-
tions of the circle.
(iii) The special orthogonal groupSO(3) ={U∈M
3(R):U
t
U=I,
detU=1}is a nonabelian compact metric group. The columns ofUform
a right-handed orthonormal tripleUe
1,Ue2,Ue3, where (e 1,e2,e3)isthe
standard orthonormal basis ofR
3
. The eigenvalues ofUare 1,e

,e
−iθ
for someθ∈[0,2π), so we can identifyUwith a rotation of the unit
sphereS
2
(1) ={x∈R
3
:ΓxΓ=1}inR
3
, where the eigenvector
corresponding to eigenvalue 1 gives the axis of rotation.
Givenx∈S
2
(1), there existsU∈SO(3) such thatx=Ue 3; indeed
whenxhas colatitudeθand longitudeφ, the coordinates ofxwith
respect to (e
1,e2,e3) form the first column of
U=


sinθcosφ−sinφ−cosθcosφ
sinθsinφcosφ−cosθsinφ
cosθ 0sin θ

. (1.2.3)
Whenx∆=±e
3, there exists a unique great circle that passes throughx
ande
3and the axis of rotation ofUis perpendicular to this great circle.
Further,Ue
3=Ve3if and only if
V
−1
U=


cosφ−sinφ0
sinψcosψ0
001

. (1.2.4)

12 Random Matrices: High Dimensional Phenomena
The mapϕ:SO(3)→S
2
(1) :UΘffUe 3induces the surface area ˆσ 2on
S
2
(1), normalized to be a probability measure, from the Haar measure
onSO(3), thusSO(3)/SO(2)

=S
2
(1). In terms of the colatitudeθand
longitudeφ,wehave
dˆσ
2=
1

sinθdφdθ. (1.2.5)
Theorem 1.2.2(Haar measure).LetGbe a compact metric group.
ThenGhas a unique Haar probability measureµ
G.
Proof.Letf∈C(G;R). For each finite listA={a
1,...,an}, possibly
with repetitions, we form the left average
m
A(f)(x)=
1
n
n

j=1
f(ajx)( x∈G). (1.2.6)
We can join listsAandB={b
1,...,bk}by concatenation to form
A◦B={a
1,...,an,b1,...,bk}; (1.2.7)
thenm
A(mB(f)) =m A◦B(f).
The oscillation offis osc(f)=supf−inff. Clearly, osc(m
A(f))≤
osc(f) since supm
A(f)≤supfand infm A(f)≥inff. Further, ifgis
continuous and not constant, then there existsAsuch that osc(m
A(g))<
osc(g). To see this, we chooseK>0 such that infg<K< supgand a
nonempty open setUsuch thatg(x)<Kfor allx∈U. By compactness,
there existsA={a
1,...,an}such thatG=

n
j=1
a
−1
j
U; hence for each
x∈Gthere existsjsuch thatx∈a
−1
j
U.Now
supm
A(g)≤
1
n

K+(n−1) sup(g)

<supg (1.2.8)
and infm
A(g)≥infg,soosc(m A(g))<osc(g).
The collection of left averagesL(f)={m
A(f):A⊆G, Afinite}
that is formed from eachf∈C(G;R) is uniformly bounded and uni-
formly equicontinuous. Givenε>0, there exists by compactness an open
neighbourhoodVofesuch that
|f(x)−f(y)|≤ε(x
−1
y∈V) (1.2.9)
and hence
|m
A(f)(x)−m A(f)(y)|≤ε(x
−1
y∈V) (1.2.10)
for allA. By the Arzel`a–Ascoli theorem [141],L(f) is relatively compact
for the supremum norm.

Metric measure spaces 13
The next observation is that the closure ofL(f) must contain a con-
stant function. We lets=inf
Aoscm A(f), and by norm compactness
can choose a sequence ofA
nsuch that oscm An
(f)→sasn→∞and
m
An
(f)→gin supremum norm asn→∞.Clearlys=osc(g); suppose
with a view to obtaining a contradiction thats>0 so that gis con-
tinuous but not constant. Then there existsAsuch that oscm
A(g)<s,
and hence we could findA
nsuch thatm A(mAn
(f))<s; but this would
contradict the definition ofs;sog is constant. We define the left mean
offto be the constant value taken byg.
Likewise we can form the right average
m
Θ
B
(f)=
1
k
k

j=1
f(xbj)(x∈G); (1.2.11)
clearly this has similar properties to the left average, so we can introduce
a constant right meang
Θ
analogously. Further, the operations of forming
left and right averages commute in the sense that
m
Θ
B
(mA(f)) =m A(m
Θ
B
(f)). (1.2.12)
Givenε>0, there existAandBsuch that
|m
A(f)(x)−g|<ε, |m
Θ
B
(f)(x)−g
Θ
|<ε (x∈G); (1.2.13)
hence we have
|g−g
Θ
|≤|m
Θ
B
(mA(f))(x)−g|+|m A(m
Θ
B
(f))(x)−g
Θ
|<2ε.(1.2.14)
Consequentlyg=g
Θ
, and we can define the left and right meanµ(f)of
fto be their common valueµ(f)=g =g
Θ
.
By the construction, the translatesf
y(x)=f (yx)andf
y
(x)=f(xy)
satisfyµ(f
y)=µ(f
y
)=µ(f); consequently
µ(m
A(f)) =µ(m
Θ
B
(f)) =µ(f) (1.2.15)
for allA, B.
We now check the axioms of Theorem 1.1.1.
(i) Clearlyµ(tf)=tµ(f) holds fort∈R.Toseethatµis additive,
we introducef,h∈C(G) and, for givenε>0 introduce listsA, Bsuch
that
|m
A(f)(x)−µ(f)|<ε,|m
Θ
B
(h)(x) −µ(h)|<ε. (1.2.16)

14 Random Matrices: High Dimensional Phenomena
Now
|m
Am
Θ
B
(f+h)(x) −µ(f)−µ(h)|<2ε, (1.2.17)
soµ(f+h)=µ( f)+µ(h).
(ii) If−1≤f(x)≤1, then−1≤m
A(f)(x)≤1, so−1≤µ(f)≤1.
Hence by the Riesz representation theorem there exists a uniqueµ
G∈
Prob(G) such that
µ(f)=
Φ
G
f(x)µ G(dx). (1.2.18)
We have already seen thatµis invariant under left and right translation.
The uniqueness ofµis immediate from the construction. ff
Definition(Integer part). For x∈R,letαxλ= max{n∈Z:n≤x}be
the integer part ofx,sox?xλis the fractional part. Further, letAbe
a finite set, and Athe cardinality ofA.
Exercise 1.2.3(Weyl’s equidistribution theorem). Letαbe irrational
and letµ
N∈Prob(T)be
µ
N=
1
N
N−1

j=0
δ
exp(2πijα) . (1.2.19)
Prove the following equivalent statements.
(i)µ
Nconverges weakly to normalized arclengthdθ/2πasN→∞.
(ii) For all intervals (a, b)⊆(0,1),
1
N
{k∈Z:0≤k≤N−1;αk?αkλβ(a, b)}→b−a(N→∞).
(1.2.20)
(iii) For all trigonometric polynomialsf(e
2πiθ
)=
Λ
n
j=−n
aje
2πijθ
,
Φ
T
fdµN→
Φ
T
f(e
2πiθ
)


(N→∞). (1.2.21)
Conversely, ifαis rational, thenµ
Ndoes not converge todθ/2π.
Definition(Monogenic). LetAbe a compact metric group. Say that
Ais monogenic if there existsg∈Asuch thatAequals the closure of
{g
n
:n∈Z}. The term topologically cyclic is also used for monogenic,
andsuchag is said to be a generator ofA.
By continuity of multiplication, any monogenic group is abelian.
Furthermore, a finite group is monogenic if and only if it is cyclic.

Metric measure spaces 15
Proposition 1.2.4(Kronecker’s theorem).
(i) The torusT
n
is monogenic for eachn=1,2,....
(ii) LetAbe a compact abelian metric group with a closed subgroup
A
0such thatA 0is the connected component that contains the
neutral element ofA, the quotient groupA/A
0is cyclic andA 0is
continuously isomorphic as a group toT
n
for somen.ThenA is
monogenic.
Proof.(i) Let{1,α
1,...,αn}be real numbers that are linearly indepen-
dent over the rationals, and considerg=(e
2πiα1
,...,e
2πiαn
). Then we
show that for anyf∈C(T
n
;C),
1
N
N−1

j=0
f(g
j
)→
Φ
T
n
f(e
2πiθ
,...,e
2πiθn
)

1

...

n

(N→∞).
(1.2.22)
Whenf=I, this is clearly true. When (m
1,...,mn)∈Z
n
\{0},and
f(e
2πiθ1
,...,e
2πiθn
) = exp(2πi
n

j=1
mjθj), (1.2.23)
we can sum the geometric series
1
N
N−1

k=0
f(e
2πikα1
,...,e
2πikαn
)=
1
N
N−1

k=0
exp
ξ
2πki
n

j=1
mjαj

=
1−exp(2πiN
Λ
n
j=1
mjαj)
N(1−exp(2πi
Λ
n j=1
mjαj))
(1.2.24)
where the denominator is non zero by rational independence of the
exponents. Hence
1
N
N−1

k=0
f(e
2πikα1
,...,e
2πikαn
)→0(N→∞).(1.2.25)
By F´ejer’s theorem, anyf∈C(T
n
;C) can be approximated in the
uniform norm by a finite sum of exp(2πi
Λ
n
j=1
mjθj); hence the limit
formula holds in general. As in Exercise 1.2.3,{g
j
:j∈Z}is dense inT
n
.
(ii) Letσ:A→A/A
0be the quotient map, and observe thatA/A 0is
compact and discrete, hence finite. By hypothesis,A/A
0is isomorphic
to the cyclic group of orderkfor some finitek, and we introducev∈A
such thatσ(v) generatesA/A
0, and by (i) we can introduce a generator
u∈A
0forA 0. Evidently the closure of the group generated byuandv

16 Random Matrices: High Dimensional Phenomena
isA. Further,v
k
∈A0,sov
k
u
−1
∈A0and we can introduces∈A 0such
thats
k
=v
k
u
−1
.Now our generator forAisg=vs
−1
; indeed,u=g
k
ands∈A 0=clγg
k
,sov=gsβγg. ff
Definition(Actions). LetGbe a compact metric group and Ω a Polish
space. An action ofGon Ω is a continuous mapα:G×Ω→Ω:
(g, ω)Θffα
g(ω) such thatα e(ω)=ω for allω∈Ω, andα h◦αg=αhg
for allg, h∈G.
Definition(Invariant measure). Letµbe a measure on Ω. Say thatµ
is invariant forαif
Φ

f(αh(ω))µ(dω)=
Φ

f(ω)µ(dω)(h∈G)
for all continuousf:Ω→ Rof compact support.
Example.LetGbe a compact metric group and letG= Ω. Then there
are three natural actions ofGonG:
(i) left translation, where (g, h)Θffgh;
(ii) right translation, where (g, h)Θffhg
−1
;
(iii) conjugation, where (g, h)Θffghg
−1
, for allg, h∈G.
Verify that Haar measure is invariant for all of these actions.
1.3 Measures on non-compact Polish spaces
In applications to random matrix theory we use Polish spaces (Ω,d)
which are non-compact, and hence there are various natural topologies
onM
b(Ω). In this section we aim to recover results which are analogous
to those of Section 1.1, often by imposing stronger hypotheses. The
contents of this section are rather technical, so the reader may wish to
skip this on a first reading and proceed to 1.4.
For technical convenience, we sometimes wish to have a standard com-
pact space into which we can map other spaces. Let

∞=[0,1]
N
={(x j)

j=1
:xj∈[0,1]} (1.3.1)
have the metric
d
∞((xj),(yj)) =


j=1
2
−j
|xj−yj|. (1.3.2)
LetB

(Ω) be the space of bounded and Borel measurable functions
f:Ω→ Rwith the supremum normΓΓ
∞.

Metric measure spaces 17
Proposition 1.3.1(Kuratowski).
(i)(Ω
∞,d∞)is compact.
(ii) For any Polish space(Ω,d), there exists a continuous and injective
functionϕ:Ω→ Ω
∞.
(iii) Let(Ω
0,d0)be a compact metric space. Then there exists an iso-
metric mapj:Ω
0→B

(Ω∞).
Proof.(i) The metric space (Ω
∞,d∞) is complete and totally bounded,
hence compact.
(ii) Let (ω
j)

j=1
be a dense sequence in (Ω,d), and let
ϕ(x)=

d(x, ω
j)
1+d(x, ω j)


j=1
. (1.3.3)
One can easily check thatϕis injective and thatd(x
n,x)→0 implies
d
∞(ϕ(x n),ϕ(x))→0asn →∞.
(iii) Letf
x:Ω0→Rbe the functionf x(y)=d 0(x, y). Thenf xis
continuous, hence Borel measurable and by the triangle inequality, the
mapxΘfff
xis an isometry Ω0→B

(Ω0). Now by (ii) there is an in-
jective and continuous mapϕ
0:Ω0→Ω ∞, which by compactness must
give a homeomorphism Ω
0→ϕ 0(Ω0). Thus there is a linear isometric
morphismB

(Ω0)→B

(Ω∞) with range equal to
{f∈B

(Ω∞):f(x)=0;x∈Ω ∞\ϕ0(Ω0)}.
The composition of these maps gives the required isometry Ω
0→
B

(Ω∞). ff
Definition(Inner regularity). LetC
c(Ω) be the space of continuous
and compactly supported functionsf:Ω→R, and letC
c(Ω)

={g∈
C
b(Ω) :g=f+tI;t∈R,f∈C c(Ω)}. Suppose that (Ω,d) has infinite
diameter and that closed and bounded subsets of Ω are compact. Then
the norm closure ofC
c(Ω) isC 0(Ω) ={f∈C b(Ω) :f(x)→0, as
d(X, X
0)→∞}for someX 0∈Ω.
Letµ∈M
+
b
(Ω) where Ω is a non-compact Polish space. Thenµis
inner regular if eachε>0 there exists a compact subsetKsuch that
µ(Ω\K)<ε.An inner regular Borel measure is called a Radon measure.
See [88].
Lemma 1.3.2(Lusin).Let(Ω,d)beaPolishspace,letµ ∈Prob(Ω).
Thenµis inner regular.
Proof.Givenε>0, let (a
n)

n=1
be a dense sequence and letB n,k={x∈
Ω:d(x, a
n)≤1/k}.Then


n=1
Bn,k=ΩsothereexistsN(k)<∞

18 Random Matrices: High Dimensional Phenomena
such that
µ

N(k)

n=1
Bn,k

>1−2
−k
ε. (1.3.4)
Now letK=


k=1

N(k)
n=1
Bn,k, and observe thatKis a closed and
totally bounded subset of a Polish space, henceKis compact. Further,
µ(Ω\K)≤


k=1
µ

Ω\
N(k)

n=1
Bn,k

< ∞

k=1
2
−k
ε≤ε. (1.3.5)

Lemma 1.3.3The topologyσ(Prob(Ω),C
b(Ω))equals the topology
σ(Prob(Ω),C
c(Ω)

); that is, forµ∈M
+
b
(Ω)asequence( µ j)

j=1
satisfies
(i)
Φ
gdµ
j→
Φ
gdµ (g∈C b(Ω)) as j→∞ (1.3.6)
if and only if
(ii)µ
j(Ω)→µ(Ω)
and
Φ
fdµ
j→
Φ
fdµasj→∞ (f∈C c(Ω)). (1.3.7)
Proof.(i)⇒(ii) This is clear.
(ii)⇒(i) Letg∈C
b(Ω) andε>0.By Lusin’s Lemma, there exists a
compact setKsuch that
Θ
Ω\K
dµ < ε; hence by Urysohn’s Lemma [141,
p. 135] there existsh∈C
c(Ω) such that 0≤h≤1andh(x)=1on K.
By hypothesis, there existsj
0such that
Θ
(1−h)dµ j<εfor allj≥j 0.
Now
Φ
gdµ
j−
Φ
gdµ=

Φ
ghdµ j−
Φ
ghdµ

+
Φ
g(1−h)dµ
j−
Φ
g(1−h)dµ(1.3.8)
whereghis of compact support, and
κ
κ
κ
Φ
g(1−h)dµ
j
κ
κ
κΛΓgΓ
Φ
(1−h)dµ
j<εΓgΓ; (1.3.9)
likewise
κ
κ
κ
Φ
g(1−h)dµ
κ
κ
κΛΓgΓ
Φ
(1−h)dµ < εΓg Γ, (1.3.10)
and hence
lim
j
Φ
gdµ
j=
Φ
gdµ. (1.3.11)

Metric measure spaces 19
Lemma 1.3.4Let(h
n)

n=1
be a sequence inC c(Ω)such that0≤h 1≤
h
2≤ ··· ≤1.Letϕ∈C b(Ω)

and suppose that0≤f≤1⇒0≤
ϕ(f)≤1forf∈C
b(Ω)andϕ(h n)→1asn→∞. Then there exists
µ∈Prob(Ω) such thatϕ(f)=
Θ
fdµ.
Proof.Letϕ
n(f)=ϕ(h nf); so thatϕ ndefines an element ofC b(Ω)

such that 0≤ϕ n(f)≤1 when 0≤f≤1. Sinceh nhas compact
support, there exists a positive compactly supported measureµ
nsuch
thatΓµ
nΓvar=ϕn(I)=ϕ( h n)andϕ n(f)=
Θ
fdµ n. Further,ϕ n→ϕ
asn→∞in [−1,1]
B
since

n(f)−ϕ(f)|=|ϕ((1−h n)f)|ΛΓfΓ ∞ϕ(1−h n)→0. (1.3.12)
Now we define, for each Borel setE,µ(E) = lim
n→∞µn(E), where
the limit exists by monotonicity. Given any sequence of disjoint subsets
(E
j)

j=1
withE=


j=1
Ej,wehaveµ n(E)=
Λ

j=1
µn(Ej); by mono-
tonicity, we deduce thatµ(E)=
Λ

j=1
µ(Ej),soµis countably additive.
By the Lemma,µ
n→µweakly, and hence
ϕ(f) = lim
n
ϕn(f) = lim
n
Φ
fdµ
n=
Φ
fdµ. (1.3.13)
We now have a satisfactory generalization of Proposition 1.1.6.ff
Theorem 1.3.5(Induced probability).Let(Ω
j,dj)forj=1,2be Polish
spaces and letϕ:Ω
1→Ω 2be continuous. Then for eachµ∈Prob(Ω 1),
there exists a uniqueν∈Prob(Ω
2)such that
Φ
Ω2
f(x)ν(dx)=
Φ
Ω1
f(ϕ(x))µ(dx)( f∈C b(Ω2)). (1.3.14)
Proof.By Lusin’s Lemma, for eachj≥2 there existsµ
j∈M
+
b
(Ω1)
such thatµ
j(E)=µ( K j∩E)andΓµ jΓvar≥1−1/j, whereK jis some
compact subset of Ω
1.Nowϕ(K j) is a compact subset of Ω2andϕ µ j
is supported onϕ(K j). Letφ j∈Cb(Ω2)

be
φ
j(f)=
Φ
Kj
f(ϕ(x))µ(dx), (1.3.15)
and letφbe a weak

cluster point of (φ j)

j=1
, as in Alaoglu’s theorem.
By Lemma 1.3.3, there existsν∈Prob(Ω
2) such that
φ(f)=
Φ
Ω2
fdν= lim
j
Φ
Kj
f(ϕ(x))µ(dx)=
Φ
Ω1
f(ϕ(x))µ(dx).(1.3.16)

20 Random Matrices: High Dimensional Phenomena
Definition(Induced probability).In Theorem 1.3.5,νis the probability
measure thatϕinduces fromµ, denotedν=ϕ µ.
Whereas Proposition 1.1.5 does not have a simple generalization to
non-compact metric spaces, the following is a usable analogue. The idea
is to introduce a weight that diverges at infinity, and then restrict at-
tention to families of probabilities that are ‘tight’ with respect to this
weight.
Proposition 1.3.6(Tightness) . Suppose thatw∈C(Ω)hasw≥0
andK
M={x∈Ω:w(x)≤M}compact for eachM<∞.Suppose
moreover that(µ
j)

j=1
is a sequence in Prob(Ω) such that
Θ
w(x)dµ j<L
forj=1,2,...for someL<∞. Then there exists µ∈Prob(Ω) and a
subsequence(µ
nk
)such thatµ nk
→µweakly.
Proof.We observe thatµ
n(Ω\K ML)≤1/Mfor alln≥1. ForM=2,
we choose a subsequence (µ
n2(k))

k=1
andµ
(2)
such thatµ n2
→µ
(2)
weakly onK 2Lwhere 1/2 ≤µ
(2)
(K2L)≤1asn 2(k)→∞.Nextwe
choose a further subsequence (µ
n3(k))

k=1
and aµ
(3)
such thatµ n3
→µ
(3)
weakly onK 4Lwhere 3/4 ≤µ
(3)
≤1asn 3(k)→∞. By Lemma 1.3.3,
the diagonal subsequence (µ
nk(k))

k=1
hasµ jj→µfor some probability
measureµ. ff
Example 1.3.7To see how measures can leak away, consider the co-
ordinate projectionϕ:R
2
→R:ϕ(x, y)=xand the measures
µ
n=δ
(0,n)onR
2
. Thenϕ µ n=δ0, whereasµ n→0intheweak
topologyσ(Prob(R
2
),Cc(R
2
)) asn→∞.
Examples.In subsequent sections we shall use the weights:
(i)w(x)=d( x, x
0)
s
wheres>0andx 0∈Ω is fixed;
(ii)w(x) = 2 log
+
|x|onR.
Definition(Lebesgue spaces). Letµbe a Radon measure on a Polish
space (Ω,d). We identify Borel-measurable functionsfandgiff(x)=
g(x) except on a set ofµmeasure zero. Then for 1≤p<∞, we define
L
p
(µ;C) to be the space of Borel-measurablef:Ω→ Csuch that
Θ

|f(x)|
p
µ(dx)<∞. ThenL
p
(µ;C) forms a Banach space for the norm
ΓfΓ
L
p=(
Θ

|f(x)|
p
µ(dx))
1/p
.
Proposition 1.3.7(H¨older’s inequality).Letµbe a positive Borel mea-
sure on a Polish spaceΩsuch thatµ(K)is finite for each compact set

Metric measure spaces 21
K.Let1 <p,q<∞satisfy1/p+1/q=1. Then for allf∈L
p
(µ;C)
andg∈L
q
(µ;C), the productfgbelongs toL
1
(µ;C)and satisfies
κ
κ
κ
Φ

f(x)g(x)µ(dx)
κ
κ
κ≤

Φ

|f(x)|
p
µ(dx)

1/p
Φ

|g(x)|
q
µ(dx)

1/q
.
(1.3.17)
Proof.See [71]. ff
Exercise 1.3.8(i) The setPof probability density functionsfon
R
n
consists of integrable functionsf:R
n
→Rsuch thatf≥0and
Θ
f(x)dx=1.There is a natural mapP→Prob(R
n
):fΘfff(x)dx,
such that the range is closed forΓ.Γ
var, but is not closed for the weak
topology. Indeed, one can check that
Γf(x)dx−g(x)dxΓ
var=
Φ
|f(y)−g(y)|dy,
so the range is norm closed sinceL
1
is complete. For eachx 0∈R
n
andj=1,2,...,there existsf j∈Psuch thatf jis supported on
{x:Γx−x
0ΓΛ1/j}; thenf j(x)dxconverges weakly to the Dirac point
massδ
x0
asj→∞. This construction gives rise to the termDirac delta
function, which is sometimes used for the measureδ
x0
.
(ii) Letf
j∈Pbe supported on{x:ΓxΓΛ1} and suppose that
Γf
jΓL
2≤Mfor someMand allj, and thatf j(x)dx→µweakly asj→
∞for someµ∈Prob(R
n
). Show thatµis absolutely continuous with
respect to Lebesgue product measure, and hence thatµ(dx)=f (x)dx
for somef∈P.
Exercise 1.3.9(i) Let 1<p,q<∞ satisfy 1/p+1/q=1.Use
H¨older’s inequality to show that, forf∈L
p
(R),g∈L
1
(R)and
h∈L
q
(R), the functionf(x)g(x−y)h(y) is integrable with respect
todxdyand
κ
κ
κ
ΦΦ
f(x)g(x−y)h(y)dxdy
κ
κ
κΛΓfΓ
L
pΓgΓ
L
1ΓhΓL
q.
(ii) Now letf,gandhbe positive and integrable with respect todx.
By grouping the functionsf(x)
1/3
,g(x−y)
1/3
andh(y)
1/3
in suitable
pairs, show that
ΦΦ
f(x)
2/3
g(x−y)
2/3
h(y)
2/3
dxdy≤

Φ
f(x)dx
Φ
g(x)dx
Φ
h(x)dx

2/3
.

22 Random Matrices: High Dimensional Phenomena
1.4 The Brunn–Minkowski inequality

The Brunn–Minkowski inequality has a functional form due to
Pr´ekopa and Leindler.

In subsequent sections, these results are used to prove concentration
inequalities.
Lebesgue product measuremonR
n
isdx=dx 1...dxn,sotherestric-
tion to [0,1]
n
is a probability measure. Letγx, y=
Λ
n
j=1
xjyjbe the
usual bilinear pairing onR
n
, and letΓxΓ=γx, x
1/2
.For a function
f:R
n
→R, write [f>λ]={x ∈R
n
:f(x)>λ}.
Definition(Sum of sets). For A, B⊆R
n
,letA+B={a+b:a∈
A, b∈B}andtA={ta:a∈A}.
Examples.(i) In many results it is convenient to use the quantity
m(A)
1/n
sincem(tA)
1/n
=tm(A)
1/n
fort>0andA compact.
(ii) LetA=B=[0,1]
n
. ThenA+B=[0,2]
n
.
(iii) For openAandB={x:ΓxΓ<ε},{x:d(x, A)<ε}=A+B.
Theorem 1.4.1(Brunn–Minkowski). For all compact subsetsAandB
ofR
n
,
m(A+B)
1/n
≥m(A)
1/n
+m(B)
1/n
. (1.4.1)
The functional form of this inequality looks like a reverse of H¨older’s
inequality.
Theorem 1.4.2(Pr´ekopa–Leindler).Letf,g,h:R
n
→R +be inte-
grable functions and suppose that
h((1−t)x+ty)≥f(x)
1−t
g(y)
t
(x, y∈R
n
) (1.4.2)
for somet∈[0,1].Then
Φ
R
n
h(x)dx≥

Φ
R
n
f(x)dx

1−t
Φ
R
n
g(x)dx

t
. (1.4.3)
Proof.The following scheme of proof is due to Keith Ball and appears
in Pisier’s book [129]. We shall prove the Brunn–Minkowski inequality
in the casen= 1, then prove by induction the Pr´ekopa–Leindler in-
equality; finally we deduce the remaining cases of the Brunn–Minkowski
inequality.
Suppose in Theorem 1.4.1 thatn= 1, and thatAandBare finite
unions of disjoint open intervals. We prove the result by induction on
the number of intervals, the result clearly being true for one or two

Metric measure spaces 23
intervals. Suppose then thatA=

N+1
j=1
(aj,bj)andB=

L
k=1
(ck,dk)
wherea<b
1≤a2<···≤b N+1. By simple geometry we have
m(A+B)≥m

N

j=1
(aj,bj)+B

+m(a N+1,bN+1),
and hence by the induction hypothesis
m(A+B)≥m(B)+
N+1

j=1
(bj−aj)=m( B)+m( A), (1.4.4)
as required.
The map (a, b) Θffa+bmaps compact sets to compact sets, and open
sets to open sets. Given compact setsAandBandε>0, sinceA+Bis
compact, there exist finite unions
˜
Aand
˜
Bof open intervals such that
˜
A⊇A,
˜
B⊇Bandm(
˜
A+
˜
B)≤m(A+B)+ε.Then by the preceding
argument, we have
m(A)+m(B)≤m(
˜
A)+m(
˜
B)≤m(
˜
A+
˜
B)≤m(A+B)+ε; (1.4.5)
hencem(A)+m(B)≤m(A+B).Furthermore, for any Lebesgue mea-
surable setAthat has finite measure, there exists a compact subsetK
ofAsuch thatm(K)≥m(A)−ε.Hence the result holds for Lebesgue
measurable setsAandBthat have finite measure.
We now consider Theorem 1.4.2 in the casen= 1. By the hypothesis
on the functions, we have [h>λ]⊇(1−t)[f>λ ]+t[g>λ], hence
m[h>λ]≥(1−t)m[f>λ]+tm[g>λ](λ>0) (1.4.6)
holds by the previous paragraph. Integrating this inequality, we obtain
by the theorem of the arithmetic and geometric means
Φ

0
m[h>λ]dλ≥(1−t)
Φ

0
m[f>λ]dλ+t
Φ

0
m[g>λ]dλ


Φ

0
m[f>λ]dλ

1−t
Φ∞
0
m[g>λ]dλ

t
.
(1.4.7)
This gives the casen= 1 of Theorem 1.4.2.
Now suppose Theorem 1.4.2 holds for somen, and considern+1. We
writex
(n)
=(x 1,...,xn), sox
(n+1)
=(x
(n)
,xn+1).By the induction

24 Random Matrices: High Dimensional Phenomena
hypothesis and the hypotheses onh,wehave
Φ
R
n
h(x
(n)
,xn+1)dx
(n)


Φ
R
n
f(y
(n)
,yn+1)dy
(n)

1−t
×

Φ
R
n
g(z
(n)
,zn+1)dz
(n)

t
(1.4.8)
wheneverx
n+1=(1−t)y n+1+tzn+1. We now apply the inequality for
n= 1 to the functions that are defined by the integrals in this inequality;
hence the result.
Finally, we can deduce the general case of Theorem 1.4.1 from Theo-
rem 1.4.2. LetAandBbe any pair of compact sets that have positive
measure, and introduce the scaled compact setsA
Θ
=A/m( A)
1/n
and
B
Θ
=B/m(B)
1/n
so thatA
Θ
andB
Θ
have unit volume. Then by The-
orem 1.4.2 applied to the indicator functionsf=I
A
ff,g=I B
ffand
h=I
tA

+(1−t)B
ff,wehave
m(A
Θ
)
t/n
m(B
Θ
)
(1−t)/n
≤m(tA
Θ
+(1− t)B
Θ
)
1/n
.
In particular, when we taket=m(A)
1/n
(m(A)
1/n
+m(B)
1/n
)
−1
,we
have
tA
Θ
+(1− t)B
Θ
=
A+B
m(A)
1/n
+m(B)
1/n
, (1.4.9)
so
m(A
Θ
)
t/n
m(B
Θ
)
(1−t)/n
≤m

A+B
m(A)
1/n
+m(B)
1/n

1/n
(1.4.10)
and by homogeneity we recover the Brunn–Minkowski inequality
1≤
m(A+B)
1/n
m(A)
1/n
+m(B)
1/n
. (1.4.11)

Corollary 1.4.3(Isoperimetric inequalities).LetB ={x:ΓxΓ<1}
and for a bounded open setA,letA
ε={y:d(y, A)<ε}.Then
lim inf
ε→0+
mn(Aε)−m n(A)
ε
≥nm
n(A)
(n−1)/n
mn(B)
1/n
.(1.4.12)
Proof.We haveA
ε⊆A+εB,so the Brunn–Minkowski inequality gives
m
n(Aε)
1/n
≥m n(A)
1/n
+εmn(B)
1/n
,
so with a little reduction one can obtain (1.4.12). ff

Metric measure spaces 25
1.5 Gaussian measures
The standardN(0,1) Gaussian probability density function onRis
γ(x)=e
−x
2
/2
/

2π.Letdx be the usual Lebesgue product measure on
R
n
, which we sometimes refer to as volume measure. Letγ,be the usual
inner product onR
n
associated with the Euclidean normΓxΓ
2
=γx, x.
Letb∈R
n
, and letAbe a real symmetric positive definite matrix, so
A
−1
is likewise. Letσ
2
1
be the largest eigenvalue of A. Then
γ
b,A(dx)=(2π)
−n/2
ξ
det(A)

−1/2
exp

?A
−1
(x−b),(x−b)

dx
(1.5.1)
is a Gaussian measure that is the joint distribution ofx=(x
1,...,xn)
with mean vectorband covariance matrixA; we writex∼N(b, A). In
particular, the standard Gaussian hasb=0andA=I. Generally we
can write
γ
b,A(dx)=(2π)
−n/2
(det(A))
−1/2
×exp


1
2
γA
−1
x, x+γA
−1
b, x?
1
2
γA
−1
b, b

dx(1.5.2)
where the dominant term in the exponent isγA
−1
x, x≥σ
−2
1
ΓxΓ
2
, hence
the integrals converge. The coordinate projectionπ
j:xΘffx jinduces a
Gaussian measure onR.
Lemma 1.5.1(Gaussian distribution).
(i) The Gaussian densityγ
A,bis correctly normalized so as to be a
probability density function, and its moment generating function is
Φ
R
n
exp(γx, y)γ A,b(dx) = exp

1
2
γAy, y+γy, b

(y∈C
n
).(1.5.3)
(ii) The standard normal distribution is invariant under orthogonal
transformationxΘffUxforx∈R
n
andU∈O(n).
Proof.(ii) Lebesgue measure is invariant under orthogonal transforma-
tion.
(i) Suppose temporarily thatb= 0. The matrixAhas eigenvalues
σ
2
1
≥σ
2
2
≥ ··· ≥σ
2
n
>0, and there exists a real orthogonal matrix
Usuch thatUAU
t
is a diagonal matrix with these eigenvalues on the
leading diagonal. The determinant ofAsatisfies (detA)
1/2
=σ1...σn.
We introduce the diagonal matrixD
1/2
= diag(σ 1,...,σn), andz=
D
−1/2
Uxso thatdz=(σ 1...σn)
−1
dxandγA
−1
x, x=ΓzΓ
2
. This
reducesγ
0,Ato the standard Gaussianγ 0,I, and one can check that
Θ
γ
0,I(dz)=1.

26 Random Matrices: High Dimensional Phenomena
Likewise, one can check the normalizing constant whenb= 0, since
Lebesgue measure in invariant under the transformationxΘffb+x.
Now we have the correct formula for the moment generating function
wheny= 0. More generally, wheny∈R
n
, we simply replacebbyb+Ay
in the exponent ofγ
b,Aand adjust the normalizing constant accordingly.
Whenyis complex, this formal transformation gives the correct result,
and may be justified rigorously by Cauchy’s theorem. ff
Definition(The Gaussian orthogonal ensemble). LetX
jk:Ω→ Rfor
1≤j≤k≤nbe mutually independent Gaussian random variables
on (Ω, P), withX
jk∼N(0,1/2n)for j<kandX jj∼N(0,1/n);
then letX
jk=X kjand form the real symmetric matrixX=[X jk].
The Lebesgue measure on a real symmetric matrixXis the product
of Lebesgue measure on the entries that are on or above the leading
diagonal,
dX=dx
11...dx1ndx22...dx2n...dxnn. (1.5.4)
LetM
s
n
(R) be the space ofn×nreal symmetric matrices, and letν nbe
the probability measure onM
s
n
(R) that is induced fromPby the map
X:Ω→ M
s
n
(R); so that for allF∈C b(M
s
n
(R);R)wehave
Φ
F(X)ν
n(dX)=2
−n/2

n
π

n(n+1)/4
Φ
M
s
n
(R)
F(X)exp
ξ

n
2
n

j,k=1
x
2
jk

dX
(1.5.5)
and in particular we have
Φ
X
2
νn(dX) = diag
ζ
n+1
2n
η
(1.5.6)
and
Φ
trace(X
2
)νn(dX)=
n+1
2n
. (1.5.7)
Lettingλ
1,...,λnbe the eigenvalues ofX=[x jk], we can rewrite the
exponent as
n

j,k=1
x
2
jk
=traceX
2
=
n

=1
λ
2

, (1.5.8)
where
Φ
1
n
n

=1
λ
2 
νn(dξ)=
1
n
n

j,k=1
Φ
X
2
j,k
νn(dX)=
n+1
2n
. (1.5.9)

Metric measure spaces 27
The Gaussian orthogonal ensemble is the spaceM
s
n
(R) of real symmetric
matrices, when endowed with this probability measureν
n. See [116].
Remark.The term ‘orthogonal’ alludes to the fact that the measure
is invariant under the conjugation action of the real orthogonal group
O(n)byXΘffUXU
t
forX∈M
s
n
(R)andU∈O(n); the measure does
not live on the space of orthogonal matrices. Whereas the matrix entries
on or above the diagonal are mutually independent, the eigenvalues are
correlated. Later we shall consider ensembles for which the entries of the
random matrices are all correlated.
Definition(Symmetric). LetS
nbe the symmetric group on{1,...,n}.
Then forx
(n)
=(x 1,...,xn)∈R
n
, writeσx
(n)
=(x
σ(1),...,x
σ(n)); now
forf:R
n
→R, writef σ(x
(n)
)=f(σx
(n)
).Say thatfis symmetric if
f
σ=ffor allσ∈S n.
Proposition 1.5.2Suppose thatF∈(M
s
n
(R);R)invariant under or-
thogonal conjugation, so thatF(UXU
t
)=F(X)for allU∈O(n)and
allX∈M
s
n
(R). Then there exists a symmetric functionf∈C b(R
n
;R)
such thatf(λ)=F(X),whereλ=(λ
1,...,λn)are the eigenvalues of
X, listed according to multiplicity.
Proof.For eachX, there existsU∈O(n) such thatUXU
t
is a diag-
onal matrix with diagonal entries in orderλ
1≤ ··· ≤λ n, namely the
eigenvalues ofX. Hence we can identify the invariant functionFwith a
function on the real diagonal matrices. By elementary operator theory,
the eigenvalues depend continuously on the matrix entries. The function
fis symmetric since we can permute the diagonal entries by suitable
orthogonal matrices. ff
In particular, whenXis chosen from the Gaussian orthogonal en-
semble, we need to considerf(λ) whenλlies close to the sphere
{x∈R
n
:ΓxΓ=
τ
n/2}.
1.6 Surface area measure on the spheres
The spheres in high dimensional Euclidean space are important:

in the historical context;

as geometrical objects on which groups act naturally;

as models for high dimensional phenomena exhibited elsewhere.
The sphereS
n
(r)={x∈R
n+1
:ΓxΓ=r}is the boundary of the open
ballB
n+1
(r)={x∈R
n+1
:ΓxΓ<r}. For an open setEinR
n+1
,we

28 Random Matrices: High Dimensional Phenomena
define the Brunn–Minkowski outer upper content of the boundary ofE
to be
σ
n(∂E) = lim sup
ε→0+
1
ε

vol
ξ
E+B
n+1
(ε)

−vol(E)

.(1.6.1)
In particular, by takingE=B
n+1
(r), we deduce that
σ
n(S
n
(r)) =r
n
σn(S
n
(1)). (1.6.2)
For an open subsetFofS
n
(r), we introduceF δ={rf:1<r<
1+δ;f∈F}and defineσ
n,r(F) = lim sup
δ→0+
vol(F δ)/δ. Thisσ n,r
gives surface area measure onS
n
(r).
Let Γ(s)=
Θ

0
t
s−1
e
−t
dtbe Euler’s Gamma function, as discussed in
[154]. Letting then
th
coordinate projection point towards the pole, we
identify
C
φ={x=(x
(n−1)
sinθ,cosθ):x
(n−1)
∈S
n−1
(1),0≤θ≤φ}(1.6.3)
with the polar cap of colatitudeφ;so
σ
n,1(dx
(n)
)=σ n−1,1(dx
(n−1)
)sin
n−1
θdθ (1.6.4)
and so
σ
n,1(S
n
(1)) =
Γ(n/2)Γ(1/2)
Γ((n+1)/2)
σ
n−1,1(S
n−1
(1)). (1.6.5)
We wish to consider the measure of polar caps in spheres asn→∞,
and soon we find that most of the measure lies close to the equator; this
is our first example of the concentration of measure phenomenon, as we
discuss in Section 3.2. To make this precise, we consider the equatorial
regions
E
ε=
˘
(x
(n−1)
sinθ,cosθ)∈S
n
(1) :x
(n−1)
∈S
n−1
(1),
π/2−ε<θ<π/2
ˇ
. (1.6.6)
Proposition 1.6.1For eachε>0,
σ
n,1(E
ε/

n)
σn,1(S
n
(1))

1


Φ
ε
−ε
e
−x
2
/2
dx (n→∞). (1.6.7)
Proof.We need to estimate
Θ
π/2
π/2−ε/

n
sin
n−1
θdθ
Θ
π/2
0
sin
n−1
θdθ
=
Θ
ε/

n
0
cos
n−1
θdθΘ
π/2
0
cos
n−1
θdθ
=
Θ
ε
0
(1−ψ
2
/2n+O(1/n
2
))
n−1


n
Θ
π/2
0
cos
n−1
θdθ
.(1.6.8)

Metric measure spaces 29
By a familiar calculation, we have
Φ
π/2
0
cos
2n
θdθ=
(2n−1)(2n −3)...1
(2n)(2n −2)...2
π
2
, (1.6.9)
where the right-hand side appears in Wallis’s product
(2n−1)(2n −3)...3
(2n−2)(2n −4)...2

π
2

n
→1(n→∞). (1.6.10)
Thus we calculate the limit of the denominator of the previous quo-
tient, while we apply the dominated convergence theorem to the numer-
ator, thus obtaining
Φ
ε
0
(1−ψ
2
/2n+O(1/n
2
))
n−1
dψ→
Φ
ε
0
e
−x
2
/2
dx (n→∞).
Thus we obtain the said limit. ff
Let
ˆσ
n−1,

n=
σ
n−1,

n
n
(n−1)/2
σn−1,1(S
n−1
(1))
, (1.6.11)
so that ˆσ
n−1,

nis the surface area measure onS
n−1
(

n) normalized to
be a probability measure, and letπ
k,n:R
n
→R
k
:πk,n(x1,...,xn)=
(x
1,...,xk) be the projection onto the firstkcoordinates.
Theorem 1.6.2(Poincar´e–McKean [115]).Asn →∞, the probability
measureπ
k,n ˆσ
n−1,

non the firstkcoordinates converges weakly to the
standard Gaussianγ
0,IonR
k
,so
Φ
S
n−1
(

n)
g(x1,...,xk)ˆσ
n−1,

n(dx)→
1
(2π)
k/2
Φ
R
k
g(x)e
− x
2
/2
dx
(1.6.12)
asn→∞for allg∈C
b(R
k
;R).
Proof.Let ˆσ
n,1be the surface area measure onS
n−1
(1),normalized to
be a probability. By geometrical considerations, we have
dˆσ
n−1,1 =|sinθ|
n−2
dˆσn−2,1
whereθis the colatitude.

30 Random Matrices: High Dimensional Phenomena
Given (ξ
1,...,ξn)∈S
n−1
(1), we introduce coordinates
ξ
1=cosθ 1,
ξ
2=sinθ 1cosθ 2,
ξ
3=sinθ 1sinθ2cosθ 3,
.
.
.
.
.
.
ξ
n−1=sinθ 1sinθ2...sinθ n−2cosθ n−1,
ξ
n=sinθ 1...sinθ n−1. (1.6.13)
The Jacobian is lower triangular, so we can easily compute
κ
κ
κ
∂(ξ
1,...,ξn−1)
∂(θ1,...,θn−1)
κ κ κ=|sin
k
θ1sin
k−1
θ2...sinθ k|.(1.6.14)
By repeatedly applying (1.6.2), we deduce that
dˆσ
n−1,1 =|sinθ 1|
n−2
|sinθ 2|
n−3
...|sinθ k|
n−k−1
dˆσn−k−1,1 (1.6.15)
where
|sinθ
1...sinθ k|=(1−ξ
2
1
−···−ξ
2
k
)
1/2
. (1.6.16)
This one proves by induction, with the induction step
sin
2
θ1...sin
2
θk=sin
2
θ1...sin
2
θk−1−ξ
2
k
=1−ξ
2
1
−···−ξ
2
k
. (1.6.17)
Now we have
|sinθ
1|
n−2
|sinθ 2|
n−3
...|sinθ k|
n−k−1
dθ1...dθk
=
ξ
|sinθ 1|
k
|sinθ 2|
k−1
...|sinθ k|

|sinθ 1...sinθ k|
n−k−2
dθ1...dθk
=
ξ
1−ξ
2
1
−···−ξ
2
k
)
(n−k−2)/2
dξ1...dξk. (1.6.18)
Now we scale up toS
n−1
(

n), and find that forgdepending only upon
ξ
1,...,ξk,
Φ
S
n−1
(

n)
g(ξ1,...,ξk)ˆσ
n−1,

n(dξ)
=C
k,n
Φ
B
k
(

n)
g(ξ1,...,ξk)

1−
1
n
k

j=1
ξ
2
j

(n−k−2)/2
dξ1...dξk
(1.6.19)
whereB
k
(

n)={(ξ j)
k
j=1
:
Λ
k
j=1
ξ
2
j
≤n}andC k,nis some positive
constant.

Metric measure spaces 31
Hence
Φ
S
n−1
(

n)
gdˆσ
n−1,

n→
1
(2π)
k/2
Φ
R
k
g(ξ1,...,ξk)
×exp
ξ

1
2
k

j=1
ξ
2
j


1...dξk(1.6.20)
asn→∞,where the limiting value ofC
n,kis given by the Gaussian
constants in Section 1.5. ff
For each unit vectore∈R
k
,xΘεγx, egives a random variable
x
eonS
n−1
(

n) with probability distribution ˆσ
n−1,

n. The statistical
interpretation is that for each collection ofkorthogonal coordinate direc-
tions, the random variables (x
e1,...,xek
) have an asymptoticN(0,I k)
distribution; in particular they are asymptotically independent.
Remark.The previous results suggest an intuitively attractive picture
of the probability measures ˆσ
n,

nconverging to some natural probability
measure on an infinite-dimensional sphere that has Gaussian marginals;
so one can think aboutS

(

∞). See [115]. InChapter 11we show how
to construct Gaussian measure on Hilbert spaceH, but also learn there
is no rotation-invariant Radon probability measure on the sphere ofH.
Nevertheless, the notion of spherical harmonics does make sense, and there is a self-adjoint operator that corresponds to the Laplace operator on the sphere. The Ornstein–Uhlenbeck process is a convenient way of uniting these ideas.
Exercise 1.6.5Consider the unit sphereS
n−1
inR
n
.LetR jk=
x
j

∂xk
−xk

∂xj
for 1≤j<k≤n,letL =
Λ
jk
R
2
jk
and Γ(f,g)=
2
−1
[L(fg)−fLg−gLf] wheref,g:S
n−1
→Rare twice continuously
differentiable functions. Show that theR
jkgenerate rotations of the
sphere, and that that Γ(f,f)=
Λ
jk
(Rjkf)
2
.
1.7 Lipschitz functions and the Hausdorff metric
Detailed discussion of the concepts in this section is available in [40,
162].
Definition(Lipschitz function). Let (Ω
j,dj)forj=1,2 be Polish
spaces. A functionf:Ω
1→Ω 2isL-Lipschitz if
d
2(f(x),f(y))≤Ld 1(x, y)(1 .7.1)

32 Random Matrices: High Dimensional Phenomena
The infimum of suchLis the Lipschitz seminormΓfΓ
Lip.Sometimes we
use Lip to stand for the space of Lipschitz function between specified
metric spaces.
Examples 1.7.1(i) Letf:R
n
→Rbe a differentiable function such
thatφηf(x)Γ
R
n≤Lfor allx∈R
n
. Thenfand|f|areL-Lipschitz.
This follows from the mean value theorem.
(ii) The functionf(x)=x/(1 +|x|) is a Lipschitz bijectionR→
(−1,1), but the inverse function is certainly not Lipschitz.
Exercise 1.7.2Letf,g:(Ω,d)→Rbe Lipschitz functions on a com-
pact metric space. We define
δ(f)(x) = lim sup
y→x
|f(x)−f(y)|
d(x, y)
(x∈Ω). (1.7.2)
Show that
δ(tf+sg)(x)≤|t|δ(f)(x)+|s|δ(g)(x)( s, t∈R;x∈Ω) (1. 7.3)
δ(fg)(x)≤|f(x)|δ(g)(x)+|g(x)|δ(f)(x). (1.7.4)
By comparison with the previous examples, one regardsδ(f)(x)asthe
analogue ofφηf(x)Γ, without attaching any particular interpretation to
∇f.
Definition(Lipschitz equivalence). Polish spaces (Ω
1,d1)and(Ω2,d2)
are said to be bi-Lipschitz equivalent if there exists a bijectionϕ:Ω
1→

2such thatϕand its inverseϕ
−1
are Lipschitz maps. The Lipschitz
distance between Ω
1and Ω2is
d
L(Ω1,Ω2)=inf
ϕ
˘
log(ΓϕΓ
LipΓϕ
−1
ΓLip)
ˇ
. (1.7.5)
One can prove that compact metric spaces (Ω
1,d1)and(Ω2,d2)are
isometric if and only ifd
L(Ω1,Ω2) = 0; further,d Ldefines a metric on
each set of compact metric spaces that are bi-Lipschitz equivalent to a
given (Ω,d).
Unfortunately, the metricd
Lis too severe to give a useful notion of
convergence for metric measure spaces with the weak topology. Gromov
therefore introduced a more subtle concept which refines the classical
notion of Hausdorff distance; see [40].
Definition(Enlargement). Given nonempty A⊆Ω, letd(b, A)=
inf{d(b, a):a∈A}be the distance ofbtoA. Then theε-enlargement
ofAis
A
ε={b∈Ω:d(b, A)≤ε}. (1.7.6)

Metric measure spaces 33
Definition(Hausdorff distance). The Hausdorff distance between sub-
setsAandBof (Ω,d )is
d
H(A, B)=inf
˘
ε>0:A⊆B ε,B⊆A ε
ˇ
. (1.7.7)
Exercise 1.7.3(Blaschke). Let (Ω,d ) be a compact metric space and
letK(Ω) ={A⊆Ω:A compact}. Prove that (K(Ω),d
H) is a compact
metric space by proving the following.
(i) LetNbe aε-net of Ω. Show that 2
N
={S:S⊆N}is aε-net of
K(Ω).
(ii) Given a Cauchy sequence (S
n)inK(Ω), letSbe the set ofx∈Ω
such that, for eachε>0, the set{y:d(x, y)<ε}intersectsS
nfor
infinitely manyn. Show thatd
H(Sn,S)→0asn→∞.
Exercise 1.7.6LetA
n(n=1,2,...)andAbe closed subsets of a com-
pact metric space (Ω,d) such thatd
H(An,A)→0asn →∞. Suppose
thatµ
n∈Prob(A n) and thatµ n→µweakly asn→∞. Show that
µ∈Prob(A).
1.8 Characteristic functions and Cauchy transforms
With a view to later generalization, we express here some basic facts
about classical probability spaces in an unfamiliar style. Let (Ω,P)be
a probability space and letX
j(j=1,...,n) be random variables on
(Ω,P). Then the (X
j)
n
j=1
are mutually independent if and only if there
existµ
j∈Prob(R) such that
Ef(X
1,...,Xn)=
Φ
...
Φ
R
n
f(x1,...,xn)µ1(dx1)...µn(dxn),(1.8.1)
for anyf∈C
b(R
n
;R). We writeµ 1⊗···⊗µ nfor the product measure
here.
Definition(Convolution). The convolution ofµ
1,µ2∈Prob(R)isthe
uniqueν∈Prob(R) such that
Φ
f(x)ν(dx)=
ΦΦ
f(x+y)µ
1(dx)µ 2(dy)(1 .8.2)
for anyf∈C
b(R). WhenX 1andX 2are independent random variables,
X
1+X2has distributionν=µ 1∗µ2.
Definition(Characteristic function). LetXbe a random variable on
(Ω,P). Then the characteristic function ofXis
ϕ
X(t)=EexpitX (t∈R). (1.8.3)

Another Random Scribd Document
with Unrelated Content

— Ja sekin on sentään kuin rukki, ja ensi viikolla sillä pitäisi
markkinoille — —. Voi yhtäkaikki, voi tuhannen kertaa! Pitääkö
minun nyt lähteä teidän tähtenne rehun ostoon.
— Ei kaiketi se meidän syymme ole, että rehu loppuu. Isäntä
vainaan aikana ei kuulunut olleen kuin kaksi hevosta ja näitä on
neljä —
— Elä venkaile, se ei sinuun kuulu vaikka olisi neljäkymmentä.
— No ei kuulukaan, ja kuolkoot vaikka nälkään minusta nähden,
mutta sen minä sanon, ett'ei niillä enään tukkia mäestä viedä.
Joppi oli niin äkäinen, että olisi käynyt renkiin kiinni, jollei toinen
renki olisi samassa sattunut tulemaan talliin.
Hän meni suutuksissaan kamariinsa takaisin ja alkoi tuumia, mitä
nyt olisi tehtävä, sillä pahalta näytti todellakin asiain tila.
Hän oli huomannut, että paljon ne olivat haaskanneet heiniä
hukkaankin, mutta se ei nyt enään siitä miksikään muuttunut. Mutta
hän ei sentään ymmärtänyt, mitä varten niin oli tehty. — Palkolliset
puolestaan olivat kyllä ymmärtäneet oman etunsa. He pääsivät
vähemmällä vaivalla, kun työnsivät aika kasan kerralla eteen. Ei
tarvinnut sitten niin usein käydä ruokkimassa, ja oli tuumittu, että
jos elukka siitä ajoi huonompia jalkoihinsakin, niin ajakoon.
Jo alkoi Joppia vähän kaduttaa, että oli niin jyrkästi seurannut
isävainaansa kauppakirjaa, ja heti eroittanut äitinsä talontoimesta
pois, erikseen asumaan. Olisihan se voinut sentään pitää jotain
hoitoa navetassa, koska karjanrehu ennenkin riitti, ja vähällä piti,
ett'ei hän lähtenyt jo äitinsä puheille. — Mutta ei, hän ei sittenkään

mene — päätti hän uudelleen. — Olisi häpeä nyt vasta häneltä apua
hakea, ja toisekseen, jos äidin yhteen liittyisi, olisi vapaus poissa
taas, niinkuin ennenkin. Siltä pitäisi aina salata ja piiloitella yhtä ja
toista. — Häpeä on ostaa rehuakin näin aikaiseen, ja kun nyt riittäisi
edes rahat sellaisiin. — Mutta kun ei ole rahaa enään tuskin muuta
kuin markkinareissuun. — Kun ei huomannut tehdä sellaista
kauppaa, että sen metsän hinnan olisi saanut kaikki yht'aikaa — —
Häntä alkoi taas tuskastuttaa talon pito.
— Kyllä tämä sittenkin on yhtä saakelin vätystystä, mutisi hän
itsekseen. — Ollapa herra, niinkuin esimerkiksi nuo virkamiehet. Ei
ole tappelemista palkollisten kanssa, eikä karjanrehuja haettava.
Joka on virkaan päässyt, se nostaa ruunulta rahaa vaan niinkuin
roskaa, ja elelee.
Mutta siinä ei nyt sillä kertaa mikään muu auttanut, kuin panna
miehet lopuilla rahoilla, joita oli markkinoita varten säästänyt,
hakemaan olkia ja kauroja, sillä heiniä hän ei päättänyt ostaa, koska
kuuluivat kallistuneen. Ja eläähän hevoset ja karja jauhosilpullakin
lopun talven, tuumaili hän.
Kun Oskari illalla palasi Jussilan metsästä, missä oli yksinään ollut
mittauksilla, pyysi Joppi häneltä pari sataa markkaa etukäteen
lainaksi. — Sillä nyt on käynyt niin tuhannen hullusti, sanoi hän,
ett'ei auta muu, ja markkinoille meno on aivan välttämätöntä,
vaihettamaan Pekka pois, koska se on saanut patin jalkaansa.
7.

Talvimarkkinain aattopäivänä — jotka markkinat tavallisesti ovat
Kallenpäivän aikaan tammikuulla — ovat Joppi ja Oskari yhdessä
matkalla Hämeenlinnaan, Joppi vaihettamaan parempaan pattijalkaa
orittansa, ja Oskari muuten markkinoita katsomaan ja hakemaan
rahaa lisää tukinajoa varten.
Joppia veti luontonsa joka markkinoille vastustamattomasti. Ne
olivat hänen lempimatkojansa ja oli hän niihin jo nuoresta pojasta
tottunut ja ihastunut. Isä oli antanut hänen mennä — ja niin oli hän
jo aikaiseen niillä matkoilla päässyt Hattelmalassa ja Punaportilla
markkinahuijauksen, kortinpelin ja hevosvaihdon makuun, eikä häntä
sittemmin koko ikänään voinut mikään voima pidättää markkinoille
menemästä. — Hän joskus kerskaili itsekin sitä, että jo sillä ijällä oli
kehittynyt niin eteväksi markkinamieheksi. Ja totta se olikin, sillä
tarkimmilleen hän tiesi, missä paikassa mikin Hollolan tai Asikkalan
hevoshuijari tavallisesti piti korttieria, missä oli paras ja turvallisin
korttia lyödä, ja mistä sai hankituksi itsellensä yönaikana pienellä
lisämaksulla nuijan viinaa, sittenkuin puodit ja ravintolat oli sulettu.
Vankkana hän istui nytkin rekensä perässä, supinahkaturkissaan,
pitkä varret saappaat kumikalosseihin pistettynä, ja ajoi varovasti
Pekkaa, ett'ei se hikaantuisi ja siitä pörröttyisi, sillä sen täytyi
huomenna olla välttämättä sileän, jos mieli saada hyvän kaupan.
Ja varma hän oli myös, että hän hyvän kaupan tekisi, sillä pattia
Pekan jalassa oli mahdoton kenenkään huomata, kun hän oli patin
päältä polttanut karvan kuumalla raudalla lyhemmäksi, ja hangannut
sen sitten poltetulla korkilla sileäksi ja kiiltäväksi. Ja kun Pekka oli
ollut laihanpuoleinen, oli Joppi syöttänyt sillä ennen lähtöään alunaa,
josta se paisui ja tuli turpeaksi. — Keino, jonka hän oli oppinut

eräältä vanhalta hollolaiselta, ja jota hän piti erityisenä
salaisuutenaan, ilmaisematta sitä edes parhaille ystävilleen.
Iltahämyssä saapuivat markkinamiehet kaupunkiin ja pysähtyivät
Myllymäkeen. Siinä oli Jopilla tuttava kauppias, jonka luona hän oli
monasti ollut ja jonka talliin nytkin sai hevosensa yöksi.
— Se on niin hyvä, kun saa hevosensa varmaan paikkaan, oli hän
sanonut Oskarille, — kyllähän miehet aina toimeen tulevat.
Itse he läksivät kaupunkiin tapailemaan markkinatuttavia ja
hakemaan ajoissa yösijaa, sillä kauppiaalla ei ollut miehille tilaa.
Lähtiessä oli Joppi sentään varmuuden vuoksi pannut rautavitjat
hevosensa kaulaan ja lukinnut sen niillä kiinni.
Kaupungin kestikievariin he suuntasivat käyntinsä, ja kun siellä
vielä oli kamaria saatavissa, niin hyyräsivät he yhden omiin
hoteihinsa, ja menivät kapakan puolelle katsomaan kylmästä tultua
jotain suuhunsa.
Siellä oli kaikki huoneet humalaisia täynnä ja pöydät lainehti
viinaa, olutta ja kaikenlaista juomatavaraa, ja vaikka palvelijoita oli
kuin "Vilkkilän kissoja", piti heidän sentään hiessäpäin juosta,
ennenkuin saivat jokaisen janoisen, juotavata huutavan
markkinamiehen suun tukituksi.
Jopin ja Oskarin piti seisaallaan ryypätä tuutingit tiskin ääressä,
sillä istuintilaa ei ollut enään.
Oskari ehdoitti, että mentäisiin omaan kamariin ja tilattaisi sinne
— mutta silloin sattui Jopin silmä joukossa erääseen

markkinamieheen, joka veti niin hänen huomionsa puoleensa, ett'ei
hän malttanutkaan lähteä.
Se oli pitkä, tummanverinen mies, lammasnahkaturkki yllä,
remelillä kiinni vedettynä, suuri karvalakki päässä ja piiska
kainalossa, kulkien ympäri aivan kuin olisi kiireellä tullut jotakuta
hakemaan. Hän näytti olevan aikalailla humalassa, päättäen
reuhtovista ja horjuvista liikkeistä — mutta kumma kyllä, hän horjui
sentään siellä toisten joukossa ja pöytien välitse niin osavasti, ett'ei
hän kertaakaan loukannut ketään eikä tehnyt mitään vahinkoa —
josta voi tulla siihen varmaan luuloon, että hän tekeytyi
humalaisemmaksi, kuin mitä oli.
Kulkien ympäri huoneita pöydästä pöytään, aivan kuin olisi ollut
tuttava koko sekalaisen seurakunnan kanssa, kyseli hän yhtenään
hollolaiseksi murtaen, että "onks' siin' pöyläss' sellaist' miest' joka
rohkenie hevoist' vaihtaa". Useassa pöydässä näytti hänellä sentään
olleen likeisiä ystäviäkin, sillä hän istahti hetkeksi jonkun polvelle
puhuen hyvin tuttavallisesti tämän kanssa suusta suuhun, tai vieden
suunsa aivan toisen korvaan, kuiskaten jotain, silmät omituisesti
vilkkuen. Mutta kauvan hän ei malttanut olla yhdessä kohti.
— Katsopa tuota miestä, sanoi Joppi Oskarille. — Se on Pennalan
Niku Hollolasta, parhaita hevosmiehiä, mitä minä tunnen. Se
velikulta pelasi minulta viime talvimarkkinoilla toista sataa
Hattalmalassa — ja tekisipä melkein mieleni niistää sitä nyt. —
Tuttava se on kaikkein kanssa — kumma kun et sinä tunne? —
Näetkö, kuinka kulkee ympäri kuin kotonaan, heh!
— Vaan näkyvätpä sille suuttuvan tuossa pöydässä, osoitti Oskari.

— Ne on niitä pöyhkeitä Hauholaisia, kyllä minä tunnen nekin
miehet — vaan ei ole pelkoa, että Niku kiikkiin joutuu, kyllä se osaa
vastata ja osaa välttää myös. Niku huomasi Jopin ja tuli luo.
— No, Joppi, vaihdatko hevosta tällä reissulla? — hyvää iltaa
sentään.
— Onko sulla hyvä, että kannattaa vastata? kysyi Joppi.
— Karva kuin silkkiä ja kintut kuin vieteriterästä —
— Missä se on?
— Tuolla on pihalla.
— Ei sitä sentään pimeinpäin vaihdeta. Missä sinulla on korttieri?
— Myllymäessä, vanhassa paikassa.
— Jaha! — Mutta jos lähtäis koettamaan —
— Vaikka paikalla, sanoi Niku ja lyödä ravahutti piiskan varrella
saapasvarteensa.
He läksivät ulos, Oskari perässä.
Pihalle tultua, ja Nikun hevosta irti päästäessä, kuiskasi Joppi
Oskarille:
— Minä menen ilman piruuttani sen rekeen — vai tuletko mukaan
sinäkin?
— Ei minua haluta, eläkä huoli sinäkään suotta lähteä nyt.

— Kyllä minä menen, päätti Joppi ja heittäytyi puoleksi pitkälleen
Nikun rekeen, ja Niku ajoi Myllymäkeen, erään syrjäisemmän,
pahanpäiväisen talorähjän pihaan, johon seisatti. Siinä korjasi hän
hevosensa jonkunlaiseen vajantapaiseen pihan perällä, jossa kuului
olleen muitakin kylmissään hörhöttäviä hevosia, ja Joppi jäi siksi
aikaa seisomaan rakennuksen eteen.
Ainoastaan yksi ovi näkyi vievän rakennukseen sisään, sen päästä,
suoraan pihalta. Sivuseinällä olevat pienet, neliruutuiset ikkunat
olivat sisäpuolelta niin peitossa, ett'ei valonsädettäkään päässyt
pihalle, vaan että siellä sisässä sentään oli elämää aikalailla, sen
saattoi päättää siitä yhtämittaisesta, epäselvästä horinasta, joka
sieltä kuului niinkuin ampiaispesästä.
Jopin vielä seistessä pihalla, aukeni äkkiä ovi, aivan kuin jonkun
sisällisen voiman pakoituksesta — tahi niinkuin aukenee höyrypillin
läppä — ja heitti hetkiseksi pihalle valonloimauksen — kovan,
sekasotkuisen äänen rähäyksen ja vaaleana usvana kylmään ilmaan
kohahtavan höyrypilven — vaan samassa kun se sulkeutui, kuului
sieltä sisältä pihalle vaan entinen tukautettu sorina.
Joku mies oli tullut ulos ja useita ääniä oli huutanut hänen
peräänsä, että: "muista nyt, pullo mieheen — —"
Joppi oli hyvästi eroittanut miehen valoisassa ovessa ja se oli
näyttänyt ihmeen tutulta, vaan kun ovi oli äkkiä taas sulkeutunut, jäi
se kohta hänen silmiinsä valon huikaisun vuoksi entistään
pimeämmäksi, eikä hän voinut ollenkaan huomata, mihin mies oli
kadonnut.
— Ihmeesti tutun näköinen mies, mutisi hän astuessaan Nikun
perässä huoneesen sisään.

Se oli jotensakin avara huone, mutta niin matala, että pitkän
miehen otti melkein pää kattoon. Sen takana mahtoi olla pari
pienempää huonetta, koska peräseinässä oli kaksi ovea, ja
useampaa huonetta ei koko rakennuksessa ollutkaan. Joka kohta
kantoi merkkiä tavattomasta rappeutumisesta ja likaisuudesta, ja piti
olla tavallista vankempi nenävärkki, että saattoi pahoinvoimatta olla
hetkenkään siellä sisällä, sillä niin täynnä oli lattian ja la'en väli
mädännyttä oluen ja tupakan hajua, kuin ainoastaan huonoimmassa
nurkkakapakassa saattaa olla. Mutta siltä istui Jopin ja Nikun
sisäänastuessa siellä joukko iloisia markkinamiehiä keskellä
huonetta, vanhanaikuisen, kuluneen, pyöreän pöydän ympärillä,
lyöden korttia ja latkien olutta. Parin tyhjän olutlaatikon päällä
seinivieressä nukkui mies, ja toinen väsähtänyt matkantekijä makasi
peräseinällä, ovien välissä olevan pienemmän pöydän alla lattialla,
kuorsaten kauheasti.
Juuri kuin ulkoa tulijat astuivat sisään, tuli toisesta peräovesta
etummaiseen huoneesen selkä edellä vanha, paljaspäinen,
kiiltävänkuluneesen sortuukkitakkiin puettu mies, riidellen
armottomasti häntä ulostyöntävän akan kanssa. Akan
"nuuskakamarin" katto oli joskus menneisyydessä painunut niin
pahasti sisään, että hänen puheensa tuli yksinomaan kurkusta, ja
vaikka sitä sentähden oli melkein mahdoton ymmärtää, saattoi
kumminkin heidän riidastaan huomata, että siinä oli talon isäntä ja
emäntä. Sillä isäntä oli pistänyt muutaman olutpullon hinnan omaan
taskuunsa, salaa akalta, ja siitä nyt oli tora, koska akka tahtoi
välttämättä pitää itseoikeutettua kasöörin virkaansa
loukkaamattomana. —
Joppi tuntui olleen hyvin tervetullut tähän hauskaan seuraan ja
useammat tervehtivät häntä kuin vanhaa tuttua.

Nikulle ei puhuttu mitään ja hän istui kuin isäntä omaan
pöytäänsä.
— Ei Koskinen ole tullut? — kysyi hän.
— Ei, vastasi yksi joukosta, — eikä se tänä iltana sieltä tulekaan.
— No nyt on piru merrassa —
— Huomenna on asia raastuvassa esissä ja haaste saatiin joka
mies.
— Kirottu poliisi!
— Ne väittävät vaan, että hevonen on varastettu ja että
tuntomerkit lyövät yhteen sen Hausjärveltä varastetun hevosen
kanssa, jota on kuuluutettu —
— Hoo pojat! — minäpä näytän toteen, että Koskinen on sen
rehellisesti tunnetulta mieheltä ostanut —
Samassa tuli joku sisään.
Joppi istui selin oveen, eikä huomannut tulijaa, ennenkuin se alkoi
puhua. Hän kääntyi hämmästyneenä oveen päin ja näki edessään
oman entisen renkinsä, Jussin, joka kaivoi pulloja taskustaan ja
asetteli pöydälle.
— Mutta kuinka on Jussi täällä?
Sillä seikalla on oma historiansa.
Jussi oli syksyllä Hinkkalasta lähdettyään kuleksinut siellä täällä
tukkimetsissä Hämeenlinnan ympäristössä, säilyttäen

satamarkkaistaan takkinsa vuorin väliin neulottuna. Vaan pitempää
aikaa ei hän yksillä paikoilla ollut viipynyt, sillä palkka hänen
mielestään ei ollut tarpeeksi riittävä missään, ja niin etsi hän
yhtenään uusia työmaita, pelaten ja tuhlaten ansionsa mitä sai, ja
jättäen pieniä ruokavelkoja jälkeensä joka paikkaan.
Satamarkkaistansa ei hän ollut tohtinut särkeä, peläten epäluuloa,
vaan kun hän lopulta oli kyllästynyt sitä povessaan kantamaan ja
takkikin oli alkanut jo niin ränstyä, että sen piilopaikkaa oli pitänyt
monasti muutella, oli hän loppiaisen jälkeen, vähää ennen
markkinoita, tullut kaupunkiin. Mutta kun sielläkään — rikkisissä
vaatteissa ollen — ei ollut rohjennut missään kauppapuodissa tai
muussa julkisessa paikassa vetää rahaansa esille, oli hän lopulta
joutunut siihen paikkaan Myllymäessä, missä hän nyt on, ja saanut
siinä hyviä, auttavia apumiehiä. Raha oli pienitty ilman suurempaa
pulaa, Jussi hankkinut itselleen parempaa vaatetta, ja jäänyt
rahoineen siihen mielestään ylen hauskaan paikkaan viettämään
joutilaita herraspäiviä.
Kun sitten markkinat lähenivät, ehdotteli talon väki yksissä tuumin
Jussille, että ruvettaisiin kaikin puulaakiin, ennenkuin hänen rahansa
loppuu, ja hankittaisi markkina-ajaksi hyvää ansiota, josta Jussillekin
lankeisi hyöty hänen osuutensa mukaan. Tuuma oli Jussista ollut
mainio ja hän oli paikalla ollut taipuvainen niin helppoon rahan
ansioon.
Hankittiin sitten viinaa niin paljon kuin saatiin, lisättiin sitä
hiukkasen vedellä, ja korkkailtiin tyhjiin olutpulloihin valmiiksi
kauppatavaraksi. Koko varasto kätkettiin vanhaan perunakuoppaan
ulkohuonerähjän takana, ja odotettiin markkinoita. Kaupan piti käydä
Jussin nimessä ja tulisi hän hoitamaan sen anniskelua. Oluen

anniskeluun oli emäntä jo kauan sitten hankkinut itselleen oikeudet
kaupungilta.
Sinne kuoppaansa oli Jussi silloin juuri ollut menossa, kun Joppi
hänen pihalla näki ulos tulevan, ja sieltä hän nyt astui pulloineen
takaisin huoneesen.
— Mistä sinä Jussi tänne — mitä sinulla niissä pulloissa on? —
kysäsi Joppi yhteen kyytiin, ja toinen vastasi, että hän on
markkinoilla niinkuin isäntäkin, kun on oma herransa nyt — ja mitä
pulloihin tulee, niin tottapa antanevat toiset maistaa.
Sitä maistettiin, kaadettiin olutlasiin vaan, ja se oli viinaa.
Kun Joppikin tahtoi puolestansa tarjota, kysyi hän Jussilta, mistä
sitä saa?
Ei sanonut Jussi sitä hänellä itsellään olevan, mutta noutamaan
hän kyllä menee vanhalle isännälleen, jos saa ryypyn vaivastaan.
— Mitä se maksaa?
— Kaksi markkaa nuija.
— Onpa tupla hinta, mutta hae häntä pari nuijaa sentään, tuossa
on raha. — Eihän sitä muualtakaan halvemmalla saa näin myöhällä
enään, sanoi Joppi.
Hierottiin siinä välillä sitten vähän hevoskauppoja, mutta kun
jokainen kehui omaansa parhaaksi ja "kaikki siat ovat pimeässä
mustia", heitettiin kaupat päättämättömiksi aamuun ja alettiin
uudelleen lyödä nakkia.

Jussi tuli peliin mukaan.
— Hyvinpä sinä olet rahoissasi, sanoi Joppi.
— On se toista, kuin renkimiehenä, vastasi Jussi, vetäen
taskustaan osansa pöytään.
Ja hän on kovin mielissään, että pääsee kerran koettamaan Jopin
kanssa isommassa seurassa. — Nyt sitä ei auta vääryyden teko —
nyt siltä on puserrettava entiset häviöt takaisin — juuri tällä paikalla
— tässä pöydässä — hän on oppinut pelaamaan hänkin jo, ja tietää
milloinka nakkaa — —
Peli kävi kiihkeällä innolla. Miesten silmät kiiluivat savuavan
lampun valossa verisinä, lihakset vavahtelivat kasvoissa, käsi puristi
kuumeentapaisesti likaisia korttia. Kuului kirouksia ja pirullista,
voitonilosta rähähtävää naurua. Jonkun oli kasvot aivan kalpeat,
toisen taas punaisenhohtavat ja hikiset. — Tuska, kiihko ja
voitonhimo olivat yhä nousemassa.
Hetken aikaa näytti onni suosivan Jussia, ja hän voitti pienempiä
summia, riemuiten sisimmässään, kun näki Jopin häviävän.
Mutta se kääntyi pian ja kohta pyyhkieli Joppi pelin toisensa
perästä pöydältä taskuunsa, leveästi nauraen ja istuen rehevänä,
kyynärpäät pöydällä, toisten pannessa rahaa kassaan.
Jussi oli saanut ison "pietin" ja sekin meni Jopin taskuun.
Hän puri hammasta ja alaleuka alkoi pahasti vavahdella. — Haa,
saatana! ajatteli hän. — Pitääkö sen nytkin voittaman minulta —

Hän pani kiihkeänä "pietinsä" uudestaan kassaan — ja huomasi
kauhukseen, että siihen ne menivätkin jo hänen rahansa kaikki.
Kortit jaettiin, vaan kukaan muu ei nakannut kuin Joppi — ja niin
joutuivat isäntä ja entinen renki uudelleen kahden peliin.
Joppi vei heti ensimmäisen tikin.
Jussi oli niin vimmatussa kiihkossa, että koko ruumis vapisi. —
Siinä oli nyt ainoat rahat menossa samalle miehelle, joka kerran
ennen jo oli tehnyt hänestä paljaan.
Joppi sai toisenkin tikin, ja Jussilta oli valtti lopussa. — Nyt se
voittaa taas — sillä on valtti kädessä varmaan —
Jussi oli koko pelinajan ollut kohona irti tuolista, pöydän yli
kumartuneena, ja tarkastanut Jopin kouraa, missä hän korttiaan piti,
niin että oli ollut pelkkänä silmänä koko mies.
Yht'äkkiä iskee hän ylitse pöydän molemmin käsin Jopin ranteesen
kiinni niin että pulloja kaatuu —
— Roisto, kirahtaa hän hammasten välistä — sinulla on liikoja
korttia taas.
— Se on vale, hellitä! Ei ole kuin valttikymppä ja voitto on minun,
— ja hän kahmasee toisella kädellään rahat pöydältä ja vie
taskuunsa.
— Elä vie rahaa ennenkuin näytät korttisi —
— Hellitä, kirottu, kyllähän sitten näet, karjasee hurjasti Joppi.

— Ei heltiä kätesi nyt, ennenkuin avaat kourasi — avatkaa sen
koura, miehet, ja syynätkää, se on ennenkin —
Mutta samassa iskee Joppi vapaalla nyrkillään Jussia rintaan niin
että tämä kaatuu ja mukana menee koko pöytä, pullot, lasit ja
lamput, kovalla helinällä, ja huone jää pilkkoisen pimeäksi.
Isäntäväki riensi kamarista, ja kun ovi aukeni, tuli uudestaan valoa
huoneesen.
Mutta Jussi oli kavahtanut lattialta ylös. Hänen ajunsa läpi on
silmänräpäyksessä kulkenut kaikki se vääryys ja vahinko, jonka hän
viime vuonna kärsi Jopin tähden — ja nyt vielä tämä viimeinen iso
tappio — suora ryöväys — ja tämä lyönti — ja hän on temmannut
puukon tupestaan ja töyttää raivossa Jopin päälle.
Sen huomasivat toiset ja karattiin Jussin käteen. Vaan ei voitu
enään ehkäistä voimakasta iskua, saatettiin se vaan harhaan — ja
puukko tunkeutui päätä myöten Nikun olkaan.
— Ai jumala! — se otti minuun, huusi Niku, vaipuen alas.
Yleinen hämmennys syntyi.
Ei ollut kukaan siinä mylläkässä kuullut, mitenkä toisessa
kamarissa avattiin akkuna, ja kaksi naisen ääntä kirkui poliisia.
Niitä liikkuu ahkerasti Myllymäen kulmalla markkina-aikana
sotilaspatrullin kanssa.
Kapakkahuoneessa toiset olivat kovana Jussille, toiset parhaallaan
vetivät puukkoa Nikun olkapäästä, mihin se oli jäänyt törröttämään,

kun ovi äkkiä temmattiin auki ja siitä näkyi kaksi painettia. Poliisi ja
kaksi sotamiestä kiväärit kädessä, kumartui matalasta ovesta sisään.
Ne tekivät pian selvän asioista.
Niku pannaan omaan rekeensä ja viedään läänin lasarettiin
sidottavaksi — tahi kuolemaan, sillä kukaan ei voinut varmaan
päättää kuinka kävisi, koska haava näytti pahalta. —
Muu joukko kaikki, jok'ikinen viedään poliisikamariin, sekin
poikkinenä emäntä, vaikka hän paraansa päästä koettaa panna
vastaan ja sättiä poliisia. Nukkuvatkin potkitaan ylös — ja he tekevät
seuraa unenpöperöisinä — silmiään hieroen, syljeskellen ja
haukotellen — vaikka eivät tajua koko tästä komennosta mitään.
Kaikki muut käyvät vapaina, mutta Jussia vartioidaan, ja vedetään
kädet selän taakse kiinni setolkkaremmillä, jonka poliisi ottaa Nikun
hevosenkaluista, muristen, että täältä jo tänään vietiin mies ja
hevonen ja nyt viedään koko roska perässä. — Kun piru virittäisi
tuleen koko hökkelin.
Poliisikamarissa manataan ja tuomitaan koko tämä matoinen
maailma pataluhaksi, mutta erittäinkin vielä nämä markkinat, jolloin
ei saa pahaakaan rauhaa yöllä eikä päivällä. Pidetään lyhyt
poliisitutkinto, jossa selviää, että Jussi yksin on ollut syypää koko
tapaukseen, ja tulee vielä sekin ilmi, että hän on kulettanut
markkinamiehille viinaa. Jussi koettaa puollustautua urheana — ei
ymmärrä ollenkaan asemansa vaarallisuutta — ja ajaa syyn Jopin ja
isäntäväen päälle. Puhuu vääryydestä korttipelissä ja yhteisestä
viinakaupasta.
Mutta joka mies väittää yhdestä suusta, ett'ei ole ollenkaan
pelattukaan, kuin leikin vuoksi. Ja isäntäkö olisi ruvennut viinaa

myömään, joka tuntee lain — pois se!
Kaikkien nimet ja kotipaikka merkitään paperiin ja koko joukko
käsketään poistumaan, vaan pitämään varansa, jos asia sattuisi jo
markkinain loppupäivinä olemaan raastuvan oikeudessa esillä.
Ainoastaan humalaisimmat pannaan yöksi putkaan.
Jussi on tähän saakka ollut jotenkin rauhallinen, mutta kun hän
näkee, että toiset alkavat poistua, eikä häntä lasketakaan, ja kun
hän kuulee kahleitten kalinaa ja huomaa, kuinka parrakas,
tuimannäköinen poliisi niitä kantaa salin poikki häntä kohden, herää
hän äkkiä tajuunsa kuin unesta ja joutuu täydelliseen epätoivoon. —
Riistävätkö ne häneltä vapauden nyt? Hän nyyhkyttää ja pyytää —
hän jo ääneensä itkee — heittäytyy polvilleen ja rukoilee pyhän
Jumalan nimessä, että hänet päästettäisiin pois — hän tulee takaisin
vaikka jo aamulla raastupaan — hän kärsii mitä siitä tulee, jos
hänelle mitään tulee siitä — sillä se oli niiden syy — kaikki oli niiden
toisten syy — hän tekee vaikka mitä, kun hän vaan saa nyt mennä
— —
Mutta hänet riuhtaistaan itkustaan huolimatta seisaalleen,
pannaan kahleet käsiin ja jalkoihin, ja viedään melkein kantamalla ja
raastamalla vielä samana yönä linnaan.
Joppi läksi käymään torin poikki ja kääntyi keskievariin tapaamaan
Oskaria, mutta ei löytänyt häntä sieltä, ei ravintolan puolelta, eikä
kamarista. Kamarin avain oli paikoillaan naulassaan ovi pielessä ja
vuoteet liikuttamattomina.
— Mihinkä ihmeesen se olisi mennyt, kun ei vielä ole palannut ja
on tuossa paikassa jo puoliyö? tuumaili Joppi. — Mutta kyllä se on
haettava käsiin, on saattanut joutua hyvinkin hauskaan seuraan.

Hän aprikoi keskievarin portilla, seisoen kadulla. Tuumaili sinne ja
tänne, mihin päin siitä alkaisi suunnata kulkuansa, että löytäisi
etsittävänsä — sillä välttämättä nyt pitäisi nähdä, millaisessa lystissä
Oskari on, ja päästä seuraan.
Äkkiä pälkähti päähän, että pitää aivan ensimmäiseksi käydä
katsomaan susiteetista, sillä siellähän oli syksylläkin oltu yhdessä sen
kasöörin kanssa, jota Oskarin nytkin oli pitänyt tavata.
Kadut olivat jo melkein tyhjät. Ainoastaan markkinakulkuvahtia
astuskeli hiljalleen siellä täällä keskikatua, parittain, kiväärit olalla,
keskenään jutellen päivän tapahtumia, ja joku markkinamies, joka oli
saanut päänsä tarpeeksi täyteen, hoippui käytävää pitkin,
muistutellen korttieripaikkaansa.
Kulkuvahti antoi heidän paraansa mukaan luovia eteenpäin,
kunhan eivät vaan pitäneet niin suurta melua, että siitä
kaupunkilaisten yörauha häiriytyisi, sillä oman mukavuutensakin
vuoksi he eivät olleet erittäin hanakoita ketään putkaan viemään, ja
toisekseen, ei siellä olisi enään ollut tilaakaan.
Joppi kiirehti askeliaan, ja juuri kun hän lähellä matkansa
päämäärää oli astumassa sen kadun poikki, joka vie torilta linnalle,
kuuli hän ääntä ja meteliä torilta päin, ja tarkemmin kuunneltuaan
huomasi hän, että siinä vietiin Jussia.
Jussi teki kovaa vastarintaa, — sillä vapaus on ihmiselle rakkain
kaikesta — mutta ei siinä auttanut. Kulkuvahti juoksi avuksi, ja Joppi
kuuli sanottavan, että "viedään se yksi mies vaikka kantamalla —
pahennat sinä tällä vaan asiaasi —"

Pieni säälin huokaus kävi Jopin rinnasta, mutta ainoastaan
ohimennen — sillä samassa hän muisti, kuinka huonosti tämä
markkinamatka olisi voinut päättyä, joll'ei isku olisi sattunut Nikuun.
— Menkööt omaa menoaan he, hänellä on myös oma tiensä, ajatteli
hän, ja jatkoi kulkuaan oikein mielihyvissään siitä, että oli toisen
kustannuksella oman nahkansa pelastanut.
Hän astui seurahuoneen isoon saliin sisään. Se oli korkea ja
kaikuva kuin kirkko, ja oli siellä vielä — puoliyön aikaan — äänekäs
seurakunta hartaassa toimessa.
Siellä istui tukkiherroja, monenlaisia liikemiehiä, kauppamiehiä,
hevosmiehiä, kaupungin virkamiehiä ja lihavia, kaksileukaisia
Hämeen rusthollaria ja mahtavia kartanon omistajia — hyllyvätä
lihaa, jota ei huoli eikä työ ollut pakoittanut missään määrätyissä
rajoissa pysymään.
Afääreistä puhuttiin ja niitä tehtiin. Keskusteltiin hevoskaupoista ja
metsäkaupoista — ja juotiin — pääasiassa vaan juotiin —
Omituista todellakin aineellisuuden kiertokulku tässä maailmassa!
Tuhansia tukkia, monta kaunista hongikkoa kokonaan, siellä nieltiin
kurkuista alas. Ne olivat ennen hiljalleen kulkea kahnustaneeet
kuukausia — kokonaisia vuosia — purojamme, jokiamme ja siintäviä
järviemme selkiä myöten maasta pois, toisinaan hyvinkin
vastahakoisesti ahtaissa uomapaikoissa — ja nyt, saatuaan toisen
olinmuodon konjakissa, viinissä, liköörissä ja korkeakaulaisissa
sampanjapulloissa — nyt ne menivät solahtaen vaan iloisten
tukkiherrain ja topakoitten maanomistajain mahaan — eikä siinä
oksat vastaan kärränneet —
Ja myöjät olivat iloisia ja ostajat iloisempia vielä.

Joppi jäi salin ovelle seisomaan, ihmeissään siitä komeudesta,
mikä täällä kohtasi silmää, ja alkoi katseineen kulkea ympäri
pöydästä pöytään, että jos näkisi Oskarin siellä. Hän oli ensi kertaa
markkinamatkoillaan, näin ison liikkeen aikana, sattunut tähän
paikkaan.
— Voi kumminkin tätä elämää — tätä iloa! ajatteli hän ja joutui
jonkunlaiseen huumaustilaan seistessään siinä ovella, niin ett'ei
kohta enään muistanutkaan asiaansa. — Vaahtopäät lasit, kauniit
pullot, häikäisevä valo, korkkien moksahdus — ja koko ilma, joka
tulvi salista vastaan ovelle, oli niin puoleensa vetävä, että hän seisoi
paikoillaan vaan hehkuvin silmin, katsellen elämää edessään niinkuin
kaunista panoraamaa — hengittäen syvästi —
Joku mies nousee siellä eräästä pöydästä ja lähtee käymään häntä
kohti — katsellen häneen — —
— Kah, Oskari! Olethan sinä täällä, sanoo Joppi iloisesti, kun
toinen jo on tullut aivan tykö. Minä jo ajattelin, että mihinkä sinä —

— Täällähän minä olen ollut koko illan — ja täällä sitä olla pitää.
Tule pois meidän pöytään, siinä on rahamiehiä Keuruulta ja
Orihvedeltä saakka — hauskoja miehiä kerrassaan — ja meidän
kasööri.
— Sopiiko sinne tulla?
— Tottakai, omallaan täällä jokainen ottaa.
Siinä oli pöydässä parhaallaan puhe metsistä.

— Vetoa minä lyön siitä, etteikö Palomäen kankaasta makseta
viisikymmentä tuhatta, sanoi eräs joukosta, lihava suurimahainen
isäntä, hienoissa verkavaatteissa, lyöden kämmentä pöytään. — Sitä
on jo kyselty ja tässä taas porilainen kyselee.
— Enhän tiedä vielä mitään luvata, sanoi se porilainen herra, jolta
Joppikin syksyllä oli rahat saanut, — mutta jos metsä on niinkuin
sanotte —
— No isompi se on kuin viime vuotinen lohko.
— Mutta Ikkola puhuu nyt turhia, sanoi toinen nuorempi,
hienomman näköinen isäntä, — koska minä koko Suonpään palstasta
en saanut kuin kolmekymmentä viisi —
— Lyödään veto; minä rohkenen. Viisi pulloa.
— Sampanjaako?
— Tietysti — ja sittepä nähdään kun herrat tulee katsomaan.
Juodaan se paikalla, kyllä minä otan, ja velon sitten Huhdasmäkeä,
kun olen voittanut.
— Ja minä maksan myös silloin kuin häviinkin, vastasi toinen
lujasti.
Tukkiherra nauroi makeasti, sytyttäen sikaria.
— Onpa hauska nähdä, kumpi herroista voittaa —
Joppi on katsellut miestä ja toista — totisena kovin — ja
kuunnellut melkein henkeään pidättäen. — Voi toisen kerran! —

Puhuukohan ne vaan totta — mutta löisikö ne pilalla noin isoa vetoa
— —.
Kyyppäri toi sampanjaa.
— Tämä Ikkola se on erinomainen tekemään hyviä kauppoja, alkoi
kolmas, pyylevä isäntä puhua, kun oli ensimmäiset lasit kallistettu. —
Tänne tullessa ennätti Tampereella ostamaan polkuhinnasta oriin,
josta minä olisin maksanut kaksituhatta empimättä.
— Ihanko totta? kysäsi joku.
— Totta vallan. Pohjanmaan parhaita juoksijoita, ja toinen
petteessä möi.
— Sillä minä jään kilpa-ajoihin ja otan ensi palkinnon, sanoi
jäykkänä Ikkola, pyyhäisten leukansa poimuja, joihin oli äsken
suupielestä valahtanut sampanjan vaahtoa —
— Pitäisipä se nähdä, sanoi tukkiherra innostuneena, minua niin
huvittaa hyvät hevoset.
— Saapi sitä katsoa, — täällä se on tallissa — vaan en minä nyt
viitsi tästä lähteä yösydännä.
Mutta toinen kiusasi itsepäisesti lähtemään ja näyttämään sitä.
— No kun tahtonet, niin haetetaan se tänne, sanoi Ikkola, jonka
päähän pisti hauska juoni. — Kyyppäri! huusi hän.
— Mitä pite olla herr Ikkolall'?
— Käske niitä renkiä tuomaan minun orittani tänne saliin, tuossa
on vaivastasi, — ja hän tarjosi takanaan seisovalle kyyppärille

viisimarkkaista olkansa ylitse.
Mies sai suuret silmät ja suu lensi auki.
— Herrijess! Ei se passaa, herr Ikkola.
— Mutta sen pitää passata. Minä tahdon tarjota pullon sampanjaa
sillekin, markkinain maistimiksi.
Joppi on kuin puusta pudonnut ja vähän säikähtänyt. — Ne on
miehiä, oih, oih! — ja hänen tekisi jo mieli työntää kyyppäri
hakemaan sitä oritta ja mennä itse auttamaan.
— Mutt' jos herr Ikkola tahto, niin me viädä sille stalliin — sotkee
kyyppäri.
Mutta Ikkola on nyt kerran saanut päähänsä, että hän näyttää
oriinsa mahtavasti, koska sitä nähdä tahdotaan, eikä anna perään,
ennenkuin saadaan ravintolan isännältä itseltään lupa tuoda hevonen
eteiseen saakka, mutta ei etemmäksi, maksosta eikä mistään.
Ori tuotiin eteiseen ja koko sali pakkautui katsomaan, kuinka sen
suuhun miesvoimalla kaadettiin sampanjaa. — Yleinen riemu,
raikuva nauru ja kättentaputus seurasi tätä näytelmää, ja Ikkola
käydä väänsi tyytyväisenä, kädet housuintaskussa, pöytään takaisin.
Tämän välinäytöksen päätyttyä veti Joppi Oskarin syrjään.
— Anna hyvä veli vielä parisataa, jos tässä hyvinkin tarvitsisi.
— Joo, se ei tule kysymykseen, mutta tehdään sentään pieni
paperi, sanoi Oskari, ja kääntyen salin ovelle, huusi hän: "Pylkkänen,
tuleppa tänne."

— Tuossahan on paperia tiskillä — tämä Hinkkala ottaa minulta
kaksi sataa — jaa, pannaan se entinen yhteen — neljäsataa
markkaa, jotka minä saan nostaa puulaakilta, hänelle tulevista
metsärahoista. Pistä sinä paperi.
— Pannaanko korkoa?
— Ei panna, tutulta mieheltä.
Juoma-ainetta tilattiin nyt joka pöytään vielä hyvät satsit, kun
tiettiin, ettei pitemmälle enään anneta näin myöhällä.
Ja kun Joppi näki, kuinka huolettomasti siinä tiliä tehtäessä
heiteltiin isoja setelirahoja pöytään, huomasi hän häpeäkseen
olevansa vielä aivan jälellä oikeista miehistä. — Mutta kyllä hänkin
vielä — tulevilla markkinoilla. — Mikä pakko hänen on nahjustella ja
lainailla pikkusummia — hän ottaa ja myö koko takapalstan —
päästä päähän, isot puut kaikki, mitä vaan puulaakille kelpaa — jää
hänelle sittenkin vielä yli oman tarpeen.
Kolme päivää he olivat Oskarin kanssa markkinoilla. Joppi vaihtoi
ja möi hevosta ja osti välillä, kunnes lopulta oli saanut mielestänsä
tarpeeksi hyvän ja virheettömän.
Kolmantena päivänä oli poliisi käynyt hakemassa häntä
raastupaan, todistamaan Myllymäen puukotusjutussa.
Lääkärin todistuksen mukaan ei Nikun haava ollut vaarallinen, sillä
puukko oli tunkeutunut pitkin pintaa. Mutta yhtäkaikki tuomittiin
Jussi vuodeksi kuritushuoneesen ja viinansakkoon, jonka hän ensin
sai suorittaa vesikopissa.

8.
On kulunut joitakuita vuosia siitä, kun Hinkkalan Joppi möi
ensimmäisen metsälohkonsa ja tuli huomaamaan, miten helpolla
siitä sai rahaa.
Sen jälestä on hän myönyt koko metsänsä, niin että tukkipuita ei
ole enään ollenkaan, eikä juuri hirsiäkään. Rahaa siitä on lähtenyt
koko joukon toistakymmentä tuhatta, mutta melkein samaa mukaa
se on kulunut kuin on tullutkin, eikä säästöä ole summiakaan enään.
Maat eivät ole entisestään parantuneet, pikemmin päinvastoin,
sillä lantaa on tehty laiskasti ja isäntä-vainaan ajoilta peltoihin jäänyt
voima on imetty loppuun. Mutta sen sijaan on talon ulkoasu,
rakennukset, muuttuneet ja komistuneet niin, että sitä tuskin
entiseksi tuntee. Ainoastaan muorin eläketupa pihan laidassa —
muori elää vielä — on ennallaan sammaltuneessa painokatossaan. Ja
sitäkin on Joppi jo usein hankkinut muuttamaan sivummalle, kun se
hänen mielestään rumentaa talon uljaan ulkonäön, vaan siihen se
sentään vielä on jäänyt paikoilleen seisomaan. Kattoa on muorin
omalla kustannuksellaan pitänyt jonkun kerran korjata, kun se on
alkanut vuotaa. Joppi ei ole tehnyt sitä, on vaan sanonut muorille,
että kuka sitä vanhaa rämää viitsii paikkailla, kun se kumminkin on
kerran siitä kokonaan muutettava pois.
Ja muutenkin oli Joppi joskus harmissaan ilmaissut
vastenmielisyytensä koko eläkkeenkin maksusta — kun on talossa
itsessäänkin menoa niin paljon — ja lausuillut, ett'ei kenenkään
hullun pidä ottaa taloa, jossa on eläke, joka saattaa hyvässä lykyssä
elää kauvemmin aikaa kuin itse talon pitäjä. — Niin hän oli monen
kuullen puhunut ja senvuoksi muorin mökki oli tainnut siihen

paikoilleen jäädäkin. Jos muori itse olisi sen siitä siirtänyt pois, olisi
Joppi tainnut olla vallan hyvillään.
Mutta asuinrakennuksen oli hän pannut herraskuntoon.
Vanhan salin päähän oli jatkettu kaksi kamaria, ja niiden alle,
kivijalkaan, tehty maitokamari. Koko rakennus oli päältä vuorattu ja
maalattu ja pantu asfalttikattoon, sillä Joppi oli sanonut sen
edullisemmaksi kuin pärekaton, nykyään kun puista maksettiin niin
hyvästi.
Ikkunat oli isonnettu, ja laitettu uusi komea eteinen, jossa seinät
olivat yhtenä lasina. Sisäpuolella oli kaikki seinät tapiseerattu, katot
ja lattiat maalattu, tehty kaakeliuunit ja hankittu kaupungista komeat
huonekalut, muhkeat ikkunan verhot ja lattiamatot, sekä salin
peräseinälle, kamarinovien väliin, iso peili, josta näki koko kuvansa,
kun astui sisään.
Sanalla sanoen, koko talo on muuttunut oikeaksi herrastaloksi,
joka ensimmäiseksi pistää silmään kylään tullessa, sillä toiset talot
ovat pysyneet jokseenkin ennallaan, yhtä harmaina. Ja Joppi niitä
aina pilkkaileekin ja sanoo olevansa ainoa mies Metsäkylässä, joka
osaa ihmisittäin elää, niinkuin muualla maailmassa, vaan että toiset
siellä kylässä vielä elävät vanhassa tyhmyydessään, sirkkain kanssa.
Eikä ole hän viitsinyt naapuriensa kanssa seurustellakaan enään,
mokomien tyhmien kollojen.
Toisen Hinkkalan kanssa on sattunut suoraa riitaakin. Se ukko on
Jopin mielestä aina niin turhan tarkka ja saivarteleva, riitelee
yhtenään raja-aidoista, yhteisistä viemäriojista ja sen semmoisista,
ett'ei ehdi kuin niitä korjailemaan ja kaivamaan. Ja poikansa opettaa
samallaisiksi — purisivat pennin halki, jos hampaat pystyisi. —

Häviisivät edes siitä talostaan! Ja eiköhän kohta pudonnekin katto
sisään, niin jääväthän loukkuun. Vaan ei ne juuttaat taida hävitä,
sillä on niillä rahoja, vähän ainakin, vaikka eivät ymmärrä niitä
käyttää. Missä homehtunevat kirstujen pohjilla. Paljoko mahtoivat
viime vuonna saada, kun jonkuisen puun möivät — kappaleluvulla.
— Ei niitä paljoa tainnut rätkähtää, kosk'eivät ole kehdanneet edes
kellekään sanoa — —
Sillä tavalla mietti Joppi naapurin väestä — ja hän suoraan
halveksi sellaisia nahjustajia.
Itse on Joppi näiden muutamain vuosien kuluessa myös
muuttunut paljon. Hän oli jo nuorena lihavanläntä ja punaverinen,
mutta nyt hän on lihonut vielä enemmän, niin että tuskin kävellä
jaksaa. Kasvot ovat turponeet ja kohistuneet, niin ett'ei ajoittaisin
juuri silmiä näy, ja saaneet tummanpunaisen värin, varsinkin nenän
seutu. Ääni on tullut entistään käheämmäksi, niin että puhe käy
raskaasti, ja yhtenään ryittää. Joppi sanoo sen rykimisen nousevan
vatsasta, kun hänelläkin on vatsakatarri — joka tauti useammin
näyttää seuraavan herrasmiehiä — ja jo oli hän eräänä kesänä
hankkinut kylpymatkallekin, vaan se jäi muiden hommain tähden
sikseen. Kaikkeen tähän tulee vielä lisäksi se muutos, että Joppi on
nainut mies, ja on ollut jo neljättä vuotta.
Hän oli kaupunkimatkoillaan poikkeillut niin usein Sillanpäähän,
että siitä lopulta oli syntynyt naimiskaupat Alman kanssa. Ja
mahtavan emännän hän oli saanutkin, sillä Alma oli tottunut
herrastamaan ja rahoilla elämään hänkin, sitten kun oli tullut täysi-
ikäiseksi ja saanut perintörahansa, viisituhatta, holhojaltaan omaan
huostaansa.

Tultuaan Hinkkalaan, oli hän heti näyttänyt kyläläisille, mistä hän
käy. Sillä vaikka Joppi jo ennen häitä oli tehnyt suuria korjauksia
talossa, vaati emäntä heti niitä vielä parantamaan. Oli pitänyt paljon
purkaa entistä ja rakentaa uutta sijaan emännän ma'un mukaan, sillä
hän oli sanonut, että "miks'ei meillä niinkuin muillakin paremmilla
ihmisillä, ei tämä tällaisena meille kelpaa".
Sitä isoisuutta oli jotkut vanhat, yksinkertaisemmat ihmiset
nauraneet ja tuumineet Jopin naimisesta, että "vakka kantensa
valitsee, vakan kansi kantimensa", mutta kun nuoren pariskunnan
korville sattui sellaiset takapuheet, olivat he yhdestä suusta
sanoneet, että se lähtee kateudesta sellainen puhe, sillä heidän
kannattaa elää ihmisiksi. Ja kannattihan heidän, sitä ei kukaan
kieltänyt, sillä rahaa oli lähtenyt metsästä, minkä vaan kääntää ehti.
Ja siltä se näytti, että heidän oli kannattanut panna rahaa moneen
muuhunkin paikkaan, ei vaan rakennuksiin ja komeihin
huonekaluihin, sillä palkollisia oli lisätty, hevosia ja karjaa oli lisätty
— ja niille kannatti ostaa rehua, kun oma ei milloinkaan riittänyt.
Joppi, kun ei "ennättänyt, eikä terveytensä vuoksi oikein
jaksanutkaan" — kuten hän sanoi — talonsa töitä johtaa, oli ottanut
voudin, jolle piti maksaa hyvä palkka. "Pehtuoriksi" sitä Joppi itse
aina kutsui, niinkuin kartanopaikoissakin oli kuullut, vaan ei se nimi
päässyt yleisempään käytäntöön. Se oli muuten miehekseen nuori,
iloinen velikulta, joka ei suinkaan antanut Jopin itseään nenästä
johtaa. Hän kyyditteli Joppia tämän juomamatkoilla ja pelasivat he
usein yhdessä Jopin humalassa ollessa korttia — jota vanhaa
tapaansa Joppi ei ollut suinkaan unohtanut. Silloin saivat rengit
tehdä töitä parhaan tahtonsa ja ymmärryksensä mukaan, ja kun
sattui käymään huonosti, syytti "pehtuori" jumalan ilmoja, huonoa

siementä, ja mitä milloinkin. Ja Joppi tuumasi, että "minkä sille voi,
hyvä vaan, ett'ei kohdannut köyhää".
Paitse tavallista talossa tarvittavaa hevoslukua, ruokki Joppi kesän
talven tallissa kahta juoksijaa, joilla oli monet kerrat koettanut kilpa-
ajoissakin palkintoa saada, kumminkaan onnistumatta vielä. — Mutta
onnistuu se kerran, oli hän väittänyt, kun oli juoksijatammastaan
saanut varsan, jota jo koulattiin. Ja ottaa hän vielä kruunun
palkinnon. Hän ei hellitä, ennenkuin näyttää palkintoherroille, kuinka
ajetaan, ja naurakoot sitten vielä, niinkuin nauroivat silloin, kun hän
tapaturmassa ajoi radalla kumoon. — Ja varsa oli ankara menemään.
Se oli jo nähty Sääksijärvellä, kirkonkylän alla, mihin pitäjään
hevosmiehet olivat teettäneet radan, ja missä sitä oli käyty voudin
kanssa opettamassa. Sitä oli jäällä ja keskievarin pihalla jo monet
kerrat ylistetty, sen onneksi oli kievarin kamarissa monta yökautta
juoda humuutettu, ja pantu sen nimeksi "Punssi".
Lehmien luku oli navetassa emännän tultua lisätty kymmenestä
viiteentoista, sillä emäntä oli paikalla taloon tultuaan kysynyt
montako on lehmää, ja kun oli vastattu, että kymmenen on, oli hän
tahtonut tietoa montako naapurissa on. Ja kun oli kuullut siellä
olevan saman määrän, oli hän sanonut, niskaansa heittäen, että "se
ei sovi, meillä pitää olla viisitoista, koska tuollaisessakin talon
rähjässä on kymmenen". Ja niin oli hankittu viisi lehmää lisää. Se
luku pysyi sitten muuttumatta, kun ei oltu lisätty naapurissakaan.
Nälkää se karjajoukko kyllä usein näki, mutta upealta kuului
sellainen kellojen solina keväällä, laitumelle ajettaissa, kun melkein
jokainen oli kellokas — niinkuin paikoin muuallakin on tapana — ja
ainahan ne jaksoivat sen verran astuskella, että kello kalahteli.
Myytävää siitä ei lähtenyt särpimen puolta, mutta tuli omiksi

tarpeiksi sentään — paitse talvella joskus, jolloin piti ostaa voita. Ja
jos olisi jotain ylijäämää tullutkin, niin myömättä se olisi jäänyt, sillä
Alma oli nauranut naapurinkin emäntää, että tämä viitsi käydä
kaupalla muutamaa voinaulaa. "Sen vertaisista lehmistään viitsivät
myödä särvintä ja näkevät itse nälkää", oli hän sanonut. Ja vaikka ei
maitoa talossa summiakaan ollut, oli Alma sentään tahtonut
laittamaan erityisen maitokamarin.

Kun Riuttan maisteri oli rakennuttanut asuinpytingin alle korkeaan
kivijalkaan meijerihuoneen, oli Hinkkalassakin tehty samalla tavalla,
ja kun ne uudet kamarit rakennuksen päässä olivat joutuneet varsin
vieruun paikkaan, tehtiin niiden alle kivijalkaan huone, pantiin siihen
ovi päädystä, ja ikkuna, varustettiin se hyllyillä ja kutsuttiin sitten
aina sitä meijeriksi. Maitoa ei siellä juuri käytetty, mutta Joppi ei
voinut sitä kyllin kiittää oluen ja muiden juomatavarain
säilytyspaikkana. Ja monta helteistä kesäpäivää oli hän siellä istunut
jonkun ystävänsä, tahi "pehtuorinsa" kanssa tupakoiden ja juoden
olutta, vaikka nykyajan meijerien ankarat puhtauden säännöt eivät
oikein taitaisi sellaista menettelyä hyväksyä.
Sen saman meijerin tähden oli Jopille sattunut ensimmäinen
ankarampi riita emäntänsä kanssa, ja kun he kerran hyvään alkuun
pääsivät, jatkui sitä sitten melkein yhtenään. Alma oli sattunut
lähtemään kylään, johonkin etemmäs, ja oli pistänyt
meijerinavaimen taskuunsa. Joppi sai sillä aikaa vieraita, eikä
auttanut muu kuin särkeä lasi.
Siitä oli tullut sellainen tora, että Joppi teetti itselleen oman
avaimen — ja jonkun aikaa kävi kaikki hyvin. Mutta sitten oli tullut
uusi riita, kun Joppi oli särkenyt kerran Alman viinipullon sinne
meijerin lattialle, ja silloin oli Alma teettänyt kokonaan uuden lukon,
hyvin juonikkaalla avaimella. Vaan selvittiin siitäkin sentään ja
lauhduttiin taas joksikuksi aikaa.
Tämä seikka olkoon sentähden ohimennen mainittu, että se oli
ollut ensimmäinen kerta, kun he olivat kumpikin koko kiivaan
luontonsa päävoimalla hyökänneet toistensa kimppuun, ja se oli
tapahtunut heti toisena vuonna heidän häittensä jälkeen. Ja alusta
pitäen olikin kyllä voinut huomata, että heidän välilleen kerran oli

puhkeava täysi sota, kun vaan sattuu joku tarpeeksi pätevä syy sen
julistamiseen, sillä he olivat kärtyisiä kumpikin ja olivat jo useasti
kiistelleet kovasti keskenään, kun oli mielipiteet jossain asiassa
sattuneet olemaan eriäviä.
Ja oli seikkoja, joissa ne aina olivat eriäviä. Sillä vaikka he
luonteeltaan molemmat olivat komeutta rakastavia ja yksimielisiä
siitä, mitä talon ja heidän itsensä isoiseen ulkoasuun tuli, niin oli
kummallakin sentään toistaan kohtaan sellainen mieli, että "kun sinä
tuhlaat yksin kaikki, niin minulle ei jää mitään". Se oli jonkunlaista
kateutta sen yhteisen komeuden alla, ja siitä riitti riitaa, kun kerran
oli alkuun päästy.
Siihen sattui sitten vielä asioita, jotka antoivat emännälle aihetta
yhä kiivaampiin kahakoihin Jopin kanssa, mutta myöskin sellaiset
aseet käsiin, että hän niillä aina Jopin nolasi ja saattoi raivoon. Ja
kun Joppi oli joutunut häviölle, läksi hän vihapäissään ajelemaan
kylille ja juomaan retuamaan, jolloin Alma tavallisesti ajoi toisella
hevosella jälestä ottamaan selkoa mihin se menee — ja tärväämään
rahaa hänkin puolestaan, siinä tuumassa, että "kun sinä kerran, niin
miks'en minäkin" — niinkuin toisensa kiusalla kumpikin.
Oli kielitty Almalle, että siellä jossain mökissä vanhempainsa
kanssa asui Iita niminen tyttö, joka Jopin nuorena miehenä ollessa
oli ollut Hinkkalassa piikana, ja josta aikoinaan oli käynyt sellainen
huhu, että Joppi oli aikonut hänet naida. — Taisi olla tuulesta
temmattu se naimajuttu, mutta se oli varma asia, että yksinään ei
Iita ollut Hinkkalasta lähtenyt, kun Joppi hänet kesken vuotta oli
ajanut talostaan pois, ja kun hänen häpeänalaisena oli sitten pitänyt
turvautua köyhäin vanhempainsa luokse asumaan, oli hän kohta sen
jälkeen saanut pojan ja nauttinut vaivaishoidon apua. Kunnanesimies

oli kyllä patistanut Iitaa käymään Hinkkalan Jopin kimppuun, mutta
se oli jäänyt tekemättä, kun ei ollut todistajia, ja kun tyttö kerran
maantiellä oli yhdyttänyt Jopin, ja pyytänyt pojalleen jotain elannon
apua, oli Joppi ollut potkaista hänet ojaan.
Tästä Iitan jutusta oli emäntä saanut sellaisen valttiässän, joka
aina sopi lyödä pöytään riidassa Jopin kanssa, ja oli hän sitä yhä
kulumattomana käyttänyt jo kolme vuotta, sillä neljäs syksy on jo
käsissä heidän naimisestaan.
* * * * *
On sateinen syyskuun päivä. Joppi on tullut kyliltä kotiin emännän
mielestä taas liiaksi humalassa ja hukannut rahaa, jota häneltä usein
"hukkui". Siitä on emäntä ruvennut löylyyttämään häntä vanhaan
nuottiin. Joppi taas on suuttunut siitä, että Alma muka piti
tarpeettoman kauvan neulojanaista töissään, tärväten hänen
kanssaan niin paljon kahvia, ett'ei ennättänyt kuin tuomaan leiviskän
toisensa perään. Molemmin puolin väitellään ensin, sitten
kiivastutaan yhä enemmän, syydetään haukkumasanoja toistensa
silmille, ja lopulta potkaisee Alma oven auki ja huutaa Jopille:
— Mene ulos, sika — mene sinne huorasi tykö, sieltä kai sitä
nytkin tulit kotiin tappelemaan. Teetä sille leninkiä — teetä silkistä —
kyllä minä tulen omillani toimeen ja teen omani kanssa mitä tahdon.
— Ett'ei jo leukasikin väsy tuota samaa asiaa tuhansiin kertoihin
jankuttamasta, sinä hassu vaivainen, sanoo vastaan Joppi. — Enkö
ole sinulle sanonut, ett'ei minulla ole mitään tekemistä kenenkään
naisen kanssa — ja nielköön tulimmaisin helvetti koko hamesakin —
— Etkä sinä väsy minulle valehtelemasta. Ulos, roikko!

— Ne ne valehtelevat, jotka niitä kielii —
— Ei säälliset ihmiset valehtele.
— Valehtelevatpa, jumal'avita, tiuskaa Joppi, käy tuolin
selkälautaan kiinni, ja paiskaa lattiaan niin että jalka kätkee tuolista.
— Herra jumala, sitä hullua, kirkaisee Alma. — Tule Fiina sieltä
kamarista joutuin hätiin, se särkee kaikki minun kauniit mööpelini,
jotka itse olen ostanut.
— Nimitäpä yksikin säällinen ihminen, joka sen on sanonut, että
minulla on leipäsusia, nimitä kerrankin yksi, hihkuu Joppi.
— Kyllä sinä pahat tekosi itsekin tiedät —
— No kun tiedän niin tiedän, se on hyvä se, sanoo Joppi ja rientää
niin tulisesti kaapin luo, ettei Alma ehdi edes kaappaamaan avainta
hänen kädestään, ja tempaa sieltä kaapin laatikosta tukon rahaa
taskuunsa, viimeistä metsärahaa.
— Hevonen puihin, pehtoori, karjasee hän porstuassa.
Alma riitelee vastaan ja "pehtoori" on vähän epätietoinen, mutta
Jopin komento on kova ja hänen kasvonsa mustuvat.
— "Punssi" valjaihin, huutaa hän toistain, kun näkee miehen
epäröivän, ja tämän täytyy valjastaa sama hevonen, jonka juuri
märkänä on päästänyt aseista. — Mutta onhan isäntää toteltava,
ajattelee hän, muistaahan se niin usein palkitakin.
Joppi lyö oritta selkään, ja lähtee hihkaisten pihasta sellaisella
voimalla, että säikäleet lentää portintolpasta ja rattaat kulkevat

pitkän matkaa toisella pyörällä vaan. Hän heitäksen koko painollaan
yläpuolelle istuinta ja on vähällä pudota, kun pyörät taas ottavat
äkkiä entisen tasapainonsa; ja kujasta kuuluu hänen kulotuksensa,
että "isolla vaan sanoi Rannanjärvi, kun ajeli vieterillä — — —"
Mutta Alma riensi hengästyneenä sisään, alkoi kiireellä pukea
päällensä ja käski voudin valjastaa toista hevosta, sillä hänen täytyy
paikalla lähteä ottamaan selkoa, mihin se riiviö taas niin kovin
humalaisena menee — hänen täytyy lähteä. — Pysyisi kotona, vaikka
sitten joisikin kuoliaaksi kerran itsensä, mutta kun aina pitää olla se
kova kiihko kylään rahoja tuhlaamaan, torui hän vielä lähtiessään
Fiinalle, neulojalle, joka melkein yhtenään oli Hoikkalassa emännän
töissä. — Katso sinä nyt lapsia, lisäsi hän ovessa.
Vouti hymähteli ja katseli emännän jälkeen, mutta kun hän tuli
porstuaan ja oli kääntymässä tuvan puolelle, kuiskasi Fiina salinoven
raosta, että "tule tänne, täällä on vielä kahvi lämpimänä — —"
Niin, Hinkkalaisilla oli lapsia myös. Sillä vaikka he olivatkin olleet
riitaisia, vallitsi heidän välillään sentään niin paljon sopua, että perhe
oli lisääntynyt vahvasti ja oli yhä lisääntymässä. Viime vuonna se oli
lisääntynyt kahdella yht'aikaa, leveäposkisia poikia molemmat,
"oikeita papan poikia", oli Jopin tapa sanoa hyvällä päällä ollessaan
joskus. Ne oleilivat vielä isoimman osan aikaansa kehdossa piian
hoidettavina, mutta kaksi vanhempaa, tyttö ja poika, teki jo
hävitysretkiä pitkin huoneita, jättäen selviä merkkiä verkaisiin
tuolinpäällystöihin, seinäpapereihin ja muihin arempiin kohtiin.
Emäntä tuli kotiin ensimmäisenä illalla. Saatuaan selvän, mihin
Joppi oli ajanut, oli hän pyörtänyt ympäri. Tuttavissaan hän sentään
oli käynyt, parilla kauppiaalla, värjärillä ja jahtivoudin matamia
katsomassa, tutkaissut kylän kuulumiset ja tehnyt selvän

Metsänkylän asioista, ostellut kangasta, kahvia, vehnäistä ja suuren
pussin karamellia lapsille, sekä muuta talouden tarvetta, ja sitten
lähtenyt vasta ajamaan kotiin.
Hän oli tiellä vakavasti päättänyt voittaa voudin puolelleen
kahakoissaan Joppia vastaan, sillä siitä voi olla apua. — Mitä sen
nytkään olisi tarvinnut hevosta valjastaa, eikä se raukka
kumminkaan itse olisi sitä kyennyt tekemään, ajatteli hän, ja kutsui
heti kotiin päästyään voudin yhteen seuraan saliin. Hän tarjosi sille
konjakkiakin teen sekaan, ja toinen oli ylen myöntyväinen emännän
esityksiin ja lupasi hyvää asiaa auttaa, mikä vaan hänen vallassaan
olisi. Juteltiin hyvillä mielin kaikki, Fiina kolmantena, ja emäntä puhui
kuulumisia kirkolta päin, lipeästi, sujuvasti, höystäen tapahtumia
aina vähän oman kuvituksensa mukaisiksi.
Lapsille oli annettu koko karamellipussi omaan haltuunsa
yht'aikaa. Ne olivat sen hajoittaneet salin lattialle ja leikkivät siinä
suutaan massutellen ja heitellen toisiaan pyöreillä, tahmeilla
makeispalluroilla, kunnes lopulta tuli riita, kummanko isompi puoli
piti oleman.
— Älä tina tika vie minun taramelliäni, sanoi toinen ja kävi kynsin
toisen käsiin kiinni, niin että veri lähti.
— Ai ai ai, toi Väinö on temmonen roitto, varattaa minulta ja repii,
parkui toinen.
Äiti eroitti riitelijät, antaen tukkapöllyä, mutta samassa hän nauroi
ja sanoi:
— Oi voi niitä lapsia! Mikä pitää kaikkia suuhunkin tuoman. On
niistäkin sentään jotain hauskuutta!

Welcome to our website – the ideal destination for book lovers and
knowledge seekers. With a mission to inspire endlessly, we offer a
vast collection of books, ranging from classic literary works to
specialized publications, self-development books, and children's
literature. Each book is a new journey of discovery, expanding
knowledge and enriching the soul of the reade
Our website is not just a platform for buying books, but a bridge
connecting readers to the timeless values of culture and wisdom. With
an elegant, user-friendly interface and an intelligent search system,
we are committed to providing a quick and convenient shopping
experience. Additionally, our special promotions and home delivery
services ensure that you save time and fully enjoy the joy of reading.
Let us accompany you on the journey of exploring knowledge and
personal growth!
ebookultra.com