Ergodic Theoretic Methods in Group Homology A Minicourse On L2 Betti Numbers in Group Theory Clara Loh Online Ebook Texxtbook Full Chapter PDF

Ergodic Theoretic Methods in Group
Homology A Minicourse on L2 Betti

Numbers in Group Theory Clara Löh
Visit to download the full and correct content document:
https://ebookmeta.com/product/ergodic-theoretic-methods-in-group-homology-a-minic
ourse-on-l2-betti-numbers-in-group-theory-clara-loh/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...
A First Course in Group Theory 1st Edition Bijan Davvaz
https://ebookmeta.com/product/a-first-course-in-group-theory-1st-
edition-bijan-davvaz/
Group A Streptococcus Methods and Protocols Thomas

Proft
https://ebookmeta.com/product/group-a-streptococcus-methods-and-
protocols-thomas-proft/
Symmetry and Group Theory in Chemistry Second Edition

S. K. Dogra
https://ebookmeta.com/product/symmetry-and-group-theory-in-
chemistry-second-edition-s-k-dogra/
On Group Analysis and Beyond Group Analysis as Meta

Theory Clinical Social Practice and Art 1st Edition
Anastassios Koukis
https://ebookmeta.com/product/on-group-analysis-and-beyond-group-
analysis-as-meta-theory-clinical-social-practice-and-art-1st-
edition-anastassios-koukis/
Group Theory 1st Edition Dinesh Khattar
https://ebookmeta.com/product/group-theory-1st-edition-dinesh-
khattar/
Group Theory 1st Edition Dinesh Khattar
https://ebookmeta.com/product/group-theory-1st-edition-dinesh-
khattar-2/
Observability A New Theory Based on the Group of

Invariance 1st Edition Agostino Martinelli
https://ebookmeta.com/product/observability-a-new-theory-based-
on-the-group-of-invariance-1st-edition-agostino-martinelli/
Drugs in Sport Taylor & Francis Group
https://ebookmeta.com/product/drugs-in-sport-taylor-francis-
group/
Group Therapy A group analytic approach 1st Edition

Nick Barwick
https://ebookmeta.com/product/group-therapy-a-group-analytic-
approach-1st-edition-nick-barwick/
SPRINGER BRIEFS IN MATHEMATICS
Clara Löh
Ergodic Theoretic
Methods in Group
Homology
A Minicourse on
2
L -Betti Numbers
in Group Theory
SpringerBriefs in Mathematics
Series Editors
Nicola Bellomo, Torino, Italy
Michele Benzi, Pisa, Italy
Palle Jorgensen, Iowa, USA
Tatsien Li, Shanghai, China
Roderick Melnik, Waterloo, Canada
Otmar Scherzer, Linz, Austria
Benjamin Steinberg, New York, USA
Lothar Reichel, Kent, USA
Yuri Tschinkel, New York, USA
George Yin, Detroit, USA
Ping Zhang, Kalamazoo, USA
SpringerBriefs in Mathematics showcases expositions in all areas of mathematics
and applied mathematics. Manuscripts presenting new results or a single new result
in a classical field, new field, or an emerging topic, applications, or bridges between
new results and already published works, are encouraged. The series is intended for
mathematicians and applied mathematicians.
Titles from this series are indexed by Web of Science, Mathematical Reviews, and
zbMATH.
More information about this series at http://www.springer.com/series/10030

Clara Löh
Ergodic Theoretic Methods

in Group Homology
A Minicourse on L2-Betti Numbers in Group
Theory
123
Clara Löh
Fakultät für Mathematik
Universität Regensburg
Regensburg, Bayern, Germany
ISSN 2191-8198 ISSN 2191-8201 (electronic)

SpringerBriefs in Mathematics
ISBN 978-3-030-44219-4 ISBN 978-3-030-44220-0 (eBook)
https://doi.org/10.1007/978-3-030-44220-0
Mathematics Subject Classification (2020): 55N25, 20J05, 37A20, 57N65, 22E40, 20P05, 46L10
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2020

This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether
the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of
illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and
transmission or information storage and retrieval, electronic adaptation, computer software, or by similar
or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are exempt from
the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, expressed or implied, with respect to the material contained
herein or for any errors or omissions that may have been made. The publisher remains neutral with regard
to jurisdictional claims in published maps and institutional affiliations.
This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
für A ∗ A
Contents
0 Introduction 1
1 The von Neumann dimension 5

1.1 From the group ring to the group von Neumann algebra 6
1.1.1 The group ring 6
1.1.2 Hilbert modules 8
1.1.3 The group von Neumann algebra 9
1.2 The von Neumann dimension 11
1.E Exercises 15
2 L2 -Betti numbers 17
2
2.1 An elementary definition of L -Betti numbers 18
2.1.1 Finite type 18
2.1.2 L2 -Betti numbers of spaces 19
2.1.3 L2 -Betti numbers of groups 20
2.2 Basic computations 20
2.2.1 Basic properties 21
2.2.2 First examples 23
2.3 Variations and extensions 24
2.E Exercises 26
3 The residually finite view: Approximation 27

3.1 The approximation theorem 28
3.2 Proof of the approximation theorem 28
3.2.1 Reduction to kernels of self-adjoint operators 29
3.2.2 Reformulation via spectral measures 29
3.2.3 Weak convergence of spectral measures 30
3.2.4 Convergence at 0 31
vii
viii Contents
3.3 Homological gradient invariants 32

3.3.1 Betti number gradients 33
3.3.2 Rank gradient 33
3.3.3 More gradients 34
3.E Exercises 35
4 The dynamical view: Measured group theory 37

4.1 Measured group theory 38
4.1.1 Standard actions 38
4.1.2 Measure/orbit equivalence 39
4.2 L2 -Betti numbers of equivalence relations 40
4.2.1 Measured equivalence relations 41
4.2.2 L2 -Betti numbers of equivalence relations 41
4.2.3 Comparison with L2 -Betti numbers of groups 43
4.2.4 Applications to orbit equivalence 46
4.2.5 Applications to L2 -Betti numbers of groups 46
4.3 Cost of groups 47
4.3.1 Rank gradients via cost 48
4.3.2 The cost estimate for the first L2 -Betti number 52
4.3.3 Fixed price 55
4.E Exercises 57
5 Invariant random subgroups 59

5.1 Generalised approximation for lattices 60
5.1.1 Statement of the approximation theorem 60
5.1.2 Terminology 61
5.2 Two instructive examples 62
5.2.1 Lattices in SL(n, R) 62
5.2.2 Why doesn’t it work in rank 1 ?! 63
5.3 Convergence via invariant random subgroups 64
5.3.1 Invariant random subgroups 64
5.3.2 Benjamini–Schramm convergence 66
5.3.3 The accumulation point 68
5.3.4 Reduction to Plancherel measures 69
5.3.5 Convergence of Plancherel measures 69
5.E Exercises 71
6 Simplicial volume 73
6.1 Simplicial volume 74
6.2 The residually finite view 75
6.3 The dynamical view 77
6.4 Basic proof techniques 79
6.4.1 The role of the profinite completion 79
6.4.2 Betti number estimates 81
6.4.3 The rank gradient/cost estimate 83
6.4.4 Amenable fundamental group 84
6.4.5 Hyperbolic 3-manifolds 87
6.4.6 Aspherical 3-manifolds 90
6.E Exercises 92
Contents ix
A Quick reference 93
A.1 Von Neumann algebras 94
A.2 Weak convergence of measures 94
A.3 Lattices 95
Bibliography 97
Symbols 109
Index 111
0
Introduction
This is an extended version of the lecture notes for a five lecture mini-
course at the MSRI summer graduate school Random and arithmetic struc-
tures in topology (organised by Alexander Furman and Tsachik Gelander) in
June 2019.
This minicourse gives a brief introduction to ergodic theoretic methods in
group homology. By now, this is a vast subject [62]. In the present course,
we will focus on L2 -Betti numbers.
The underlying fundamental observation is that taking suitable coefficients
for group (co)homology allows us to connect homological invariants with
ergodic theory; good candidates are coefficients related to non-commutative
measure theory, i.e., to von Neumann algebras and dynamical systems. This
interaction works in both directions:
• On the one hand, (co)homology with coefficients based on dynamical

systems leads to orbit/measure equivalence invariants.
• On the other hand, many homological gradient invariants can be ac-

cessed through the dynamical system given by the profinite completion.
Overview of this minicourse
• Chapter 1. We will start our gentle introduction to L2 -Betti numbers by

introducing the von Neumann dimension. In order to keep the technical
overhead at a minimum, we will work with an elementary approach to
von Neumann dimensions and L2 -Betti numbers.
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 1

C. Löh, Ergodic Theoretic Methods in Group Homology,
SpringerBriefs in Mathematics, https://doi.org/10.1007/978-3-030-44220-0_1
2 Introduction
• Chapter 2. The von Neumann dimension allows us to define L2 -Betti

numbers. We will explore basic computational tools and calculate L2 -
Betti numbers in simple examples.
• Chapter 3. On the one hand, L2 -Betti numbers are related to classical

Betti numbers through approximation. This residually finite view has
applications to homological gradient invariants.
• Chapter 4. On the other hand, L2 -Betti numbers are also related to
measured group theory. We will compare this dynamical view with the
residually finite view.
• Chapter 5. Via invariant random subgroups, ergodic theory also gives a

new way of obtaining approximation results for normalised Betti num-
bers of lattices/locally symmetric spaces.
• Chapter 6. Finally, we will use the dynamical approach to L2 -Betti

numbers and related invariants as a blueprint to prove approximation
results for simplicial volume.
None of the material in this book is original; in particular, Chapters 1–3

are covered thoroughly in textbooks on L2 -Betti invariants [105, 84] and large
parts of Chapter 4 can be found in surveys on measured group theory [65, 62].
Prerequisites. I tried to keep things as elementary as reasonably possible;
this means that a basic background in algebraic topology (fundamental
group, covering theory, (co)homology, classical group (co)homology), func-
tional analysis (bounded operators, measure theory), and elementary homo-
logical algebra should be sufficient to follow most of the topics of the book.
Chapter 5 requires some knowledge on lattices/locally symmetric spaces and
Chapter 6 requires moderate familiarity with the topology and geometry of
manifolds.
Some basic notions on von Neumann algebras, weak convergence of mea-
sures, and lattices are collected in Appendix A.
Exercises. Each chapter ends with a small selection of exercises; moreover,
small on-the-fly exercises are marked with “(check!)” in the text.
Conventions. The set N of natural numbers contains 0. All rings are unital
and associative (but very often not commutative). By default, modules are
left modules (if not specified otherwise).
Additional material
This brief introduction is in no way exhaustive, but covers only a few selected
topics. I hope that these notes inspire the reader to get involved with more
systematic treatments and the original research literature.
Introduction 3
Textbooks on L2 -invariants
• H. Kammeyer. Introduction to 2 -invariants, Springer Lecture Notes in
Mathematics, 2247, 2019.
• W. Lück. L2 -Invariants: Theory and Applications to Geometry and
K-Theory, Ergebnisse der Mathematik und ihrer Grenzgebiete, 44,
Springer, 2002.
Recommended further reading
• M. Abért, N. Bergeron, I. Biringer, T. Gelander, N. Nikolov, J. Raim-
bault, I. Samet. On the growth of L2 -invariants for sequences of lattices
in Lie groups, Ann. of Math. (2), 185(3), pp. 711–790, 2017.
• M. Abért, N. Nikolov. Rank gradient, cost of groups and the rank versus
Heegaard genus problem, J. Eur. Math. Soc., 14, 1657–1677, 2012.
• M. W. Davis. The Geometry and Topology of Coxeter Groups, Lon-
don Mathematical Society Monographs, 32, Princeton University Press,
2008.
• B. Eckmann. Introduction to l2 -methods in topology: reduced l2 -
homology, harmonic chains, l2 -Betti numbers, notes prepared by Guido
Mislin, Israel J. Math., 117, pp. 183–219, 2000.
• A. Furman. A survey of measured group theory. In Geometry, Rigidity,
and Group Actions (B. Farb, D. Fisher, eds.), 296–347, The University
of Chicago Press, 2011.
• D. Gaboriau. Coût des relations d’équivalence et des groupes, Invent.
Math., 139(1), 41–98, 2000.
• D. Gaboriau. Invariants 2 de relations d’équivalence et de groupes,
Inst. Hautes Études Sci. Publ. Math., 95, 93–150, 2002.
• T. Gelander. A view on invariant random subgroups and lattices, Pro-
ceedings of the International Congress of Mathematicians. Volume II,
pp. 1321–1344, World Sci. Publ., 2018.
• A. S. Kechris, B. D. Miller. Topics in Orbit Equivalence, Springer Lec-
ture Notes in Mathematics, 1852, 2004.
• D. Kerr, H. Li. Ergodic theory. Independence and dichotomies, Springer
Monographs in Mathematics, Springer, 2016.
• W. Lück. Approximating L2 -invariants by their finite-dimensional ana-
logues, Geom. Funct. Anal., 4(4), pp. 455–481, 1994.
• J. Raimbault. Blog, https://perso.math.univ-toulouse.fr/jraimbau/
• R. Sauer. Amenable covers, volume and L2 -Betti numbers of aspherical
manifolds, J. Reine Angew. Math., 636, 47–92, 2009.
Videos. The videos of the original lectures within the MSRI summer grad-
uate school are available at https://www.msri.org/people/23160 . Moreover,
material for the other two minicourses of this summer school is available at
https://www.msri.org/summer schools/853 .
Errata. Comments and corrections for these notes can be submitted by email
to [email protected]; errata will be collected at
http://www.mathematik.uni-r.de/loeh/l2 book/errata.pdf .
4 Introduction
Acknowledgements. I would like to thank Alex Furman and Tsachik Ge-

lander for organising the summer school and inviting me as a lecturer. Special
thanks go to Gil Goffer, who skillfully taught the discussion sessions for this
course.
Moreover, I am grateful to Roman Sauer for sharing his notes on Chapter 3,
to Damien Gaboriau for valuable comments on Chapter 4, to Miklos Abért
for a lesson on IRS in the Schwarzwald, and to Marco Moraschini for patient
proof-reading.
This book project has been partially supported by the MSRI (through
funding of the summer school) and the SFB 1085 Higher Invariants (funded
by the DFG, Universität Regensburg).
Regensburg, February 2020 Clara Löh

1
The von Neumann dimension
Betti numbers are dimensions of (co)homology groups. In the presence of a

group action, we can alternatively also use an equivariant version of dimen-
sion; this leads to L2 -Betti numbers.
In this chapter, we will introduce such an equivariant version of dimension,
using the group von Neumann algebra. In Chapter 2, this dimension will allow
us to define L2 -Betti numbers of groups and spaces.
Overview of this chapter.
1.1 From the group ring to the group von Neumann algebra 6
1.2 The von Neumann dimension 11
1.E Exercises 15
Running example. the additive group Z, finite (index sub)groups

6 1. The von Neumann dimension
1.1 From the group ring

to the group von Neumann algebra
Examples of fundamental invariants in algebraic topology are Betti num-
bers of spaces and groups, which are numerical invariants, extracted from
(co)homology by taking dimensions of homology groups.
We will now pass to an equivariant setting: Let Γ X be a contin-
uous group action on a topological space X. Then the singular chain com-
plex C∗sing (X; C) and the singular homology H∗ (X; C) of X inherit a Γ-action,
and thus consist of modules over the group ring CΓ.
Unfortunately, the group ring CΓ, in general, does not admit an accessi-
ble module/dimension theory. We will therefore pass to completions of the
group ring: 2 Γ (for the modules) and the von Neumann algebra N Γ (for
the morphisms), which lead to an appropriate notion of traces and thus to a
Γ-dimension. We will now explain this in more detail.
1.1.1 The group ring

The group ring of a group Γ is an extension of the ring C with new units
coming from the group Γ:
Definition 1.1.1 (group ring). Let Γ be a group. The (complex) group ring
of Γ is the C-algebra CΓ (sometimes also denoted by C[Γ] to avoid misun-
derstandings)
• whose underlying C-vector space is g∈Γ C, freely generated by Γ (we

denote the basis element corresponding to g ∈ Γ simply by g),
• and whose multiplication is the C-bilinear extension of composition
in Γ, i.e.:
· : CΓ × CΓ −→ CΓ

ag · g, bg · g −→ ah · bh−1 ·g · g
g∈Γ g∈Γ g∈Γ h∈Γ
(where all sums are “finite”, i.e., all but finitely many coefficients are 0).
Example 1.1.2 (group rings).
• The group ring of “the” trivial group 1 is just C[1] ∼
=Ring C.
• The group ring C[Z] of the additive group Z is isomorphic to C[t, t−1 ],
the ring of Laurent polynomials over C (check!).
1.1. From the group ring to the group von Neumann algebra 7
• Let n ∈ N>0 . Then we have C[Z/n] ∼

=Ring C[t]/(tn − 1) (check!).
• In general, group rings are not commutative. In fact, a group ring CΓ

is commutative if and only if the group Γ is Abelian (check!). Hence,
for example, the group ring C[F2 ] of “the” free group F2 of rank 2 is
not commutative.
Caveat 1.1.3 (notation in group rings). When working with elements in group
rings, some care is required. For example, the term 4 · 2 in C[Z] might be
interpreted in the following different ways:
• the product of 4 times the ring unit and 2 times the ring unit, or
• 4 times the group element 2.
We will circumvent this issue in C[Z], by using the notation “t” for a gen-
erator of the additive group Z and viewing the infinite cyclic group Z as
multiplicative group. Using this convention, the first interpretation would be
written as 4 · 2 (which equals 8) and the second interpretation would be writ-
ten as 4 · t2 . Similarly, also in group rings over other groups, we will try to
avoid ambiguous notation.
Proposition 1.1.4 (group ring, universal property). Let Γ be a group. Then the
group ring CΓ, together with the canonical inclusion map i : Γ −→ CΓ (as
standard basis) has the following universal property: For every C-algebra R
and every group homomorphism f : Γ −→ R× , there exists a unique C-algebra
homomorphism Cf : CΓ −→ R with Cf ◦ i = f .
Γ
f
/ R× incl 6/ R
i
∃! Cf

CΓ
Proof. This is a straightforward calculation (check!).
Outlook 1.1.5 (Kaplansky conjecture). The ring structure of group rings is

not well understood in full generality. For example, the following versions of
the Kaplansky conjectures are still open: Let Γ be a torsion-free group.
• Then the group ring CΓ is a domain (?!).
• The group ring CΓ does not contain non-trivial idempotents (?!).

(I.e., if x ∈ CΓ with x2 = x, then x = 1 or x = 0).
However, a positive solution is known for many special cases of groups [38,
117, 44][105, Chapter 10] (such proofs often use input from functional analysis
or geometry) and no counterexamples are known.
1.1.2 Hilbert modules

Homology modules are quotient modules. In the presence of an inner prod-
uct, quotients of the form A/B (by closed subspaces B) can be viewed as
submodules of A (via orthogonal complements). Therefore, we will pass from
the group ring CΓ to the completion 2 Γ:
Definition 1.1.6 (2 Γ). Let Γ be a group. Then
· , · : CΓ × CΓ −→ C

ag · g, bg · g −→ a g · bg
g∈Γ g∈Γ g∈Γ
is an inner product on CΓ. The completion of CΓ with respect to this inner

product is denoted by 2 Γ (which is a complex Hilbert space). More con-
cretely, 2 Γ is the C-vector space of 2 -summable functions Γ −→ C with the
inner product
· , · : 2 Γ × 2 Γ −→ C

ag · g, bg · g −→ a g · bg .
g∈Γ g∈Γ g∈Γ
Example 1.1.7.
• If Γ is a finite group, then 2 Γ = CΓ.
• If Γ = Z = t | , then Fourier analysis shows that
F : 2 Γ −→ L2 [−π, π], C

1
an · tn −→ x → √ · an · ei·n·x
n∈Z
2π n∈Z
is an isomorphism of C-algebras (with inner product).

Remark 1.1.8 (countability and separability). In order to avoid technical com-
plications, in the following, we will always work with countable groups; then,
the associated 2 -space will be separable.
Definition 1.1.9 (Hilbert modules). Let Γ be a countable group.
• A Hilbert Γ-module is a complex Hilbert space V with a C-linear iso-
metric (left) Γ action such that there exists an n ∈ N and an isometric
Γ-embedding V −→ (2 Γ)n . Here, we view 2 Γ as a left CΓ-module via
Γ × 2 Γ −→ 2 Γ
(g, f ) −→ x → f (x · g) .

1.1. From the group ring to the group von Neumann algebra 9
• Let V and W be Hilbert Γ-modules. A morphism V −→ W of Hilbert

Γ-modules is a Γ-equivariant bounded C-linear map V −→ W .
In a complex Hilbert space, we have the following fundamental equality
for (closed) submodules A (check!):
C-dimension of A = trace of the orthogonal projection onto A.
For Hilbert Γ-modules, we will use this description of the dimension as a def-
inition. Therefore, we first need to be able to describe orthogonal projections
and we need a suitable notion of trace. Both goals can be achieved by means
of the group von Neumann algebra.
1.1.3 The group von Neumann algebra

Let Γ be a countable group and let a ∈ CΓ. Then the right multiplication
map Ma : 2 Γ −→ 2 Γ by a is a (left) Γ-equivariant isometric C-linear map.
Similarly, matrices A over CΓ induce morphisms MA between finitely gener-
ated free 2 Γ-modules.
However, morphisms of Hilbert Γ-modules, in general, will not be of this
simple form: We will need more general matrix coefficients.
Definition 1.1.10 (group von Neumann algebra). Let Γ be a countable group.
• Let B(2 Γ) be the C-algebra of bounded linear operators 2 Γ −→ 2 Γ.
• The group von Neumann algebra of Γ is the weak closure of CΓ (acting
by right multiplication on 2 Γ) in B(2 Γ).
Remark 1.1.11 (alternative descriptions of the group von Neumann algebra).
Let Γ be a countable group. Then the group von Neumann algebra N Γ is
a von Neumann algebra (Definition A.1.1) and thus can equivalently be de-
scribed as follows (Theorem A.1.2):
• N Γ is the strong closure of CΓ (acting by right multiplication on 2 Γ).
• N Γ is the bicommutant of CΓ (acting by right multiplication on 2 Γ).
• N Γ is the subalgebra of B(2 Γ) consisting of all bounded operators that
are left CΓ-equivariant.
Theorem 1.1.12 (von Neumann trace). Let Γ be a countable group and let
trΓ : N Γ −→ C
a −→ e, a(e) ,
where e ∈ CΓ ⊂ 2 Γ denotes the atomic function at e ∈ Γ. Then trΓ satisfies

the following properties:
1. Trace property. For all a, b ∈ N Γ, we have trΓ (a ◦ b) = trΓ (b ◦ a).
2. Faithfulness. For all a ∈ N Γ, we have trΓ (a∗ ◦ a) = 0 if and only

if a = 0. Here, a∗ denotes the adjoint operator of a.
3. Positivity. For all a ∈ N Γ with a ≥ 0, we have trΓ a ≥ 0. Here, a ≥ 0

if and only if x, a(x) ≥ 0 for all x ∈ 2 Γ.
Proof. Ad 1. A straightforward computation shows that the trace property

holds on the subalgebra CΓ (check!). By construction, trΓ is weakly contin-
uous. Therefore, the trace property also holds on N Γ.
Ad 2. For the non-trivial implication, let a ∈ N Γ with trΓ (a∗ ◦ a) = 0.
Then, by definition, we have
0 = trΓ (a∗ ◦ a) = e, a∗ ◦ a(e) = a(e), a(e)
and thus a(e) = 0. Because a is Γ-linear, we also obtain a(g · e) = g · a(e) = 0.

Continuity of a therefore shows that a = 0.
Ad 3. This is clear from the definition of positivity and the trace.
Example 1.1.13 (some von Neumann traces).
• If Γ is a finite group, then N Γ = CΓ. The von Neumann trace is
trΓ : N Γ = CΓ −→ C

ag · g −→ ae .
g∈Γ
• If Γ = Z = t | , then we obtain [105, Example 1.4]: The group

von Neumann algebra N Γ is canonically isomorphic to L∞ [−π, π], C
(as can be seen via the Fourier transform) and the action on 2 Γ ∼ =
L2 ([−π, π], C) is given by pointwise multiplication; under this isomor-
phism, the trace trΓ on N Γ corresponds to the integration map
L∞ [−π, π], C −→ C

1
f −→ · f dλ.
2π [−π,π]
In view of the previous example (and Theorem A.1.3), the abstract theory
of von Neumann algebras is also sometimes referred to as non-commutative
measure theory.
Remark 1.1.14 (extension of the trace to matrices and morphisms). As in linear

algebra, we can extend the trace from the group von Neumann algebra to
matrices: Let Γ be a countable group and let n ∈ N. Then we define the trace
1.2. The von Neumann dimension 11
trΓ : Mn×n (N Γ) −→ C
n
A −→ trΓ Ajj .
j=1
This trace also satisfies the trace property, is faithful, and positive (check!).
Moreover, every bounded (left) Γ-equivariant map (2 Γ)n −→ (2 Γ)n is
represented by a matrix in Mn×n (N Γ) (by the last characterisation in Re-
mark 1.1.11). Therefore, every bounded Γ-equivariant map (2 Γ)n −→ (2 Γ)n
has a trace.
1.2 The von Neumann dimension

We can now define the von Neumann dimension of Hilbert modules via the
trace of projections:
Proposition and Definition 1.2.1 (von Neumann dimension). Let Γ be a count-
able group and let V be a Hilbert Γ-module. Then the von Neumann Γ-di-
mension of V is defined as
dimN Γ V := trΓ p,
where i : V −→ (2 Γ)n (for some n ∈ N) is an isometric Γ-embedding and

p : (2 Γ)n −→ (2 Γ)n is the orthogonal Γ-projection onto i(V ). This is well-
defined (i.e., independent of the chosen embedding into a finitely generated
free 2 Γ-module) and dimN Γ V ∈ R≥0 .
Proof. As a first step, we note that i(V ) is a closed subspace of (2 Γ)n (be-
cause V is complete and i is isometric). Hence, there indeed exists an orthog-
onal projection p : (2 Γ)n −→ im i.
The trace is independent of the embedding: Let j : V −→ (2 Γ)m also
be an isometric Γ-embedding and let q : (2 Γ)m −→ im j be the orthogonal
projection. Then we define a partial isometry u : (2 Γ)n −→ (2 Γ)m by tak-
ing j ◦ i−1 on im i and taking 0 on (im i)⊥ . By construction j = u ◦ i. Taking
adjoints shows that q = p ◦ u∗ and hence
trΓ q = trΓ (j ◦ q) = trΓ (u ◦ i ◦ q) = trΓ (u ◦ i ◦ p ◦ u∗ )

= trΓ (i ◦ p ◦ u∗ ◦ u) (trace property)
= trΓ (i ◦ p ◦ p) = trΓ (i ◦ p) = trΓ p.
The von Neumann dimension is non-negative: Let P ∈ Mn×n (N Γ) be the

matrix representing p. Because p (as an orthogonal projection) is a positive
operator (check!), all the diagonal entries Pjj ∈ N Γ of P are also positive
operators (check!). Therefore, positivity of the von Neumann trace (Theo-
n
rem 1.1.12) shows that dimN Γ V = trΓ p = j=1 trΓ Pjj ≥ 0.
Example 1.2.2 (von Neumann dimension).

• Let Γ be a finite group and let V be a Hilbert Γ-module. Then
1
dimN Γ V = · dimC V,
|Γ|
as can be seen from a direct computation (check!) or by applying the

restriction formula (Theorem 1.2.3).
• Let Γ = Z = t | . We will use the description of 2 Γ and N Γ from
Example 1.1.13. Let A ⊂ [−π, π] be a measurable set. Then V :=
{f · χA | f ∈ L2 ([−π, π], C)} is a Hilbert Γ-module and the 1 × 1-
matrix (χA ) ∈ M1×1 (N Γ) describes the orthogonal projection onto V .
Hence,
π
1 1
dimN Γ V = trΓ χA = · χA dλ = · λ(A)
2π −π 2π
and thus every number in [0, 1] occurs as the von Neumann dimension
of a Hilbert N Γ-module (!).
Theorem 1.2.3 (basic properties of the von Neumann dimension). Let Γ be a
countable group.
1. Normalisation. We have dimN Γ 2 Γ = 1.
2. Faithfulness. For every Hilbert Γ-module V , we have dimN Γ V = 0 if
and only if V ∼
=Γ 0.
3. Weak isomorphism invariance. If f : V −→ W is a morphism of Hilbert
Γ-modules with ker f = 0 and im f = W , then dimΓ V = dimΓ W .
4. Additivity. Let 0 / V i / V π / V / 0 be a weakly ex-

act sequence of Hilbert Γ-modules (i.e., i is injective, im i = ker π and
im π = V ). Then
dimN Γ V = dimN Γ V + dimN Γ V .
5. Multiplicativity. Let Λ be a countable group, let V be a Hilbert Γ-

module, and let W be a Hilbert Λ-module. Then the completed tensor
product V ⊗C W is a Hilbert Γ × Λ-module and
dimN (Γ×Λ) (V ⊗C W ) = dimN Γ V · dimN Λ W.
6. Restriction. Let V be a Hilbert Γ-module and let Λ ⊂ Γ be a subgroup

of finite index. Then
dimN Λ ResΓΛ V = [Γ : Λ] · dimN Γ V.

1.2. The von Neumann dimension 13
Proof. Ad 1. This is clear from the definition (we can take id2 Γ as embedding
and projection).
Ad 2. In view of faithfulness of the von Neumann trace (Theorem 1.1.12),
it follows that the von Neumann trace of a projection is 0 if and only if the
projection is 0 (check!).
Ad 3. This is a consequence of polar decomposition: Let f = u ◦ p be the
polar decomposition of f into a partial isometry u and a positive operator p
with ker u = ker p. We now show that u is a Γ-isometry between V and W :
As f is injective, we have ker u = ker p = 0. Moreover, as a partial isometry,
u has closed image and so im u = im u = im f = W . Hence, u is an isom-
etry. Moreover, the uniqueness of the polar decomposition shows that u is
Γ-equivariant. Therefore, dimN Γ V = dimN Γ W .
Ad 4. The von Neumann dimension is additive with respect to direct sums
(check!). Moreover,
V −→ im i ⊕ V
x −→ p(x), π(x)

is a weak isomorphism of Hilbert Γ-modules (check!), where p : V −→ im i

denotes the orthogonal projection. Therefore, weak isomorphism invariance
of dimΓ shows that
dimΓ V = dimΓ (im i ⊕ V ) = dimΓ im i + dimΓ V = dimΓ V + dimΓ V .
Ad 5. The key observation is that 2 (Γ × Λ) is isomorphic (as a Hilbert

Γ × Λ-module) to 2 Γ ⊗C 2 Λ [105, Theorem 1.12].
Ad 6. This is Exercise 1.E.4.
Moreover, the von Neumann dimension also satisfies inner and outer reg-
ularity [105, Theorem 1.12].
Outlook 1.2.4 (the extended von Neumann dimension). The above hands-on
construction of the von Neumann dimension is convenient for simple com-
putations. However, this approach only works for the category of Hilbert
modules, which in general is only additive but not Abelian. Several exten-
sions of the von Neumann dimension are available, e.g., by Cheeger and
Gromov [36], Farber [51], and Lück [104, 105]. We will briefly outline Lück’s
algebraic version:
Let Γ be a countable group. Then the category of Hilbert Γ-modules canon-
ically embeds into the category of N Γ-modules (Exercise 1.E.2); moreover,
this construction can be refined to a C-linear equivalence F from the cat-
egory of Hilbert Γ-modules to the category of finitely generated projective
N Γ-modules that satisfies F (2 Γ) = N Γ and preserves (weak) exactness [104,
Theorem 1.8].
If P is a finitely generated projective N Γ-module, then we set
pdimN Γ P := trN Γ p := trΓ A ∈ R≥0 ,
where p : (N Γ)n −→ (N Γ)n is a projection with P = ∼N Γ im p and associ-

ated matrix A ∈ Mn×n (N Γ); this definition is independent of the chosen
projection p [105, p. 238f].
For a general N Γ-module V one then defines

dimN Γ V := sup pdimN Γ P P is a finitely generated projective
N Γ-submodule of V ∈ R≥0 ∪ {∞}.

It turns out that this definition provides a well-behaved notion of dimension

for N Γ-modules that coincides via F with the von Neumann dimension of
Hilbert Γ-modules [104, Theorem 0.6]. One of the key ingredients is the obser-
vation that the ring N Γ is semi-hereditary (i.e., that every finitely generated
submodule of a projective N Γ-module is projective).
Furthermore, the N Γ-modules of dimension 0 form a Serre subcategory of
the category of N Γ-modules; this allows us to efficiently use standard tools
from homological algebra when working with von Neumann dimensions [123].
In fact, the same construction works for every finite von Neumann algebra,
not only for the group von Neumann algebra N Γ [104, 123]. In Chapter 4,
we will use such an extended von Neumann dimension in the context of
equivalence relations.
Outlook 1.2.5 (Atiyah conjecture). The Atiyah question/conjecture comes in

many flavours (originally formulated in terms of closed Riemannian mani-
folds). One version is:
Let Γ be a torsion-free countable group, let n ∈ N, and let A ∈

Mn×n (CΓ). Then dimN Γ ker MA ∈ Z (?!)
This version of the Atiyah conjecture is known to hold for many classes of
groups (and no counterexample is known so far); however, more general ver-
sions of the Atiyah conjecture are known to be false [12, 72]. One interesting
aspect of the Atiyah conjecture is that it implies the Kaplansky zero-divisor
conjecture (Exercise 1.E.5).
1.E. Exercises 15
1.E Exercises
Exercise 1.E.1 (the “trivial” Hilbert module). For which countable groups Γ
is C (with the trivial Γ-action) a Hilbert Γ-module? Which von Neumann
dimension does it have?
Exercise 1.E.2 (Hilbert modules as modules over the von Neumann algebra).
Let Γ be a countable group and let V be a Hilbert Γ-module. Show that the
left Γ-action on V extends to a left N Γ-action on V . How can this construction
be turned into a functor?
Hints. This fact is the reason why Hilbert Γ-modules are often called Hilbert
N Γ-modules.
Exercise 1.E.3 (kernels and cokernels). Let Γ be a countable group, let V and
W be Hilbert Γ-modules, and let ϕ : V −→ W be a morphism of Hilbert
Γ-modules.
1. Show that ker ϕ (with the induced inner product and Γ-action) is a
Hilbert Γ-module
2. Show that W/im ϕ (with the induced inner product and Γ-action) is a
Hilbert Γ-module.
Hints. Orthogonal complement!
Exercise 1.E.4 (restriction formula for the von Neumann dimension [105, Theo-
rem 1.12(6)]). Let Γ be a countable group, let V be a Hilbert Γ-module, and
let Λ ⊂ Γ be a finite index subgroup. Show that
dimN Λ ResΓΛ V = [Γ : Λ] · dimN Γ V.
Exercise 1.E.5 (Atiyah =⇒ Kaplansky [105, Lemma 10.15]). Let Γ be a count-

able torsion-free group that satisfies the Atiyah conjecture (Outlook 1.2.5).
Show that CΓ is a domain.
2
L2-Betti numbers
We will now use the von Neumann dimension to define L2 -Betti numbers of
groups and spaces. We will study basic properties of L2 -Betti numbers, and
we will compute some simple examples. Furthermore, we will briefly outline
generalisations of L2 -Betti numbers beyond our elementary approach.
2.1 An elementary definition of L2 -Betti numbers 18

2.2 Basic computations 20
2.3 Variations and extensions 24
2.E Exercises 26
Running example. the additive group Z, finite (index sub)groups, free groups

18 2. L2 -Betti numbers
2.1 An elementary definition of L2-Betti numbers

L2 -Betti numbers are an equivariant version of ordinary Betti numbers. For
simplicity, we will only consider L2 -Betti numbers of free Γ-CW-complexes
of finite type.
2.1.1 Finite type
Definition 2.1.1 (equivariant CW-complex). Let Γ be a group.

• A free Γ-CW-complex is a CW-complex X together with a free Γ-action
such that:
– the Γ-action permutes the open cells of X and
– if e is an open cell of X and g ∈ Γ is non-trivial, then g · e ∩ e = ∅.
• A morphism of Γ-CW-complexes is a Γ-equivariant cellular map.
Definition 2.1.2 (finite type).
• A CW-complex is of finite type if for each n ∈ N, there are only finitely
many open n-cells.
• Let Γ be a group. A (free) Γ-CW-complex is of finite type if for each
dimension n ∈ N, there are only finitely many Γ-orbits of open n-cells.
• A group Γ is of finite type if it admits a classifying space of finite
type (equivalently, a classifying space whose universal covering with
the induced free Γ-CW-structure is a Γ-CW-complex of finite type).
Remark 2.1.3 (more on groups of finite type). Let Γ be a group.
• If Γ is of finite type, then Γ is finitely presented (as the fundamental
group of a CW-complex of finite type); in particular, Γ is countable.
• If Γ is finitely presented, then Γ is of finite type if and only if C (with the
trivial Γ-action) admits a projective resolution over CΓ that is finitely
generated in each degree [30, Chapter VIII].
Example 2.1.4 (groups of finite type).
• The group Z is of finite type: We can take the circle (with our favourite
CW-structure) as a classifying space.
• If Γ and Λ are of finite type, then so is Γ × Λ (we can take the product
of finite type models as a model for the classifying space of Γ × Λ).
2.1. An elementary definition of L2 -Betti numbers 19
• If Γ and Λ are of finite type, then so is Γ ∗ Λ (we can take the wedge of
finite type models as a model for the classifying space of Γ ∗ Λ).
• In particular, free Abelian groups of finite rank and free groups of finite
rank are of finite type.
• Let g ∈ N≥2 and let Σg be “the” oriented closed connected surface of

genus g. Then π1 (Σg ) is of finite type (because Σg is a (finite) model
for the classifying space of Σg ).
• More generally: If M is an oriented closed connected hyperbolic mani-

fold, then π1 (M ) has finite type (because M is a (finite) model for the
classifying space of π1 (M )).
• If n ∈ N≥2 , then Z/n is of finite type (check!), but there is no finite

model for the classifying space of Z/n [30, Corollary VIII.2.5].
• More generally: All finite groups are of finite type (for instance, one can
use the simplicial Γ-resolution as a blueprint to construct a contractible
free Γ-CW-complex and then take its quotient).
• There exist finitely presented groups that are not of finite type [20].
2.1.2 L2 -Betti numbers of spaces
Definition 2.1.5 (L2 -Betti numbers of spaces). Let Γ be a countable group

and let X be a free Γ-CW-complex of finite type.
• The cellular L2 -chain complex of X is the twisted chain complex

(2)
C∗ (Γ X) := 2 Γ ⊗CΓ C∗ (X).
Here, C∗ (X) denotes the cellular chain complex of X (with C-coeffi-

cients) with the induced Γ-action and 2 Γ carries the left CΓ-module
structure given by right translation on Γ.
• Let n ∈ N. The (reduced) L2 -homology of X in degree n is defined by
(2)
Hn(2) (Γ X) := ker ∂n(2)

im ∂n+1 ,
(2) (2)
where ∂∗ denotes the boundary operator on C∗ (Γ X).
• The n-th L2 -Betti number of X is defined by
b(2) (2)
n (Γ X) := dimN Γ Hn (Γ X),
where dimN Γ is the von Neumann dimension (Definition 1.2.1). It

(2)
should be noted that Hn (Γ X) is indeed a Hilbert Γ-module (this
follows from Exercise 1.E.3).
Notation 2.1.6. Moreover, we use the following abbreviation: If X is a CW-
complex of finite type with fundamental group Γ and universal covering X
(with the induced free Γ-CW-complex structure), then we write
b(2) (2)
n (X) := bn (Γ X).

(2)
It should be noted that in the literature the notation bn (X)
can also be found
(2) However, we prefer the notation b(2)
as an abbreviation for bn (Γ X). n (X)
(2)
as it is less ambiguous (what is bn (H2 ) ?!).
Remark 2.1.7 (homotopy invariance). Let Γ be a countable group, let X and
Y be free Γ-CW-complexes, and let n ∈ N. If f : X −→ Y is a (cellular)
Γ-homotopy equivalence, then
b(2) (2)
n (Γ X) = bn (Γ Y ),
because f induces a CΓ-chain homotopy equivalence C∗ (X) CΓ C∗ (Y ) and

(2) (2)
thus a chain homotopy equivalence C∗ (Γ X) −→ C∗ (Γ Y ) in the
category of chain complexes of Hilbert Γ-modules.
2.1.3 L2 -Betti numbers of groups

Let Γ be a group and let X and Y be models for the classifying space of Γ.
Then the universal coverings X
and Y with the induced Γ-CW-structures are
(2) (2)
(cellularly) Γ-homotopy equivalent. Therefore, bn (X) = bn (Y ) for all n ∈ N
(Remark 2.1.7). Hence, the following notion is well-defined:
Definition 2.1.8 (L2 -Betti numbers of groups). Let Γ be a group of finite type
and let n ∈ N. Then the n-th L2 -Betti number of Γ is defined by
b(2) (2)
n (Γ) := bn (X),
where X is a model for the classifying space of Γ of finite type.

Similarly, we could also define/compute L2 -Betti numbers of groups by
tensoring finite type projective resolutions of C over CΓ with 2 Γ (check!).
2.2 Basic computations

For simplicity, in the following, we will focus on L2 -Betti numbers of groups.
Similar statements hold for L2 -Betti numbers of spaces [105, Theorem 1.35].
2.2. Basic computations 21
2.2.1 Basic properties
Proposition 2.2.1 (degree 0). Let Γ be a group of finite type.

(2)
1. If Γ is finite, then b0 (Γ) = 1/|Γ|.
(2)
2. If Γ is infinite, then b0 (Γ) = 0.
Both cases can conveniently be summarised in the formula
(2) 1
b0 (Γ) = .
|Γ|
(2)
Proof. Classical group homology tells us that b0 (Γ) = dimN Γ V , where
V = 2 Γ SpanC {x − g · x | x ∈ 2 Γ, g ∈ Γ}.

If Γ is finite, then 2 Γ = CΓ, whence V ∼

=Γ C (with the trivial Γ-action).
Therefore (Exercise 1.E.1),
(2) 1
b0 (Γ) = dimN Γ C = .
|Γ|
If Γ is infinite, then it suffices to show that V ∼=Γ 0: To this end, we only

need to show that every bounded C-linear functional V −→ C is the zero
functional. Equivalently, we need to show that every Γ-invariant bounded C-
linear functional f : 2 Γ −→ C satisfies f |Γ = 0 (check!). As Γ is infinite (and
countable), we can enumerate Γ as (gn )n∈N . The element x := n∈N 1/n · gn
lies in 2 Γ and the computation
1
f (x) = · f (gn ) (continuity and linearity of f )
n
n∈N
1
= · f (e) (Γ-invariance of f )
n
n∈N
shows that f (e) = 0 (otherwise the series would not converge). Hence f (g) =
f (e) = 0 for all g ∈ Γ (by Γ-invariance), as desired.
Theorem 2.2.2 (inheritance properties of L2 -Betti numbers). Let Γ be a group
of finite type and let n ∈ N.
1. Dimension. If Γ admits a finite model of the classifying space of dimen-
(2)
sion less than n, then bn (Γ) = 0.
2. Restriction. If Λ ⊂ Γ is a subgroup of finite index, then Λ is also of
finite type and
n (Λ) = [Γ : Λ] · bn (Γ).
b(2) (2)
3. Künneth formula. If Λ is a group of finite type, then

n
(2) (2)

n (Γ × Λ) =
b(2) bj (Γ) · bn−j (Λ).
j=0
4. Additivity. If Λ is a group of finite type, then

(2) (2) (2) (2) (2)
b1 (Γ ∗ Λ) = b1 (Γ) + b1 (Λ) + 1 − b0 (Γ) + b0 (Λ)

and, if n > 1, then
n (Γ ∗ Λ) = bn (Γ) + bn (Λ).
b(2) (2) (2)
5. Poincaré duality. If Γ admits a classifying space that is an oriented

closed connected d-manifold, then
(2)
b(2)
n (Γ) = bd−n (Γ).
Proof. Ad 1. This is clear from the definition.

Ad 2. If X is a finite type model for the classifying space of Γ, then the
covering space Y associated with the subgroup Λ ⊂ Γ is a model for the
classifying space of Λ; moreover, Y is of finite type because the covering
degree is [Γ : Λ], which is finite. Algebraically, on the cellular chain complex
of the universal covering space X = Y (and whence also on its reduced
cohomology with 2 -coefficients), this corresponds to applying the restriction
functor ResΓΛ . Then, we only need to apply the restriction formula for the
von Neumann dimension (Theorem 1.2.3).
Ad 3. Let X and Y be finite type models for the classifying space of Γ
and Λ, respectively. Then X × Y is a model for the classifying space of Γ ×
Λ, which is of finite type. One can now use a Künneth argument and the
multiplicativity of the von Neumann dimension (Theorem 1.2.3) to prove the
claim [105, Theorem 1.35].
Ad 4. Let X and Y be finite type models for the classifying space of Γ and
Λ, respectively. Then the wedge X ∨ Y is a model for the classifying space
of Γ ∗ Λ, which is of finite type. One can now use a (cellular) Mayer–Vietoris
argument to prove the additivity formula [105, Theorem 1.35].
Ad 5. The main ingredients are twisted Poincaré duality (applied to the
coefficients 2 Γ) and the fact that the L2 -Betti numbers can also be computed
in terms of reduced cohomology [105, Theorem 1.35].
Proposition 2.2.3 (Euler characteristic). Let Γ be a group that admits a finite

classifying space. Then
2.2. Basic computations 23

χ(Γ) = (−1)n · b(2)
n (Γ).
n∈N
Proof. This follows (as in the classical case) from the additivity of the
von Neumann dimension (Exercise 2.E.2).
2.2.2 First examples
Example 2.2.4 (finite groups). Let Γ be a finite group. Then Γ is of finite

type (Example 2.1.4) and, for all n ∈ N, we have

1
1 if n = 0
(2)
bn (Γ) = · dimC Hn (Γ; CΓ) = |Γ|
|Γ| 0 if n > 0.
Example 2.2.5 (the additive group Z). There are many ways to see that the
L2 -Betti numbers of the additive group Z are all equal to 0. For instance: Let
n ∈ N. For k ∈ N>1 , we consider the subgroup k · Z ⊂ Z of index k. Then the
restriction formula (Theorem 2.2.2) shows that
n (Z) = bn (k · Z)
b(2) (2)
(because k · Z ∼
=Group Z)
=k· b(2)
n (Z) (restriction formula)
(2)
and so bn (Z) = 0.
groups). Let r ∈ N≥1 and let Fr be “the” free group of

Example 2.2.6 (free
r 1
rank r. Then Xr := S is a model of the classifying space of Fr .
(2)
• Because Fr is infinite, we have b0 (Fr ) = 0 (Proposition 2.2.1).
(2)
• Because dim Xr = 1, we have bn (Fr ) = 0 for all n ∈ N≥2 .
(2)
• It thus remains to compute b1 (Fr ). Because the Euler characteristic
can be calculated via L2 -Betti numbers (Proposition 2.2.3), we obtain
(2) (2)
b1 (Fr ) = −χ(Fr ) + b0 (Fr ) = −χ(Xr ) + 0 = r − 1.
Example 2.2.7 (surface groups). Let g ∈ N≥1 and let Γg := π1 (Σg ), where Σg
is “the” oriented closed connected surface of genus g. Then Σg is a model of
the classifying space of Σg .
(2)
• Because Γg is infinite, we have b0 (Γg ) = 0 (Proposition 2.2.1).
• Because dim Σg = 2, we obtain from Poincaré duality (Theorem 2.2.2)

(2) (2) (2)
that b2 (Γg ) = b0 (Γg ) = 0 and that bn (Γg ) = 0 for all n ∈ N≥3 .
(2)
• It thus remains to compute b1 (Σg ). Because the Euler characteristic
can be calculated via L2 -Betti numbers (Proposition 2.2.3), we obtain
(2) (2) (2)
b1 (Γg ) = −χ(Γg ) + b0 (Γg ) + b2 (Γg ) = −χ(Σg ) + 0 = 2 · (g − 1).
Outlook 2.2.8 (hyperbolic manifolds). More generally: Let Γ be the funda-

mental group of an oriented closed connected hyperbolic manifold M of di-
mension d.
(2)
• If d is odd, then bn (Γ) = 0 for all n ∈ N.
(2)
• If d is even, then bn (Γ) = 0 for all n ∈ N \ {d/2}. Moreover,
(2)
bd/2 (Γ) = 0.
The proof is based on the fact that L2 -Betti numbers can be computed in
terms of spaces of L2 -harmonic forms [46][105, Chapter 1.4] and the ex-
plicit computation of L2 -harmonic forms of hyperbolic manifolds [47][105,
Theorem 1.62]. (A cellular version of harmonic forms is discussed in Exer-
cise 3.E.2.)
Outlook 2.2.9 (Singer conjecture). The Singer conjecture predicts that L2 -

Betti numbers of closed aspherical manifolds are concentrated in the middle
dimension:
Let M be an oriented closed connected aspherical manifold of dimen-

sion d. Then
∀n∈N\{d/2} b(2)

n π1 (M ) = 0 (?!)
No counterexample is known. However, the analogue of the Singer conjecture

for rationally aspherical manifolds is false [13] and not much is known about
L2 -Betti numbers of “exotic” closed aspherical manifolds [41].
2.3 Variations and extensions

• Analytic definition. Originally, Atiyah defined L2 -Betti numbers (of
closed smooth manifolds) in terms of the heat kernel on the universal
covering [11]. Dodziuk proved that these analytic L2 -Betti numbers
admit a combinatorial description (in terms of 2 -chain complexes of
simplicial/cellular complexes of finite type) [46].
• Singular definition. Cheeger and Gromov [36], Farber [51], Lück [104,
105] extended the definition of the von Neumann dimension to all mod-
ules over the von Neumann algebra (Outlook 1.2.4); in particular, this
2.3. Variations and extensions 25
allows for a definition of L2 -Betti numbers of spaces in terms of singular

homology with twisted coefficients in 2 π1 or N π1 .
• Extension to equivalence relations. Gaboriau extended the definition

of L2 -Betti numbers of groups to standard equivalence relations [64].
We will return to this point of view in Chapter 4.
• Extension to topological groups. Petersen gave a definition of L2 -Betti

numbers of locally compact, second countable, unimodular groups [119].
• Version for von Neumann algebras. In a slightly different direction,
Connes and Shlyakhtenko introduced a notion of L2 -Betti numbers for
tracial von Neumann algebras [40]. However, it is unknown to what
extent these L2 -Betti numbers of group von Neumann algebras coincide
with the L2 -Betti numbers of groups (which would be helpful in the
context of the free group factor isomorphism problem).
It should be noted that Popa and Vaes showed that Thom’s continuous
version of the corresponding L2 -cohomology for tracial von Neumann
algebras is always trivial in degree 1 [121].
2.E Exercises
Exercise 2.E.1 (products).
(2)
1. Let Γ and Λ be infinite groups of finite type. Show that b1 (Γ × Λ) = 0.
(2)
2. Conclude: If Γ is a group of finite type with b1 (Γ) = 0, then Γ does
not contain a finite index subgroup that is a product of infinite groups.
Exercise 2.E.2 (Euler characteristic). Let Γ be a group that admits a finite
classifying space. Show that

χ(Γ) = (−1)n · b(2)
n (Γ).
n∈N
Exercise 2.E.3 (an explicit 2 -cycle [84, Figure 1.1]). Give an explicit example
(2)
of a non-zero 1-cycle in C∗ (F2 T ), where T is “the” regular 4-valent tree.
Exercise 2.E.4 (QI?! [135]). Let Γ be a group of finite type and let r ∈ N≥2 .
1. Compute all L2 -Betti numbers of Fr ∗ Γ in terms of r and the L2 -Betti
numbers of Γ.
(2) (2)
2. Let k ∈ N≥2 . Conclude that the quotient b1 /bk is not a quasi-
isometry invariant.
3. Show that the sign of the Euler characteristic is not a quasi-isometry
invariant.
4. Use these results to prove that there exist groups of finite type that are
quasi-isometric but not commensurable.
Hints. If s ∈ N≥2 , then it is known that Fr and Fs are bilipschitz equivalent
and thus that Fr ∗ Γ and Fs ∗ Γ are quasi-isometric [115, 135].
Exercise 2.E.5 (deficiency [102]).

1. Let Γ be a group of finite type and let S | R be a finite presentation
of Γ. Show that
(2) (2) (2)
|S| − |R| ≤ 1 − b0 (Γ) + b1 (Γ) − b2 (Γ).
Taking the maximum of all these differences thus shows that the defi-
ciency def(Γ) of Γ is bounded from above by the right-hand side.
2. Let Γ ⊂ Isom+ (H4 ) be a torsion-free uniform lattice. Show that
3
def(Γ) ≤ 1 − χ(Γ) = 1 − · vol(Γ \ H4 ).
4 · π2
3
The residually finite view:
Approximation
The L2 -Betti numbers are related to classical Betti numbers through ap-
proximation by the normalised Betti numbers of finite index subgroups/finite
coverings.
We explain the (spectral) proof of this approximation theorem and briefly
discuss the relation with other (homological) gradient invariants.
This residually finite view will be complemented by the dynamical view in
Chapter 4 and the approximation theorems for lattices in Chapter 5.
3.1 The approximation theorem 28

3.2 Proof of the approximation theorem 28
3.3 Homological gradient invariants 32
3.E Exercises 35
Running example. free Abelian groups, free groups

28 3. The residually finite view: Approximation
3.1 The approximation theorem

In the residually finite view, one approximates groups by finite quotients/finite
index subgroups and spaces by finite coverings.
Definition 3.1.1 (residual chain, residually finite group). Let Γ be a finitely
generated group.
• A residual chain for Γ is a sequence (Γn )n∈N
of finite index normal
subgroups of Γ with Γ0 ⊃ Γ1 ⊃ Γ2 ⊃ . . . and n∈N Γn = {e}.
• The group Γ is residually finite if it admits a residual chain.
Example 3.1.2 (residually finite groups).
• All finitely generated linear groups are residually finite [107, 113].
In particular: Fundamental groups of closed hyperbolic manifolds are
residually finite. In contrast, it is unknown whether all finitely gener-
ated (Gromov-)hyperbolic groups are residually finite.
• There exist finitely presented groups that are not residually finite (e.g.,
each finitely presented infinite simple group will do).
Theorem 3.1.3 (Lück’s approximation theorem [103]). Let X be a connected
CW-complex of finite type with residually finite fundamental group Γ, let
(Γn )n∈N be a residual chain for Γ, and let k ∈ N. Then
(2) bk (Xn )
bk (X) = lim .
n→∞ [Γ : Γn ]
Here, Xn denotes the finite covering of X associated with the subgroup Γn ⊂ Γ

and bk is the ordinary k-th Q-Betti number (which equals the C-Betti num-
ber).
Corollary 3.1.4 (approximation theorem, for groups). Let Γ be a residually
finite group of finite type, let (Γn )n∈N , and let k ∈ N. Then
(2) bk (Γn )
bk (Γ) = lim .
n→∞ [Γ : Γn ]
3.2 Proof of the approximation theorem

The proof of the approximation theorem (Theorem 3.1.3) is of a spectral
nature; we will roughly follow unpublished notes of Sauer.
3.2. Proof of the approximation theorem 29
3.2.1 Reduction to kernels of self-adjoint operators

As first step, we reduce the approximation theorem to a statement about
von Neumann dimensions of kernels of self-adjoint operators:
• On the one hand, if Λ Y is a free Λ-CW-complex (with a countable

group Λ), then the (combinatorial) Laplacian Δ∗ of the 2 -chain com-
plex of Λ Y is a positive self-adjoint operator on a Hilbert Λ-module
that satisfies (Exercise 3.E.2)
(2)
bk (Λ Y ) = dimN Λ ker Δk .
• On the other hand, the right-hand side in the approximation theorem

(Theorem 3.1.3), can also be written as a von Neumann dimension: For
each n ∈ N, the finite group Γ/Γn is of finite type (Example 2.1.4), we
have (Example 1.2.2)
bk (Xn ) (2)
= bk (Γ/Γn Xn ),
[Γ : Γn ]
(2)
and the boundary operator [Laplacian] on C∗ (Γ/Γn Xn ) is the
(2)
reduction of the boundary operator [Laplacian] on C∗ (Γ X) mod-
ulo Γn .
Therefore, Theorem 3.1.3 is a consequence of the following, slightly more

algebraically looking, version (because the cellular Laplacian is defined over
the integral group ring):
Theorem 3.2.1 (approximation theorem for kernels). Let Γ be a finitely gener-

ated residually finite group with a residual chain (Γn )n∈N , let m ∈ N, and let
A ∈ Mm×m (ZΓ) be self-adjoint and positive. Then
dimN Γ ker MA : (2 Γ)m → (2 Γ)m

= lim dimN (Γ/Γn ) ker MAn : (2 (Γ/Γn ))m → (2 (Γ/Γn ))m ,

n→∞
where An ∈ Mm×m (Z[Γ/Γn ]) denotes the reduction of A modulo Γn .
We will now prove Theorem 3.2.1.
3.2.2 Reformulation via spectral measures

We reformulate the claim of Theorem 3.2.1 in terms of spectral measures: Let
μA be the spectral measure on R (with the Borel σ-algebra) of the self-adjoint
operator A. This measure has the following properties [23, Chapter 6][84,
Chapter 5.2]:
• The measure μA is supported on the compact set [0, a], where a := A.
• If f : [0, ∞] −→ R is a measurable bounded function, then the bounded
linear operator f (MA ), defined by functional calculus, satisfies
f dμA = trΓ f (MA ).

R
• In the same way, for each n ∈ N, the spectral measure μAn of the
reduction An of A is also supported on [0, a] (because An ≤ A).
Therefore, we obtain
dimN Γ ker MA = trΓ (orthogonal projection onto ker MA )

= trΓ χ{0} (MA )
= μA {0})

and, for all n ∈ N,
dimN (Γ/Γn ) ker MAn = μAn {0} .

Hence, the claim of Theorem 3.2.1 is equivalent to the following property of

the spectral measures: μA ({0}) = limn→∞ μAn ({0}). We will now prove this
statement on spectral measures.
3.2.3 Weak convergence of spectral measures
We first establish weak convergence of the spectral measures (Definition A.2.1):

Lemma 3.2.2 (weak convergence of spectral measures). In this situation, the
sequence (μAn )n∈N of measures on R weakly converges to μA , i.e., for all
continuous functions f : [0, a] −→ R, we have
lim f dμAn = f dμA .

n→∞ R R
Proof. The measures μAn with n ∈ N and the measure μA are all supported
on a common compact set (namely [0, a]). Therefore, by the Weierstraß ap-
proximation theorem, it suffices to take test functions of the form (x → xd )
with d ∈ N.
Thus, let d ∈ N and f := (x → xd ). We then have
f dμA = trΓ f (MA ) = trΓ (Ad )

R
3.2. Proof of the approximation theorem 31

and R f dμAn = trΓ/Γn Adn .
Let F ⊂ Γ be the support of Ad (i.e., the set of all elements of Γ that
occur with non-zero coefficient in A). Because F is finite and (Γn )n∈N is a
residual chain, there exists an N ∈ N with
∀n∈N≥N F ∩ Γn ⊂ {e}.
Then (by definition of the trace; check!)
trΓ/Γn Adn = trΓ Ad
for all n ∈ N≥N . This shows weak convergence.
3.2.4 Convergence at 0
In general, weak convergence does not imply convergence of the measures
on {0} (Exercise 3.E.3). But by the portmanteau theorem (Theorem A.2.2),
we at least obtain the following inequalities from Lemma 3.2.2:
lim sup μAn {0}) ≤ μA {0}

n→∞
∀ε∈R>0 lim inf μAn (−ε, ε) ≥ μA (−ε, ε) .

n→∞
(2)
The first inequality already gives lim supn→∞ 1/[Γn : Γ]bk (Xn ) ≤ bk (X),
the Kazhdan inequality [83].
In order to show the missing lower bound lim inf n→∞ μAn ({0}) ≥ μ({0}),
we will use integrality of the coefficients of A. More precisely, we will show:
Lemma 3.2.3. In this situation, for all n ∈ N and all ε ∈ (0, 1), we have
m · ln(C)
μAn (0, ε) ≤

,
ln(ε)
where C := max A, 1 (which does not depend on n).

Proof. Let n ∈ N and let d := [Γ : Γn ]. Computing μAn ((0, ε)) amounts to

counting eigenvalues. We can view An as a matrix in Mm×m (Z[Γ/Γn ]) ⊂
Md·m×d·m (Z). In this view, An is symmetric and positive semi-definite
(check!); let
0 = λ1 ≤ · · · ≤ λz = 0 < λz+1 ≤ · · · ≤ λd·m
be the eigenvalues of An (listed with multiplicities). Then the characteristic
polynomial of An is of the form T z · q with q ∈ Z[T ]. In particular, q(0) = 0
and thus (because of integrality!)
λz+1 · · · · · λd·m = q(0) ≥ 1.

For ε ∈ (0, 1), let M (ε) be the number of eigenvalues of An in (0, ε). Then
1 ≤ λz+1 · · · · · λd·m ≤ εM (ε) · MAn d·m−z−M (ε) ≤ εM (ε) · C d·m ,
and so M (ε) ≤ d · m · ln C/| ln ε|. Therefore, we obtain

μAn (0, ε) = dimN (Γ/Γn ) all eigenspaces of An for eigenvalues in (0, ε)
M (ε)
=
d
m · ln C
≤ .
ln(ε)
We can now complete the proof of Theorem 3.2.1 as follows: We have
lim inf μAn {0} = lim inf μAn [0, ε] − μAn (0, ε)

n→∞ n→∞
m · ln C
≥ lim inf μAn (−ε, ε) −

(Lemma 3.2.3)
n→∞ ln(ε)
m · ln C
≥ μA (−ε, ε) −

(portmanteau theorem)
ln(ε)
m · ln C
≥ μA {0} − .
ln(ε)
Taking ε → 0 yields the desired estimate lim inf n→∞ μAn ({0}) ≥ μA ({0}).
This finishes the proof of Theorem 3.2.1 and whence also of the approximation
theorem (Theorem 3.1.3).
3.3 Homological gradient invariants

If I is a numerical invariant of (finitely generated residually finite) groups,
then one can consider the associated gradient invariant I: If Γ is a finitely
generated residually finite group and Γ∗ is a residual chain in Γ, then
I(Γn )
I(Γ,
Γ∗ ) := lim .
n→∞ [Γ : Γn ]
This raises the following questions:
• Does the limit exist?
• Does I(Γ,
Γ∗ ) depend on the residual chain Γ∗ of Γ ?
• Does I have a different interpretation?

3.3. Homological gradient invariants 33
3.3.1 Betti number gradients

For the gradient invariant associated to the ordinary Betti numbers, the ap-
proximation theorem (Theorem 3.1.3) gives a satisfying answer for finitely
presented residually finite groups (of finite type).
Caveat 3.3.1. There exist finitely generated residually finite groups Γ with a
residual chain (Γn )n∈N such that the limit limn→∞ b1 (Γn )/[Γ : Γn ] does not
exist [50].
In Chapter 5, we will discuss convergence of ordinary Betti numbers when
moving from residual chains to BS-convergent sequences. Generalisations of
the more classical version of the approximation theorem are surveyed in the
literature [106, 91].
For Fp -Betti number gradients the situation is much less understood. There
are known positive examples of convergence/independence, but good candi-
dates for alternative interpretations of the limits are rare.
3.3.2 Rank gradient

A non-commutative version of the first Betti number gradient is the rank
gradient, introduced by Lackenby [93]:
Definition 3.3.2 (rank gradient). Let Γ be a finitely generated infinite resid-
ually finite group.
• For a finitely generated group Λ, we write d(Λ) for the minimal size of
a generating set of Λ.
• If Γ∗ is a residual chain, then we define the rank gradient of Γ with
respect to Γ∗ by
d(Γn ) − 1
rg(Γ, Γ∗ ) := lim .
n→∞ [Γ : Γn ]
• Moreover, the (absolute) rank gradient of Γ is defined as
d(Λ) − 1
rg Γ := inf ,
Λ∈F(Γ) [Γ : Λ]
where F(Γ) denotes the set of all finite index subgroups of Γ.

Remark 3.3.3. If Γ is a finitely generated group and Λ ⊂ Γ is a finite index
subgroup, then the rank estimate of the Nielsen–Schreier theorem shows that
d(Λ) − 1 ≤ [Γ : Λ] · d(Γ) − 1 .

Hence, the limit in the definition of the rank gradient indeed exists (and is
equal to the infimum).
Remark 3.3.4 (rank gradient via normal subgroups). Let Γ be a finitely gener-
ated infinite group. Then
d(Λ) − 1 d(Λ) − 1
rg Γ = inf = inf ,
Λ∈F(Γ) [Γ : Λ] Λ∈NF(Γ) [Γ : Λ]
where NF(Γ) denotes the set of all finite index normal subgroups of Γ (because
every finite index subgroup of Γ contains a finite index subgroup that is
normal in Γ).
Corollary 3.3.5 (rank gradient estimate for the first L2 -Betti number). Let Γ
be a finitely presented infinite residually finite group (of finite type). Then
(2)
b1 (Γ) ≤ rg Γ.
Proof. If Λ is a finitely generated group, then there exists a classifying space

of Λ with d(Λ) one-dimensional cells. Therefore, b1 (Λ) ≤ d(Λ). Applying the
approximation theorem (Theorem 3.1.3) to a residual chain Γ∗ of Γ shows
that
(2) b1 (Γn ) d(Γn ) d(Γn ) − 1

b1 (Γ) = lim ≤ lim = lim = rg(Γ, Γ∗ ).
n→∞ [Γ : Γn ] n→∞ [Γ : Γn ] n→∞ [Γ : Γn ]
Taking the infimum over all residual chains, we obtain with Remark 3.3.3
and Remark 3.3.4 that
(2) d(Λ) − 1
b1 (Γ) ≤ inf = rg Γ.
Λ∈NF(Γ) [Γ : Λ]
(2)
Alternatively, one can also show directly that b1 (Λ) ≤ d(Λ) for all finitely
(2)
generated groups and then use multiplicativity of b1 under finite coverings
(Theorem 2.2.2).
It remains an open problem to determine whether the rank gradient de-
pends on the residual chain; in all known cases, the inequality in Corol-
lary 3.3.5 is an equality and the absolute rank gradient can be computed
by every residual chain. This is related to the fixed price problem (Out-
look 4.3.13).
The Betti number-rank estimate can be improved to estimates for the min-
imal size of normal generating sets in terms of the first L2 -Betti number; this
gives lower bounds on the girth of certain Cayley graphs [131, Theorem 5.1].
3.3.3 More gradients

Further examples of gradient invariants are:
• Homology log-torsion gradients (which conjecturally might be related
to L2 -torsion?!) [106].
• Simplicial volume gradients (Chapter 6).
3.E. Exercises 35
3.E Exercises
Exercise 3.E.1 (surface groups, free groups). Prove the approximation theo-
rem for surface groups and free groups by direct computation of the right-
hand side. Compute the (absolute) rank gradient of surface groups and free
groups.
Exercise 3.E.2 (Laplacian [105, Lemma 1.18]). Let Γ be a countable group, let
C∗ be a chain complex of Hilbert Γ-modules (with boundary operators ∂∗ ),
and let Δ∗ be the Laplacian of C∗ , i.e., for each n ∈ N, we set
∗
Δn := ∂n+1 ◦ ∂n+1 + ∂n∗ ◦ ∂n .
Show that there exists an isomorphism
ker Δn −→ ker ∂n

im ∂n+1
of Hilbert Γ-modules.
⊥
Hints. Consider the orthogonal projection onto im ∂n+1 .
Exercise 3.E.3 (weak convergence). Give an example of a sequence (μn )n∈N
of probability measures on [0, 1] (with the Borel σ-algebra) that weakly con-
verges to a probability measure μ on [0, 1], but that satisfies
lim μn {0} = μ {0} .

n→∞
Exercise 3.E.4 (rank gradients of products). Let Γ and Λ be finitely generated

infinite residually finite groups. Compute rg(Γ × Λ).
Exercise 3.E.5 (self-maps). Let M be an oriented closed connected aspherical
manifold with residually finite fundamental group Γ. Moreover, we suppose
that M admits a self-map f : M −→ M with |deg f | ≥ 2.
1. Give examples of this situation.
2. Show that rg(Γ) = 0.
(2)
3. Show that bk (Γ) = 0 for all k ∈ N [105, Theorem 14.40].
4. Challenge: Does the vanishing of L2 -Betti numbers of Γ also hold with-
out any residual finiteness or Hopficity condition on Γ ?! (This is an
open problem.)
Hints. Covering theory shows that im f has finite index in Γ. Moreover, it
is useful to know that residually finite groups are Hopfian, i.e., every self-
epimorphism is an automorphism.
4
The dynamical view:
Measured group theory
The theory of von Neumann algebras can be viewed as a model of non-com-

mutative measure theory. Therefore, it is plausible that L2 -Betti numbers
can be computed in terms of probability measure preserving actions.
• On the one hand, this leads to an additional way of computing L2 -Betti

numbers of groups.
• On the other hand, in this way, L2 -Betti numbers provide orbit equiv-
alence invariants.
We will first recall some basic terminology from measured group theory.
Then we will study L2 -Betti numbers of standard equivalence relations. More-
over, we will discuss cost and its relation with the first L2 -Betti number and
rank gradients.
4.1 Measured group theory 38

4.2 L2 -Betti numbers of equivalence relations 40
4.3 Cost of groups 47
4.E Exercises 57
Running example. amenable groups, free groups

38 4. The dynamical view: Measured group theory
4.1 Measured group theory

Measured group theory is the theory of dynamical systems, i.e., of (probabil-
ity) measure preserving actions of groups. We briefly introduce some of the
terminology. More information can be found in the literature [62, 65, 89, 90].
4.1.1 Standard actions
Definition 4.1.1 (standard action). Let Γ be a countable group.

• A standard action of Γ is an action of Γ on a standard Borel probability
space by measure preserving Borel automorphisms.
• A standard action Γ (X, μ) is essentially free if μ-almost every point

has trivial stabiliser group.
• A standard action Γ (X, μ) is ergodic if every measurable subset A ⊂

X with Γ · A = A satisfies μ(A) ∈ {0, 1}.
A standard Borel space is a measurable space that is isomorphic to a Pol-
ish space with its Borel σ-algebra. Standard Borel spaces form a convenient
category for measure theory [87].
Example 4.1.2 (Bernoulli shift). Let
Γ be a countable group. Then the shift
action of Γ on the product space Γ {0, 1} (with the product σ-algebra and
the product of the uniform distribution on {0, 1}) is a standard action of Γ.
Moreover, if Γ is infinite, this action is essentially free and ergodic [126,
Lemma 3.37][90, Chapter 2.3.1].
Example 4.1.3 (finite quotients). Let Γ be a countable group and let Λ ⊂ Γ
be a finite index subgroup. Then the translation action of Γ on the coset
space Γ/Λ (with the discrete σ-algebra and the uniform distribution) is a
standard action. It is ergodic, but apart from pathological cases, not essen-
tially free.
Example 4.1.4 (profinite completion). Let Γ be a finitely generated group. We
then consider the profinite completion of Γ, defined by
Γ
:= lim Γ/Λ,
←−
Λ∈NF(Γ)
where NF(Γ) denotes the set of all finite index normal subgroups of Γ. Then
is a group with the induced composition and the diagonal map Γ −→ Γ
Γ
is a group homomorphism, which leads to a Γ-action on Γ (by translation of

Another random document with
no related content on Scribd:
still is seen in Silenia. There the same muscular partition exists, but
the branchial lamellae on either side have disappeared, the slits
between the two chambers, which occur in Poromya, still persisting,
but separated into three groups. Cuspidaria represents the last stage
in the development. In the ventral chamber there appears nothing at
all corresponding to a branchia; the surface of the partition appears
perfectly uniform, but on careful examination three little separate
orifices, remains of the three groups of orifices in Silenia, are
observed.[276]
Relation between Branchiae and Heart.—The object of the
branchiae being, as has been already stated, to aerate the blood on
its way to the heart, we find that the heart and the branchiae stand in
very important structural relations to one another. When the
branchiae are in pairs, we find that the auricles of the heart are also
paired, the auricle on the right and left sides being supplied by the
right and left branchiae respectively. This is the case with the
Dibranchiate Cephalopods (Argonauta, Octopus, Loligo, etc.), the
Zygobranchiate Prosobranchs (Fissurella, Haliotis), and all
Pelecypoda. In the Amphineura (Chiton, etc.) there are two auricles
corresponding to the two sets of multiple branchiae. In the case of
the Tetrabranchiate Cephalopods (Nautilus) there are four auricles
corresponding to each of the four branchiae. Compare Fig. 79, A, B,
C, D, E.
On the other hand, when the branchia is single, or when both
branchiae are on the same side, and one is aborted and
functionless, the auricle is single too, and on the same side as the
branchia. This is the case with the Tectibranchiate Opisthobranchs
(Philine, Scaphander, etc.), all the Pectinibranchiate Prosobranchs
(Rachiglossa, Taenioglossa, and Ptenoglossa), and the other
Azygobranchiate Prosobranchs (Trochidae, Neritidae, etc.). In the
last case the right auricle exists, as well as the left, but is simply a
closed sac, the coalescing of the two gills on the left side having
thrown all the work upon the left auricle. Compare Fig. 79, F, G, H.
Fig. 79.—Diagram illustrating the relations between
branchiae, heart, and aorta in the Mollusca: A, In
Chiton; B, Pelecypoda; C, Dibranchiate
Cephalopoda; D, Tetrabranchiate Cephalopoda; E,
Prosobranchiata Zygobranchiata; F,
Prosobranchiata Azygobranchiata; G,
Prosobranchiata Monotocardia; H,
Opisthobranchiata Tectibranchiata: 1, Ventricle; 2,
Auricle; 3, Aorta; 3a, Cephalic aorta; 3b, Visceral
aorta; 3c, Posterior aorta. (From A. Lang.)
Circulatory System
All Mollusca, without exception, possess a circulatory system of
more or less complexity. The centre of the system is the heart, which
receives the aerated blood from the breathing organs, and propels it
to every part of the body. In the Scaphopoda alone there appears to
be no distinct heart.
The heart may consist simply of a single auricle and ventricle, and
an aorta opening out of the ventricle. From the aorta the blood is
conveyed to the various parts of the body by arteries. Veins convey
the blood back to the breathing organs, after passing over which it
returns by the branchial or pulmonary vein to the heart, thus
completing the circuit.
As regards position, the heart is situated within the pericardium, a
separate chamber which in the Pelecypoda, Cephalopoda, and the
bilaterally symmetrical Gasteropoda lies on the median line, while in
the asymmetrical Gasteropoda it is on one or other of the sides of
the body, usually the right. The veins connected with the branchiae,
and consequently the auricle into which they open, are situated
behind the ventricle in the Opisthobranchiata (whence their name),
while in the Prosobranchiata they are situated in front of the
ventricle.
The number of auricles corresponds to the number of branchiae.
Thus there is only one auricle in the great majority of
Prosobranchiata (which are accordingly classified as Monotocardia),
and also in the Opisthobranchiata, while the Pulmonata have a
single auricle corresponding to the pulmonary chamber. There are
two auricles in the Amphineura, in a small group of Gasteropoda,
hence known as Diotocardia, in all Pelecypoda, and in the
Dibranchiate Cephalopoda. In the Tetrabranchiate Cephalopoda
alone there are four auricles corresponding to the four branchiae.
A single aorta occurs only in the Amphineura and in the
Tetrabranchiate Cephalopoda. In all the other groups there are two
aortae, leading out of the anterior and posterior ends of the ventricle
in Pelecypoda and Dibranchiate Cephalopoda, while a single aorta
leads out of the posterior end alone, and subsequently bifurcates, in
most of the Gasteropoda. One aorta, the cephalic, supplies the front
part of the body, the oesophagus, stomach, mantle, etc.; the other,
the visceral aorta, supplies the posterior part, the liver and sexual
organs.
The general circulatory system in the Mollusca has not yet been
thoroughly investigated. As a general rule, the blood driven from the
ventricle through the aorta into the arteries, passes, on reaching the
alimentary canal and other adjacent organs, into a number of
irregular spaces called lacunae. These in their turn branch into
sinuses, or narrow tubes covered with muscular tissue, which
penetrate the body in every direction. In the Dibranchiate
Cephalopoda true capillaries are said to occur, which in some cases
form a direct communication between the arteries and veins.
According to some authorities[277] capillaries and veins exist in
certain Pelecypoda in connexion with the intestinal lacunae, but this
again is regarded by others as not established. A similar difference
of opinion occurs with regard to the precise function of the foot-pore
which occurs in many Mollusca, some holding that it serves as a
means for the introduction of water into the blood-vascular system,
while others regard it as a form of secretion gland, the original
purpose of which has perhaps become lost.
Blood.—As a rule, the blood of the Mollusca—i.e. not the
corpuscles but the liquor sanguinis—is colourless, or slightly tinged
with blue on exposure to the air. This is due to the presence of a
pigment termed haemocyanin, in which are found traces of copper
and iron, the former predominating. Haemoglobin, the colouring
matter of the blood in Vertebrates, is, according to Lankester,[278] of
very restricted occurrence. It is found—(1) in special corpuscles in
the blood of Solen legumen (and Arca Noae); (2) in the general
blood system of Planorbis; (3) in the muscles of the pharynx and
jaws of certain Gasteropoda, e.g. Limnaea, Paludina, Littorina,
Chiton, Aplysia. This distribution of haemoglobin is explained by
Lankester in reference to its chemical activity; whenever increased
facilities for oxidisation are required, then it may be present to do the
work. The Mollusca, being as a rule otiose, do not possess it
generally diffused in the blood, as do the Vertebrata. The actively
burrowing Solen possesses it, and perhaps its presence in Planorbis
is to be explained from its respiring the air of stagnant marshes. Its
occurrence in the pharyngeal muscles and jaws of other genera may
be due to the constant state of activity in which these organs are
kept.[279]
According to Tenison-Woods[280] a species of Arca (trapezia
Desh.) and two species of Solen, all Australian, have red blood. It is
suggested that in these cases the habits of the animal (the Solen
burrowing deeply in sand, the Arca in mud) require some highly
oxidising element, surrounded as the creature is by ooze. In Arca
pexata (N. America) the blood is red, the animal being familiarly
known as the ‘bloody clam.’ Burrowing species, however, are not all
distinguished by this peculiarity. Tenison-Woods finds red fluids in
the buccal mass of many Gasteropoda, e.g. in species of Patella,
Acmaea, Littorina, Trochus, Turbo, giving the parts the appearance
of raw meat.
The Mantle
On the dorsal side of the typical molluscan body, between the
visceral sac and the shell, lies a duplicature of the integument,
generally known as the mantle. The depending sides of the mantle,
which are usually somewhat thickened, enclose between themselves
and the body mass a chamber of varying size and shape, called the
mantle cavity, which communicates freely with the external air or
water, and encloses and furnishes a protection for the organ or
organs of respiration. On its upper or dorsal surface the mantle is
closely applied to the shell throughout its whole extent, the cells with
which it is furnished secreting the materials from which the shell is
formed (see p. 255). The whole mantle is capable, to some degree,
of secreting shelly matter, but the most active agent in its production
is the mantle edge or margin.
In the Prosobranchiata the mantle cavity, for reasons which have
already been explained, is found on the left side of the animal, its
front portion being in many cases produced into a tubular siphon.
Within the mantle cavity are found, besides the branchia, the anus,
the apertures of the kidneys, and the osphradium. In the pulmonata
the mantle fold encloses a so-called lung-cavity. The front edge of
the mantle coalesces with the integument of the neck in such a way
as to enclose the cavity very completely, the only communication
with the outer air being by means of the contractile breathing or
pulmonary aperture on the right side. In the Tectibranchiate
Opisthobranchs the mantle fold is inconsiderable, and is usually not
of sufficient extent to cover the branchia, while in the Nudibranchs,
which have no true branchiae, it disappears altogether.
In the Pelecypoda the mantle cavity is equally developed on each
side, enclosing the two sets of branchiae. The mantle may thus be
regarded as consisting of two equal portions, which form a sort of
lining to the two valves. The lower or ventral portion of the mantle
edges may be simple, or provided with ocelli (Pecten, Arca),
tentacles, cilia (Lima, Lepton), or doubled folds. The two portions of
the mantle touch one another along the whole line of the edge of the
two valves, and, although thus in contact, may remain completely
separate from one another, or else become permanently united at
one or more points. This fusion of the mantle edges corresponds to
important changes in the organisation of the animal as a whole. The
anal and branchial siphons are no more than prolongations of the
mantle edges on the posterior side into a tubular form. These
‘siphons’ exhibit the siphonal form more distinctly according as the
adjacent portions of the mantle become more definitely fused
together.
Fig. 80.—Diagram illustrating the various stages in the closing of the mantle in
Pelecypoda: A, mantle completely open; B, rudiments of siphons, mantle still
completely open; C, mantle closed at one point; D, mantle closed at two
points, with complete formation of siphonal apertures; E, development of
siphons, ventral closure more extended; F, mantle closed at three points, with
fourth orifice: f, foot; s.a, s.b, anal and branchial siphons; 1, 2, 3, first, second,
and third points of closure of mantle. (After A. Lang.)
This progressive fusion of the mantle edges may be taken as
indicating definite stages in the development of the Pelecypoda. A
perfectly free mantle edge, joined at no point with the edge of the
adjacent mantle, occurs in Nucula, Arca, Anomia, and Trigonia (see
Fig. 80, A, B). Here there is nothing in the nature of a siphon, either
anal or branchial; in other words, no contrivance exists to prevent the
spent water which has passed over the branchiae from becoming
mixed with the fresh water which is to reach them. When the mantle
edges are fused at one point only, this is invariably on the middle
part of the posterior side, thus separating off an anal opening which
may become prolonged into a tube-like form. At the same time the
adjacent underlying portions of the mantle edges draw together,
without actually coalescing, to form an opening for the incurrent
stream of water, the rudiments of the ‘branchial siphon’ (Fig. 80, C).
This is the case with most Mytilidae (see Fig. 75) with Cardita,
Astarte, and Pisidium. In the next stage the branchial opening is
separated off by the concrescence of the mantle edges beneath it,
and we have the mantle united in two places, thus forming three
openings, the ventral of which is the opening for the protrusion of the
foot (Fig. 80, D). This is the case in Yoldia, Leda, the majority of the
Eulamellibranchiata (e.g. Lucina, Cyrena, Donax, Psammobia,
Tellina, Venus, Cardium, Mactra), and all Septibranchiata. In Chama
and Tridacna the fused portions of the mantle become more
extended, and in Pholas, Xylophaga, Teredo, Pandora, and Lyonsia
this concrescence takes place over the greater length of the whole
mantle edge, so that the mantle may be regarded as closed, with the
exception of the three apertures for the foot and the two siphons
(Fig. 80, E).
In certain genera there occurs, besides these three apertures, a
fourth, in the line of junction between the pedal and branchial
orifices. It appears probable that this fourth orifice (which has been
regarded by some as an inlet for water when the siphons are
retracted), stands in relation to the byssal apparatus (Fig. 80, F). In
Lyonsia, for instance, a thick byssus protrudes through the orifice,
which is large and open. In Solen, Lutraria, Glycimeris,
Cochlodesma, Thracia, Aspergillum, and a few more genera, which
have no byssus, the orifice is very small and narrow. It is possible
that in these latter cases, the byssal apparatus having become
atrophied, the orifice has been correspondingly reduced in size.[281]
Mantle Reflected over the Shell.—It is sometimes the case that
the mantle edges tend to double back over the external surface of
the shell, and to enclose it to a greater or less extent. When this
process is carried to an extreme, the edges of the reflected mantle
unite, and the shell becomes completely internal. We see an
incipient stage of this process in Cypraea and Marginella, where the
bright polish on the surface of the shell is due to the protection
afforded by the lobes of the mantle. A considerable portion of the
shell of Scutus is concealed in a similar way, while in Cryptochiton,
Lamellaria, and Aplysia the shell is more or less completely
enclosed. Among Pulmonata, it is possible that in forms like Vitrina,
Parmacella, Limax, and Arion, we have successive stages in a
process which starts with a shell completely external, as in Helix,
and ends, not merely by enveloping the shell in the mantle, but by
effecting its disappearance altogether. In Vitrina and some allied
genera we have a type in which the mantle lobes are partly reflected
over the shell, which at the same time exhibits rather less of a spiral
form than in Helix. In the stage represented by Parmacella, the
mantle edges have coalesced over the whole of the shell, except for
a small aperture immediately over the spire; the nucleus alone of the
shell is spiral, the rest considerably flattened. In Limax the shell has
become completely internal, and is simply a flat and very thin plate,
the spiral form being entirely lost, and the nucleus represented by a
simple thickening at one end of the plate. In Arion, the final stage, we
find that the shell, being no longer needed as a protection to the vital
organs, has either become resolved into a number of independent
granules, or else has entirely disappeared.
Some indications of a similar series of changes occur in the
Pelecypoda. The mantle edge of Lepton is prolonged beyond the
area of the valves, terminating in some cases in a number of
filaments. In Galeomma and Scintilla the valves are partially
concealed by the reflected mantle lobes, and in a remarkable form
recently discovered by Dall[282] (Chlamydoconcha) the shell is
completely imbedded in the mantle, which is perforated at the
anterior end by an orifice for the mouth, and at the posterior end by a
similar orifice for the anus. In all these cases, except Lepton, it is
interesting to notice that the hinge teeth have completely
disappeared, the additional closing power gained by the external
mantle rendering the work done by a hinge unnecessary. It is quite
possible, on the analogy of the Gasteropoda mentioned above, and
also, it may be added, of the Cephalopoda and other groups, that we
have here indicated the eventual occurrence of a type of Pelecypoda
altogether deprived of valves, a greatly thickened mantle performing
the part of a shell.[283]
The following works will be found useful for further study of this
portion of the subject:—
F. Bernard, Recherches sur les organes palléaux des
Gastéropodes prosobranches: Ann. Sc. Nat. Zool. (7) ix. (1890), pp.
89–404.
G. Cuvier, Le Régne animal (ed. V. Masson); Mollusca, Text and
Atlas.
C. Grobben, Beiträge zur Kenntniss des Baues von Cuspidaria
(Neaera) cuspidata Olivi, nebst Betrachtungen über das System der
Lamellibranchiaten: Arb. Zool. Inst. Wien, x. (1893), pp. 101–146.
E. Ray Lankester, Encyclopaedia Britannica, 9th ed., vol. xvi.
(1883), Art. ‘Mollusca.’
A. Ménégaux, Recherches sur la circulation des Lamellibranches
marins: Besançon, 1890.
K. Mitsukuri, On the structure and significance of some aberrant
forms of Lamellibranchiate gills: Q. Journ. Micr. Sc., N.S. xxi. (1881),
pp. 595–608.
H. L. Osborn, On the gill in some forms of Prosobranchiate
Mollusca: Stud. Biol. Lab. Johns Hopk. Univ. iii. (1884), pp. 37–48.
R. Holman Peck, The structure of the Lamellibranchiate gill: Q.
Journ. Micr. Sc., N.S. xvii. (1877), pp. 43–66.
P. Pelseneer, Contributions à l’étude des lamellibranches: Arch.
Biol. xi. (1891), pp. 147–312.
CHAPTER VII
ORGANS OF SENSE: TOUCH, SIGHT, SMELL, HEARING—THE FOOT—THE
NERVOUS SYSTEM
Organs of Sense: I. Touch

Tactile organs, although occurring in some of the Mollusca, do not
appear to attain special or marked development, except in a few
cases. The whole surface of the skin, and particularly of the foot, is
very sensitive to the slightest impression. Nearly all Gasteropoda are
furnished with at least two cephalic tentacles, projecting like horns
from each side of the fore part of the head. At or near the base of
these are generally situated the eyes. In the Helicidae the eyes are
situated, not at the base, but at the apex of the tentacles, and in that
case—except in Vertigo—a second pair of shorter tentacles appears
beneath the longer pair. It frequently happens that several senses
are centred in a single organ; thus the upper tentacles of snails not
only carry the eyes and serve to a certain extent as tactile organs,
but they also carry the organs of smell.
The edges of the mantle, which are sometimes specialised into
lobes, appear to be keenly sensitive to touch in all Gasteropoda.
In Cypraea (Fig. 81) these lobes, or tentaculae, are a prominent
feature of the animal, and also in certain genera of the Trochidae
(Fig. 82). In most of the carnivorous land Pulmonata—e.g.
Testacella, Rhytida, Ennea—there are developed, under the lower
pair of tentacles, and close to the mouth, large labial palps or feelers.
These are connected with the cerebral ganglion by a very large
nerve, and may therefore be supposed to be of extreme
sensitiveness. In some of the large carnivorous forms (Glandina,
Aerope, compare Fig. 21, p. 54) these palpae are of great size, and
curl upwards like an enormous pair of moustaches. When a
Glandina seizes its prey, the palpae (see Fig. 83) appear to enfold it
and draw it in towards the mouth.
Fig. 81.—Cypraea moneta L., showing tentaculae
at edge of mantle, which partly envelops the
shell: Si, siphon; M, M, mantle; F, foot; T´,
tentaculae at edge of mantle. (After Quoy and
Gaimard.) × 3/2.
Fig. 82.—Monodonta
canalifera Lam., New
Ireland, showing mantle
lobes. (After Quoy and
Gaimard.)
Fig. 83.—Glandina seizing
its prey, with buccal
papillae turned back.
(Strebel.)
It is in the Opisthobranchiata that the organs of touch attain their
maximum development. Many of this group are shell-less or possess
a small internal shell, and accordingly, in the absence of this special
form of defence, a multiplied sense of touch is probably of great
service. Thus we find, besides the ordinary cephalic tentacles,
clusters or crowns of the same above the head of many
Nudibranchiata, with lobe-like prolongations of the integument, and
tentacular processes in the neighbourhood of, or surrounding the
branchiae (see Figs. 58 and Fig. 84, or even projecting from the
whole upper surface of the body (Fig. 5, C).
In the Pelecypoda, the chief organs of touch are the foot, which is
always remarkably sensitive, especially towards its point, the labial
palps on each side of the mouth, and the siphons. In certain cases
the mantle border is prolonged into a series of threads or filaments.
These are particularly noticeable in Pecten, Lepton, and Lima (Fig.
85), the mantle lobes of the common L. hians of our own coasts
being very numerous, and of a bright orange colour. In many genera
—e.g. Unio, Mactra—this sensibility to touch appears to be shared
by the whole mantle border, although it is not furnished with any
special fringing. The ‘arms’ of the Cephalopoda appear to be keenly
sensitive to touch, and this is particularly the case with the front or
tentacular pair of arms, which seem to be employed in an especial
degree for exploration and investigation of strange objects.
Fig. 84.—Idalia Leachii A. and H., British seas;
br, branchiae. (After Alder and Hancock.)
Fig. 85.—Lima squamosa Lam.,

Naples, showing tentacular lobes
of mantle (t, t); a, anus; ad.m,
adductor muscle; br, br,
branchiae; f, foot; sh, shell.
Taste.—The sense of taste is no doubt present, to a greater or
less extent, in all the head-bearing Mollusca. In many of these a
special nerve or nerves has been discovered in the pharynx,
connecting with the cerebral ganglion; this no doubt indicates the
seat of the faculty of taste. The Mollusca vary greatly in their likings
for different kinds of food. Some seem to prefer decaying and highly
odoriferous animal matter (Buccinum, Nassa), others apparently
confine themselves to fresh meat (Purpura, Natica, Testacella),
others again, although naturally vegetarian, will not refuse flesh on
occasion (Limax, Helix).
Mr. W. A. Gain[284] has made some interesting experiments on
the taste of British land Mollusca, as evidenced by the acceptance or
rejection of various kinds of food. He kept twelve species of Arion
and Limax, and eight species of Helix in captivity for many months,
and tried them with no less than 197 different kinds of food,
cannibalism included. Some curious points came out in his table of
results. Amalia gagates appears to be surprisingly omnivorous, for
out of 197 kinds of food it ate all but 25; Arion ater came next, eating
all but 40. Limax arborum, on the other hand, was dainty to a fault,
eating only seven kinds of food, and actually refusing Swedes, which
every other species took with some avidity. Certain food was
rejected by all alike, e.g. London Pride, Dog Rose, Beech and
Chestnut leaves, Spruce Fir, Common Rush, Liverwort, and Lichens;
while all, or nearly all, ate greedily of Potatoes, Turnips, Swedes,
Lettuces, Leeks, Strawberries, Boletus edulis, and common grasses.
Few of our common weeds or hedgerow flowers were altogether
rejected. Arion and Limax were decidedly less particular in their food
than Helix, nearly all of them eating earthworms and puff-balls, which
no Helix would touch. Arion ater and Limax maximus ate the slime
off one another, and portions of skin. Cyclostoma elegans and
Hyalinia nitida preferred moist dead leaves to anything else.
II. Sight
Position of Eyes.—In the majority of the head-bearing Mollusca

the eyes are two in number, and are placed on, or in the immediate
neighbourhood of the head. Sometimes they are carried on
projecting tentacles or ‘ommatophores,’ which are either simple (as
in Prosobranchiata) or capable of retraction like the fingers of a glove
(Helix, etc.). Sometimes, as in a large number of the marine
Gasteropoda, the eyes are at the outer base of the cephalic
tentacles, or are mounted on the tentacles themselves, but never at
the tip (compare Fig. 60, p. 153 and Fig. 98, p. 199). In other cases
they are placed somewhat farther back, at the sides of the neck. The
Pulmonata are usually subdivided into two great groups,
Stylommatophora and Basommatophora (Fig. 86), according as the
eyes are carried on the tip of the large tentacles (Helix, and all non-
operculate land shells), or placed at the inner side of their base
(Limnaea, Physa, etc.). In land and fresh-water operculates, the
eyes are situated at the outer base of the tentacles.
Fig. 86.—A, Limnaea peregra Müll.; e, e, eyes; t, t, tentacles; B,

Helix nemoralis Müll.; e, e, eyes; t, t, tentacles; p.o,
pulmonary orifice.
In the Helicidae, careful observation will show that the eyes are
not placed exactly in the centre of the end of the tentacle, but on its
upper side, inclining slightly outwards. The eye is probably pushed
on one side, as it were, by the development of the neighbouring
olfactory bulb. The sense of smell being far more important to these
animals than the sense of sight, the former sense develops at the
expense of the latter.
Organisation of the Molluscan Eye.—The eye in Mollusca
exhibits almost every imaginable form, from the extremely simple to
the elaborately complex. It may be, as in certain bivalves, no more
than a pigmented spot on the mantle, or it may consist, as in some of
the Cephalopoda, of a cornea, a sclerotic, a choroid, an iris, a lens,
an aqueous and vitreous humour, a retina, and an optic nerve, or of
some of these parts only.
In most land and fresh-water Mollusca the eye may be regarded,
roughly speaking, as a ball connected by an exceedingly fine thread
(the optic nerve) with a nerve centre (the cerebral ganglion). In
Paludina this ball is elliptic, in Planorbis and Neritina it is drawn out
at the back into a conical or pear shape. In Helix (Fig. 87) there is a
structureless membrane, surrounding the whole eye, a lens, and a
retina, the latter consisting of a nervous layer, a cellular layer, and a
layer of rods containing pigment, this innermost layer (that nearest
the lens) being of the thickness of half the whole retina.
Fig. 87.—Eye of Helix pomatia

L., retracted within the
tentacle; c, cornea; ep,
epithelial layer; l, lens; op.n,
optic nerve; r, retina. (After
Simroth.)
Comparing the eyes of different Gasteropoda together, we find
that they represent stages in a general course of development. Thus
in Patella the eye is scarcely more than an invagination or
depression in the integument, which is lined with pigmented and
retinal cells. The next upward stage occurs in Trochus, where the
depression becomes deeper and bladder-shaped, and is filled with a
gelatinous or crystalline mass, but still is open at the top, and
therefore permits the eye to be bathed in water. Then, as in Turbo,
the bladder becomes closed by a thin epithelial layer, which finally,
as in some Murex, becomes much thicker, while the ‘eyeball’
encloses a lens (Fig. 88), which probably corresponds with the
‘vitreous humour’ of other types.
Fig. 88.—Eyes of Gasteropoda, showing arrest of

development at successive stages: A, Patella; B,
Trochus; C, Turbo; D, Murex; ep, epidermis; l, lens;
op.n, optic nerve; r, retina; v.h, vitreous humour.
(After Hilger.)
In Nautilus the eye is of a very simple type. It consists of a cup-
shaped depression, with a small opening which is not quite closed by
the integument. The retina consists of cells which line the interior of
the depression, and which communicate directly with the branches of
the optic nerve, there being no iris or lens. This type of eye, it will be
observed, corresponds exactly with that which occurs in Patella. It
appears also to correspond to a stage in the development of eyes in
the Dibranchiata (e.g. Octopus, Sepia, Loligo). Lankester has
shown[285] that in Loligo the eye first appears as a ridge, enclosing
an oval area in the integument. By degrees the walls of this area
close in, and eventually join, enclosing the retinal cells within the
chamber in which the lens is afterwards developed (Fig. 89). It thus
appears that in some cases the development of the eye is arrested
at a point which in other cases only forms a temporary stage towards
a higher type of organisation.
Fig. 89.—Three stages in the development of

the eye of Loligo; r, r, ridge, enclosing
p.o.c, primitive optic chamber; or, orifice
between the closing ridges; s.o.c,
secondary optic chamber; ci, ci, ciliary
body; l, rudimentary lens; R, retina. (After
Lankester.)
Fig. 90.—Eye in A, Loligo; B, Helix or Limax; C, Nautilus: a.o.c, anterior optic
chamber; c, cornea; int, integument; ir, iris; l, lens; l´, external portion of
lens; op.n, optic nerve; op.g, optic ganglion; p.o.c, posterior optic
chamber; r, retina. (After Grenacher.)
The developed eye in the dibranchiate Cephalopods consists of a
transparent cornea, which may or may not be closed over the front of
the lens. Behind the cornea is a narrow chamber (the anterior optic
chamber) which is continued for three parts round the whole circle of
the eye, and into which project the front portion of the lens and the
folds of the iris. Throughout its whole extent, the anterior optic
chamber is lined by the integument, the portion of which on the inner
side is the choroid. The lens is divided into an outer and inner
segment by a thin membrane, and is supported by the ciliary body
which forms a continuation of the retina. The main portion of the lens
lies within the posterior optic chamber, at the back and sides of
which is found the retina (Grenacher).
There can be no doubt that the Cephalopoda use their eyes to
observe, but there is nothing to show that any other Mollusca use
their eyes for this purpose, the sense of smell in their case largely
taking the place of visual observation. Madame Jeannette Power
once saw[286] the Octopus in her aquarium holding a fragment of
rock in one of its arms, and watching a Pinna which was opening its
valves. As soon as they were perfectly open, the Poulpe, with
incredible address and promptitude, placed the stone between the
valves, preventing the Pinna from closing again, upon which it set
about devouring its victim. The next day the Poulpe was seen, after
crushing some Tellina, to stretch himself down close by a Triton
nodiferus, and watch it attentively. After four hours the Triton
emerged from its shell, when the Octopus sprang upon it, and
surrounded it with its arms.
Powers of Vision in Land Mollusca.—The Helicidae are
undoubtedly very short-sighted. Seldom emerging from their retreats
except in twilight and darkness, they are naturally myopic, and see
better in a subdued than in a bright light. Experiment has shown that
a Helix can perceive an object better at 6 centimetres distance in a
weak light than at 4 or 5 millimetres in a strong one. Cyclostoma
elegans and Paludina vivipara are comparatively long-sighted,
perceiving objects at a distance of 20 to 30 centimetres.[287] The
increased power of vision is due, in these two cases, to increased
elaboration in the construction of the eye, Paludina possessing a
large and almost spherical lens, to which the vitreous humour closely
adheres, while in Cyclostoma the lens is remarkably hard, and the
aqueous humour very abundant. According to V. Willem,[288] the
Pulmonata are very sensitive to the slightest movement of the air or
jarring of the surface on which they crawl, but are so short-sighted as
only to perceive a confused image of a large object at about 1 cm.,
and to distinguish the form of objects at not more than 1 or 2 mm.
The senses of touch and smell are far more active than that of sight.
A bean-pod enclosed in a narrow glass case and placed before a
hungry snail was not noticed, but when taken out of the case and
placed 8 cm. behind the snail, the latter at once turned towards it to
devour it.
Some interesting experiments were conducted by the same
author with the view of ascertaining whether snails avoid or court the
light. He placed a number of species in different wooden boxes,
which were divided into a light and a dark compartment, having
previously well soaked the boxes in water to secure a humid
atmosphere and surface, and so induce the snails to move about.

Ergodic Theoretic Methods in Group Homology A Minicourse On L2 Betti Numbers in Group Theory Clara Loh Online Ebook Texxtbook Full Chapter PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Ergodic Theoretic Methods in Group Homology A Minicourse On L2 Betti Numbers in Group Theory Clara Loh Online Ebook Texxtbook Full Chapter PDF

Uploaded by

Copyright:

Available Formats

Ergodic Theoretic Methods in Group

Homology A Minicourse on L2 Betti

A First Course in Group Theory 1st Edition Bijan Davvaz

Group A Streptococcus Methods and Protocols Thomas

Symmetry and Group Theory in Chemistry Second Edition

On Group Analysis and Beyond Group Analysis as Meta

Group Theory 1st Edition Dinesh Khattar

Observability A New Theory Based on the Group of

Drugs in Sport Taylor & Francis Group

Group Therapy A group analytic approach 1st Edition

More information about this series at http://www.springer.com/series/10030

Ergodic Theoretic Methods

ISSN 2191-8198 ISSN 2191-8201 (electronic)

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2020

1 The von Neumann dimension 5

3 The residually ﬁnite view: Approximation 27

3.3 Homological gradient invariants 32

4 The dynamical view: Measured group theory 37

5 Invariant random subgroups 59

• On the one hand, (co)homology with coeﬃcients based on dynamical

• On the other hand, many homological gradient invariants can be ac-

Overview of this minicourse

• Chapter 1. We will start our gentle introduction to L2 -Betti numbers by

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 1

• Chapter 2. The von Neumann dimension allows us to deﬁne L2 -Betti

• Chapter 3. On the one hand, L2 -Betti numbers are related to classical

• Chapter 5. Via invariant random subgroups, ergodic theory also gives a

• Chapter 6. Finally, we will use the dynamical approach to L2 -Betti

None of the material in this book is original; in particular, Chapters 1–3

Acknowledgements. I would like to thank Alex Furman and Tsachik Ge-

Regensburg, February 2020 Clara Löh

Betti numbers are dimensions of (co)homology groups. In the presence of a

Overview of this chapter.

Running example. the additive group Z, ﬁnite (index sub)groups

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 5

1.1 From the group ring

1.1.1 The group ring

• Let n ∈ N>0 . Then we have C[Z/n] ∼

• In general, group rings are not commutative. In fact, a group ring CΓ

• 4 times the group element 2.

Proof. This is a straightforward calculation (check!).

Outlook 1.1.5 (Kaplansky conjecture). The ring structure of group rings is

• Then the group ring CΓ is a domain (?!).

• The group ring CΓ does not contain non-trivial idempotents (?!).

1.1.2 Hilbert modules

is an inner product on CΓ. The completion of CΓ with respect to this inner

is an isomorphism of C-algebras (with inner product).

• Let V and W be Hilbert Γ-modules. A morphism V −→ W of Hilbert

C-dimension of A = trace of the orthogonal projection onto A.

1.1.3 The group von Neumann algebra

where e ∈ CΓ ⊂ 2 Γ denotes the atomic function at e ∈ Γ. Then trΓ satisﬁes

1. Trace property. For all a, b ∈ N Γ, we have trΓ (a ◦ b) = trΓ (b ◦ a).

2. Faithfulness. For all a ∈ N Γ, we have trΓ (a∗ ◦ a) = 0 if and only

3. Positivity. For all a ∈ N Γ with a ≥ 0, we have trΓ a ≥ 0. Here, a ≥ 0

Proof. Ad 1. A straightforward computation shows that the trace property

0 = trΓ (a∗ ◦ a) = e, a∗ ◦ a(e) = a(e), a(e)

and thus a(e) = 0. Because a is Γ-linear, we also obtain a(g · e) = g · a(e) = 0.

Example 1.1.13 (some von Neumann traces).

• If Γ is a ﬁnite group, then N Γ = CΓ. The von Neumann trace is

• If Γ = Z = t | , then we obtain [105, Example 1.4]: The group

Remark 1.1.14 (extension of the trace to matrices and morphisms). As in linear

1.2 The von Neumann dimension

where i : V −→ (2 Γ)n (for some n ∈ N) is an isometric Γ-embedding and

0 = trΓ (a∗ ◦ a) = e, a∗ ◦ a(e) = a(e), a(e)

• If Γ = Z = t | , then we obtain [105, Example 1.4]: The group

4. Additivity. Let 0 / V i / V π / V / 0 be a weakly ex-

dimN Γ V = dimN Γ V + dimN Γ V .

dimΓ V = dimΓ (im i ⊕ V ) = dimΓ im i + dimΓ V = dimΓ V + dimΓ V .

because f induces a CΓ-chain homotopy equivalence C∗ (X) CΓ C∗ (Y ) and