Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

i i

“Forum Mathematicum, Verlag Walter de Gruyter GmbH & Co. KG” — 2010/2/12 — 13:42 — page 241 — #1

i i

Forum Math. 22 (2010), 241–253 Forum


DOI 10.1515/FORUM.2010.13 Mathematicum
© de Gruyter 2010

Lie theoretic significance of the measure topologies


associated with a finite trace
Daniel Beltiţă
(Communicated by Karl Strambach)

Abstract. Let M be a W ∗ -algebra with a faithful normal finite trace and think of the unitary group UM
endowed with the ultraweak topology. By using the measure topology associated with the aforemen-
tioned trace, we prove that UM is a topological group with Lie algebra in the sense of K.H. Hofmann
and S.A. Morris (2005). We discuss the corresponding exponential map and find that it is never locally
injective in the case of factors of type II1 .
2000 Mathematics Subject Classification: 22A10; 46L51, 46H70, 22E65.

1 Introduction
The category of topological groups with Lie algebras —set forth in the paper [HM05] and in
the monograph [HM07]— provides a framework where the relationship between Lie groups
and algebras carries over in a remarkable way. We are going to prove that in every von
Neumann algebra with a finite trace the unitary group endowed with the ultraweak topology
is a topological group with Lie algebra. The corresponding topological Lie algebra can
be identified with the Lie algebra of all skew-adjoint measurable operators endowed with
the measure topology associated with the finite trace under consideration; see Theorem 3.6
below. This fact can turn out important from the standpoint of the endeavor to understand
to what extent Lie theory extends beyond the classical setting of Lie groups; see [HM05],
[Ne06], and [HN08], and also [CY74], [Gl02], and [LR04] for details in this connection.
It is well known that the unitary groups of C ∗ -algebras endowed with the norm topology
are Banach-Lie groups, and this leads to several geometric features of operator algebras;
see for instance [Up85], [BR07], and [Bel06]. In the present paper we point out that a
weaker topology still makes many of these unitary groups into topological groups with Lie
algebras. However, by investigating their exponential maps we find a rather pathological
behavior: they are never locally injective in the case of von Neumann factors of type II1
(Corollary 4.4). To make things worse, the corresponding topological Lie algebras are never
locally convex (Remark 2.6). Nevertheless, the situation could have been even worse since
not every topological group has a Lie algebra: the unitary group on an infinite-dimensional
Hilbert space fails badly to be a topological group with Lie algebra when it is endowed with
the strong operator topology or, which is the same, with the ultraweak one; see Example 2.1
Brought to you by | Instytut Matematyczny Pan
Authenticated
Download Date | 9/30/16 6:25 PM

i i

i i
i i
“Forum Mathematicum, Verlag Walter de Gruyter GmbH & Co. KG” — 2010/2/12 — 13:42 — page 242 — #2

i i

242 D. Beltiţă

below for some more details. It would be interesting to determine precisely for which W ∗ -
algebras it is true that the unitary group endowed with the ultraweak (or w∗ -)topology is a
topological group with Lie algebra.
On the other hand, let us mention that the present research is related to an old theme of
the theory of operator algebras, namely that some of the basic properties of von Neumann
algebras are encoded in properties of their unitary groups thought of as topological groups
with respect to the ultraweak topology. See for instance the results obtained in the classical
papers [Dy53] and [dlH79] and Corollary 3.8 below.
The structure of our paper is as follows: In §2 we recall some basic notions and terminol-
ogy necessary for the presentation to be reasonably self-contained. The main topics are the
topological groups with Lie algebras and the theory of noncommutative integration with re-
spect to a semifinite trace. In §3 we obtain our main result on Lie algebras of unitary groups
thought of as topological groups with respect to the ultraweak topology (Theorem 3.6). Fi-
nally, in §4 we investigate continuity properties and injectivity properties of the exponential
maps of unitary groups.

2 Preliminaries
Topological groups with Lie algebras
We shall briefly recall the notion of topological group with Lie algebra in the sense of the
paper [HM05]; see also Section IV in [Ne06].
Let G be a topological group with the set of neighborhoods of 1 ∈ G denoted by VG (1).
Define

L(G) = {X : R → G continuous | (∀t, s ∈ R) X (t + s) = X (t)X (s)}

and for every n ∈ N and U ∈ VG (1) denote

W n,U = {(X , Y) ∈ L(G) × L(G) | (∀t ∈ [−n, n]) Y(t)X (t)−1 ∈ U}.

Then {W n,U | n ∈ N, U ∈ VG (1)} is a set of so-called entourages of the diagonal in


L(G) × L(G) defining a uniform structure on the set L(G) (see e.g., Chapter 6 in [Kel57]).
We shall think of L(G) endowed with the corresponding topology, which is exactly the
topology of uniform convergence on the compact subsets of R.
We say that G is a topological group with Lie algebra if L(G) has a structure of topological
Lie algebra (over R) whose scalar multiplication, vector addition and bracket satisfy the
following conditions for all t ∈ R and X , Y ∈ L(G):

(2.1) (∀s ∈ R) (t · X )(s) = X (ts);

(2.2) (X + Y)(t) = lim (X (t/n)Y(t/n))n ;


n→∞

[X , Y](t 2 ) = lim (X (t/n)Y(t/n)X (−t/n)Y(−t/n))n .


2
(2.3)
n→∞
Brought to you by | Instytut Matematyczny Pan
Authenticated
Download Date | 9/30/16 6:25 PM

i i

i i
i i
“Forum Mathematicum, Verlag Walter de Gruyter GmbH & Co. KG” — 2010/2/12 — 13:42 — page 243 — #3

i i

Lie theoretic significance of the measure topologies 243

In addition, the mapping expG : L(G) → G, X → X (1) is called the exponential map
of G.
All Banach-Lie groups are topological groups with Lie algebras in the above sense. More
generally, this property is shared by every locally exponential Lie group modeled on a locally
convex space; see Lemma IV.1.17 and the comment after Definition IV.1.23 in [Ne06].
On the other hand, let us discuss an example of familiar topological group that fails to
have a Lie algebra; see also page 119 in [HM07].

Example 2.1. Let H be a complex infinite-dimensional separable Hilbert space with the
corresponding group of unitary operators denoted by U(H), and think of U(H) as a topo-
logical group with respect to the strong operator topology (see also Lemma 3.4(1) below).
Then U(H) is not a topological group with Lie algebra.
In fact, it follows by Stone’s theorem and perturbation theory for strongly continuous
1-parameter groups of unitary operators that L(U(H)) is homeomorphic to the set of all
(possibly unbounded) skew-adjoint operators in H endowed with the topology of strong re-
solvent convergence (see [Kat66]). Now Proposition 5.8 in [Ch74] shows a specific example
of X , Y ∈ L(U(H)) such that the limit in (2.2) does not exist for almost every t ∈ R. On the
other hand, it follows by Proposition 4.1 and Theorem 6.4 of [Ch74] that if X ∈ L(U(H))
has the property that the limit in (2.2) exists for every Y ∈ L(U(H)) uniformly on the
compact subsets of R, then X corresponds to a bounded operator, that is, X : R → U(H)
is continuous with respect to the norm operator topology. Moreover, the addition may not
be associative: Proposition 5.3 of [Ch74] shows that there exist X , Y, Z ∈ L(H) such that
all of the sums X + Y, (X + Y) + Z , Y + Z , X + (Y + Z ) ∈ L(U(H)) are well defined by
means of (2.2), and yet (X + Y) + Z = X + (Y + Z ). Finally, the addition need not be even
separately continuous: Proposition 5.5 in [Ch74] shows that there exist X ∈ L(U(H)) and a
sequence {Y n }n≥1 convergent to 0 ∈ L(U(H)) such that the sequence {X + Y n }n≥1 is well
defined and converges in L(U(H)) to some limit different from X .

Noncommutative integration. We refer to [Ta79], [SZ79], and [Da96] for general in-
formation on operator algebras, and to [Se53], [St59], [Nl74], [Te81], [FK86], [Ci90], and
[Ta03] for noncommutative integration in W ∗ -algebras with semifinite traces. Some inter-
esting recent results can be found in the papers [Bir04] and [Bir06].

Notation 2.2. Unless otherwise mentioned, we shall be working in the setting provided
by a W ∗ -algebra M endowed with a faithful normal semifinite tracial weight τ : M+ →
[0, ∞], where M+ = {x ∈ M | x ≥ 0}. That is, τ is a function with the following
properties:
• we have τ (λ1 x 1 +λ2 x 2 ) = λ1 τ (x 1 )+λ2 τ (x 2 ) whenever x 1 , x 2 ∈ M+ and λ1 , λ2 ∈ [0, ∞)
(with the convention 0 · ∞ = 0);
• the set pτ := {x ∈ M+ | τ (x) < ∞} spans a w∗ -dense linear subspace of M;
• for every x ∈ M+ \ {0} we have τ (x) > 0;
• we have τ (supi∈I x i ) = supi∈I τ (x i ) for every bounded increasing net {x i }i∈I in M+ ;
• for every x ∈ M we have τ (x ∗ x) = τ (xx ∗ ).
In addition, we shall need the following objects associated with the weight τ :
Brought to you by | Instytut Matematyczny Pan
Authenticated
Download Date | 9/30/16 6:25 PM

i i

i i
i i
“Forum Mathematicum, Verlag Walter de Gruyter GmbH & Co. KG” — 2010/2/12 — 13:42 — page 244 — #4

i i

244 D. Beltiţă

• nτ := {x ∈ M | x ∗ x ∈ pτ }, which is a two-sided self-adjoint ideal of M;


• mτ = span {y ∗ x | x, y ∈ nτ } ⊆ nτ , which is again a two-sided self-adjoint ideal of M
such that mτ ∩ M+ = pτ ;
• N τ := {x ∈ M | τ (x ∗ x) = 0} ⊆ nτ ;
• Hτ the Hilbert space obtained as the completion of nτ /N τ with respect to the scalar
product defined by
nτ /N τ × nτ /N τ → C, (x + N τ , y + N τ ) → τ (y ∗ x),
where we denote again by τ : mτ → C the unique linear functional that extends the
function τ |pτ ; one has τ (ax) = τ (xa) for all x ∈ M and a ∈ mτ ;
• ητ : nτ → nτ /N τ → Hτ , x → x + N τ , which is a linear map with Ran ητ dense in
Hτ ; actually even ητ (mτ ) is dense in Hτ ;
• πτ : M → B(Hτ ) the semi-cyclic representation of M associated with τ , which is a
faithful w∗ -continuous ∗-representation uniquely determined by the condition that for all
a ∈ M and x ∈ nτ we have π(a)η(x) = η(ax).
If τ is a finite trace in the sense that τ (1) < ∞, then the corresponding semi-cyclic rep-
resentation is just the Gelfand-Naı̆mark-Segal (GNS) representation associated with τ . We
refer to §1 in Chapter VII of [Ta03] and subsections C.7.3 and 10.14 in [SZ79] for more
details on these constructs.

We shall use the notation PM = {p ∈ M | p = p 2 = p ∗ } for the set of all projections in M.

Definition 2.3. Assume that M → B(H) is a faithful w∗ -continuous representation. A


closed densely-defined linear operator a : D(a) ⊆ H → H is said to be affiliated to the von
Neumann algebra M if for every y ∈ M we have ya ⊆ ay. The set of all operators affiliated
to M will be denoted by M. Moreover, the set of all τ -measurable operators is defined by

 = {a ∈ M | (∀ε > 0)(∃p ∈ PM ) Ran p ⊆ D(a) and τ (1 − p ) ≤ ε}.


M


See [SZ79], [Nl74], and [Te81] for spectral characterizations of the operators in M and M,
respectively.

Remark 2.4. Let M → B(H) be a faithful w∗ -continuous representation. The set of all
τ -measurable operators M is a ∗-algebra with respect to the strong operator sum, strong
operator product, and the adjoint operation. Moreover the function

 → [0, ∞),
ρτ : M ρτ (a) = inf max{ap , τ (1 − p )}
p ∈P M
Ran p ⊆D(a)

is an F-norm and the corresponding topology defined by the metric (a, b) → ρτ (a − b)


makes M into a complete metrizable topological ∗-algebra such that M is dense in M.
 See
[Se53], [St59], [Nl74], and Theorem 28 in [Te81] for more details.
Brought to you by | Instytut Matematyczny Pan
Authenticated
Download Date | 9/30/16 6:25 PM

i i

i i
i i
“Forum Mathematicum, Verlag Walter de Gruyter GmbH & Co. KG” — 2010/2/12 — 13:42 — page 245 — #5

i i

Lie theoretic significance of the measure topologies 245

Definition 2.5. In the setting of Remark 2.4, assume that a and aj for j ∈ J are τ -measurable
operators on H. We say that the net {aj }j ∈J converges in measure to the operator a if
 If this is the case, then we write aj −→ τ
limj ∈J aj = a in the topological algebra M. a.
j ∈J
Accordingly, the topology of the algebra M  of Remark 2.4 will be called the τ -measure
topology.
We shall say that the net {aj }j ∈J converges locally in measure to a if for every p ∈ PM
τ τ loc
with τ (p ) < ∞ we have aj p −→ ap , which we denote by aj −→ a. The corresponding
j ∈J j ∈J
topology on M  will be called the localized τ -measure topology.
Finally, the net {aj }j ∈J converges weakly locally in measure to a if for every p ∈ PM
τ τ weak
with τ (p ) < ∞ we have p aj p −→ p ap , which we denote by aj −→ a. The corresponding
j ∈J j ∈J
 is the weak localized τ -measure topology.
topology on M

Remark 2.6. If M = B(H) endowed with the usual trace τ = tr, then it is easy to see that
M = B(H) and in this case the τ -measure topology, the localized τ -measure topology, and
the weak localized τ -measure topology are equal to the norm topology, the strong operator
topology (so-topology), and the weak operator topology (wo-topology), respectively.
Nevertheless, the τ -measure topology of M  is fails to be locally convex in general (see
also Section 6 in [Ci90]). This follows by Exercise 24 in §5, Chapter IV of [Bo65] if
M = L ∞ ([0, 1]). More generally, let us assume that M is a von Neumann algebra of type
II1 (see subsection 4.16 in [SZ79]) with a finite trace τ . For arbitrary ε > 0 we shall prove
that the open ball V ε = {a ∈ M  | ρτ (a) < ε} has the property C1 ⊆ {a1 + · · · + an |
ε
a1 , . . . , a nε ∈ V ε } for a suitable integer nε ≥ 1; it then easily follows that the 1-dimensional
subspace C1 is contained in every convex neighborhood of 0 ∈ M,  and then the Hausdorff
topological vector space M  could not possibly be locally convex. Now, to prove the above
assertion, let ε > 0 arbitrary and pick an integer N ≥ 1 such that 1/2N < ε. Since M is
of type II1 , it follows by Proposition 4.11 in [SZ79] that there exist mutually orthogonal
projections e1 , . . . , e2N ∈ PM with e1 ∼ · · · ∼ e2N and e1 + · · · + e2N = 1. Then for every
α ∈ C we get α1 = a1 + · · · + a2N , where for j = 1, . . . , 2N we have aj := αej ∈ V ε .
(Indeed, for p := 1−ej we have (αej )p = 0 and τ (1−p ) = 1/2N < ε, hence ρτ (αej ) < ε.)
Thus the above assertion holds with nε = 2N .

3 Lie theory for unitary groups with the ultraweak topology


In this section we prove the main result of the present paper (Theorem 3.6). The technical
ingredients are a sort of perturbation properties of the measure topologies; see Corollary 3.3
and Corollary 3.7 below. Statements like that are well known in perturbation theory of linear
operators (see for instance [Kat66]), where they are proved by means of the integral formulas
relating the resolvent and the 1-parameter group generated by some operator. However, this
method of proof does not work in the case of (localized) measure topologies associated with
a trace. The reason for this is that the measure topologies are not locally convex in general
Brought to you by | Instytut Matematyczny Pan
Authenticated
Download Date | 9/30/16 6:25 PM

i i

i i
i i
“Forum Mathematicum, Verlag Walter de Gruyter GmbH & Co. KG” — 2010/2/12 — 13:42 — page 246 — #6

i i

246 D. Beltiţă

(Remark 2.6 above), and this leads to a pathological behavior of integrals of functions with
values in the space of measurable operators (see [PR66]). Therefore we shall need to set up
another method of proof, whose starting point is the following result.

Lemma 3.1. Let M be a W ∗ -algebra with a faithful normal semifinite tracial weight τ and
assume M → B(Hτ ) by means of the corresponding semi-cyclic representation. Then the
following assertions hold:
(1) The so-topology is stronger than the localized τ -measure topology, and these two topolo-
gies coincide on every norm-bounded subset of M.
(2) If the trace τ is finite, then the τ -measure topology and the so-topology coincide on the
norm-bounded subsets of M.

Proof. Let x ∈ M and {x j }j ∈J an arbitrary net in M.



so
If x j −→ x, then for every h ∈ Hτ we have limj ∈J (x j − x)h2 = 0. In particular, for
j ∈J
p ∈ PM with τ (p ) < ∞ we have p ∈ nτ → Hτ and then we get limj ∈J (x j − x)p 2 = 0,
τ
that is, limj ∈J τ (|(x j − x)p |2 ) = 0. Now x j p −→ xp by Theorem 3.2 in [St59]. Since this
j ∈J
τ loc
holds for arbitrary p ∈ PM with τ (p ) < ∞, we see that x j −→ x.
j ∈J
For the second part of assertion (1), we have to check that if R := supj ∈J x j  < ∞,
then the following implication holds true:

τ loc so
(3.1) x j −→ x ⇒ x j −→ x.
j ∈J j ∈J

Let us note that, since the net {x j }j ∈J is uniformly bounded and mτ is dense in Hτ (see
Notation 2.2), it will be enough to check that for arbitrary a ∈ mτ we have limj ∈J (x j −
x)a2 = 0. For this purpose let ε > 0 arbitrary and denote p := l(a) = χ(0,∞) (|a|) the
left support of a. Then p ∈ PM and, since a ∈ mτ , it easily follows that τ (p ) < ∞. Since
τ loc τ
x j −→ x, it then follows that x j p −→ xp , hence there exists a net of projections {p j }j ∈J in
j ∈J j ∈J
M such that supj ∈J (x j − x)p p j  < ε and supj ∈J τ (1 − p j ) < ε. Now note that for every
j ∈ J we have

(x j − x)a2 = (x j − x)p a2


≤ (x j − x)p p j a2 + (x j − x)p (1 − p j )a2
≤ εa2 + |τ (a∗ (1 − p j )p (x j − x)∗ (x j − x)p (1 − p j )a)|1/2
= εa2 + |τ (aa∗ (1 − p j )p |x j − x|2 p (1 − p j ))|1/2
= εa2 + |(1 − p j | p |x j − x|2 p (1 − p j )aa∗ )|1/2
≤ εa2 + (1 − p j 2 · p |x j − x|2 p (1 − p j )aa∗ 2 )1/2
= εa2 + τ (1 − p j )1/4 · p |x j − x|2 p (1 − p j )aa∗ 2 .
1/2
Brought to you by | Instytut Matematyczny Pan
Authenticated
Download Date | 9/30/16 6:25 PM

i i

i i
i i
“Forum Mathematicum, Verlag Walter de Gruyter GmbH & Co. KG” — 2010/2/12 — 13:42 — page 247 — #7

i i

Lie theoretic significance of the measure topologies 247

Since p |x j − x|2 p (1 − p j )aa∗ 2 ≤ |x j − x|2 (1 − p j ) · aa∗ 2 ≤ 4R 2aa∗ 2 , we get


lim supj ∈J (x j −x)a2 ≤ εa2 +ε1/4 ·4R 2aa∗ 2 for arbitrary ε > 0, hence limj ∈J (x j −
x)a2 = 0, and this concludes the proof of (3.1).
Assertion (2) follows by assertion (1) since the localized τ -measure topology coincides
with the τ -measure topology in the case when the trace τ is finite. This completes the
proof.

Remark 3.2. The proof of the implication (3.1) is inspired from the classical method of
proof of a dominated convergence theorem; see for instance Theorem 4.9 in [St59].

Corollary 3.3. Let M be a W ∗ -algebra with a faithful normal semifinite tracial weight τ , and
 are skew-adjoint τ -measurable operators. Then the following
assume that a, a1 , a 2 , . . . ∈ M
assertions hold:
τ loc
(1) If there exists λ0 ∈ C with Re λ0 > 0 and (λ0 − an )−1 −→ (λ0 − a)−1 , then for every
n→∞
τ loc
compact set K ⊆ R we have exp(tan ) −→ exp(ta) uniformly for t ∈ K.
n→∞
τ loc
(2) Conversely, if for each compact set K ⊆ R we have exp(tan ) −→ exp(ta) uniformly
n→∞
τ loc
for t ∈ K, then (λ − an )−1 −→ (λ − a)−1 whenever λ ∈ C with Re λ > 0.
n→∞

Proof. Use Theorem A.6 in [Ta03] and the above Lemma 3.1.

Lemma 3.4. Let H be a complex Hilbert space and M a von Neumann subalgebra of B(H)
with the unitary group UM .
(1) The so-topology, the wo-topology, and the w∗ -topology induce the same topology on
UM , thereby making UM into a topological group.
(2) If M has a faithful normal semifinite tracial weight τ , then the localized τ -measure
topology, the weak localized τ -measure topology, and the w∗ -topology induce the same
topology on UM .
(3) If M has a faithful normal tracial state τ , then the τ -measure topology and the w∗ -
topology induce the same topology on UM .

Proof. (1) This has been well known for a long time (see for instance Lemma 1 in [Dy53]).
It is easy to check that the convergence in the wo-topology is equivalent to the convergence
in the so-topology on the unitary group UM . On the other hand, it is well known that the
wo-topology and the w∗ -topology coincide on the bounded subsets of M (see [Ta79]).
In addition, a straightforward reasoning shows that the multiplication and the inversion
are continuous on UM with respect to the so-topology.
(2) The localized τ -measure topology and the weak localized τ -measure topology coin-
cide on UM by Lemma 3.7(ii) in [Bir06].
Now let M → B(Hτ ) be the semi-cyclic representation associated with τ . According
to Proposition 3.1(2), the localized τ -measure topology coincides with the so-topology
Brought to you by | Instytut Matematyczny Pan
Authenticated
Download Date | 9/30/16 6:25 PM

i i

i i
i i
“Forum Mathematicum, Verlag Walter de Gruyter GmbH & Co. KG” — 2010/2/12 — 13:42 — page 248 — #8

i i

248 D. Beltiţă

on UM . Moreover, the so-topology coincides with the w∗ -topology on UM by the preceding


assertion (1), and this completes the proof.
(3) This is a special case of assertion (2), since the localized τ -measure topology is equal
to the τ -measure topology if the trace τ is finite.

Remark 3.5. It is not always true that the τ -measure topology coincides with the w∗ -
topology on the unitary group. For instance, if M = B(H) with the usual trace, then the
corresponding measure topology on B(H) is the norm topology, which does not coincide
with the w∗ -topology on the unitary group of B(H).

Theorem 3.6. Let M be a W ∗ -algebra with a faithful normal tracial state τ , and think of UM
endowed with the topology induced by the w∗ -topology. Then UM is a topological group with
Lie algebra, and the corresponding topological Lie algebra is isomorphic to the algebra of
skew-adjoint τ -measurable operators on the GNS representation space associated with τ .

Proof. To begin with, denote G = UM and g = {a ∈ M  | a ∗ = −a}, where M  stands


for the set of all τ -measurable operators on the GNS representation space Hτ associated
with τ . Recall from §2 that if the topological group G has a Lie algebra L(G), then the latter
consists of the set of all continuous group homomorphisms X : R → G. Now define

(3.2) Ψ : g → L(G), a → Ψa ,

where for every a ∈ g we denote Ψa : R → G, Ψa (t) = exp(ta). We are going to check that
the mapping (3.2) is a homeomorphism that preserves the operations of sum, bracket, and
scalar multiplication. Then the conclusion will follow since g is already a topological Lie
algebra.
To see that Ψ preserves the aforementioned operations, just recall the Trotter formulas for
one-parameter unitary groups of the sum and the commutator of two skew-adjoint operators
(see for instance [Kat66]), and compare them with the way the scalar multiplication, the
sum, and the bracket in L(G) were defined in equations (2.1)–(2.3).
Next note that the mapping Ψ is surjective as a consequence of Stone’s theorem and of the
 consists of all operators affiliated with the realization of M as a von Neumann
fact that M
algebra on Hτ , since the trace τ is finite (see for instance [Nl74]). The mapping Ψ is also
injective since for a = −a∗ ∈ M  we have Ψa = 0 if and only if for every t ∈ R we have
exp(ta) = 1, which is equivalent to a = 0.
Finally, Ψ is a homeomorphism since the measure topology is metrizable (Remark 2.4)
and for any sequence {an }n∈N in g and a ∈ g we have
τ τ
(3.3) an −→ a ⇐⇒ (1 − an )−1 −→(1 − a)−1 ⇐⇒ lim Ψan = Ψa .
n∈N n∈N n∈N

The first of these equivalences is an easy consequence of Theorem 2.1 in [Pad79], and the
latter equivalence follows by Corollary 3.3(1) along with Lemma 3.4(3). (In fact, recall from
§2 that convergence in the Lie algebra L(G) means uniform convergence on the compact
intervals in R.)

Brought to you by | Instytut Matematyczny Pan


Authenticated
Download Date | 9/30/16 6:25 PM

i i

i i
i i
“Forum Mathematicum, Verlag Walter de Gruyter GmbH & Co. KG” — 2010/2/12 — 13:42 — page 249 — #9

i i

Lie theoretic significance of the measure topologies 249

The main technical point of the above proof is interesting on its own, and therefore we record
it as follows:

Corollary 3.7. Let M be a W ∗ -algebra with a faithful normal tracial state τ , and assume
τ
that {an }n∈N is a sequence of self-adjoint τ -measurable operators. Then an −→ a for some
n∈N
w ∗
 if and only if for every compact interval K in R we have exp(itan ) −→
a ∈ M exp(ita)
n∈N
uniformly for t ∈ K.

Proof. This follows from Lemma 3.4(3) and the fact that the above mapping (3.2) is a
homeomorphism. Equivalently, use (3.3) from the proof of Theorem 3.6.

Corollary 3.8. For j = 1, 2 let Mj be a W ∗ -algebra with a faithful normal tracial state τj
and let uM 
 j be the topological Lie algebra of all skew-adjoint operators in Mj . If the

topological groups UM1 and UM2 with the respective w -topologies are isomorphic, then
uM 1 and uM 2 are isomorphic as topological Lie algebras.

Proof. Use Theorem 3.6 above along with Proposition 3.7(ii) in [HM05].

The topological groups with Lie algebras exhibited by the next corollary play an interesting
role in certain areas of topological group theory and of mathematical physics; see [GP03]
and [CG04], respectively. Genuine Lie theory for the smaller groups consisting of essentially
bounded matrix-valued functions can be found in the paper [Gl03]. As usual, for an arbitrary
integer n ≥ 1 we shall denote by U(n) the unitary group of n × n complex matrices, and by
u(n) the corresponding Lie algebra of skew-adjoint matrices of size n.

Corollary 3.9. Assume that (S, Σ, μ) is a finite-measure space and n ≥ 1 is an integer. Let
L 0(S, U(n)) be the set of equivalence classes consisting of μ-almost everywhere identical,
measurable functions defined on S with values into the group U(n), and L 0 (S, u(n)) the set of
equivalence classes of u(n)-valued measurable functions on S. If we endow both L 0 (S, U(n))
and L 0(S, u(n)) with the topology of entry-wise convergence in measure, then L 0 (S, U(n))
is a topological group with Lie algebra and the corresponding topological Lie algebra is
isomorphic to L 0(S, u(n)).

Proof. Denote M = L ∞ (S, Σ, μ; M n (C))  L ∞ (S, Σ, μ) ⊗ M n (C), the set of equiva-


lence classes of essentially bounded measurable functions from S into the full matrix alge-
bra M n (C). Let trn : M n (C) → C, (aj k )1≤j ,k≤n → n1 nj=1 aj j , the normalized trace on
M n (C), and define

1
τ : M → C, τ (a) = trn (a)dμ.
μ(S)
S

Then τ is a faithful normal tracial state on the W ∗ -algebra M, and Theorem 3.6 applies.

Brought to you by | Instytut Matematyczny Pan


Authenticated
Download Date | 9/30/16 6:25 PM

i i

i i
i i
“Forum Mathematicum, Verlag Walter de Gruyter GmbH & Co. KG” — 2010/2/12 — 13:42 — page 250 — #10

i i

250 D. Beltiţă

4 The exponential maps of unitary groups


This short section is devoted to discussing some continuity properties of the exponential
maps of the topological groups with Lie algebras constructed in Theorem 3.6. We shall also
see that in general these exponential maps fail to be locally injective (Corollary 4.4) and we
shall construct natural continuous cross-sections —“logarithms”— for them over suitable
subsets of the unitary groups under consideration; see Corollary 4.2.

Proposition 4.1. Let M ⊆ B(H) be a von Neumann algebra and consider the Banach-Lie
group G = UM with the Lie algebra g = uM = {a ∈ M | a∗ = −a}. Then the following
assertions hold:
(1) The exponential map expG : g → G is surjective and the unitary group G is the union
of its one-parameter subgroups.
(2) For arbitrary δ ∈ (0, π), if we denote Tδ = {exp(it) | t ∈ (−δ, δ)}, Uδ = {g ∈ G |
σ(g) ⊆ T δ }, and V δ = {a ∈ g | σ(a) ⊆ (−iδ, iδ)}, then we have a real analytic
diffeomorphism

expG |V δ : V δ → Uδ

which is a homeomorphism when both sets Uδ and V δ are endowed with the strong
operator topology.

Proof. (1) This assertion is a straightforward application of the Borel functional calculus
for unitary operators; see for instance the introductory comments to Problem 131 in [Ha82].
(2) Note that
Uδ = {g ∈ G | 1 − g < 2 sin(δ/2)} and V δ = {a ∈ g | a < δ}
as easy consequences of the spectral theorem for normal operators. By using the functional
calculus for normal operators we see that expG |V δ : V δ → Uδ is bijective and its inverse is
(expG |V δ )−1 = logG,Uδ : Uδ → V δ .
In particular expG |V δ : V δ → Uδ is a real analytic diffeomorphism when we think of Uδ
and V δ as open subsets of the Banach manifolds G and g, respectively.
It follows by Theorem 4.2 in [Kad68] that this diffeomorphism is also a homeomorphism
with respect to the strong operator topology.

Corollary 4.2. Let M be a W ∗ -algebra with a faithful normal semifinite tracial weight τ .
For arbitrary δ ∈ (0, π) denote T δ = {exp(it) | t ∈ (−δ, δ)}, and also Uδ = {g ∈ UM |
σ(g) ⊆ Tδ } and V δ = {a ∈ M | a = −a∗ and σ(a) ⊆ (−iδ, iδ)}. Then the exponential map
induces a homeomorphism expUM |V δ : V δ → Uδ when both sets Uδ and V δ are endowed
with the localized τ -measure topology.

Proof. Use Proposition 4.1(2) and Proposition 3.1(1).

Results related to Corollary 4.2 can be found in the papers [Pad79] and [Tik87].
Brought to you by | Instytut Matematyczny Pan
Authenticated
Download Date | 9/30/16 6:25 PM

i i

i i
i i
“Forum Mathematicum, Verlag Walter de Gruyter GmbH & Co. KG” — 2010/2/12 — 13:42 — page 251 — #11

i i

Lie theoretic significance of the measure topologies 251

Proposition 4.3. Let M be a W ∗ -algebra with a faithful normal semifinite tracial weight τ
and think of uM endowed with the measure topology. If the exponential map exp : uM → UM
is injective on some neighborhood of 0 ∈ uM then 0 ∈ R is an isolated point of the set τ (PM ).

Proof. If {p j }j ∈J is a net in PM such that limj ∈J τ (p j ) = 0, then limj ∈J 2πip j 2 = 0,


τ
hence 2πip j −→ 0 by Theorem 3.2 in [St59]. On the other hand, for every j ∈ J we have
j ∈J
2πip j ∈ uM and exp(2πip j ) = 1.

Corollary 4.4. Let M be a factor of type II1 with the canonical trace denoted by τ and think
of the unitary group UM endowed with the τ -measure topology. Then UM is a topological
group with Lie algebra and the corresponding exponential map is not locally injective.

Proof. One can use Theorem 3.6 and Proposition 4.3 since it is well known that τ (PM ) =
[0, 1]. (We have already seen in Remark 2.6 that any finite trace on a von Neumann algebra
of type II1 takes arbitrarily small positive values on the orthogonal projections.)

Acknowledgment. I am very grateful to Professor Hendrik Grundling for drawing my


attention to the paper [CG04]; Professor Airat Bikchentaev and Professor Oleg Tikhonov
for sending me upon my request some of their papers; and finally the Referee for a number
of really helpful and illuminating remarks, questions, and references.
Financial support from the CNCSIS grant 23GR/11.06.2008 (code 813) is also acknow-
ledged.

References
[Bel06] Beltiţă D.: Smooth Homogeneous Structures in Operator Theory. Monogr. and Surveys
in Pure and Appl. Math. 137. Chapman & Hall/CRC Press, Boca Raton–London–New
York–Singapore 2006
[BR07] Beltiţă D., Ratiu T. S.: Geometric representation theory for unitary groups of operator
algebras. Adv. Math. 208 (2007), 299–317
[Bir04] Bikchentaev A. M.: Minimality of the convergence in measure topology on finite von
Neumann algebras. (Russian) Mat. Zametki 75 (2004), 342–349; translation in Math.
Notes 75 (2004), 315–321
[Bir06] Bikchentaev A. M.: Local convergence in measure on semifinite von Neumamn algebras.
(Russian) Tr. Mat. Inst. Steklova. Funkts. Prostran., Teor. Priblizh., Nelinein. Anal. 255
(2006), 41–54; translation in Proc. Steklov Inst. Math. 255 (2006), 41–54
[Bo65] Bourbaki N.: Éléments de Mathématique. Fasc. XIII. Livre VI: Intégration. Chapitres 1,
2, 3 et 4. Deuxième édition revue et augmentée. Actualités Scientifiques et Industrielles,
No. 1175, Hermann, Paris 1965
[CG04] Carey A., Grundling H.: On the problem of the amenability of the gauge group. Lett. Math.
Phys. 68 (2004), 113–120
[Ch74] Chernoff P. R.: Product formulas, nonlinear semigroups, and addition of unbounded op-
erators. Memoirs of the Amer. Math. Soc. 140. Amer. Math. Soc., Providence, RI 1974
[CY74] Chen S. S., Yoh R. W.: The category of generalized Lie groups. Trans. Amer. Math. Soc.
199 (1974), 281–294
Brought to you by | Instytut Matematyczny Pan
Authenticated
Download Date | 9/30/16 6:25 PM

i i

i i
i i
“Forum Mathematicum, Verlag Walter de Gruyter GmbH & Co. KG” — 2010/2/12 — 13:42 — page 252 — #12

i i

252 D. Beltiţă

[Ci90] Ciach L. J.: Subadditive measures on projectors of a von Neumann algebra. Dissert.
Math. 288 (1990)
[Da96] Davidson K. R.: C∗ -Algebras by Example. Fields Institute Monographs 6.American Math-
ematical Society, Providence, RI 1996
[Dy53] Dye H. A.: The unitary structure in finite rings of operators. Duke Math. J. 20 (1953),
55–69
[FK86] Fack T., Kosaki H.: Generalized s-numbers of τ -measurable operators. Pacific J. Math.
123 (1986), 269–300
[GP03] Gao S., Pestov V.: On a universality property of some abelian Polish groups. Fund. Math.
179 (2003), 1–15
[Gl02] Glöckner H.: Real and p -adic Lie algebra functors on the category of topological groups.
Pacific J. Math. 203 (2002), 321–368
[Gl03] Glöckner H.: Lie groups of measurable mappings. Canad. J. Math. 55 (2003), 969–999
[Ha82] Halmos P. R.: A Hilbert Space Problem Book. Second edition. Graduate Texts in Mathe-
matics 19. Encyclopedia of Mathematics and its Applications 17. Springer-Verlag, New
York – Berlin 1982
[dlH79] de la Harpe P.: Moyennabilité du groupe unitaire et propriété P de Schwartz des algèbres
de von Neumann. In: Algèbres d’Opérateurs (Sém., Les Plans-sur-Bex, 1978). Lecture
Notes in Math. 725, pp. 220–227. Springer Berlin 1979
[HM05] Hofmann K. H., Morris S. A.: Sophus Lie’s third fundamental theorem and the adjoint
functor theorem. J. Group Theory 8 (2005), 115–133
[HM07] Hofmann K. H., Morris S. A.: The Lie Theory of Connected Pro-Lie Groups. A Structure
Theory for Pro-Lie Algebras, Pro-Lie Groups, and Connected Locally Compact Groups.
EMS Tracts in Mathematics 2. European Mathematical Society (EMS), Zürich 2007
[HN08] Hofmann K. H., Neeb K.-H.: Pro-Lie groups which are infinite dimensional Lie groups.
Math. Proc. Cambridge Philos. Soc. (to appear)
[Kad68] Kadison R. V.: Strong continuity of operator functions. Pacific J. Math. 26 (1968), 121–129
[Kat66] Kato T.: Perturbation Theory for Linear Operators. Die Grundlehren der mathematischen
Wissenschaften 132. Springer-Verlag, Berlin–Heidelberg–New York 1966
[Kel57] Kelley J. L.: General Topology. D. Van Nostrand Company, New Jersey 1957
[LR04] Lech J., Rybicki T.: Abstract Lie groups and locally compact topological groups. Ann.
Acad. Pedagog. Crac. Stud. Math. 4 (2004), 123–141
[Ne06] Neeb K.-H.: Towards a Lie theory of locally convex groups. Japanese J. Math. 1 (2006),
291–468
[Nl74] Nelson E.: Notes on non-commutative integration. J. Functional Analysis 15 (1974), 103–
116
[Pad79] Padmanabhan A. R.: Probabilistic aspects of von Neumann algebras. J. FunctionalAnalysis
31 (1979), 139–149
[PR66] Przeworska-Rolewicz D., Rolewicz S.: On integrals of functions with values in a complete
linear metric space. Studia Math. 26 (1966), 121–131
[Se53] Segal I. E.: A non-commutative extension of abstract integration. Ann. of Math. (2) 57
(1953), 401–457
[St59] Stinespring W. F.: Integration theorems for gages and duality for unimodular groups. Trans.
Amer. Math. Soc. 90 (1959), 15–56
[SZ79] Strătilă Ş., Zsidó L.: Lectures on von Neumann Algebras. Revision of the 1975 original.
Editura Academiei, Bucharest; Abacus Press, Tunbridge Wells 1979
[Ta79] Takesaki M.: Theory of Operator Algebras I. Springer-Verlag, New York–Heidelberg 1979
Brought to you by | Instytut Matematyczny Pan
Authenticated
Download Date | 9/30/16 6:25 PM

i i

i i
i i
“Forum Mathematicum, Verlag Walter de Gruyter GmbH & Co. KG” — 2010/2/12 — 13:42 — page 253 — #13

i i

Lie theoretic significance of the measure topologies 253

[Ta03] Takesaki M.: Theory of Operator Algebras II. Encyclopaedia of Mathematical Sciences
125. Operator Algebras and Non-commutative Geometry 6. Springer-Verlag, Berlin 2003
[Te81] Terp M.: L p spaces associated with von Neumann algebras. Rapport no. 3 (June 1981)
100 pp., Matematisk Institut, Københavns Universitet, 1981
[Tik87] Tikhonov O. E.: Continuity of operator functions in topologies connected with a trace on
a von Neumann algebra. (Russian) Izv. Vyssh. Uchebn. Zaved. Mat. (1987), 77–79
[Up85] Upmeier H.: Symmetric Banach Manifolds and Jordan C∗ -algebras. North-Holland Math-
ematics Studies 104. Notas de Matemàtica 96. North-Holland Publishing Co., Amsterdam
1985

Received 15 April 2008; in final form 27 May 2008


Institute of Mathematics “Simion Stoilow” of the Romanian Academy, P.O. Box 1-764, Bucharest,
Romania
Daniel.Beltita@imar.ro

Brought to you by | Instytut Matematyczny Pan


Authenticated
Download Date | 9/30/16 6:25 PM

i i

i i

You might also like