ALGEBRAIC GELFAND COMPACTNESS

FRANK MURPHY-HERNANDEZ AND JAIME GARCÍA-VILLEDA

Abstract.

Introduction
In this paper the rings are associative but we do not ask them to have unit.

1. Preliminaries
[?] [?]
If A and B are commutative C ∗ -algebras, and f : A −→ B is an ∗-algebra mor-
phism such that the linear span of {f (a)b | a ∈ A, b ∈ B} is dense in B, then f is
called non-degenerate We denote by C ∗ the category of commutative C ∗ -algebras
and non-degenerate morphisms.
If X and Y are topological spaces and α : X −→ Y is a continuous map such
that the inverse image of compact subsets of Y are compact subsets of X, then we
call α a proper map. We denote by T the category of Hausdorff locally compact
topological spaces and proper continuous maps. If α : X −→ C is a continuous
map such that for every  > 0 there is K compact subset of X with |α(x)| <  for
all x ∈ X \ K , then we say that α vanishes at infinity. We denote by C0 (X) the
algebra of all continuous maps that vanishes at infinity. We denote by C00 (X) the
subalgebra of C0 (X) which contains all the functions with compact support.
Proposition 1.1 (Non-unital Gelfand Duality). There is an equivalence of cate-
gories: T op ∼
= C∗.
It is well known that, a topological space X is compact if and only if C0 (X) is
unital.
We recommend the books as reference for C ∗ -algebras [?] and [?].

2. Enough Idempotents
Definition 2.1. Let X ∈ T and A ⊆ X. We define the chaarateristic function of
A as χA : X −→ C given by:
(
1, if x ∈ A
χA (x) =
0, if x ∈
/A

Proposition 2.1. Let X ∈ T and e ∈ C0 (X) idempotent. Then e = χK for some

K ⊆ X compact clopen.

Date: May 20, 2020.

2000 Mathematics Subject Classification. Primary ****, ****; Secondary ****, ****.
Key words and phrases. ,,
1
2 FRANK MURPHY-HERNANDEZ AND JAIME GARCÍA-VILLEDA

Proof. As e is idempotent, we have that e(x)2 = e(x) for any x ∈ X. Then

im(e) ⊆ {0, 1}. Thus e = χK with K = e−1 (1). As {1} is closed in C and e is
continuous, we get that K is closed. Analogous, we obtain that X \ K is closed. In
this way we deduce that K is clopen. By hypothesis e vanishes at infinity, so for
 = 12 we have a L ⊆ X compact subset such that for any x ∈ X \ L, |e(x)| < .
Then K ⊆ L. As K is closed in compact set L, we have that K is compact. 
Definition 2.2. Let R be a ring. We say that R has enough idempotents, if there
a family {eλ }L
exists L λ∈Λ of pairwise orthogonal idempotents of elements in R with
R = λ∈Λ Reλ = λ∈Λ eλ R. In this case {eλ }λ∈Λ is called a complete family
of idempotents in R. This concept was introduced by Fuller in [?]. We say that a
commutative C ∗ -algebra A has enough idempotents, if there exists a family {eλ }λ∈Λ
L
of pairwise orthogonal idempotents of elements in A with A = λ∈Λ Aeλ .
Example 2.1. Ejemplo que la version algebraica es trivial
Proposition 2.2. LetFX ∈ T . Then C0 (X) is an C ∗ -algebra with enough idempo-
tents if and only X = i∈I Xi with Xi compact for all i ∈ I.
Proof. ⇒)
⇐) 

3. Local Units
Definition 3.1. Let R be a ring. We say that R is locally unital, if for any finitely
many a1 , . . . , an ∈ R there is an idempotent e ∈ R with a1 , . . . , an ∈ eRe. This
definition is due Ánh and Márki [?]. We say that a commutative C ∗ -algebra is
locally unital, if as a ring is locally unital.
The following example give us that the definition of locally unital is not trivial
since not all commutative C ∗ -algebras are locally unital
Example 3.1. Let X = n∈Z [2n, 2n + 1], and f ∈ C0 (X) given by f (x) = e−x
F
for all x ∈ X. If there is an idempotent e ∈ C0 (X) with f e = f = ef , then
X = sup(f ) ⊆ sup(e), and this way we have that sup(e) = X. Now, since e is
idempotent, there is compact clopen K with e = χK . Thus K = X, which cannot
be since X is not compact. Therefore C0 (X) is not locally unital.
Proposition 3.1. Let X be a discrete space. Then C0 (X) is locally unital if and
only X is finite.
Proof.
P∞ −i ⇒) If X is not finite, then there is a family {xi }∞i=1 in X. We define f :=
i=1 2 χ{x i } . Since C 0 (X) is complete, we get that f ∈ C0 (X). By hypothesis
there is an idempotent e ∈ C0 (X) with f e = f = ef . So we have that sup(f ) ⊆
sup(e). Also, there is a compact clopen K ⊆ X with e = χK . It follows that
{xi }∞
i=1 ⊆ K, which is a contradiction as all compact discrete space are finite.
Therefore X is finite. 
Proposition 3.2. Let X ∈ T . Then C0 (X) is locally unital if and only X is
compact.
Proof.
F ⇒) Let {Xi }i∈I be the connected components of X. We have that X =
i∈I Xi and Xi is a clopen for all i ∈ I. Let j ∈ I and x ∈ Xj , so there is a
compact subset K of X and an open neighbourhood U of x with x ∈ K ⊆ U .
 ALGEBRAIC GELFAND COMPACTNESS 3

Then by Tietze extension theorem there function f ∈ C00 (X) with f |K = 1. By

hypothesis, there is an idempotent element e ∈ C0 (X) with f e = f = ef . Thus
there is a compact clopen L with e = χL . So x ∈ K ⊆ sup(f ) ⊆ sup(e) = L. It
follows that Xj ⊆ L, and Xj is compact. Therefore Xi is compact for all i ∈ I.
If {Xi }i∈I is not finite, then there is a enumerable
F∞ subset {Xn }∞ . Without loss
n=1P
∞
of generality we may assume that X = n=1 Xn . We define g := n=1 χXn 2−n .
Since C0 (X) is complete, we get that g ∈ C0 (X). Again, by hypothesis there is
an idempotent h ∈ C0 (X) with gh = g = hg and there is a compact clopen M
such that h = χM . We deduce that X = sup(g) ⊆ sup(h) = M . Afterwards
M = X, which is a contradiction as X is not compact because {Xn }∞ n=1 is an open
covering which does not have a finite subcovering. Therefore I is finite and X is
compact. 
From this we conclude that in the world of commutative C ∗ -algebras, by the
Gelfand duality, the notion of being unital and locally unital coincide.
Definition 3.2. Let R be a ring. We say that R has local units, if for any a ∈ R
then there is e ∈ R idempotent with ae = a = ea. This definition was made by
Abrams in [?]. We say that a commutative C ∗ -algebra has local units, if as a ring
has local units.
If a commutative C ∗ -algebra has local units then is locally unital.
Proposition 3.3. Let X ∈ T . Then C0 (X) is a locally unital if and only C0 (X)
has local units.
Proof. 
Definition 3.3. Let R be a ring. We say that R is a s-unital ring, if for any finitely
many x1 , . . . , xn ∈ R there is y ∈ R with xi y = xi = yxi for i = 1, . . . , n. This
definition was introduced by Tominaga in [?] and [?]. We say that a commutative
C ∗ -algebra is s-unital, if as a ring is s-unital.
Proposition 3.4. Let X ∈ T . Then C0 (X) is an C ∗ -algebra is s-unital if and
only X is compact.
Proof. ⇒) Let f ∈ C0 (X). Then there is an idempotent e ∈ C0 (X) with f e = f =
ef . So sup(f ) ⊆ sup(e), and e|sup(f ) = 1. As e vanishes at infinity, for  = 21 ,
there is a compact subset K of X with e−1 ( 21 , ∞) ⊆ K. Thus sup(f ) ⊆ K, and
f ∈ C00 (X). It follows that C00 (X) = C0 (X). Therefore X is compact. 

4. Idempotent and Firm C ∗ -algebra

Definition 4.1. Let R be a ring. We say that R is idempotent, if R2 P = R, that
n
is, that for any x ∈ R there are x1 , . . . , xn , y1 , . . . , yn ∈ R such that x = i=1 xi yi .
We say that a commutative C ∗ -algebra is idempotent, if as a ring is idempotent.
Proposition 4.1. Let X ∈ T , and f ∈ C0 (X). Then there is g ∈ C0 (X) such that
g 2 = f . Moreover, sup(g) = sup(f ).
Proof. Jaime 
√
We denote g by f .... Jaime
Proposition 4.2. Let X ∈ T . Then C0 (X) is an idempotent commutative C ∗ -
algebra.
4 FRANK MURPHY-HERNANDEZ AND JAIME GARCÍA-VILLEDA

√ √ 2
Proof. Let f ∈ C0 (X). Then there is f ∈ C0 (X) with f = f . Therefore C0 (X)
is an idempotent commutative C ∗ -algebra. 
Definition 4.2. Let R be a ring. We say that R is a firm ring, if the canonical
morphism µ : R ⊗R R −→ R given by µ(r ⊗ s) = rs is an isomorphism.
We observe that the condition of µ being surjective is equivalent to being idem-
potent.
Proposition 4.3. Let X ∈ T . Then C0 (X) is a firm commutative C ∗ -algebra.
Proof. Let f, g ∈ C0 (X) with f g = 0. Then sup(f ) ∩ sup(g) = ∅. So there is h ∈
C0 (X) with h2 = g and sup(g) = sup(h). It folloes that f ⊗ g = f h ⊗ h = 0 ⊗ h = 0.
Thus ker(µ) = 0 and µ is injective. Therefore C0 (X) is a firm commutative C ∗ -
algebra. 
In the context of rings, we have that firm rings are idempotent rings, but not
conversely. But for commutatives C ∗ -algebras both concepts coincide.

5. Compact Dimension
We like to define a dimension that measures how far a topological space is from
being compact. By the Gelfand duality this is equivalent to measures how far a
C ∗ -algebra from being unitary.
Definition 5.1. Let A be a commutative C ∗ -algebra, and {eλ }λ∈Λ a directed family
of self adjoint elements of A. We say that {eλ }λ∈Λ is an approximate identity, if
xeλ → x, for all x ∈ A.
It is well know that any commutative C ∗ -algebra has an approximate identity.
In fact there is a canonical approximate identity given by the family of all positive
self adjoint elements with norm less or equal than one with its natural order.
Definition 5.2. Let A be a commutative C ∗ -algebra. If κ is least cardinal such
that exists an approximate unit of A with cardinality κ, then We define the compact
dimension of A as κ if κ is a transfinite cardinal, and as κ − 1 if it is a finite
ordinal. We denote this cardinal by dimC (A).
Since any commutative C ∗ -algebra has an approximate identity and the cardi-
nality of an approximate unit is always greater or equal to 1, we have the compact
dimension is well defined. Also we have that a commutative C ∗ -algebra has unit if
and only if dimC (A) = 0.
If we have that dimC (A) = n is a natural, then it must be 0. Since an approxi-
mate identity is a net when it is finite what we have is a unit.
Proposition 5.1. Let A be non-unital commutative C ∗ -algebra and B be a C ∗ -
subalgebra of A. Then dimC (B) ≤ dimC (A).
Proof. 
∗
Proposition 5.2. Let A be non-unital commutative C -algebra and I be a a closed
ideal of A. Then dimC (A/I) ≤ dimC (A).
Proof. 
Corollary 5.1. Let f : A −→ B be a surjective proper morphism. Then dimC (B) ≤
dimC (A).
 ALGEBRAIC GELFAND COMPACTNESS 5

∗
PropositionL5.3. Let {A
Pi }i∈I be a family of non-unital commutative C -algebras.
Then dimC ( i∈I Ai ) = i∈I dimC (Ai ).
Proof. 
Proposition 5.4. Let A and B be non-unital commutative C ∗ -algebras. Then
dimC (A ⊗ B) = dimC (A) dimC (B).
Proof. 
Proposition 5.5. Let A be non-unital commutative C ∗ -algebra. Then dimC (A) ≤
|A|
Proof. 
Proposition 5.6. Let κ be a transfinite cardinal. Then there is an non-unital
commutative C ∗ -algebra A with dimC (A) = κ.
Proof. 

Facultad de Ciencias, UNAM, Mexico City

Email address: [email protected]

Instituto de Matemticas, UNAM, Mexico City