This paper characterizes closed ideals and Lie ideals of minimal tensor products of certain C*-algebras, linking their structure to properties of the component algebras and the underlying space.
Contribution
It provides a characterization of closed ideals in $C_0(X,A)$ for algebras with finitely many ideals and describes the structure of Lie ideals in these tensor products.
Findings
01
Closed ideals of $C_0(X,A)$ are characterized via ideals of $A$ and subspaces of $X$.
02
Closed ideals of $C_0(X) ensor^{ ext{min}} A$ are finite sums of product ideals.
03
The centre-quotient property of $C_0(X,A)$ is equivalent to that of $A$.
Abstract
For a locally compact Hausdorff space X and a C∗-algebra A with only finitely many closed ideals, we discuss a characterization of closed ideals of C0(X,A) in terms of closed ideals of A and certain (compatible) closed subspaces of X. We further use this result to prove that a closed ideal of C0(X)⊗minA is a finite sum of product ideals. We also establish that for a unital C∗-algebra A, C0(X,A) has centre-quotient property if and only if A has centre-quotient property. As an application, we characterize the closed Lie ideals of C0(X,A) and identify all closed Lie ideals of C0(X)⊗minB(H), H being a separable Hilbert space.
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Full text
Closed ideals and Lie ideals of minimal tensor product of certain C∗-algebras
For a locally compact Hausdorff space X and a C∗-algebra A with only finitely many closed ideals, we discuss a characterization of closed ideals of C0(X,A) in terms of closed ideals of A and certain (compatible) closed subspaces of X. We further use this result to prove that a closed ideal of C0(X)⊗minA is a finite sum of product ideals.
We also establish that for a unital C∗-algebra A, C0(X,A) has centre-quotient property if and only if A has
centre-quotient property.
As an application, we characterize the closed Lie ideals of C0(X,A) and identify all closed Lie ideals of C0(X)⊗minB(H), H being a separable Hilbert space.
The first named author was supported by a Junior Research Fellowship
of CSIR with file number 09/045(1442)/2016-EMR-I
1. Introduction
Let A⊗αB denote the completion of the algebraic tensor product of two C∗-algebras A and B under an algebra cross norm ∥⋅∥α. A natural question arises whether the closed ideals of A⊗αB can be identified in terms of the closed ideals of A and B or not. It is known that the closed ideals of the Banach algebras A⊗hB, A⊗γB and A⊗B are directly related to the closed ideals of A and B, where ⊗h,⊗γ and ⊗ are the Haagerup tensor product, Banach space projective tensor product and operator space projective tensor product, respectively. Interestingly, if either A or B possesses finitely many closed ideals then every closed ideal of these spaces is a finite sum of product ideals (see, for instance, [1, 5, 7]). However, in 1978, Wassermann [16] established an astonishing result that not every closed ideal of B(H)⊗minB(H) is a finite sum of product ideals, ⊗min being the minimal C∗-tensor norm.
Recall that for every x∈A⊗B, ∥x∥min=sup{∥(π1⊗π2)(x)∥}, where π1 and π2 run over all representations of A and B respectively (see [14] for details).
In the present article, we prove that this anamoly can be removed by assuming one of the C∗-algebras to be commutative with the help of a totally different technique than those used for ⊗h, ⊗γ and ⊗.
A Banach algebra B naturally imbibes a Lie algebra structure with the Lie bracket given by [a,b]=ab−ba for every a,b∈B.
A closed subspace L of B is said to be a Lie ideal if [B,L]⊆L where [B,L]=span{[b,l]:b∈B,l∈L}. The closed Lie ideals for C∗-algebras are extensively studied, one may refer to the expository article [11] for details. Recently some research has been done to identify the closed Lie ideals for the various tensor products of C∗-algebras. In [4, Section 5], [6, Section 4]) the closed Lie ideals of C0(X)⊗minA have been characterized in terms of closed subspaces of X, X being a locally compact Hausdorff space and A being a simple C∗-algebra with at most one tracial state. However, if A is not simple nothing is known about the closed Lie ideals of such spaces. In this article, we discuss a characterization of closed Lie ideals of C0(X,A), for any C∗-algebra A.
We present a brief summary of the main results of the article. In Section 2, we first establish an appropriate (surjective) correspondence between a class of closed subspaces of X and the closed ideals of C0(X,A), for any C∗-algebra A. Interestingly, this correspondence turns out to be bijective if A has finitely many closed ideals. We use this correspondence to identify the image of a closed ideal of C0(X,A) in C0(X)⊗minA under the canonical isomorphism. This will pave our way to establish that every closed ideal of C0(X)⊗minA is a finite sum of product ideals, where X is a locally compact Hausdorff space and A is a C∗-algebra with finitely many closed ideals.
In Section 3, we characterize the closed Lie ideals of C0(X,A) in terms of some closed subspaces of X.
To obtain a better picture of closed Lie ideals of C0(X)⊗minA, we prove that if A is unital then the
centre-quotient property of A passes to C0(X,A) and vice-versa.
As an application we establish an interesting result that a closed subspace L of C0(X)⊗minB(H) is a Lie ideal
if and only if there exist closed subspaces S1,S2 of X with S1⊆S2 and a closed subspace K of
C0(X)⊗C1 such that L=J(S1)⊗K(H)+J(S2)⊗B(H)+K, where H is a separable Hilbert space and
for F⊆X, J(F):={f∈C0(X):f(F)⊆{0}}.
2. Closed ideals of C0(X)⊗minA
Let X be a locally compact Hausdorff space and A be any C∗-algebra. It is a well known fact that there is a bijective correspondence between the closed subspaces of X and the closed ideals of C0(X) given by F↔J(F). However, if we move from complex valued functions to the vector valued functions, such a correspondence is not known. Although, in the literature, it is established that every closed ideal of C0(X,A) is of the form {f∈C0(X,A):f(x)∈Ix,∀x∈X} where for every x∈X, Ix is a closed ideal of A [12, V.26.2.1], but this description fails to be fruitful while moving from C0(X,A) to C0(X)⊗minA in order to determine the closed ideals.
We first generalize the former notion to the continuous vector valued functions by establishing a correspondence between the closed subspaces of X and closed ideals of C0(X,A). This correspondence will further enable us to characterize closed ideals of C0(X)⊗minA in terms of closed ideals of A and subspaces of X, when A has finitely many closed ideals. This is due to the fact that there exists an isometric ∗-isomorphism φ~:C0(X)⊗minA→C0(X,A), which takes f⊗a to af for every f∈C0(X) and a∈A, where (af)(x)=f(x)a (see, [14, Theorem 4.14 (iii)], [9, Proposition 1.5.6]).
Let us first fix some notations for further use. For any t∈N, the set {1,2,3,…,t} is be denoted by Nt.
The spaces Cb(X,A) and Cc(X,A), as usual, denote the css of all bounded continuous functions and compactly supported continuous functions, respectively, from X to A endowed with sup norm. For a non-unital C∗-algebra A, A~ will denote its unitization.
For a locally compact Hausdorff space X and any function g∈C0(X), we define g^∈C0(X,A~) (resp., g^∈C0(X,A)) by g^(x)=g(x)1, where 1 is the unit of A~ (resp., of A) if A is non unital (resp., if A is unital).
For any a∈A, we denote by a′ the constant function in Cb(X,A) such that a′(x)=a for every x∈X.
For an indexing set Δ, let S={Si}i∈Δ and T={Ti}i∈Δ be collections of subspaces of some sets Y and Z.
We define S* to be compatible with T* if whenever for some subset γ of Δ, ∩j∈γTj=Ti for some i∈Δ, then ∩j∈γSj=Si.
For a locally compact Hausdorff space X, α∈Δ and T as above, define a collection TTα:={S={Si}i∈Δ:Si is a closed subspace of X for every i∈Δ,Sis compatible withTandSα=X}.
Theorem 2.1**.**
Let X be a locally compact Hausdorff space, A be a C∗-algebra and I={Ii}i∈Δ be the collection of all closed ideals of A with Iβ=A.
Then there exists a surjection θ from TIβ into K, the set of all closed ideals of C0(X,A).
Proof.
Define θ:TIβ→K by θ(S)=J(S) where J(S)={f∈C0(X,A):f(Si)⊆Ii,∀i∈Δ}.
Clearly θ is well defined.
To see that θ is onto, consider J∈K.
For each i∈Δ, set Si=∩f∈Jf−1(Ii).
If Ii=∩j∈γIj for some subset γ of Δ, then
[TABLE]
so that S={Si}i∈Δ is compatible with I. Since, Sβ=∩f∈Jf−1(Iβ)=∩f∈Jf−1(A)=X, we obtain S∈TIβ. Clearly J⊆J(S), so it is sufficient to prove that J is dense in J(S).
For this, let f∈J(S) be non-zero and ϵ>0 be arbitrary.
We first claim that for any x∈X, there exists an element hx∈J such that ∥hx(x)−f(x)∥≤3ϵ.
Indeed, if γ′={i∈Δ:x∈Si}, then x∈∩i∈γ′Si and f(x)∈∩i∈γ′Ii=Ij for some j∈Δ.
Let Ir denote the closed ideal of A generated by the set {g(x):g∈J}.
Then x∈Sr which implies that f(x)∈Ir. Hence ∥∑i=1kaigi(x)bi−f(x)∥≤31ϵ for some a1,a2,…ak,b1,b2,…bk in A (resp., in A~) if A is unital (resp., if A is non unital). From [8, Corollary 4.2.10], we know that J is a closed ideal of C0(X,A~), thus we have a function hx=∑i=1nai′gibi′ in J which satisfies the required condition.
Denote by X~=X∪{∞}, the one point compactification of X.
For each x∈X, let h~x and f~ be the continuous extensions of hx and f to X~ which take ∞ to [math], and let h~x0 be the continuous extension of hx0, the zero function, to X~ such that h~x0(∞)=0.
As h~x0(∞)=f~(∞)=0, there exists a neighbourhood Vx0 of ∞ in X~ such that ∥h~x0(y)−f(y)∥≤31ϵ for every y∈Vx0. Further f~ and h~x are continuous at x∈X, so there exists a neighbourhood Vx of x such that ∥hx(y)−f(y)∥≤ϵ for every y∈Vx.
The collection {Vx0,Vx}x∈X~ is an open cover for X~.
As X~ is compact, there is a finite subcover {Vx0,Vx1,Vx2,…,Vxk} of X~.
Then there exists a partition of unity {g0,g1,g2,…gk} subordinate to this subcover such that for every i∈{0,1,2,…,k}, gi∈C(X~), 0≤gi≤1, supp(gi)⊆Vxi and for each x∈X~, ∑i=0kgi(x)=1.
Now for h~=∑i=0kgi^h~xi∈C(X~,A), its restriction to X, given by h=∑i=0kgi^∣Xhxi, is an element in J.
Hence for any y∈X, we have
[TABLE]
thus ∥h−f∥≤ϵ.
∎
Remark 2.2**.**
Note that the notation J(S) given in the above theorem coincides with the previous notation of J(F) by taking S={F,X} and I={{0},A=C}.
It is interesting to note that if A possesses only finitely many closed ideals, then the above defined mapping θ turns out to be injective, hence providing a nice characterization of the closed ideals of C0(X,A). We would like to mention that the case when A is a simple C∗-algebra is discussed in [6, Proposition 4.1], wherein we proved that every closed ideal of C0(X,A) is of the form
̃{f∈C0(X,A):f(x)=0,∀x∈F}, for some closed subspace F of X.
Theorem 2.3**.**
Let X be a locally compact Hausdorff space, A be a C∗-algebra and I={I1,I2,…,In=A}, n∈N, be the set of all closed ideals of A.
Then the map θ from TIn to the set of all closed ideals of C0(X,A) is a bijection. In particular, every closed ideal of C0(X,A) is precisely of the form J(S)={f∈C0(X,A):f(Si)⊆Ii,∀i∈Nn}, for a unique S={Si} compatible with I.
Proof.
Let S={Si}i∈Nn and S′={Si′}i∈Nn be two distinct elements of TIn so that Si=Si′ for some i∈Nn∖{n}.
Without loss of generality, we may assume that there exists x∈Si∖Si′.
Define a non-empty subset γ={j∈Nn:x∈/Sj′} of Nn.
By Urysohn’s Lemma [13, Theorem 2.12] , as Vx=(∪i∈γSi′)c is an open set containing x, there exists g∈Cc(X) such that g(x)=1, g(X)⊆[0,1] and supp(g)⊆Vx.
It is easy to see that (∩j∈γcIj)∖Ii=∅, because if ∩j∈γcIj⊆Ii then ∩j∈γcSj′⊆Si′, which is not true as x∈∩j∈γcSj′ but x∈/Si′.
Let a∈(∩j∈γcIj)∖Ii.
Consider the function h=a′g^∈C0(X,A).
Observe that h∈/J(S), since x∈Si but h(x)=a′(x)g^(x)=a∈/Ii. However, we assert that h∈J(S′) which proves J(S)=J(S′). For k∈γ, y∈Sk′ implies y∈Vxc, so that h(y)=ag^(y)=0∈Ik. Also, if k∈γc, then h(y)=ag^(y)∈(∩j∈γcIj)⊆Ik for every y∈Sk′.
∎
In the quest of proving the main result regarding the characterization of closed ideals of C0(X)⊗minA, we require few more ingredients.
Lemma 2.4**.**
Let X be a locally compact Hausdorff space and A be a C∗-algebra.
Then for a closed subspace C of A and a closed ideal J(Y) of C0(X), Y⊆X being closed, there is an isometric isomorphism of Banach spaces
[TABLE]
Proof.
Denote by J the closed subspace {f∈C0(X,A):f(Y)={0},f(X)⊆C} of C0(X,A), and I=J(Y).
Let φ denote the restriction of φ~ to I⊗Cmin, where ϕ~:C0(X)⊗minA→C0(X,A) is the isometric ∗-isomorphism as discussed earlier.
Then for ∑j=1nfj⊗cj∈I⊗C, we have φ(∑j∈Nnfj⊗cj)(Y)={0}, so that φ(I⊗C)⊆J.
Since φ is an isometry, it is sufficient to prove that φ(I⊗C) is dense in J.
Let g∈J and ϵ>0 be arbitrary.
Since J is also a closed subspace of C0(X,C) and Cc(X,C) is dense in C0(X,C), there exists a function h∈Cc(X,C) such that ∥g−h∥<ϵ/2.
Let K:=supp(h),
Br(b):={c∈C:∥c−b∥<r} and Br×(b):=Br(b)∖{0}, where b∈C and r>0.
Since ∥h(y)∥=∥g(y)−h(y)∥<ϵ/2 for every y∈Y, the collection {h−1(B2ϵ×(h(x))∖h(Y)):x∈K∖Y}∪h−1(Bϵ(0)) forms an open cover of the compact set K.
Fix a finite subcover, say, h−1(Bϵ(0))∪{h−1(B2ϵ×(h(xi))∖h(Y)):1≤i≤n}.
Since K is a compact subspace of a locally compact Hausdorff space X, there exists a partition of unity subordinate to this finite subcover, i.e. there exist functions f0,f1,…,,fn in Cc(X) such that 0≤fi≤1 for all 0≤i≤n, supp(f0)⊆U0:=h−1(Bϵ(0)), supp(fi)⊆Ui:=h−1(B2ϵ×(h(xi))∖h(Y)) for all 1≤i≤n and ∑i=0nfi(x)=1 for x∈K(see [13, Theorem 2.13]).
Let V=(∑i=0nfi)−1(0,3/2).
Then V∩(∪i=0nUi) is an open set containing K.
Pick f~∈Cc(X) such that f~ is 1 on K, supp(f~)⊆V∩(∪i=0nUi) and 0≤f~≤1.
Then for fi~=f~fi, supp(fi~)⊆V∩Ui because supp(fi)⊆Ui and supp(f~)⊆V.
Now for x∈K, we have
[TABLE]
Also notice that 0≤∑i=0nfi~≤3/2 because for x∈V∩(∪i=0nUi), ∑i=0nfi~(x)=∑i=0nf~(x)fi(x)≤∑i=0nfi(x)=3/2 and for x∈(V∩(∪i=0nUi))c, we have ∑i=0nfi~(x)=0 .
Now for 1≤i≤n, the open set Ui, and thus V∩Ui is disjoint from Y so that ∑i=1nfi~⊗h(xi)∈I⊗C.
Fix x0∈Kc, then for each x∈X
[TABLE]
Hence we obtain ∥g−φ(∑i=1nfi~⊗h(xi))∥<23ϵ, proving that φ(I⊗C) is dense in J.
∎
As a consequence of the above result, we have an interesting observation which identifies certain closed ideals of C0(X,A) with some closed ideals of C0(X)⊗minA.
Corollary 2.5**.**
Let Y be a closed subspace of a locally compact Hausdorff space X. For any closed ideal I of a C∗-algebra A, we have
[TABLE]
Proof.
Let I={Ii}i∈Δ be the set of all closed ideals of A with Iβ=A, for some β∈Δ, and set I=It.
If t=β then the result is trivial.
Otherwise, let J1=C0(X)⊗minIt+J(Y)⊗minA and J2={f∈C0(X,A):f(Y)⊆It}.
Then, by Theorem 2.1, there exist elements S={Si}i∈Δ and S′={Si′}i∈Δ in TIβ such that J1=J(S) and J2=J(S′). It is sufficient to prove that S=S′.
We first mention a common trick used in the proof.
For any x∈X and a closed subspace F of X with x∈/F, Urysohn’s Lemma implies that there exists f∈Cc(X) such that f(x)=1 and f(F)=0.
Then for any fixed a∈A and any y∈X, there exists a function g(y):=f(y)a in C0(X,A) such that g(x)=a and g vanishes on F.
We now claim that St=∩f∈J1f−1(It)=Y.
For f∈J1, f=f1+f2 for some f1∈C0(X)⊗minIt and f2∈J(Y)⊗minA. Thus, for any y∈Y, f(y)=f1(y)+f2(y)∈It as f2(y)=0 by 2.4, so that Y⊆St. For the reverse containment assume that Y⊊St. Pick a∈A∖It and x∈St∖Y, then there exists a function in C0(X,A) which vanishes on Y and maps x to a which is a contradiction to the definition of St.
On the similar lines, using the fact that St′=∩g∈J2g−1(It), one can easily deduce that St=Y=St′.
Now fix i∈Δ with i=β,i=t. Note that J2=∩i∈Δ{f∈C0(X,A):f(Si′)⊆Ii}, so that Gt:={f∈C0(X,A):f(St′)⊆It}⊆{f∈C0(X,A):f(Si′)⊆Ii}=Gi (say), for every i∈Δ.
Case(i): Ii⊊It, then Si′=∅=Si.
Because for y∈Si′⊆St′ and a∈It∖Ii, there exists a function in C0(X,A) which takes y to a. Then such a function is in Gt but not in Gi. Also, if there exists an x∈Si⊆St=Y, then a′g as defined above will be a function in C0(X)⊗minIt⊂J1 which takes an element x of Si to a which does not belong to Ii, which is a contradiction to the definition of Si.
Case(ii): It⊊Ii, then Si=St=St′=Si′.
To see this, if St′ is properly contained in Si′, then for x∈Si′∖St′ and a∈/Ii, there exists a function in C0(X,A) which takes x to a and St′ to 0. This function belongs to Gt but does not belong to Gi, which is a contradiction. Similarly, if St is properly contained in Si, then for x∈Si∖St and a∈/Ii, there is a function in J(Y)⊗minA⊂J1 which takes x outside Ii which contradicts the definition of Si.
Case(iii): Ii is neither a subset nor a superset of It, then we claim that Si=Si′=∅. If Ij=It∩Ii, then Ij⊊It so that by Case(i), St∩Si=Sj=∅=Sj′=St′∩Si′. Now, for x∈Si′, x∈/St′, as argued in Case (ii), we obtain that Gt is not contained in Gi, which is a contradiction, thus Si′=∅.
Similarly, for x∈Si, x is not a member of Y=St. So for any a∈It∖Ii, applying the technique mentioned in the beginning, we get a function g in J(Y)⊗minA⊆J1 such that g(x)∈/Ii, a contradiction.
This proves that S=S′ and hence J1=J2.
∎
We are now ready to prove the main result of this section. Note that a product ideal is a closed ideal of the form I⊗minJ, where I and J are closed ideals of A and B, respectively.
Theorem 2.6**.**
Let X be a locally compact Hausdorff space and A be a C∗-algebra with finitely many closed ideals, say, I1,I2,…,In with I1={0} and In=A. Then for any closed ideal K of C0(X)⊗minA, there exists S={Si}i∈Nn∈TIn, where I={Ii}i∈Nn, such that
[TABLE]
where γj={i∈Nn:Ij⊈Ii}, for every j∈{2,3,…,n}.
In particular, every closed ideal of C0(X)⊗minA is a finite sum of product ideals.
Proof.
By Theorem 2.3, there exists S={Si}i∈Nn∈TIn such that K=J(S)={f∈C0(X,A):f(Si)⊆Ii,i∈Nn}. Set K′=∑j=2nJ(∪k∈γjSk)⊗minIj, then by 2.4, K′ can be considered as a closed ideal of C0(X,A).
By virtue of Theorem 2.1, it is sufficient to prove that Si=∩f∈K′f−1(Ii) for every i∈Nn.
It is clear that Sn=X=∩f∈K′f−1(A). Fix i∈Nn−1 and consider any x∈Si.
For f∈K′, f=f2+f3+⋯+fn, where fr∈J(∪k∈γrSk)⊗minIr for every r∈{2,3,…,n}.
Then for any such r, either i∈γr or i∈γrc.
If i∈γr then fr(x)=0∈Ii.
If i∈γrc then fr(x)∈Ir⊆Ii.
These two conclusions together imply that Si⊆∩f∈K′f−1(Ii).
Next, pick x∈/Si and define αi={j∈Nn:Ij⊈Ii}. Note that αi is non empty as n∈αi.
It is sufficient to prove the existence of a function f∈K′ such that f(x)∈/Ii.
We shall actually prove that such a function exists in the subset ∑r∈αiJ(∪k∈γrSk)⊗minIr of K′. It is further enough to prove that there exists an r∈αi such that x∈/∪k∈γrSk, so that the required function f exists in J(∪k∈γrSk)⊗minIr.
In fact, by Urysohn’s Lemma there exists a function g∈Cc(X) such that 0≤g≤1, g(x)=1 and g(∪k∈γrSk)={0}. Then by 2.4, for a∈Ir∖Ii (since Ir⊈Ii), the function a′g^ serves the purpose.
We claim that ∩r∈αi(∪k∈γrSk)=Si, which will ensure the existence of such an r.
When r∈αi, we have i∈γr and hence Si⊆∩r∈αi(∪k∈γrSk).
We now prove the reverse inclusion.
Set αi={r1,r2,…,rq} and for each rj∈αi, let there be prj number of elements in γrj, say γrj={rj,trj:1≤trj≤prj}.
So
[TABLE]
We have obtained that ∩r∈αj(∪k∈γrSk) is a union of Πj=1qprj objects, each of which is an intersection of q objects which looks like Sr1,tr1∩Sr2,tr2∩⋯∩Srq,trq.
Pick an ideal Ijr1,tr1∩Ijr2,tr2∩⋯∩Ijrq,trq.
Then there exists an m∈Nn such that Im=Ir1,tr1∩Ir2,tr2∩⋯∩Irq,trq, and hence Sm=Sr1,tr1∩Sr2,tr2∩⋯∩Srq,trq.
If Sm⊆Si, we are done.
Otherwise, Sm⊈Si will imply Im⊈Ii and hence m∈αi.
Thus m=rl for some l∈{1,2,…k}.
Then rl,trl∈γrl which implies (Im=)Irl⊈Irl,trl, which is a contradiction to the fact that Im=Ir1,tr1∩Ir2,tr2∩⋯∩Irq,trq.
∎
Let us demonstrate the above theorem with the help of few examples.
Example 2.7**.**
Let H be a separable Hilbert space and A=B(H)⊕B(H).
For I={I1,…,I9=A}, the lattice of closed ideals of A is given in the diagram below.
[TABLE]
Looking at the diagram, we obtain that γ2={1,3,6}, γ3={1,2,4}, γ4={1,2,3,5,6,8}, γ5={1,2,3,4,6}, γ6={1,2,3,4,5,7}, γ7={1,2,3,4,5,6,8}, γ8={1,2,3,4,5,6,7} and γ9={1,2,3,4,5,6,7,8}.
Hence, every closed ideal of C0(X)⊗minA is of the form:
J(S6)⊗minI2+J(S4)⊗minI3+J(S8)⊗minI4+J(S4∪S6)⊗minI5+J(S7)⊗minI6+J(S8)⊗minI7+J(S7)⊗minI8+J(S7∪S8)⊗minI9, for a unique S=(S1,S2,…,S9)∈TI9, where a pictorial representation of S is the following (here an arrow from Sj to Sk means that Sj⊆Sk, for j,k∈N9):
[TABLE]
We next discuss the precise form of closed ideals of C0(X)⊗minB(H), in terms of product ideals. Note that for a Hilbert space H, the set of all closed ideals forms a chain (see, [10, Corollary 6.2]). In the following, w0 denotes the cardinality of the set of all natural numbers and for every i∈N, let wi=2wi−1 so that w1 is continuum.
Example 2.8**.**
Let X be a locally compact Hausdorff space and H be a Hilbert space with wn (n∈N) as its Hilbert dimension.
Then the closed ideals of C0(X)⊗minB(H) are of the form ∑j=2n+3J(Sj−1)⊗minIj , where Ij’s are closed ideals of B(H) and Sj’s are some closed subspaces of X.
To see this, let {0}=I1⊊I2⊊I3⊊⋯⊊In+3=B(H) be the chain of closed ideals of B(H) and J be a closed ideal of C0(X)⊗minB(H).
By Theorem 2.1, there exists n+3 closed subspaces S1⊆S2⊆S3⊆⋯⊆Sn+3=X such that J=J(S) where S={Si}i=1n+3.
As in Theorem 2.6, for j∈Nn+3, ∪k∈γjSk=Sj−1 and hence J=∑j=2n+3J(Sj−1)⊗minIj.
3. Closed Lie ideals of C0(X,A)
The Lie normalizer of a subspace S of a Lie algebra A is defined by N(S):={a∈A:[a,A]⊆S}.
It can be easily verified that N(I) is a closed subalgebra of A for a closed ideal I in A.
The Lie normalizer plays an important role in determining the Lie ideals of A (for instance, see [4],[6]). We identify the Lie normalizer of ideals of C0(X,A) and use this identification to characterize its closed Lie ideals.
Theorem 3.1**.**
Let X be a locally compact Hausdorff space and A be a C∗-algebra with I={Ii}i∈Δ as the collection of all closed ideals such that Iβ=A.
Then a closed subspace L of C0(X,A) is a closed Lie ideal if and only if there is an element S={Si}i∈Δ∈TIβ such that
[TABLE]
Proof.
We know that a closed subspace L of the C∗-algebra C0(X,A) is a Lie ideal if and only if there exists a closed ideal J⊆C0(X,A) such that [J,C0(X,A)]⊆L⊆N(J) ([3, Proposition 5.25, Theorem 5.27]).
By Theorem 2.1, J=J(S) for some S={Si}i∈Δ∈TIβ.
Since for any fixed a∈A and x∈X, there is an element in C0(X,A) which takes x to a, we have
[TABLE]
We know from [3, Proposition 5.25] that [I,B]=I∩[B,B] for closed ideal I of a C∗-algebra B.
This fact, along with 2.4 gives
[TABLE]
Hence the result.
∎
From the last result one can observe that if N(I)=I+Z(A), Z(A) being the centre of A, for every closed ideal I of A, then for a closed ideal J of
C0(X,A), N(J)={f∈C0(X,A):f(Si)⊆Ii+Z(A),∀i∈Δ} for some {Si}i∈Δ∈TIβ.
The question that comes next is the following: Can we write every element of N(J) as g+h such that g(Si)⊆Ii for every i∈Δ and h is Z(A)-valued?
We shall observe in 3.6 that a positive answer of this question will help us to obtain a better representation of all closed Lie ideals of C0(X,A).
Recall that a C∗-algebra A is said to have the centre-quotient property if Z(A/I)=(Z(A)+I)/I for every closed ideal I of A (see [2] for details). Using the fact that for the natural quotient map
π:A→A/I, N(I)=π−1(Z(A/I)), we have a nice relation between Lie normalizer and the
centre-quotient property.
Lemma 3.2**.**
A C∗-algebra A has centre-quotient property if and only if N(I)=I+Z(A) for every closed ideal I in A.
A unital C∗-algebra A is called weakly central if the continuous surjection ψ:Max(A)→Max(Z(A)) given by ψ(I)=I∩Z(A) is an injection, where Max(B) denotes the space of all maximal ideals of a C∗-algebra B endowed with the hull-kernel topology.
It is well known that a unital C∗-algebra with unique maximal ideal must have one dimensional centre [2, Lemma 2.1].
Since weak centrality and centre-quotient property are equivalent in unital css (see, [15, Theorem 1 and 2]),
presence of unique maximal ideal in a unital C∗-algebra A implies that A has centre-quotient property.
In [6, Lemma 4.6] it was observed that for a simple unital C∗-algebra A and a closed ideal I⊆A, N(I)=I+C0(X,C1).
Theorem 3.3**.**
Let X be a locally compact Hausdorff space and A be a unital C∗-algebra with unique maximal ideal.
Then N(J)=J+C0(X,C1) for any closed ideal J of C0(X,A).
Proof.
Note that Z(C0(X,A))=C0(X,Z(A))=C0(X,C1) and hence J+C0(X,C1)⊆N(J).
Let I={Ii}i∈Δ be the collection of all closed ideals of A with A=Iβ and let Iβ′ be the unique maximal ideal of A, β,β′∈Δ.
Then, Theorem 2.1, there exists an element S={Si}i∈Δ∈TIβ such that J=J(S).
Since A has centre-quotient property, by 3.2, N(Ii)=Ii+C1 for every i∈Δ.
Let f∈N(J)={g∈C0(X,A):g(Si)⊆Ii+C1,∀i∈Δ} as noted in Theorem 3.1.
Since Iβ′ is the unique maximal ideal of A and S∈TIβ, we have Sα⊆Sβ′ for every α∈Δ∖{β}.
On Sβ′, write f=g+h which satisfy h(Sβ′)⊆C1 and g(Si)⊆Ii for every i∈Δ∖{β}.
This is possible as no proper ideal of A can intersect C1.
Since Iβ′∩C1={0}, by Hahn-Banach Theorem, there exists T∈A∗ such that ∥T∥=1, T(Iβ′)={0} and T(λ1)=λ for λ∈C.
Then Tf=Th on Sβ′.
Also f vanishes at infinity and ∥T∥=1, so we obtain that Tf vanishes at infinity because ∥Tf∥≤∥f∥.
Since ∥Th∥=∥h∥, Th and hence h is a continuous function vanishing at infinity.
So g=f−h is continuous on Sβ′ and is vanishing at infinity.
By [6, Theorem 4.5], there exists an h′∈C0(X) such that h∣Sβ′′=h.
For x∈Sβ′c, define g′(x)=f(x)−h′(x).
Then f=g′+h′ with g′∈J(S) and h′∈C0(X,C1) and we are done.
∎
It is observed in the previous result that if A is a unital C∗-algebra with a unique maximal ideal then C0(X,A) has the centre-quotient property.
We now generalize this result and prove that the centre-quotient property of A passes to C0(X,A). We would like to point out that the proof given above does not work when A has more than one maximal ideal because in this case a closed ideal may intersect the centre non-trivially, that is, I∩Z(A)={0}, and this is where the proof will fail.
As an intermediate step towards this generalization, we provide the following result.
Proposition 3.4**.**
Let X be a compact Hausdorff space and A be a unital C∗-algebra having centre-
quotient property. Then C(X,A) has centre-quotient property.
Proof.
It is sufficient to prove that C(X,A) is weakly central. Consider maximal ideals J1 and J2 of C(X,A) such that J1∩Z(C(X,A))=J2∩Z(C(X,A)).
For i=1,2, there exist maximal ideals Ii of A and xi∈X such that Ji={f∈C(X,A):f(xi)∈Ii} ([12, Corollary V.26.2.2]). With the help of Urysohn’s Lemma, one can easily verify that Ii and xi are unique. Since Z(C(X,A))=C(X,Z(A)), we have Ji∩Z(C(X,A))={f∈C(X,Z(A)):f(xi)∈Ii∩Z(A)}. By the uniqueness, x1=x2 and I1∩Z(A)=I2∩Z(A). Since A is weakly central, we obtain I1=I2 and hence J1=J2.
∎
Theorem 3.5**.**
If X is a locally compact Hausdorff topological space and A is a C∗-algebra.
If C0(X,A) has centre-quotient property then A has centre-quotient property.
Converse is true if A is unital.
Proof.
In order to prove that A has centre-quotient property, it is sufficient to prove that for a closed ideal I of A, N(I)⊆I+Z(A). For a fixed x∈X, consider
a closed ideal J of C0(X,A) given by {f∈C0(X,A):f(x)∈I}. Now, for a∈N(I), let f∈C0(X,A) such that f(x)=a.
For any g∈C0(X,A), (fg−gf)(x)=ag(x)−g(x)a∈I, which implies that f∈N(J).
Since C0(X,A) has centre-quotient property, f=g+h, for some g∈J and h∈C0(X,Z(A)).
Thus a=f(x)=g(x)+h(x)∈I+Z(A).
Conversely, suppose that A is unital and has centre-quotient property.
If X~ denotes the one point compactification of X, then we have a natural inclusion C0(X,A)⊆C(X~,A). For a closed ideal J of C0(X,A), we claim that N(J)C0(X,A)⊆N(J)C(X~,A), where N(J)B represents the Lie normalizer in B. Let f∈N(J)C0(X,A) and {fμ} be a quasi central approximate identity of the closed ideal C0(X,A) of C(X~,A).
Then for any g∈C(X~,A),
[TABLE]
since {fμ} is quasi central approximate identity and gf∈C0(X,A).
Note that fμg∈C0(X,A) and f∈N(J)C0(X,A) together imply that ffμg−fμgf∈J for every μ.
Thus fg−gf∈J which gives f∈N(J)C(X~,A).
Since C(X~,A) has centre-quotient property (Theorem 3.5) and
J is also a closed ideal of C(X~,A), we have N(J)C(X~,A)=J+C(X~,Z(A)).
Now, for f∈N(J)C0(X,A), there exist g∈J and h∈C(X~,Z(A)) such that f=g+h.
If X~=X∪{∞}, then f(∞)=0=g(∞) which gives h(∞)=f(∞)−g(∞)=0 so that h∈C0(X,Z(A)). Thus N(J)C0(X,A)⊆J+C0(X,Z(A)) and hence the result holds.
∎
We now characterize the Lie ideals of a class of C∗-algebras. Recall that a bounded linear functional f on a C∗-algebra A is said to be a tracial state if f is positive of norm 1 and f([a,b])=0 for every a,b∈A.
Corollary 3.6**.**
Let X be a locally compact Hausdorff space and A be a unital C∗-algebra with centre-quotient property and no tracial states.
Then a closed subspace L of C0(X,A) is a Lie ideal if and only if it is of the form J+K for some closed ideal J of C0(X,A) and a closed subspace K of C0(X,Z(A)).
Proof.
Since A has no tracial states, from [6, Lemma 2.4], C0(X,A) has no tracial states. Thus, by [3, Proposition 5.25], [J,A]=J for every closed ideal J of C0(X,A).
From [3, Theorem 5.27] and Theorem 3.5, a closed subspace L of C0(X,A) is a Lie ideal if and only if there exists a closed ideal J of C0(X,A) such that J⊆L⊆J+C0(X,Z(A)).
Hence L must be of the form J+K for some closed subspace K of C0(X,Z(A)).
∎
As a consequence, we can now characterize all closed ideals of C0(X)⊗minB(H).
Corollary 3.7**.**
For a separable Hilbert space H and a locally compact Hausdorff space X, a closed subspace L of C0(X)⊗minB(H) is a Lie ideal if and only if there exist two closed subspaces S1⊆S2 of X and a closed subspace K of C0(X)⊗C1 such that
[TABLE]
Proof.
Since B(H) has no
tracial states and has a unique maximal ideal, the result is an easy consequence of Example 2.8 and 3.6.
∎
Bibliography16
The reference list from the paper itself. Each links out to its DOI / PubMed record.
1[1] S. D. Allen, A. M. Sinclair and R. R. Smith, The ideal structure of the Haagerup tensor product of C ∗ superscript 𝐶 ∗ C^{\ast} -algebras, J. Reine Angew. Math. 442 (1993), 111–148.
2[2] R. J. Archbold, L. Robert and A. Tikuisis, The Dixmier property and tracial states for C ∗ superscript 𝐶 C^{*} -algebras, J. Funct. Anal. 273 (8) (2017), 2655–2718.
3[3] M. Brešar, E. Kissin and V. S. Shulman, Lie ideals: from pure algebra to C ∗ superscript 𝐶 C^{*} -algebras, J. Reine Angew. Math. 623 (2008), 73–121.
4[4] V. P. Gupta and R. Jain, On closed Lie ideals of certain tensor products of C ∗ superscript 𝐶 C^{*} -algebras, Math. Nachr. 291 (8-9) (2018), 1297–1309.
5[5] V. P. Gupta and R. Jain, On Banach space projective tensor product of C ∗ superscript 𝐶 C^{*} -algebras, To appear in Banach J. Math. Anal.
6[6] V. P. Gupta, R. Jain and B. Talwar, On closed Lie ideals of certain tensor products of C ∗ superscript 𝐶 C^{*} -algebras II, Math. Nachr. 293 (2020), 1-14.
7[7] R. Jain and A. Kumar, Ideals in operator space projective tensor product of C ∗ superscript 𝐶 C^{*} -algebras, J. Aust. Math. Soc. 91 (2) (2011), 275–288.
8[8] R. V. Kadison and J. R. Ringrose, Fundamentals of the theory of operator algebras, Vol. I , Pure and Applied Mathematics, 100 , Academic Press, Inc., New York, 1983.