Maximal Algebraic Ideals in Nonunital $C^{*}$ -Algebras

Zhichao Liu Address: School of Mathematical Sciences, Dalian University of Technology, Dalian, 116024, China Email address: lzc.12@outlook.com and Xin Ma Address: Institute for Advanced Study in Mathematics, Harbin Institute of Technology, Harbin, China, 150001 Email address: xma17@hit.edu.cn

Abstract.

Motivated by Ozawa’s question of whether every maximal algebraic two-sided ideal in a $C^{*}$ -algebra must be closed, we study the existence of maximal algebraic two-sided ideals in nonunital $C^{*}$ -algebras. We formulate singular-distribution estimates intrinsically through lower semicontinuous $2$ -quasitraces and apply them to control algebraic ideal membership. This allows to develop a novel criterion—the admissible quasitracial projection scale—for establishing the nonexistence of maximal ideals in nonunital $C^{*}$ -algebras. This criterion applies to a wide class of simple $C^{*}$ -algebras, including

(i)

all $A\otimes\mathcal{K}$ where $A$ is unital, simple, stably finite, $\operatorname{QT}_{2}^{1}(A)\neq\varnothing$ , and the radius of comparison $\operatorname{rc}(A)$ is finite, as well as all their hereditary $C^{*}$ -subalgebras whenever $A$ satisfies some further minor assumptions.
(ii)

all nonunital, simple, separable, stably finite, $\mathcal{Z}$ -stable $C^{*}$ -algebras $A$ that have an approximate identity consisting of increasing projections $(p_{n})$ and for which the simplex $\operatorname{QT}_{2}^{1}(A,p_{1})$ of normalized traces at $p_{1}$ has finitely many extreme points;

We also show that a large class of $C^{*}$ -algebras $E$ constructed from extensions of $C^{*}$ -algebras above such that $E$ still has no maximal ideals. In particular, for these classes, Ozawa’s question could be settled in an unexpected manner.

Key words and phrases:

Maximal ideal, Cuntz semigroup, generalized singular number, Pedersen ideal

2020 Mathematics Subject Classification

Primary 46L05, 46L35; Secondary 16D25, 19K14

1. Introduction

An algebraic ideal of a $C^{*}$ -algebra is a complex-linear two-sided ideal which need not be norm closed. Such ideals are less rigid than norm-closed ideals, but they arise naturally and have been studied since Pedersen’s work on measure theory and the work of Pedersen and Petersen on ideals closed under roots [20, 22]. In particular, every $C^{*}$ -algebra $A$ has a smallest dense ideal, the Pedersen ideal $\operatorname{Ped}(A)$ [19, Section 5.6]. In a nonunital algebra this ideal can be proper, even when $A$ is simple, which already shows that algebraic ideals may behave quite differently from their closed counterparts.

It is known that if all algebraic ideals in $A$ are modular (e.g., $A$ is unital), every maximal algebraic ideal is closed. In addition, it is not hard to see all maximal algebraic ideals in $C_{0}(X,A)$ are closed provided that $X$ is locally compact Hausdorff space and $A$ is a unital $C^{*}$ -algebra. Ozawa asked whether this remains true for a general $C^{*}$ -algebra [18]; see also [14, Question 2.8] and [15, Remark 5.15]. The question remains open in general. Gardella and Thiel proved that every proper algebraic ideal is contained in a prime ideal [14]. Gardella, Kitamura, and Thiel subsequently showed that semiprime ideals are positively spanned, self-adjoint, hereditary, and closed under positive roots [15]. Throughout the paper, ideals are algebraic complex-linear two-sided ideals unless specified otherwise. A proper ideal $J$ of $A$ satisfies $0\subsetneq J\subsetneq A$ , and a maximal ideal is maximal among these ideals. Recall that a $C^{*}$ -algebra is said to be simple if it has no nonzero proper closed ideals and $A$ is said to be algebraically simple if $A$ has no nonzero proper algebraic ideals. It is well-known that $A$ is algebraically simple if and only if $A$ is simple and $\operatorname{Ped}(A)=A$ . An obvious example illustrating the difference between these two concepts is the compact operators $\mathcal{K}$ , which is simple and its Pedersen ideal consists of the finite-rank operators.

For a simple $C^{*}$ -algebra $A$ , every nonzero algebraic ideal is dense; hence any maximal ideal in $A$ , if it exists, must be non-closed. In this case, the mere existence of such an ideal would already answer Ozawa’s question in the negative. For non-simple $C^{*}$ -algebras, by contrast, proper closed ideals may exist, yet it remains unknown whether a maximal ideal—closed or not—always exists. The existence question is therefore fundamental: if no such ideal exists, then the issue of whether it is closed becomes vacuous, and Ozawa’s question has no object to which it can apply. This naturally leads to the following question, which may be regarded as a preliminary counterpart to Ozawa’s.

Question 1.1.

Does every nonunital $C^{*}$ -algebra admit a maximal ideal?

Zorn’s lemma cannot be applied directly here, since in a nonunital algebra an increasing chain of proper ideals may exhaust the algebra. The main technical difficulty in answering this question is to decide membership in maximal ideals. Ordinary Cuntz subequivalence alone does not yield such a conclusion, even when the ideal is maximal. Our main technical contribution is to use generalized singular functions to control membership in nonclosed ideals (Theorem 4.7). These functions are defined through the cutdown distributions $t\mapsto d_{\tau}((|x|-t)_{+})$ associated with a lower semicontinuous $2$ -quasitrace $\tau$ (Definition 3.3). Their role is to retain quantitative information about spectral decay and to make this information compatible with quasitracial comparison and projection packing. For a faithful trace, this notion coincides with the generalized singular numbers introduced by Fack and Kosaki [13] in the associated semifinite von Neumann algebra. Combining exact comparison with a slowly decreasing diagonal factorization argument, we rule out dense maximal algebraic ideals in algebras that admit an admissible quasitracial projection scale (Definition 4.2). The criterion depends primarily on a set $\mathcal{T}$ of certain lower semicontinuous $2$ -quasitraces on $A$ , each of which may be infinite or finite; the two cases require distinct modes of control. We denote the infinite and finite parts by $\mathcal{T}_{\inf}$ and $\mathcal{T}_{\mathrm{fin}}$ , respectively. This separation is what gives the criterion its flexibility and allows it to reach a notably broad range of nonunital settings. Our main result is as follows.

Main theorem (Theorem 5.3 and Corollary 5.4). A $C^{*}$ -algebra admitting an admissible quasitracial projection scale has no dense maximal ideals. In particular, a simple $C^{*}$ -algebra admitting such a scale has no maximal ideals.

This theorem is then applied to establish the nonexistence of maximal ideals in a substantial range of stable and nonstable classes. These results may be viewed as partial evidence that Ozawa’s question admits an affirmative answer: for a large class of simple nonunital $C^{*}$ -algebras $A$ , they have no maximal ideals at all.

Corollary 1.2 (Theorem 6.9).

Let $D$ be unital, simple, and stably finite, with $\operatorname{QT}^{1}_{2}(D)\neq\varnothing$ . If $\operatorname{rc}_{\mathrm{p}}(D)<\infty$ , then $D\otimes\mathcal{K}$ has no maximal ideals and its Pedersen ideal is proper. In particular, this holds when $\operatorname{rc}(D)<\infty$ , hence when $D$ has strict comparison.

Note that $\mathcal{T}_{\rm fin}$ is empty for $D\otimes\mathcal{K}$ in Corollary 1.2. The following applies to the cases that both $\mathcal{T}_{\inf}$ and $\mathcal{T}_{\rm fin}$ may be non-empty. For example, if $A$ is assumed to satisfy real rank $0$ and has finitely many extreme traces, then it is demonstrated in Corollary 6.15 that every non-zero hereditary $C^{*}$ -subalgebra still lacks maximal ideals. As a consequence, we present in Example 6.16 a concrete AF example that both $\mathcal{T}_{\inf}$ and $\mathcal{T}_{\rm fin}$ are non-empty.

Corollary 1.3 (Corollary 6.14).

Let $A$ be simple, separable, nonunital, stably finite, and $\mathcal{Z}$ -stable, with an increasing projection approximate identity $(p_{n})$ . If $\operatorname{QT}^{1}_{2}(A,[p_{1}])$ has finitely many extreme points, then $A$ has no maximal ideals.

We refer to Theorem 6.13 for a stronger version of Corollary 1.3, which particularly includes nonunital simple AF algebras with finitely many traces. The finiteness assumption on the set of extreme normalized quasitraces in Corollary 1.3 is imposed only for convenience and is by no means essential: it can be replaced by a possibly weaker condition that supplies uniform control over all quasitraces in $\mathcal{T}_{\rm fin}$ .

In the nonsimple setting, there are still many $C^{*}$ -algebras without maximal ideals. Denote by $\mathcal{C}$ the class of non-unital $C^{*}$ -algebras $A$ without maximal ideal and $\operatorname{Ped}(A)\neq A$ . For example, the class $\mathcal{C}$ includes $C^{*}$ -algebras $A$ with $\operatorname{Ped}(A)\neq A$ satisfying the main theorem, which, among others, particularly includes that in Corollaries 1.2 and 1.3. It is shown in Theorem 7.2 that $\mathcal{C}$ is preserved by $C^{*}$ -algebra extensions. Therefore, starting from main theorem, one constructs many $C^{*}$ -algebras, which are not algebraically simple and have no maximal ideals.

Outline of the paper

The paper is organized as follows. Section 2 collects preliminaries on Cuntz subequivalence and algebraic two-sided ideals in $C^{*}$ -algebras. In Section 3, we introduce generalized singular functions for $C^{*}$ -algebras. Section 4 develops the admissible quasitracial projection scales and establishes a key technical result (Theorem 4.7) for later use. In Section 5, we prove the main theorem. In Section 6, we apply this theorem to construct a class of simple examples without maximal ideals. Finally, in Section 7, we use extensions of $C^{*}$ -algebras to obtain non-simple $C^{*}$ -algebras without maximal ideals.

2. Preliminaries

2.1. Cuntz comparison

Definition 2.1.

Let $A$ be a $C^{*}$ -algebra. Let $p$ and $q$ be two projections in $A$ . Recall that $p$ is Murray–von Neumann equivalent to $q$ in $A$ , written $p\sim_{\rm MvN}q$ , if there exists $x\in A$ such that $x^{*}x=p$ and $xx^{*}=q$ . We will write $p\preceq q$ if $p$ is equivalent to some subprojection of $q$ . If $A$ is not unital, let us denote the minimal unitization of $A$ by $A^{\sim}$ .

Definition 2.2.

(Cuntz semigroup) (see, e.g., [8, 10]) Denote by $A_{+}$ the positive cone of $A$ . Let $a,b\in A_{+}$ . One says that $a$ is Cuntz subequivalent to $b$ , denoted by $a\precsim_{\rm Cu}b$ , if there exists a sequence $(r_{n})$ in $A$ such that $r_{n}^{*}br_{n}\rightarrow a$ . One says that $a$ is Cuntz equivalent to $b$ , denoted by $a\sim_{\rm Cu}b$ , if $a\precsim_{\rm Cu}b$ and $b\precsim_{\rm Cu}a$ . The Cuntz semigroup of $A$ is defined as ${\rm Cu}(A)=(A\otimes\mathcal{K})_{+}/\sim_{\rm Cu}$ . We will denote the class of $a\in(A\otimes\mathcal{K})_{+}$ in ${\rm Cu}(A)$ by $[a]$ . Note that ${\rm Cu}(A)$ is a positively ordered abelian semigroup with zero (or monoid) when equipped with the addition: $[a]+[b]=[a\oplus b]$ , and the relation:

[a]\leq[b]\Leftrightarrow a\precsim b,\quad a,b\in(A\otimes\mathcal{K})_{+}.

Let $x\in A_{+}$ and $t\in\mathbb{R}_{+}$ , and set $e_{t}(x)=\max\{x-t,0\}=(x-t)_{+}$ .

Lemma 2.3 ([23]).

Let $A$ be a C*-algebra, let $a,b\in A_{+}$ , and let $p,q$ be projections. Then

(i) $a\precsim b$ if and only if $(a-\varepsilon)_{+}\precsim b$ for all $\varepsilon>0$ ;

(ii) if $\|a-b\|<\varepsilon$ , then $(a-\varepsilon)_{+}\precsim b$ ;

(iii) if $0\leq a\leq b$ , then $a\precsim b$ ;

(iv) $p\preceq q$ if and only if $p\precsim q$ .

Definition 2.4.

For $a\in A_{+}$ , let $A_{a}:=\overline{aAa}$ be the hereditary algebra generated by $a$ and write $p_{a}=1_{(0,\infty)}(a)\in A^{**}$ for the support (spectral) projection of $a$ .

The following lemma is well-known.

Lemma 2.5.

Let $p\in A$ be a projection and $a\in A_{+}$ . If $p\precsim a$ , then there exists a partial isometry $v\in A$ such that $p=v^{*}v$ and $vv^{*}\in A_{a}$ .

Definition 2.6.

([4, 3]) A (bounded) quasitrace on a $C^{*}$ -algebra $A$ is a function $\tau:A\rightarrow\mathbb{C}$ such that:

(i) $0\leq\tau\left(x^{*}x\right)=\tau\left(xx^{*}\right)$ for all $x$ in $A$ ;

(ii) $\tau$ is linear on commutative ^∗-subalgebras of $A$ ;

(iii) If $x=a+ib$ with $a,b$ self-adjoint, then $\tau(x)=\tau(a)+i\tau(b).$

If $\tau$ extends to a quasitrace on $M_{2}(A)$ , then $\tau$ is called a 2-quasitrace. A linear quasitrace is a trace.

If $A$ is unital and $\tau(1)=1$ , then we say $\tau$ is normalized. Denote by $\operatorname{QT}^{1}_{2}(A)$ the space of all the normalized 2-quasitraces on $A$ and by $T(A)$ the space of all the tracial states on $A$ .

Remark 2.7.

It is an open question whether every 2-quasitrace on a $C^{*}$ -algebra is a trace (asked by Kaplansky). A theorem of Haagerup [16] says that if $A$ is exact and unital then every bounded 2-quasitrace on $A$ is a trace. This theorem can be extended to obtain that every lower semicontinuous 2-quasitrace (not necessarily bounded) on an exact $C^{*}$ -algebra must be a trace (see [4, Remark 2.29(i)]). Brown and Winter [7] presented a short proof of Haagerup’s result in the finite nuclear dimension case.

Definition 2.8.

Let $\tau$ be a lower semicontinuous $2$ -quasitrace on $A$ , extended canonically to $A\otimes\mathcal{K}$ . We define a map $d_{\tau}:(A\otimes\mathcal{K})_{+}\rightarrow[0,\infty]$ by

d_{\tau}(a)=\lim_{n\rightarrow\infty}\tau(a^{1/n}).

It has the following properties:

(1) if $a\precsim b$ , then $d_{\tau}(a)\leq d_{\tau}(b)$ ;

(2) if $a$ and $b$ are mutually orthogonal, then $d_{\tau}(a+b)=d_{\tau}(a)+d_{\tau}(b)$ ;

(3) $d_{\tau}((a-\varepsilon)_{+})\rightarrow d_{\tau}(a)$ ( $\varepsilon\rightarrow 0$ ).

This map depends only on the Cuntz equivalence class of $a\in A\otimes\mathcal{K}$ . Hence, we will also $d_{\tau}$ to denote the induced functional on ${\rm Cu}(A)$ .

Definition 2.9.

The dimension-finite ideal of $\tau$ is

\mathcal{D}_{\tau}(A)=\operatorname{span}\{a\in A_{+}:d_{\tau}(a)<\infty\}.

The quasitrace $\tau$ is densely finite if $d_{\tau}(a)<\infty$ for every $a\in\operatorname{Ped}(A)_{+}$ , equivalently $\operatorname{Ped}(A)\subseteq\mathcal{D}_{\tau}(A)$ . For lower semicontinuous $2$ -quasitraces this is equivalent to being finite on a dense ideal, the usual definition of dense finiteness; see [4, 12]. It is not hard to see $\mathcal{D}_{\tau}(A)$ is an algebraic ideal with positive cone $\{a\in A_{+}:d_{\tau}(a)<\infty\}$ and is contained in $N_{\tau}=\{a\in A:\tau(a^{*}a)<\infty\}$ .

2.2. Algebraic ideals

Definition 2.10.

A $C^{*}$ -algebra $A$ is simple if it has no nontrival closed ideals. It is algebraically simple if it contains no nontrival ideals. An ideal $J$ of $A$ is semiprime if

\{xy:x,y\in I\}\subseteq J\quad\Longrightarrow\quad I\subseteq J

for every algebraic two-sided ideal $I\lhd A$ .

A simple unital $C^{*}$ -algebra is algebraically simple. A nonunital simple $C^{*}$ -algebra need not be algebraically simple.There is one especially important and well-behaved ideal in any $C^{*}$ - algebra; see [19, Section 5.6].

Theorem 2.11.

Let $A$ be a ${C}^{*}$ -algebra, there is a dense ideal ${\rm Ped}(A)$ in $A$ which is contained in every dense ideal of $A$ . ${\rm Ped}(A)$ is called the Pedersen ideal of $A$ , and has the following properties:

(i) it contains all projections in $A$ ;

(ii) it is self-adjoint, hereditary, and spanned by its positive elements.

(iii) it is strongly invariant: $x^{*}x\in{\rm Ped}(A)\Leftrightarrow xx^{*}\in{\rm Ped}(A).$

Remark 2.12.

If $A$ is simple, let $p\in A$ be a nonzero projection. Then $p$ is a full projection, and the ideal generate by $p$ is exactly $\operatorname{Ped}(A).$

Definition 2.13.

For a nonzero $C^{*}$ -algebra $A$ , denote by $\Sigma(A)$ the algebraic quotient algebra $A/\operatorname{Ped}(A).$ In general, $\operatorname{Ped}(A)$ need not be norm closed, so $\Sigma(A)$ need not be a $C^{*}$ -algebra.

Theorem 2.14.

Let $A$ be a nonzero $C^{*}$ -algebra and let $M$ be a maximal ideal. Exactly one of the following holds.

(i)

$M$ is dense, $\operatorname{Ped}(A)\subseteq M$ , and $\Sigma(A)/(M/\operatorname{Ped}(A))$ is nonzero and algebraically simple;
(ii)

$M$ is norm closed and $A/M$ is algebraically simple.

Conversely, ideals $N\lhd\Sigma(A)$ with $\Sigma(A)/N$ nonzero and algebraically simple lift to dense maximal ideals of $A$ , and a nonzero proper closed ideal $J$ is maximal algebraically whenever $A/J$ is algebraically simple.

Proof.

If $M$ is dense, Theorem 2.11 gives $\operatorname{Ped}(A)\subseteq M$ , and the correspondence theorem gives the assertion in the quotient. Conversely, the inverse image of such an ideal $N$ is maximal, contains the nonzero ideal $\operatorname{Ped}(A)$ , and is dense.

Suppose that $M$ is not dense. Then $\overline{M}$ is a proper algebraic ideal containing $M$ . Maximality gives $M=\overline{M}$ . Ideals of $A/M$ lift to ideals of $A$ containing $M$ , so $A/M$ is algebraically simple. The last assertion follows from the same ideal correspondence. ∎

Corollary 2.15.

A nonzero $C^{*}$ -algebra $A$ has no maximal ideals if and only if both of the following conditions hold:

(i)

$\Sigma(A)$ has no nonzero algebraically simple quotient.
(ii)

there is no nonzero proper closed ideal $J\lhd A$ such that $A/J$ is algebraically simple.

Proposition 2.16.

([15, Proposition 5.16]) Every maximal ideal of a $C^{*}$ -algebra is prime and thus semiprime.

We use the following exact consequences of [15, Theorem 5.2 and Corollary 5.5].

Proposition 2.17.

Let $J$ be a semiprime algebraic ideal of a $C^{*}$ -algebra $A$ . Then:

(i)

$J$ is idempotent, positively spanned, and self-adjoint;
(ii)

$J$ is hereditary: $0\leq x\leq y\in J_{+}$ implies $x\in J$ ;
(iii)

$J$ is strongly invariant: $x^{*}x\in J$ if and only if $xx^{*}\in J$ ;
(iv)

$y\in J_{+}$ implies $y^{1/2}\in J$ .

Remark 2.18.

Let $M$ be a maximal ideal and let $a,b\in A_{+}$ with $a\precsim b\in M$ . For every $\varepsilon>0$ , there exist $\delta>0$ and $x\in A$ such that

(a-\varepsilon)_{+}=x^{*}(b-\delta)_{+}x.

Proposition 2.17 gives $(b-\delta)_{+}\in M$ , and hence $(a-\varepsilon)_{+}\in M$ . One cannot pass to the norm limit and conclude $a\in M$ . We provide a concrete example here and this issue will be addressed in Theorem 4.7. Let $A=\mathcal{K}(\ell^{2})$ and $J=\bigcup_{m=1}^{\infty}S_{m}$ , in which $S_{m}$ denotes the Schatten $m$ -class. Note that $J_{+}$ is closed under the square-root and thus $J$ is semiprime by [15, Theorem A]. Now, let $a=\operatorname{diag}(2^{-n})$ and $b=\operatorname{diag}(\frac{1}{\log(n+1)})$ . Then it can be verified that $a\notin J,b\in J$ and $[a]=[b]$ in $\operatorname{Cu}(A)$ .

3. Generalized singular functions

In this section, we first introduce some generalized comparison results, and then we construct the generalized singular functions for $C^{*}$ -algebras and establish similar properties.

Lemma 3.1.

For $x\in A$ and $\alpha\geq 0$ , there is $x_{\alpha}\in A$ such that

|x_{\alpha}|=(|x|-\alpha)_{+},\qquad\lVert x-x_{\alpha}\rVert\leq\alpha.

Proof.

For $\alpha=0$ , take $x_{0}=x$ . For $\alpha>0$ , let $x=v|x|$ be the polar decomposition in $A^{**}$ and put

g_{\alpha}(t)=\begin{cases}(t-\alpha)_{+}/t,&t>0,\\ 0,&t=0.\end{cases}\qquad x_{\alpha}=xg_{\alpha}(|x|).

Then $x_{\alpha}=v(|x|-\alpha)_{+}\in A$ , and the assertions follow from functional calculus. ∎

Lemma 3.2.

Let $x,y\in A$ , let $u,v\in A^{\sim}$ , and let $\alpha,\beta,\delta>0$ . Then we have

(i) $|uxv|\precsim|x|$ ;

(ii) $\|a-b\|<\eta$ $\Rightarrow$ $(|a|-\eta)_{+}\precsim|b|$ ;

(iii) $(|x+y|-\alpha-\beta-\delta)_{+}\precsim(|x|-\alpha)_{+}\oplus(|y|-\beta)_{+};$

(iv) $(|uxv|-\lVert u\rVert\lVert v\rVert\alpha-\delta)_{+}\precsim(|x|-\alpha)_{+}.$

Proof.

(i) We shall also use the basic facts $c\sim c^{2}$ and $z^{*}z\sim zz^{*}$ ( $c\in A_{+}$ , $z\in A$ ).

Since we have

|uxv|^{2}=v^{*}x^{*}u^{*}uxv\leq\lVert u\rVert^{2}v^{*}x^{*}xv\sim v^{*}|x|^{2}v,

and

v^{*}|x|^{2}v=(|x|v)^{*}(|x|v)\sim(|x|v)(|x|v)^{*}=|x|vv^{*}|x|\leq\lVert v\rVert^{2}|x|^{2}\precsim|x|^{2},

Now we conclude that

|uxv|\sim|uxv|^{2}\precsim|x|^{2}\sim|x|.

(ii) Let us view $A$ as a $C^{*}$ -subalgebra of its enveloping von Neumann algebra $A^{**}$ , and consider a faithful representation of $A^{**}$ as a subset of $B(\mathcal{H})$ .

Denote

\varepsilon=\|a-b\|,\quad z=(|a|-\eta)_{+},\quad p=\mathds{1}_{(\eta,\infty)}(|a|).

For every $\xi\in p\mathcal{H}$ ,

\|a\xi\|=\||a|\xi\|\geq\eta\|\xi\|.

Consequently,

\|b\xi\|\geq\|a\xi\|-\|(a-b)\xi\|\geq(\eta-\varepsilon)\|\xi\|.

Thus,

pb^{*}bp\geq(\eta-\varepsilon)^{2}p.

Since $z^{1/2}=pz^{1/2},$ then multiplying $z^{1/2}$ on both sides gives

z^{1/2}b^{*}bz^{1/2}\geq(\eta-\varepsilon)^{2}z.

Therefore,

z\precsim z^{1/2}b^{*}bz^{1/2}\sim bzb^{*}\leq\|z\|bb^{*}\sim b^{*}b=|b|^{2}\sim|b|.

Finally,

(|a|-\eta)_{+}\precsim|b|.

(iii) Let $x_{\alpha}$ and $y_{\beta}$ be the elements supplied by Lemma 3.1. Thus

|x_{\alpha}|=(|x|-\alpha)_{+},\qquad|y_{\beta}|=(|y|-\beta)_{+},

and

\lVert x-x_{\alpha}\rVert\leq\alpha,\qquad\lVert y-y_{\beta}\rVert\leq\beta.

It follows that

\lVert(x+y)-(x_{\alpha}+y_{\beta})\rVert\leq\alpha+\beta.

Applying (ii) gives

(|x+y|-\alpha-\beta-\delta)_{+}\precsim|x_{\alpha}+y_{\beta}|.

To compare the other part, applying (i) to

\begin{pmatrix}1&1\\ 0&0\end{pmatrix}\begin{pmatrix}x_{\alpha}&0\\ 0&y_{\beta}\end{pmatrix}\begin{pmatrix}1&0\\ 1&0\end{pmatrix}=\begin{pmatrix}x_{\alpha}+y_{\beta}&0\\ 0&0\end{pmatrix}

yields that

|x_{\alpha}+y_{\beta}|\oplus 0\precsim|x_{\alpha}|\oplus|y_{\beta}|.

Now we have

(|x+y|-\alpha-\beta-\delta)_{+}\precsim(|x|-\alpha)_{+}\oplus(|y|-\beta)_{+}.

(iv) For the product estimate, if $0=\lVert u\rVert\lVert v\rVert.$ the inequality is trivial. Otherwise,

\lVert uxv-ux_{\alpha}v\rVert\leq\lVert u\rVert\lVert v\rVert\lVert x-x_{\alpha}\rVert\leq\lVert u\rVert\lVert v\rVert\alpha.

Another application of (ii) gives

(|uxv|-\lVert u\rVert\lVert v\rVert\alpha-\delta)_{+}\precsim|ux_{\alpha}v|.

Using (i) and the definition of $x_{\alpha}$ , we obtain

|ux_{\alpha}v|\precsim|x_{\alpha}|=(|x|-\alpha)_{+}.

Therefore,

(|uxv|-\lVert u\rVert\lVert v\rVert\alpha-\delta)_{+}\precsim(|x|-\alpha)_{+}.

∎

Definition 3.3.

Let $\tau$ be a lower semicontinuous $2$ -quasitrace on $A$ . For any fixed $x\in A$ , define the $\tau$ -distribution function of $x$ by

\Delta_{x,\tau}(t)=d_{\tau}((|x|-t)_{+}),\qquad t\geq 0.

For any parameter $s>0$ , define

\nu_{s}^{\tau}(x)=\inf\{t\geq 0:\Delta_{x,\tau}(t)\leq s\}.

The value $\nu_{s}^{\tau}(x)$ is finite because $\Delta_{x,\tau}(\lVert x\rVert)=0$ .

Proposition 3.4.

Let $\tau$ be a lower semicontinuous $2$ -quasitrace on $A$ . For $x,y\in A$ , $r,s>0$ , $u,v\in A^{\sim}$ , $\lambda\in\mathbb{C}$ , and $a\in A_{+}$ , one has

(i) $\nu_{r+s}^{\tau}(x+y)\leq\nu_{r}^{\tau}(x)+\nu_{s}^{\tau}(y);$

(ii) $\nu_{s}^{\tau}(uxv)\leq\lVert u\rVert\lVert v\rVert\nu_{s}^{\tau}(x);$

(iii) $\nu_{s}^{\tau}(\lambda x)=|\lambda|\nu_{s}^{\tau}(x);$

(iv) $\nu_{s}^{\tau}(a^{1/2})=\nu_{s}^{\tau}(a)^{1/2}.$

Moreover, $s\mapsto\nu_{s}^{\tau}(x)$ is decreasing, and $0\leq a\leq b$ implies $\nu_{s}^{\tau}(a)\leq\nu_{s}^{\tau}(b)$ .

Proof.

Choose $\alpha>\nu_{r}^{\tau}(x)$ and $\beta>\nu_{s}^{\tau}(y)$ , then

d_{\tau}((|x|-\alpha)_{+})\leq r,\quad d_{\tau}((|y|-\beta)_{+})\leq s,

Applying $d_{\tau}$ to Lemma 3.2 (iii) gives

\Delta_{x+y,\tau}(\alpha+\beta+\delta)\leq r+s.

Taking infima proves (i).

For (ii), it can be similarly deduced from Lemma 3.2 (iv); (iii) is trivial.

For $t>0$ , since we have

(a-t^{2})_{+}=(a^{1/2}-t)_{+}(a^{1/2}+t),

then

(a^{1/2}-t)_{+}\sim(a-t^{2})_{+}.

Hence, $\Delta_{a^{1/2},\tau}(t)=\Delta_{a,\tau}(t^{2})$ , which proves (iv). The remaining statements follow directly from the definition and $(a-t)_{+}\precsim(b-t)_{+}$ for $0\leq a\leq b$ . ∎

Proposition 3.5.

Let $\tau$ be a faithful, densely finite, lower semicontinuous trace on $A$ , and compute generalized singular numbers in the semifinite von Neumann algebra generated by the GNS representation of $\tau$ . Then

\nu_{s}^{\tau}(x)=\mu_{s}^{\tau}(x)\qquad(x\in A,\ s>0).

Proof.

For $t\geq 0$ ,

d_{\tau}((|x|-t)_{+})=\tau(\mathds{1}_{(t,\infty)}(|x|)).

The generalized-inverse formula for $\mu_{s}^{\tau}$ therefore agrees with Definition 3.3; see [13]. ∎

Corollary 3.6.

Let $b\in A_{+}$ and let $\tau$ be a lower semicontinuous $2$ -quasitrace. If $L,\gamma>0$ and

\nu_{s}^{\tau}(b)\geq\gamma\qquad(0<s<L),

then, for every $0\leq\eta<\gamma$ , $d_{\tau}((b-\eta)_{+})\geq L$ .

Proposition 3.7.

Let $(q_{n})$ be pairwise orthogonal nonzero projections in $A$ , and put $p_{n}=\sum_{j=1}^{n}q_{j}$ , $p_{0}=0$ . Let $(\gamma_{n})$ be a decreasing sequence of positive real numbers such that $\lim\limits_{n\to\infty}\gamma_{n}=0$ . Assume $0<\tau(p_{n})<\infty$ . Set $c=\sum_{n=1}^{\infty}\gamma_{n}q_{n},$ then

\nu_{s}^{\tau}(c)=\gamma_{k},\quad\text{when}\quad\sum_{n=1}^{k-1}\tau(q_{n})\leq s<\sum_{i=1}^{k}\tau(q_{n}).

For $N\geq 1$ ,

\nu_{s}^{\tau}(\sum_{n=N+1}^{\infty}\gamma_{n}q_{n})=\nu_{s+\tau(p_{N})}^{\tau}(c).

Proof.

For $t>0$ ,

(c-t)_{+}=\sum_{\gamma_{n}>t}(\gamma_{n}-t)q_{n},\qquad\Delta_{c,\tau}(t)=\sum_{\gamma_{n}>t}d_{\tau}(q_{n})=\sum_{\gamma_{n}>t}\tau(q_{n}).

If $\sum_{n=1}^{k-1}\tau(q_{n})\leq s<\sum_{i=1}^{k}\tau(q_{n})$ , we have

\nu_{s}^{\tau}(c)=\inf\{t\geq 0:\sum_{\gamma_{n}>t}\tau(q_{n})\leq s\}=\gamma_{k}.

The second equality is a simple translation. ∎

4. Admissible quasitracial projection scale

In this section, we assume that $A$ is a nonunital $C^{*}$ -algebra but has an increasing approximate unit consisting of projections $\{p_{n}\}$ , with $p_{0}=0$ . We write $q_{n}=p_{n}-p_{n-1}$ and require $q_{n}\neq 0$ for $n\geq 1$ .

Definition 4.1.

Let $\mathcal{T}$ be a nonempty family of faithful, densely finite, lower semicontinuous $2$ -quasitraces on $A$ . Every $\tau\in\mathcal{T}$ is normalized by $d_{\tau}(p_{1})=1$ . Define the sets

\mathcal{T}_{\mathrm{fin}}=\{\tau\in\mathcal{T}:\sup_{n\geq 1}\tau(p_{n})<+\infty\},\qquad\mathcal{T}_{\mathrm{inf}}=\{\tau\in\mathcal{T}:\sup_{n\geq 1}\tau(p_{n})=+\infty\}.

These two subsets are disjoint and their union is $\mathcal{T}$ .

Definition 4.2.

Let $\rho\geq 0$ , $\varepsilon_{0}>0$ , and $e\in\operatorname{Ped}(A)_{+}$ , we say the tuple

\left(\mathcal{T},\{p_{n}\},\rho,\mathcal{T}_{\mathrm{inf}},(\mathcal{T}_{\mathrm{fin}},e,\varepsilon_{0})\right)

is an admissible quasitracial projection scale if it satisfies the following properties:

(1)

for every $n\geq 1$ and every $c\in A_{+}$ ,

$d_{\tau}(q_{n})+\rho<d_{\tau}(c)\quad\text{for all }\tau\in\mathcal{T}\quad\Longrightarrow\quad q_{n}\precsim c;$

(2)

if $\mathcal{T}_{\mathrm{inf}}\neq\varnothing$ , then

\sup_{\tau\in\mathcal{T}_{\mathrm{inf}}}\tau(p_{n})<+\infty\quad{\rm and}\quad\lim_{n}\inf_{\tau\in\mathcal{T}_{\mathrm{inf}}}\tau(p_{n})=+\infty;

(3)

if $\mathcal{T}_{\mathrm{fin}}\neq\varnothing$ , then

$\inf_{\tau\in\mathcal{T}_{\mathrm{fin}}}d_{\tau}((e-\varepsilon_{0})_{+})>\rho$

and

$\sup_{\tau\in\mathcal{T}_{\mathrm{fin}}}\bigl(\sup_{j\geq 1}\tau(p_{j})-\tau(p_{n})\bigr)\longrightarrow 0.$

If $\mathcal{T}_{\mathrm{fin}}$ or $\mathcal{T}_{\mathrm{inf}}$ is empty, then we assume that the corresponding components naturally satisfy the condition. Particularly, if $\mathcal{T}_{\mathrm{fin}}=\varnothing$ , the tuple reduces to $\left(\mathcal{T},\{p_{n}\},\rho,\mathcal{T}_{\mathrm{inf}},\varnothing\right).$

In a simple $C^{*}$ -algebra, every nonzero lower semicontinuous quasitrace is faithful. In the non-simple setting, faithfulness is a genuine hypothesis.

Lemma 4.3.

Suppose that $\mathcal{T}_{\mathrm{inf}}\neq\varnothing$ and condition (2) of Definition 4.2 holds, then there is a strict increasing map $\Theta:\mathbb{N}\to\mathbb{N}$ , with $\Theta(n)>n$ , such that

\tau(p_{\Theta(n)})\geq 2\tau(p_{n})+\rho,\quad\tau\in\mathcal{T}_{\mathrm{inf}}.

Proof.

For every $n$ , by condition (2), there exists $N_{n}\geq n$ such that

\inf_{\tau\in\mathcal{T}_{\mathrm{inf}}}\tau(p_{N_{n}})>2\sup_{\tau\in\mathcal{T}_{\mathrm{inf}}}\tau(p_{n})+\rho.

Now we define

\Theta(n)=N_{1}+N_{2}+\cdots+N_{n}+n.

Then it is obvious that $\Theta$ satisfies

\Theta(n)>n,\quad\Theta(n+1)>\Theta(n)

and

\tau(p_{\Theta(n)})\geq 2\tau(p_{n})+\rho,\quad\tau\in\mathcal{T}_{\mathrm{inf}}.

∎

Proposition 4.4.

Suppose that $\mathcal{T}_{\mathrm{inf}}$ is nonempty and satisfies condition (2) of Definition 4.2. Then, for every $x\in A$ ,

\sup_{\tau\in\mathcal{T}_{\mathrm{inf}}}\nu_{s}^{\tau}(x)\longrightarrow 0\qquad(s\to\infty).

Proof.

Given $\delta>0$ , choose $N$ with

\lVert x-p_{N}xp_{N}\rVert<\delta.

By Lemma 3.2 (ii), we have

(|x|-\delta)_{+}\precsim|p_{N}xp_{N}|\precsim p_{N}.

Hence,

\Delta_{x,\tau}(\delta)\leq\tau(p_{N}),\quad\tau\in\mathcal{T}_{\mathrm{inf}}.

Then for every $s\geq\sup_{\tau\in\mathcal{T}_{\mathrm{inf}}}\tau(p_{N})<\infty$ and every $\tau\in\mathcal{T}_{\mathrm{inf}}$ , we have

\nu_{s}^{\tau}(x)\leq\delta.

∎

Lemma 4.5.

Let $h\in A_{+}$ , let $p\in A$ be a projection with $p\leq p_{h}$ in $A^{**}$ , and put $c=(1-p)h(1-p)$ . Then $p_{c}=p_{h}-p$ and

[h]=[p]+[c]\quad\text{in }\operatorname{Cu}(A).

Consequently, $d_{\tau}(h)=d_{\tau}(p)+d_{\tau}(c)$ for every lower semicontinuous $2$ -quasitrace $\tau$ .

Proof.

Consider the universal representation of $A$ on a Hilbert space $\mathcal{H}$ , and view $A^{**}$ as a subalgebra of $B(\mathcal{H})$ .

Note that

c=(1-p)p_{h}hp_{h}(1-p)=(p_{h}-p)h(p_{h}-p).

Consequently,

p_{c}\leq p_{h}-p.

Then for any $\xi\in(p_{h}-p)\mathcal{H}$ , if we assume that $c\xi=0$ , then

0=\langle c\xi,\xi\rangle=\langle(1-p)h(1-p)\xi,\xi\rangle=\langle h\xi,\xi\rangle=||h^{1/2}\xi||^{2}.

Thus, $h^{1/2}\xi=0.$ Since $\ker h^{1/2}=(1-p_{h})\mathcal{H}$ , it follows that $\xi=0$ . Thus

p_{c}=p_{h}-p.

Since $pc=cp=0,$ and $p+c$ has support $p_{h}$ . Hence $p+c$ and $h$ generate the same hereditary subalgebra, and hence $p+c\sim h$ . Now we have $[h]=[p]+[c]$ .

Applying $d_{\tau}$ to this equality gives

d_{\tau}(h)=d_{\tau}(p)+d_{\tau}(c).

∎

Definition 4.6.

Let $\Theta:\mathbb{N}\to\mathbb{N}$ be a strictly increasing map satisfying $\Theta(n)>n$ and fix $N\geq 0$ . Let $(\alpha_{n})$ be a decreasing sequence of positive numbers tending to $0$ and let

a=\sum_{n=N+1}^{\infty}\alpha_{n}q_{n}.

We say $a$ is a $(\Theta,N)$ -regular tail if there exists $L_{\Theta,N}>0$ such that

\sup_{n\geq N+1}\frac{\alpha_{n}}{\alpha_{\Theta(n)}}\leq L_{\Theta,N}.

Theorem 4.7.

Suppose that $A$ has an admissible quasitracial projection scale

\left(\mathcal{T},\{p_{n}\},\rho,\mathcal{T}_{\mathrm{inf}},(\mathcal{T}_{\mathrm{fin}},e,\varepsilon_{0})\right).

Let $a$ be a $(\Theta,N)$ -regular tail as in Definition 4.6. Let $b\in A_{+}$ and $C>0$ .

Assume that we have the followings:

(i) if $\mathcal{T}_{\mathrm{inf}}\neq\varnothing$ , suppose also that $\Theta$ satisfies the conclusions in Lemma 4.3 and

\nu_{s}^{\tau}(a)\leq C\nu_{s}^{\tau}(b),\quad s>0,\ \tau\in\mathcal{T}_{\mathrm{inf}}.

(ii) if $\mathcal{T}_{\mathrm{fin}}\neq\varnothing$ , suppose also that $\frac{\alpha_{N+1}}{2C}<\varepsilon_{0}$ and

\sup_{j\geq 1}\tau(p_{j})-\tau(p_{N})+\rho<d_{\tau}((b-\varepsilon_{0})_{+}),\quad\tau\in\mathcal{T}_{\mathrm{fin}}.

Then there exists $x\in A$ such that

x^{*}x=a^{2},\qquad xx^{*}\leq 4C^{2}L_{\Theta,N}^{2}\lVert b\rVert b.

Moreover, if $J$ is a semiprime ideal of $A$ and $b\in J_{+}$ , then $a\in J$ .

Proof.

For $n\geq 1$ , set

\delta_{n}=\frac{\alpha_{\Theta(N+n)}}{2C},\qquad h_{n}=(b-\delta_{n})_{+},\qquad p_{h_{n}}=\mathds{1}_{(0,\infty)}(h_{n})\in A^{**}.

The sequence $(p_{h_{n}})$ is increasing.

Now we recursively construct mutually orthogonal projections $r_{n}\in A$ $(n\geq 1)$ satisfying

r_{n}\sim_{\rm MvN}q_{N+n},\quad f_{n}:=r_{1}+\cdots+r_{n}\leq p_{h_{n}}.

Set $r_{0}=f_{0}=0$ , suppose that $r_{0},r_{1},\ldots,r_{n-1}$ have been constructed, since $f_{n-1}\leq p_{h_{n}}$ $(n\geq 1)$ , put $c_{n}=(1-f_{n-1})h_{n}(1-f_{n-1})$ .

Applying Lemma 4.5 for $c_{n}$ , $f_{n-1}$ (in place of $p$ ), $h_{n}$ yields $p_{c_{n}}=p_{h_{n}}-f_{n-1}$ ,

[h_{n}]=[f_{n-1}]+[c_{n}]\quad{\rm and}\quad d_{\tau}(h_{n})=d_{\tau}(f_{n-1})+d_{\tau}(c_{n}).

For any $\tau\in\mathcal{T}_{\mathrm{inf}}$ , if $0<s<\tau(p_{\Theta(N+n)})-\tau(p_{N})$ , Proposition 3.7 implies

\nu_{s}^{\tau}(a)\geq\alpha_{\Theta(N+n)}.

Then we have

2\delta_{n}=\frac{\alpha_{\Theta(N+n)}}{C}\leq\frac{\nu_{s}^{\tau}(a)}{C}\leq\nu_{s}^{\tau}(b).

Now applying Lemma 3.6 for $\delta_{n}$ yields

d_{\tau}(h_{n})=d_{\tau}((b-\delta_{n})_{+})\geq\tau(p_{\Theta(N+n)})-\tau(p_{N}).

By the inductive hypothesis, $\tau(r_{i})=\tau(q_{N+i})$ for $1\leq i<n$ , then

d_{\tau}(f_{n-1})=\sum_{i=1}^{n-1}\tau(r_{i})=\sum_{i=N+1}^{N+n-1}\tau(q_{i})=\tau(p_{N+n-1})-\tau(p_{N})<\infty.

Now we have

	$\displaystyle d_{\tau}(c_{n})$	$\displaystyle=d_{\tau}(h_{n})-d_{\tau}(f_{n-1})$
		$\displaystyle\geq\tau(p_{\Theta(N+n)})-\tau(p_{N+n-1})$
		$\displaystyle=\tau(q_{N+n})+\tau(p_{\Theta(N+n)})-\tau(p_{N+n})$
		$\displaystyle>d_{\tau}(q_{N+n})+\rho,$

where the last inequality uses Lemma 4.3.

Now let $\tau\in\mathcal{T}_{\mathrm{fin}}$ . Since

\delta_{n}=\frac{\alpha_{\Theta(N+n)}}{2C}\leq\frac{\alpha_{N+1}}{2C}<\varepsilon_{0},

one has

[(b-\varepsilon_{0})_{+}]\leq[h_{n}].

Now we have

	$\displaystyle d_{\tau}(c_{n})$	$\displaystyle=d_{\tau}(h_{n})-d_{\tau}(f_{n-1})$
		$\displaystyle\geq d_{\tau}((b-\varepsilon_{0})_{+})-(\tau(p_{N+n-1})-\tau(p_{N}))$
		$\displaystyle>\sup_{j\geq 1}\tau(p_{j})-\tau(p_{N+n-1})+\rho$
		$\displaystyle\geq d_{\tau}(q_{N+n})+\rho.$

Finally, we obtain

d_{\tau}(q_{N+n})+\rho<d_{\tau}(c_{n})\quad{\rm for\,\,all}\,\,\tau\in\mathcal{T}_{\mathrm{inf}}\cup\mathcal{T}_{\mathrm{fin}}.

Hence, $q_{N+n}\precsim c_{n}$ . By Lemma 2.5, there exist a projection $r_{n}\in A_{c_{n}}$ and $\omega_{n}\in A$ such that

q_{N+n}=\omega_{n}^{*}\omega_{n}\sim_{\rm MvN}\omega_{n}\omega_{n}^{*}=r_{n}.

Then $r_{n}\leq p_{h_{n}}-f_{n-1}$ , this completes the induction.

By the inductive procedure, we have constructed a sequence of partial isometries $\{\omega_{n}\}$ . Let

x=\sum_{n=1}^{\infty}\alpha_{N+n}\omega_{n},

then $x\in A$ and

x^{*}x=\sum_{n=1}^{\infty}\alpha_{N+n}^{2}q_{N+n}=a^{2},\qquad xx^{*}=\sum_{n=1}^{\infty}\alpha_{N+n}^{2}r_{n}.

Since

\sum_{n=1}^{\infty}\alpha_{N+n}^{2}r_{n}=\sum_{n=1}^{\infty}(\alpha_{N+n}^{2}-\alpha_{N+n+1}^{2})(r_{1}+r_{2}\cdots+r_{n}).

Now we have

xx^{*}=\sum_{n=1}^{\infty}(\alpha_{N+n}^{2}-\alpha_{N+n+1}^{2})f_{n}\leq\sum_{n=1}^{\infty}(\alpha_{N+n}^{2}-\alpha_{N+n+1}^{2})p_{h_{n}}=g(b),

where $\gamma_{n}=\alpha_{N+n}^{2}-\alpha_{N+n+1}^{2}\geq 0$ and

g(t)=\sum_{n=1}^{\infty}\gamma_{n}\mathds{1}_{(\delta_{n},\infty)}(t).

For every $t>0$ , let $n_{0}$ be the first index with $\alpha_{\Theta(N+n_{0})}<2Ct$ . Then

g(t)=\sum_{n=n_{0}}^{\infty}\gamma_{n}=\alpha_{N+n_{0}}^{2}\leq L_{\Theta,N}^{2}\alpha_{\Theta(N+n_{0})}^{2}<4C^{2}L_{\Theta,N}^{2}t^{2}.

Borel functional calculus in $A^{**}$ and $g(0)=0$ yield

xx^{*}\leq 4C^{2}L_{\Theta,N}^{2}b^{2}\leq 4C^{2}L_{\Theta,N}^{2}\lVert b\rVert b.

If $J$ is semiprime, by Proposition 2.17, one has $xx^{*}\in J$ . Strong invariance gives $a^{2}=x^{*}x\in J$ , and positive root closure gives $a\in J$ . ∎

5. The quasitracial-scale theorem

In this section, we retain the conventions established at the beginning of Section 4.

Lemma 5.1.

Let $\Theta:\mathbb{N}\to\mathbb{N}$ be a strictly increasing map satisfying $\Theta(n)>n$ . If $M$ is a maximal ideal of $A$ , then there is a decreasing sequence of positive numbers $(\beta_{n})$ converging to $0$ such that

\beta_{\Theta(n)}\geq\frac{1}{2}\beta_{n}

for every $n\geq 1$ and

b:=\sum_{n=1}^{\infty}\beta_{n}q_{n}\in A_{+}\setminus M.

Proof.

Propositions 2.16 and 2.17 imply that $M$ is positively spanned. Choose $d\in A_{+}\setminus M$ . Set $N_{1}=1$ . Having chosen $N_{k}$ , choose $N_{k+1}$ such that

N_{k+1}>\Theta(N_{k})\quad{\rm and}\quad\lVert d(1-p_{N_{k}-1})\rVert\leq\frac{1}{k\cdot 2^{k}},

Denote

Q_{k}=p_{N_{k+1}-1}-p_{N_{k}-1}=\sum_{n=N_{k}}^{N_{k+1}-1}q_{n},

and let

b=\sum_{k=1}^{\infty}\frac{1}{k}Q_{k},\quad c=\sum_{k=1}^{\infty}k\,dQ_{k}

Then $b,c\in A$ and $cb=d$ . Hence, $b\notin M$ .

Set

\beta_{n}=\begin{cases}1,&\mbox{if }1\leq n<N_{2}\\ 1/2,&\mbox{if }N_{2}\leq n<N_{3}\\ 1/3,&\mbox{if }N_{3}\leq n<N_{4}\\ \vdots&\mbox{\vdots}\end{cases},

Then

b=\sum_{n=1}^{\infty}\beta_{n}q_{n}.

For any $n\geq 1$ , if $N_{k}\leq n<N_{k+1}$ , then

n<\Theta(n)\leq\Theta(N_{k+1})<N_{k+2}.

and

\beta_{\Theta(n)}\geq\frac{1}{k+1}\geq\frac{1}{2}\beta_{n}.

∎

Remark 5.2.

Suppose that $\mathcal{T}_{\mathrm{inf}}$ is nonempty and satisfies condition (2) of Definition 4.2. Then Lemma 4.3 gives a map $\Theta:\mathbb{N}\to\mathbb{N}$ such that

\tau(p_{\Theta(n)})\geq 2\tau(p_{n}),\quad n\geq 1,\,\,\tau\in\mathcal{T}_{\mathrm{inf}}.

Let $b$ be the element satisfy the properties in Lemma 5.1. Given $m,r\in\mathbb{N}$ with $2^{r}\geq m\geq 2$ . If there exists $n\geq n_{0}$ such that

\tau(p_{n_{0}})\leq\tau(p_{n-1})\leq s<\tau(p_{n}),

then

ms<m\tau(p_{n})\leq 2^{r}\tau(p_{n})\leq\tau(p_{\Theta^{r}(n)}),

where $\Theta^{r}$ is the $r$ -fold composition of $\Theta$ .

Now Proposition 3.7 yields

\nu_{ms}^{\tau}(b)\geq\beta_{\Theta^{r}(n)}\geq\dfrac{\beta_{n}}{2^{r}}=\dfrac{\nu_{s}^{\tau}(b)}{2^{r}}.

Then we have

\nu_{ms}^{\tau}(b)\geq\frac{\nu_{s}^{\tau}(b)}{2^{r}},\quad s\geq\tau(p_{n_{0}}),\ \tau\in\mathcal{T}_{\mathrm{inf}}.

Theorem 5.3.

If $A$ has an admissible quasitracial projection scale

\left(\mathcal{T},\{p_{n}\},\rho,\mathcal{T}_{\mathrm{inf}},(\mathcal{T}_{\mathrm{fin}},e,\varepsilon_{0})\right).

Then $A$ has no dense maximal ideal.

Proof.

Assume that $M$ is a dense maximal ideal of $A$ , Theorem 2.11 gives

p_{n}\in M\,\,(n\geq 1),\qquad\operatorname{Ped}(A)\subseteq M.

If $\mathcal{T}_{\mathrm{inf}}\neq\varnothing$ , choose $\Theta$ as in Lemma 4.3; otherwise take $\Theta(n)=n+1$ . By Lemma 5.1, there exists an element

b=\sum_{n=1}^{\infty}\beta_{n}q_{n}\in A_{+}\setminus M.

For every $k\geq 1$ , set

a_{k}=b^{1/2}(1-p_{k})=\sum_{n=k+1}^{\infty}\alpha_{n}q_{n},\qquad\alpha_{k}=\sqrt{\beta_{k}}.

Then

\frac{\alpha_{k}}{\alpha_{\Theta(k)}}=\left(\frac{\beta_{k}}{\beta_{\Theta({k})}}\right)^{1/2}\leq\sqrt{2}.

Then $a_{k}$ is a $(\Theta,k)$ -regular tail and we can choose $L_{\Theta,k}=\sqrt{2}$ for all $k$ .

We proceed with the proof by considering several cases.

Case 1: $\mathcal{T}_{\mathrm{inf}}=\varnothing$ .

In this case, $\mathcal{T}=\mathcal{T}_{\mathrm{fin}}$ . Let

\varepsilon=\inf_{\tau\in\mathcal{T}_{\mathrm{fin}}}d_{\tau}((e-\varepsilon_{0})_{+})-\rho>0

Definition 4.2 (3) gives an $n_{0}$ such that

\sup_{j\geq 1}\tau(p_{j})-\tau(p_{n_{0}})<\varepsilon,\,\,\,\,n\geq n_{0},\,\,\tau\in\mathcal{T}_{\mathrm{fin}}.

Since $\beta_{n}\to 0$ , then there exists $N\geq n_{0}$ such that

\frac{\alpha_{N+1}}{2}=\frac{\sqrt{\beta_{N+1}}}{2}<\varepsilon_{0}.

We also have

\sup_{j\geq 1}\tau(p_{j})-\tau(p_{n_{0}})+\rho<d_{\tau}((e-\varepsilon_{0})_{+}).

Apply Theorem 4.7 for $a_{N}$ (in place of $a$ ), $e$ (in place of $b$ ), $C=1$ , $M$ (in place of $J$ ), we obtain $a_{N}\in M$ . Therefore

b(1-p_{N})=a_{N}^{2}\in M,\qquad bp_{N}\in M,

and hence $b\in M$ .

Case 2: $\mathcal{T}_{\mathrm{inf}}\neq\varnothing$ .

The map $\Theta$ has already been chosen to satisfy Lemma 4.3.

Since $M$ is maximal and $b\notin M$ , then the ideal generated by $M$ and $b$ must be $A$ . Hence, there are $z\in M$ , $k\geq 1$ , and $x_{j},y_{j}\in A^{\sim}$ $(1\leq j\leq k)$ such that

b^{1/2}=z+\sum_{j=1}^{k}x_{j}by_{j}.

Set $K=\sum_{j=1}^{k}\lVert x_{j}\rVert\lVert y_{j}\rVert,$ and choose $r\geq 1$ with $2^{r}\geq k+1$ . By Proposition 3.4, we have

\nu_{(k+1)s}^{\tau}(b^{1/2})\leq\nu_{s}^{\tau}(z)+K\cdot\nu_{s}^{\tau}(b),\qquad s>0,\ \tau\in\mathcal{T}_{\mathrm{inf}}.

Choose $n_{0}$ as in Remark 5.2. Then we have

\frac{\nu_{s}^{\tau}(b)}{2^{r}}\leq\nu_{(k+1)s}^{\tau}(b),\quad s\geq\tau(p_{n_{0}}),\,\,\tau\in\mathcal{T}_{\mathrm{inf}}.

Proposition 3.4 (iv) gives

\frac{\nu_{s}^{\tau}(b)^{1/2}}{2^{r/2}}\leq\nu_{s}^{\tau}(z)+K\cdot\nu_{s}^{\tau}(b).

After applying Proposition 4.4, then

\sup_{\tau\in\mathcal{T}_{\rm inf}}\nu_{s}^{\tau}(b)\to 0,\,\,\,\,(s\to+\infty).

Now there exists $s_{*}\geq\sup_{\tau\in\mathcal{T}_{\mathrm{inf}}}\tau(p_{n_{0}})$ such that

K\cdot\nu_{s}^{\tau}(b)\leq\frac{\nu_{s}^{\tau}(b)^{1/2}}{2^{1+r/2}},\qquad s\geq s_{*},\ \tau\in\mathcal{T}_{\mathrm{inf}}.

Now we have

\frac{\nu_{s}^{\tau}(b)^{1/2}}{2^{1+r/2}}\leq\nu_{s}^{\tau}(z),\qquad s\geq s_{*},\quad\tau\in\mathcal{T}_{\mathrm{inf}}.

Now we claim that $z\neq 0$ . Otherwise, $\nu_{s}^{\tau}(b)=0$ for large $s$ . This contradicts with Proposition 3.7 and the fact that

\tau(p_{n})\to\infty,\quad{\rm for\,\,every}\,\,\tau\in\mathcal{T}_{\mathrm{inf}}.

Then we separate the following two possible situations.

Case 2a: $\mathcal{T}_{\mathrm{fin}}=\varnothing$ .

By Definition 4.2(2), we have

\inf_{\tau\in\mathcal{T}_{\mathrm{inf}}}\tau(p_{n})\to\infty.

There exists $N$ such that

\inf_{\tau\in\mathcal{T}_{\mathrm{inf}}}\tau(p_{N})>s_{*},

Then applying Proposition 3.7 gives

\nu_{s}^{\tau}(a_{N})=\nu_{s+\tau(p_{N})}^{\tau}(b^{1/2})\leq 2^{1+r/2}\nu_{s+\tau(p_{N})}^{\tau}(z)\leq 2^{1+r/2}\nu_{s}^{\tau}(|z|),\qquad s>0,\ \tau\in\mathcal{T}_{\mathrm{inf}}.

Now we apply Proposition 2.17 and Theorem 4.7, then we have $|z|\in M_{+}$ , and hence, $a_{N}\in M$ . As in Case 1, we have

b=bp_{N}+a_{N}^{2}\in M.

Case 2b: $\mathcal{T}_{\mathrm{fin}}\neq\varnothing$ .

Since $|z|+e\in M_{+}$ and $e\leq|z|+e$ , we have

\inf_{\tau\in\mathcal{T}_{\mathrm{fin}}}d_{\tau}((|z|+e-\varepsilon_{0})_{+})>\rho.

Put $C=2^{1+r/2}$ . There exists $N$ such that

\frac{\sqrt{\beta_{N+1}}}{2C}<\varepsilon_{0},\quad{}\inf_{\tau\in\mathcal{T}_{\mathrm{inf}}}\tau(p_{N})>s_{*}

and

\sup_{j\geq 1}\tau(p_{j})-\tau(p_{N})+\rho<d_{\tau}((e-\varepsilon_{0})_{+}),\qquad\tau\in\mathcal{T}_{\mathrm{fin}}.

For $\tau\in\mathcal{T}_{\mathrm{inf}}$ , similar calculation gives

\nu_{s}^{\tau}(a_{N})\leq 2^{1+r/2}\nu_{s}^{\tau}(|z|)\leq 2^{1+r/2}\nu_{s}^{\tau}(|z|+e),\quad s>0.

Now we apply Theorem 4.7 for $a_{N}$ (in place of $a$ ), $|z|+e$ (in place of $b$ ), $C=2^{1+r/2}$ and $M$ (in place of $J$ ), we have $a_{N}\in M$ , and again,

b=bp_{N}+a_{N}^{2}\in M.

All cases lead to a contradiction. Therefore $A$ has no dense maximal ideal. ∎

Corollary 5.4.

If a simple $C^{*}$ -algebra admits an admissible quasitracial projection scale, then it has no maximal ideals.

Proof.

A maximal ideal in a simple $C^{*}$ -algebra is dense, so Theorem 5.3 applies. ∎

Corollary 5.5 (Nonsimple form).

Suppose that $A$ admits an admissible quasitracial projection scale. If no nonzero proper closed ideal $J\lhd A$ has algebraically simple quotient $A/J$ , then $A$ has no maximal ideals.

Proof.

Theorem 5.3 excludes the dense alternative in Theorem 2.14; the stated hypothesis excludes the closed alternative. ∎

6. Main applications

6.1. Stablization of unital simple algebras

Definition 6.1 ([24]).

Let $A$ be unital, simple, and stably finite, with $\operatorname{QT}^{1}_{2}(A)\neq\varnothing$ . For $r\geq 0$ , we say that $A$ has $r$ -comparison if

d_{\tau}(a)+r<d_{\tau}(b)\quad(\tau\in QT^{1}_{2}(A))\quad\Longrightarrow\quad a\precsim b

for all $a,b\in M_{\infty}(A)_{+}$ . The radius of comparison is

\operatorname{rc}(A)=\inf\{r\geq 0:A\text{ has }r\text{-comparison}\}.

We say that $A$ has strict comparison if it has $0$ -comparison. Equivalently, the strict dimension inequality implies comparison for positive elements of $A\otimes\mathcal{K}$ .

The infimum of an empty set of comparison buffers is $\infty$ .

Definition 6.2.

For $A$ as above, the restricted projection radius $\operatorname{rc}_{\mathrm{p}}(A)$ is the infimum of the numbers $r\geq 0$ such that

d_{\tau}(q)+r<d_{\tau}(b)\quad(\tau\in\operatorname{QT}^{1}_{2}(A))\quad\Longrightarrow\quad q\precsim b

for every projection $q\in M_{\infty}(A)$ and every $b\in M_{\infty}(A)_{+}$ . Only the element on the left is required to be a projection.

It follows directly that

0\leq\operatorname{rc}_{\mathrm{p}}(D)\leq\operatorname{rc}(D)\leq\infty.

The restricted radius supplies exactly the comparison of $q_{n}$ with positive elements required in Definition 4.2.

Using Cuntz-semigroup formulation, it follows from [6, Section 3] that the definition of radius of comparison is extended to non-unital $C^{*}$ -algebras. Let $A$ be a $C^{*}$ -algebra. We denote by $\operatorname{QT}_{2}(A)$ the collection of all extended lower semicontinuous $2$ -quasitrace on $A$ . It follows from [11, Proposition 4.2] that the map from $\operatorname{QT}_{2}(A)$ to $F(\operatorname{Cu}(A))$ defined by $\tau\mapsto d_{\tau}$ , is bijective, in which $F(\operatorname{Cu}(A))$ denotes the set of all functionals on $\operatorname{Cu}(A)$ . Thus, one obtains the following definition.

Definition 6.3.

([6, Definition 3.3.2]) Let $A$ be a $C^{*}$ -algebra and $e\in(A\otimes\mathcal{K})_{+}$ full. The radius of comparison of $A$ relative to $[e]$ is defined to be the infimum of the numbers $r>0$ satisfying

d_{\tau}([a])+rd_{\tau}([e])\leq d_{\tau}([b])\quad(\tau\in\operatorname{QT}_{2}(A))\quad\Longrightarrow\quad a\precsim b.

We denote this number by $\rm rc(\operatorname{Cu}(A),[e])$ . Similarly, one may define $\rm rc_{p}(\operatorname{Cu}(A),[e])$ .

Remark 6.4.

It is shown in [6, Proposition 3.2.3] that for a unital $C^{*}$ -algebra $A$ that all the quotients are stably finite (e.g., $A$ is unital, stably finite and simple), the radius of comparison $\rm rc(A)$ in the sense of Definition 6.1 is equivalent to $\rm rc(\operatorname{Cu}(A),[1_{A}])$ . Similarly, one also has $\rm rc_{p}(A)=\rm rc_{p}(\operatorname{Cu}(A),[1_{A}])$ .

Definition 6.5.

Let $A$ be a $C^{*}$ -algebra and $[e]\in\operatorname{Cu}(A)$ . We denote by $\operatorname{QT}_{2}(A,[e])$ the set

\operatorname{QT}^{1}_{2}(A,[e])=\{\tau\in\operatorname{QT}_{2}(A):d_{\tau}([e])=1\}.

Remark 6.6.

Let $A$ be a $\sigma$ -unital simple $C^{*}$ -algebra and $p\in A$ be a non-zero projection. For the full hereditary $C^{*}$ -algebra $B=pAp$ , it follows from the Brown stable isomorphism in [5, Theorem 2.8] that $B\otimes\mathcal{K}\simeq A\otimes\mathcal{K}$ and $\operatorname{Cu}(B)\simeq\operatorname{Cu}(A)$ via an isomorphism $\Phi$ with $\Phi([1_{B}]_{B})=[p]_{A}$ . Moreover, it is not hard to see $\operatorname{QT}^{1}_{2}(A,[p])$ can be identified by $\operatorname{QT}^{1}_{2}(B)$ by the natural extension.

Proposition 6.7.

Let $A$ be a simple, stably finite and $\sigma$ -unital $C^{*}$ -algebra. Let $q\in A$ be a non-zero projection. Define $B=qAq$ . Then $\rm rc(B)=\rm rc(\operatorname{Cu}(A),[q])$ and $\rm rc_{p}(B)=\rm rc_{p}(\operatorname{Cu}(A),[q])$ . In addition, for any $\rho>\rm rc_{p}(B)$ , any projection $e$ and positive element $c$ in $A\otimes\mathcal{K}$ , one has

d_{\tau}(e)+\rho\leq d_{\tau}(c)\quad(\tau\in\operatorname{QT}^{1}_{2}(A,[q]))\quad\Longrightarrow\quad e\precsim c.

Proof.

First, it follows from Remark 6.6 that $F(\operatorname{Cu}(A))=F(\operatorname{Cu}(B))$ induced by an isomorphism $\Phi:\operatorname{Cu}(B)\to\operatorname{Cu}(A)$ such that $\Phi([1_{B}])=[q]$ . This implies that

\rm rc(\operatorname{Cu}(A),[q])=\rm rc(\operatorname{Cu},[1_{B}])=\rm rc(B)

and

\rm rc_{p}(\operatorname{Cu}(A),[q])=\rm rc_{p}(\operatorname{Cu},[1_{B}])=\rm rc_{p}(B)

by Remark 6.4.

For the last claim, we first shows that in the definition of $\rm rc(\operatorname{Cu}(A),[q])$ , it suffices to look at all functionals $d_{\tau}\in F(\operatorname{Cu}(A))$ where $\tau\in\operatorname{QT}^{1}_{2}(A,[q])$ . Indeed, for $\tau\in\operatorname{QT}_{2}(A)$ , if $d_{\tau}([q])=\tau(q)=0$ . Then for any $[a]\in\operatorname{Cu}(A)$ , since $A$ is simple, for any $\varepsilon>0$ , there exists $n\in\mathbb{N}$ such that $[(a-\varepsilon)_{+}]\leq n[q]$ . This implies that $d_{\tau}([(a-\varepsilon)_{+}])=0$ . Therefore, one has $d_{\tau}([a])=0$ . If $d_{\tau}([q])=\tau(q)=\infty$ , then because $[q]$ is a compact element in $\operatorname{Cu}(A)$ and $A$ is simple, for any non-zero $[a]\in\operatorname{Cu}(A)$ , there exists an $n\in\mathbb{N}$ such that $[q]\leq n[a]$ . This implies that $d_{\tau}([a])=\infty$ for any non-zero $[a]\in\operatorname{Cu}(A)$ . In this two cases, the inequality

d_{\tau}([a])+rd_{\tau}([q])\leq d_{\tau}([b])

always hold. Then in the case $0<\tau(q)<\infty$ , we define $\bar{\tau}(\cdot)=\tau(\cdot)/\tau(q)\in\operatorname{QT}_{2}^{1}(A,[q])$ . This thus implies that the two conditions

(1)

d_{\tau}([a])+rd_{\tau}([q])\leq d_{\tau}([b])\quad(\tau\in\operatorname{QT}_{2}(A))

and

(2)

d_{\tau}([a])+r\leq d_{\tau}([b])\quad(\tau\in\operatorname{QT}^{1}_{2}(A,[q]))

are equivalent. Thus, the final claim is established. ∎

Remark 6.8.

In the setting of Proposition 6.7, in order to verify $d_{\tau}(e)+\rho\leq d_{\tau}(c)$ for all $\tau\in\operatorname{QT}^{1}_{2}(A,[q])$ , it suffices to verify the inequality for all $\tau\in\partial_{e}\operatorname{QT}_{2}^{1}(A,[q])$ . This is a standard application of the Choquet simplex theory. Nevertheless, we prefer to provide a brief explanation to our $\rm rc_{p}(A,[q])$ case. Indeed, note that $f:\operatorname{QT}_{2}^{1}(A)\to(-\infty,+\infty]$ defined by $f(\tau)=d_{\tau}(c)-d_{\tau}(e)-\rho$ is lower semicontinuous and affine because $d_{\tau}(e)=\tau(e)$ . Moreover, since $A\otimes\mathcal{K}$ is simple, one has $[e]\leq n[q]$ for some $n\in\mathbb{N}$ . This implies that $f$ is bounded below. If f attains a negative value on $\operatorname{QT}_{2}^{1}(A,[q])$ , then the set of all the minimizers forms a non-empty compact face whose extreme points also belong to $\partial_{e}\operatorname{QT}_{2}^{1}(A,[q])$ . But this is a contradiction.

The following is the first main application of Corollary 5.4.

Theorem 6.9.

Proof.

Put $A=D\otimes\mathcal{K}$ and use

q_{n}=1_{D}\otimes e_{nn},\qquad p_{n}=\sum_{j=1}^{n}q_{j},\qquad\mathcal{T}=\operatorname{QT}_{2}^{1}(A,[q_{1}])

that can be identified by $\operatorname{QT}_{2}^{1}(D)$ by Remark 6.6. Then $\tau(p_{1})=1$ , $d_{\tau}(q_{n})=1$ , and $\tau(p_{n})=n$ for all $\tau\in\mathcal{T}$ . Hence $\mathcal{T}_{\mathrm{fin}}$ is empty, and

\sup_{\tau\in\mathcal{T}}\tau(p_{n})=n<\infty,\qquad\inf_{\tau\in\mathcal{T}}\tau(p_{n})=n\longrightarrow\infty.

Choose $\rho>\operatorname{rc}_{\mathrm{p}}(D)$ . Note that Lemma 6.7 verifies Definition 4.2 (1); the displayed formulas above verify Definition 4.2 (2). Finally, since $\mathcal{T}_{\rm fin}$ is empty, there is no need to verify Definition 4.2 (3). Now Corollary 5.4 proves the first assertion.

For properness of the Pedersen ideal, let $h=\sum_{n\geq 1}2^{-n}q_{n}$ and fix $\tau\in\mathcal{T}$ . The dimension-finite ideal $\mathcal{D}_{\tau}(A)$ of Definition 2.9 contains every finite matrix corner, so it is dense and contains $\operatorname{Ped}(A)$ by Theorem 2.11. But $d_{\tau}(h)=\sup_{n}\tau(p_{n})=\infty$ , so $h\notin\mathcal{D}_{\tau}(A)$ and $h\notin\operatorname{Ped}(A)$ . ∎

We end this section by demonstrating that in many cases, $\rm rc_{p}(A)$ and $\rm rc(A)$ are actually equal.

Proposition 6.10.

Let $A$ be a unital simple and stably finite $C^{*}$ -algebra with $\operatorname{QT}_{2}^{1}(A)\neq\varnothing$ as in Definition 6.1. Suppose $A$ has real rank zero, or $\operatorname{Cu}(A)$ has weak cancellation in the sense that $x+z\ll y+z$ implies $x\ll y$ in $\operatorname{Cu}(A)$ (see e.g., [26, Definition 10.1]). Then one has

\operatorname{rc}_{\mathrm{p}}(A)=\operatorname{rc}(A).

In particular, the equality holds when $A$ has stable rank one.

Proof.

It suffices to show $\operatorname{rc}(A)\leq\operatorname{rc}_{\mathrm{p}}(A)$ . Suppose the $\rm rc_{p}(A)$ is finite. Fix $r>\operatorname{rc}_{\mathrm{p}}(D)$ , and let $a,b\in M_{n}(A)_{+}$ satisfy $d_{\tau}(a)+r<d_{\tau}(b)$ for every normalized quasitrace.

If $A$ has real rank zero, the hereditary subalgebra generated by $a$ , denoted by $(M_{n}(A))_{a}$ , has an approximate identity of projections by [25, Theorem 2.6]. Given $\varepsilon>0$ , choose one of these projections $q$ with $\|a-qaq\|<\varepsilon$ . Lemma 2.3 gives

(a-\varepsilon)_{+}\precsim qaq\precsim q\precsim a.

Thus $d_{\tau}(q)+r<d_{\tau}(b)$ , and restricted comparison gives $q\precsim b$ . It follows that $(a-\varepsilon)_{+}\precsim b$ for every $\varepsilon>0$ , whence $a\precsim b$ by Lemma 2.3.

Now suppose that $\operatorname{Cu}(A)$ has weak cancellation. Without loss of generality, one may assume that $a$ is contraction. Denote by $1_{n}$ the unit for $M_{n}(A)$ . For $0<\varepsilon<1$ , put

a_{\varepsilon}=(a-\varepsilon)_{+},\qquad c=(1_{n}-a/\varepsilon)_{+}.

The elements $a_{\varepsilon}$ and $c$ are orthogonal. The scalar function $t+(1-t/\varepsilon)_{+}$ is strictly positive on $[0,1]$ , so $a+c$ is invertible in $M_{n}(A)$ . Therefore

[a_{\varepsilon}]+[c]=[a_{\varepsilon}+c]\leq[1_{n}]=[a+c]\leq[a]+[c].

Applying $d_{\tau}$ gives $d_{\tau}(1_{n})+r<d_{\tau}(b)+d_{\tau}(c)$ , which gives $[1_{n}]\leq[b]+[c]$ . Since $[1_{n}]\ll[1_{n}]$ , the preceding inequalities imply

[a_{\varepsilon}]+[c]\ll[b]+[c].

Weak cancellation yields $[a_{\varepsilon}]\ll[b]$ , hence $(a-\varepsilon)_{+}\precsim b$ . Lemma 2.3 gives $a\precsim b$ . In both cases $A$ has $r$ -comparison for every $r>\operatorname{rc}_{\mathrm{p}}(A)$ , which proves equality of the radii. The stable rank one assertion follows from [26, Theorem 10.3], which is originally due to Rørdam and Winter [28]. ∎

6.2. $\mathcal{Z}$ -stability and comparison

Our principal application in this subsection is Theorem 6.13 and Theorem 6.14. We first give the finite matrix and scale arguments used in the proof.

Lemma 6.11.

If $M_{m}(A)$ has no maximal ideals for some $m\geq 1$ , then $A$ has no maximal ideals.

Proof.

Suppose that $M$ is a maximal ideal of $A$ and put $S=A/M$ . Then $S$ is nonzero and algebraically simple, and $S^{2}=S$ because $A^{2}=A$ by Cohen factorization theorem [9]. In addition, note that left and right annihilators of $S$ are ideals. However since $S^{2}\neq 0$ , neither of them can equal $S$ . Therefore, both of them are zero. Consequently, if $0\neq x\in S$ , there exist $s,t\in S$ with $sxt\neq 0$ . This implies that the nonzero ideal $\operatorname{span}(SxS)$ has to equal $S$ .

Let $J$ be a nonzero ideal of $M_{m}(S)$ and choose an arbitrary $X\in J$ with a nonzero entry $X_{ij}=x$ for some $1\leq i,j\leq n$ . If $E_{ki}(s)$ denotes the matrix with entry $s$ at $(k,i)$ and all other entries zero, then a simple calculation shows

E_{kl}(sxt)=E_{ki}(s)XE_{jl}(t)\in J.

Since $\operatorname{span}(SxS)=S$ , this entails that every elementary matrix over $S$ in $J$ and therefore $J=M_{m}(S)$ . It follows that $M_{m}(A)/M_{m}(M)\cong M_{m}(S)$ is algebraically simple, so the nonzero ideal $M_{m}(M)$ is maximal, a contradiction. ∎

Lemma 6.12.

Let $A$ be simple and nonunital, with an increasing approximate unit consisting of projections $\{p_{n}\}$ , with $p_{0}=0$ . Define $D=p_{1}Ap_{1}$ and assume that $D$ is stably finite, $\operatorname{QT}^{1}_{2}(D)\neq\varnothing$ , and $\operatorname{rc}_{\mathrm{p}}(D)<\infty$ . Let $\mathcal{T}$ consists of extreme points of $\operatorname{QT}^{1}_{2}(A,[p_{1}])\simeq\operatorname{QT}^{1}_{2}(D)$ . Suppose that its infinite-scale part $\mathcal{T}_{\rm{inf}}$ satisfies Definition 4.2 (2) and that

\sup_{\tau\in\mathcal{T}_{\mathrm{fin}}}\left(\sup_{j}\tau(p_{j})-\tau(p_{n})\right)\longrightarrow 0

when $\mathcal{T}_{\mathrm{fin}}\neq\varnothing$ . Then $A$ has no maximal ideals.

Proof.

Choose $\rho>\operatorname{rc}_{\mathrm{p}}(D)$ and an integer $m>\rho$ . Define $B=M_{m}(A)$ and set

P_{n}=p_{n}\otimes 1_{m},\qquad Q_{n}=q_{n}\otimes 1_{m},\qquad\widehat{\tau}=\frac{1}{m}\tau^{(m)}\quad(\tau\in\mathcal{T}).

Here $\tau^{(m)}$ is the canonical matrix extension. For every $n$ , note that

\widehat{\tau}(P_{n})=\tau(p_{n}),\qquad\widehat{\tau}(Q_{n})=\tau(q_{n}).

Note that

\operatorname{QT}_{2}^{1}(B,[P_{1}])=\{\hat{\tau}:\tau\in\operatorname{QT}_{2}^{1}(A,[p_{1}])\}

and therefore $\hat{\mathcal{T}}:=\{\hat{\tau}:\tau\in\mathcal{T}\}$ is the extreme boundary of $\operatorname{QT}_{2}^{1}(B,[P_{1}])$ . Moreover, one has $\hat{\mathcal{T}}_{\inf}=\{\hat{\tau}:\tau\in\mathcal{T}_{\inf}\}$ and $\hat{\mathcal{T}}_{\rm fin}=\{\hat{\tau}:\tau\in\mathcal{T}_{\rm fin}\}.$

For $c\in B_{+}$ , suppose

d_{\widehat{\tau}}(Q_{n})+\rho/m<d_{\widehat{\tau}}(c)\quad(\tau\in\mathcal{T})

holds. After multiplying by $m$ , this implies

d_{\tau^{(m)}}(Q_{n})+\rho<d_{\tau^{(m)}}(c)\quad(\tau\in\mathcal{T}).

Since all functional in $F(\operatorname{Cu}(A))$ can also be written in the form $d_{\eta^{(m)}}$ for $\eta\in\operatorname{QT}_{2}(A)$ , Lemma 6.7 and Remark 6.8 show $Q_{n}\precsim_{B}c$ . This verifies Definition 4.2 (1) with the constant $\rho/m$ . Definition 4.2 (2) holding for $\hat{\mathcal{T}}_{\inf}$ directly follows from the assumption that $\mathcal{T}_{\inf}$ satisfies Definition 4.2 (2).

For $\hat{\mathcal{T}}_{\rm fin}$ , we take $e=P_{1}$ and $\varepsilon_{0}=1/2$ . The projection $P_{1}$ belongs to $\operatorname{Ped}(B)_{+}$ , and

d_{\widehat{\tau}}((P_{1}-1/2)_{+})=\widehat{\tau}(P_{1})=1>\rho/m.

Finally, by the assumption on $\mathcal{T}_{\rm fin}$ , it is not hard to see

\sup_{\hat{\tau}\in\hat{\mathcal{T}}_{\mathrm{fin}}}\left(\sup_{j}\hat{\tau}(P_{j})-\hat{\tau}(P_{n})\right)\longrightarrow 0.

This establishes Definition 4.2 (3). It then follows from Corollary 5.4 that $B$ has no maximal ideals. Then Lemma 6.11 provides the conclusion for $A$ . ∎

In the finitely many normalized extreme $2$ -quasitraces case, we have the following.

Theorem 6.13.

Let $A$ be simple and nonunital, with an increasing projection approximate identity $(p_{n})$ . Suppose that $D=p_{1}Ap_{1}$ is stably finite, $\operatorname{rc}_{\mathrm{p}}(D)<\infty$ , and $\operatorname{QT}^{1}_{2}(D)$ , (equivalently, $\operatorname{QT}_{2}^{1}(A,[p_{1}])$ ) is nonempty with finitely many extreme points. Then $A$ has no maximal ideals.

Proof.

Take $\mathcal{T}=\partial_{e}\operatorname{QT}^{1}_{2}(A,[p_{1}])$ and partition it into $\mathcal{T}_{\inf}$ and $\mathcal{T}_{\rm fin}$ . Since $A$ is simple, each $p_{n}$ is subequivalent to a finite direct sum of copies of the full projection $p_{1}$ . Thus, $\tau(p_{n})$ is finite for every $\tau\in\mathcal{T}$ . Therefore, one has

\sup_{\tau\in\mathcal{T}_{\inf}}\tau(p_{n})=\max_{\tau\in\mathcal{T}_{\inf}}\tau(p_{n})<\infty.

Now let $\tau\in\mathcal{T}_{\mathrm{inf}}$ . Then each sequence $\tau(p_{n})$ increases to $\infty$ by definition. Given $R>0$ , choose $N_{\tau}$ with $\tau(p_{N_{\tau}})>R$ and define $N_{0}=\max_{\tau\in\mathcal{T}_{\inf}}N_{\tau}$ . Then $\inf_{\mathcal{T}_{\mathrm{inf}}}\tau(p_{n})>R$ for all $n>N_{0}$ . Therefore, one has

\lim_{n\to\infty}\inf_{\tau\in\mathcal{T}_{\inf}}\tau(p_{n})=\infty.

For $\tau\in\mathcal{T}_{\mathrm{fin}}$ , the number $\sup_{j}\tau(p_{j})<\infty$ . Since $\{p_{n}:n=0,1\dots,\}$ is increasing, the number $\sup_{j}\tau(p_{j})-\tau(p_{n})\to 0$ as $n\to\infty$ . This implies

\sup_{\tau\in\mathcal{T}_{\mathrm{fin}}}\left(\sup_{j}\tau(p_{j})-\tau(p_{n})\right)\longrightarrow 0

because $\mathcal{T}_{\rm fin}$ is finite. Now the result follows from Lemma 6.12. ∎

This first covers nonunital simple AF algebras with finitely many trace. As another corollary, we have the following for $\mathcal{Z}$ -stable $C^{*}$ -algebras.

Corollary 6.14.

Proof.

Define $B=p_{1}Ap_{1}$ . The projection $p_{1}$ is full and therefore $B\otimes\mathcal{K}\simeq A\otimes\mathcal{K}$ . Therefore, $\operatorname{Cu}(B)$ is isomorphic to $\operatorname{Cu}(A)$ . Since $A$ is $\mathcal{Z}$ -stable, the Cuntz semigroup $\operatorname{Cu}(A)$ is almost unperforated by [27], and so is $\operatorname{Cu}(B)$ . This implies that $\operatorname{rc}_{\mathrm{p}}(B)=0$ . Theorem 6.13 now applies to $A$ . ∎

6.3. Hereditary subalgebras

Corollary 6.15 (Real rank zero).

Let $D$ be separable, unital, simple, stably finite, and of real rank zero, with $\operatorname{rc}(D)<\infty$ . If $\operatorname{QT}^{1}_{2}(D)$ has finitely many extreme points, then every nonzero hereditary subalgebra of $D\otimes\mathcal{K}$ has no maximal ideals.

Proof.

First, any hereditary subalgebra in $D\otimes\mathcal{K}$ is separable, and thus is of the form $A=\overline{h(D\otimes\mathcal{K})h}$ for some positive $h\in A$ . If $A$ is unital, simplicity of $A$ gives the algebraic simplicity of $A$ . Otherwise, it follows from [25] that real rank zero of $D\otimes\mathcal{K}$ provides an increasing projection approximate identity $(p_{n})$ in $A$ . Write $B=p_{1}Ap_{1}$ , which is also a hereditary $C^{*}$ -algebra of $D\otimes\mathcal{K}$ . Then the inclusion $\iota:B\to D\otimes\mathcal{K}$ induce an order embedding $\Phi:\operatorname{Cu}(A)\to\operatorname{Cu}(D\otimes\mathcal{K})=\operatorname{Cu}(D)$ such that $\Phi([1_{B}]_{B})=[p_{1}]_{D}$ (see, e.g., [1, Lemma 5.11]). Moreover, since $B$ is full in $D\otimes\mathcal{K}$ , then $\Phi$ has to be an isomorphism. This implies that $\operatorname{QT}_{2}^{1}(B)$ can be identified by $\operatorname{QT}_{2}^{1}(D\otimes\mathcal{K},[p_{1}])$ .

Then we work in $\operatorname{Cu}(D)$ . First, by simplicity, there exists $n,m\in\mathbb{N}$ such that $[p_{1}]\leq n[1_{D}]$ and $[1_{D}]\leq m[p_{1}]$ . This implies that for any $\tau\in\operatorname{QT}_{2}^{1}(D\otimes\mathcal{K},[p_{1}])$ , one has

\frac{1}{n}\leq d_{\tau}([1_{D}])=\tau(1_{D})\leq m.

In addition, note that $f:\operatorname{QT}_{2}^{1}(D\otimes\mathcal{K},[p_{1}])\to\operatorname{QT}_{2}^{1}(D\otimes\mathcal{K},[1_{D}])$ , defined by $\tau\mapsto\tau/\tau(1_{D})$ , is bijective. Now, let $\rho>\rm rc(D)$ and suppose

d_{\tau}([a])+\rho\tau(1_{D})<d_{\tau}([b])

holds for any $\tau\in\operatorname{QT}_{2}^{1}(D\otimes\mathcal{K},[p_{1}])=\operatorname{QT}_{2}^{1}(B)$ . Then, this implies that

d_{\eta}([a])+\rho<d_{\eta}([b])

holds for any $\eta\in\operatorname{QT}_{2}^{1}(D\otimes\mathcal{K},[1_{D}])=\operatorname{QT}^{1}_{2}(D)$ and thus $a\precsim b$ . This implies $\rm rc(B)\leq m\rho<\infty$ .

Finally note that $f^{-1}:\operatorname{QT}_{2}^{1}(D)\to\operatorname{QT}_{2}^{1}(B)$ preserve the extreme points. Thus, if $\operatorname{QT}^{1}_{2}(D)$ has finitely many extreme points, so does $\operatorname{QT}^{1}_{2}(B)$ . Now apply Theorem 6.13 to $A$ . ∎

We end this subsection by providing a concrete AF example satisfying Corollary 6.15 such that both $\mathcal{T}_{\inf}$ and $\mathcal{T}_{\rm fin}$ are non-empty. This should be compared to the case in Theorem 6.9, in which only $\mathcal{T}_{\inf}$ is non-empty.

Example 6.16.

Consider the ordered group

G=\mathbb{Q}^{2},\qquad G_{+}=\{(0,0)\}\cup\{(x,y):x>0,\ y>0\},\qquad u=(1,1).

The normalized states form the segment with endpoints $\rho_{1}(x,y)=x$ and $\rho_{2}(x,y)=y$ . Then using [11], one obtains a simple unital AF algebra $D_{0}$ with ordered $K_{0}$ -group $(G,G_{+},u)$ and these two extremal traces. Choose orthogonal projections $q_{n}\in D_{0}\otimes\mathcal{K}$ with $[q_{n}]=(2^{-n},1)$ , and put

h=\sum_{n=1}^{\infty}2^{-n}q_{n},\qquad A=(D_{0}\otimes\mathcal{K})_{h}.

Then $A$ is simple, nonunital, nonstable, and AF. Its Pedersen ideal is proper, and it has no maximal ideals.

Proof.

Projection realization in the stable AF algebra gives the $q_{n}$ ; placing representatives in disjoint finite matrix blocks makes them orthogonal. Recall a standard fact that lower semi-continuous traces can be identified by states on the dimension group. This allows to identify $\rho_{1},\rho_{2}$ to be extreme traces on $D_{0}$ . Use $p_{n}=\sum_{j=1}^{n}q_{j}$ and normalize $\tau_{1}=2\rho_{1}$ , $\tau_{2}=\rho_{2}$ at $p_{1}=q_{1}$ . Moreover, one has

\tau_{1}(p_{n})=\sum_{i=1}^{n}\tau_{1}(q_{i})=2(1-2^{-n})

and

\tau_{2}(p_{n})=\sum_{i=1}^{n}\tau_{2}(q_{i})=n.

Define $\mathcal{T}=\{\tau_{1},\tau_{2}\}$ has $\mathcal{T}_{\mathrm{fin}}=\{\tau_{1}\}$ and $\mathcal{T}_{\mathrm{inf}}=\{\tau_{2}\}$ .

The order on $K_{0}(D_{0})$ gives strict comparison of projections. This implies the strict comparison of positive element because $D_{0}\otimes\mathcal{K}$ is AF and has real rank 0. Therefore, one has $\rm rc(D_{0})=0$ . Then it follows from Corollary 6.15 that $A$ has no maximal ideals.

On the other hand, since $A$ has a finite trace $\tau_{1}$ , it is not stable. Finally, the properness of $\operatorname{Ped}(A)$ in $A$ follows from [2, Remark 4.6(c)] because $A$ has an infinite trace $\tau_{2}$ . ∎

6.4. $C^{*}$ -algebras containing a properly infinite projection

In this section, we show that if there exists a non-trivial properly infinite projection in a simple non-unital $C^{*}$ -algebra $A$ with an approximate identities of increasing projections, then $A$ is algebraically simple. Recall that a nonzero projection $p$ is properly infinite if $p\oplus p\precsim p$ .

Lemma 6.17.

Let $(p_{n})$ be an increasing projection approximate identity of $A$ , with $p_{0}=0$ and $q_{n}=p_{n}-p_{n-1}$ . Suppose that, for a projection $P\in M_{m}(A)$ , there are mutually orthogonal projections $r_{n}\leq P$ equivalent to $q_{n}\otimes e_{11}$ . Then $\operatorname{Ped}(A)=A$ .

Proof.

Choose $u_{n}\in M_{m}(A)$ with $u_{n}^{*}u_{n}=q_{n}\otimes e_{11}$ and $u_{n}u_{n}^{*}=r_{n}$ . Orthogonality gives $u_{n}^{*}u_{m}=0$ for $n\neq m$ . For $a\in A_{+}$ set

y_{N}=\Bigl(\sum_{n=1}^{N}u_{n}\Bigr)(a^{1/2}\otimes e_{11}).

If $M>N$ , then

\|y_{M}-y_{N}\|^{2}=\|a^{1/2}\sum_{n=N+1}^{M}(q_{n}\otimes e_{11})a^{1/2}\|=\|a^{1/2}(p_{M}-p_{N})a^{1/2}\|\longrightarrow 0,

because $(p_{n})$ is an approximate identity. Thus $y_{N}\to y$ in norm for some $y\in M_{m}(A)$ . Since $Pu_{n}=Pr_{n}u_{n}=r_{n}u_{n}=u_{n}$ , one has $Py=y$ . In addition, one has

y^{*}Py=\lim_{N}y_{N}^{*}y_{N}=\lim_{N}(a^{1/2}p_{N}a^{1/2})\otimes e_{11}=a\otimes e_{11}.

By Theorem 2.11(i), $P\in\operatorname{Ped}(M_{m}(A))=M_{m}(\operatorname{Ped}(A))$ (see e.g., [19, Section 5.6]) and therefore every entry $P_{ij}$ of $P$ comes from $\operatorname{Ped}(A)$ . Taking the first diagonal entry of the exact identity above gives

a=\sum_{i,j=1}^{m}y_{i1}^{*}P_{ij}y_{j1}\in\operatorname{Ped}(A).

All positive elements belong to $\operatorname{Ped}(A)$ , and hence $\operatorname{Ped}(A)=A$ . ∎

Theorem 6.18.

Let $A$ be simple and $\sigma$ -unital, with an approximate identity of increasing projections. If $A$ contains a nonzero properly infinite projection, then $\operatorname{Ped}(A)=A$ .

Proof.

The unital case follows from simplicity, so assume $A$ is nonunital. Let $(p_{n})$ be the increasing projections serving as the approximate identity of $A$ and define $q_{n}=p_{n}-p_{n-1}$ as before. Let $p\in A$ be a non-zero properly infinite projection. Since $p$ is full and $[q_{n}]$ is compact, there exists a $k_{n}\in\mathbb{N}$ such that,

[q_{n}]\leq k_{n}[p]\leq[p].

We next construct mutually orthogonal projections $s_{n}\leq p$ , each equivalent to $p$ . Proper infiniteness gives orthogonal subprojections $s_{1},t_{1}\leq p$ with $s_{1}\sim_{\rm MvN}p\sim_{\rm MvN}t_{1}$ . The projection $t_{1}$ is again properly infinite. Split it into orthogonal $s_{2},t_{2}$ , each equivalent to $t_{1}$ , and continue. At step $n$ , split the unused projection $t_{n-1}$ ; hence all the $s_{n}$ are orthogonal and all are equivalent to $p$ . Since $q_{n}\precsim s_{n}$ , Lemma 2.5 gives $u_{n}\in A$ with

u_{n}^{*}u_{n}=q_{n},\qquad r_{n}=u_{n}u_{n}^{*}\leq s_{n}.

The $r_{n}$ are orthogonal subprojections of $p$ . Now, Lemma 6.17, with $m=1$ and $P=p$ , implies $\operatorname{Ped}(A)=A$ . ∎

We remark that if $A$ is $\sigma$ -unital, simple and purely infinite, then it follows from [17, Theorem 2.11] that any positive element is a positive combination of projections. Since $\operatorname{Ped}(A)$ contains all projections, one also has $\operatorname{Ped}(A)=A$ .

7. Non-simple $C^{*}$ -algebras

When the $C^{*}$ -algebra is not simple, one looks at a nonzero closed ideal $I$ and the short exact sequence $0\to I\to E\to E/I\to 0$ . The next proposition shows how the preceding simple examples can be used in the setting.

Proposition 7.1.

Consider the short exact sequence of $C^{*}$ -algebras

0\longrightarrow I\longrightarrow E\overset{\pi}{\longrightarrow}B\longrightarrow 0,\qquad I\neq 0.

Suppose that $I$ has no maximal ideals and is not algebraically simple. Then every maximal ideal of $E$ contains $I$ , and

N\longmapsto\pi^{-1}(N)

is a bijection from the collection of ideals $N\lhd B$ for which $B/N$ is nonzero and algebraically simple onto the family of maximal ideals of $E$ . Consequently, $E$ has no maximal ideals if and only if $B$ has no nonzero algebraically simple quotient.

Proof.

Let $M$ be a maximal ideal of $E$ . Suppose first that $I\nsubseteq M$ . The ideal $I+M$ strictly contains $M$ , so maximality of $M$ implies $I+M=E$ . The quotient map restricts to a surjection $I\to E/M$ , and hence

I/(I\cap M)\cong E/M,

which is nonzero and algebraically simple. Now, If $I\cap M=0$ , then $I$ is algebraically simple, which is a contradiction to the hypothesis. Otherwise, $I\cap M$ has to be a maximal ideal of $I$ , which is also a contradiction to the hypothesis. Thus $I\subseteq M$ necessarily holds.

For the second claim. Let $N\lhd B$ such that $B/N$ is nonzero and algebraically simple. Then $\pi^{-1}(N)$ has to be maximal in $E$ and thus contains $I$ . Moreover, $\pi^{-1}(N)\neq E$ . For the converse, let $M$ be maximal in $E$ and define $N=\pi(M)\lhd B$ . The induced quotient homomorphism $\tilde{\pi}:E/M\to B/N$ has to be an isomorphism because $E/M$ is algebraically simple. Note that this two assignments are inverse, proving the bijection and the last assertion. ∎

The hypotheses on $I$ hold for $I=D\otimes\mathcal{K}$ whenever $D$ satisfies Theorem 6.9, in particular for $I=\mathcal{K}$ . If $B$ is simple, the condition on $B$ says precisely that it has no maximal ideals and $\operatorname{Ped}(B)\neq B$ . The mixed examples above therefore provide choices of $B$ for which the middle algebra $E$ has no maximal ideals. On the other hand, an algebraically simple quotient makes $I$ maximal, as in the sequence $0\to\mathcal{K}\to\mathcal{K}^{\sim}\to\mathbb{C}\to 0$ .

As an application, we have the following. Denote by $\mathcal{C}$ the class of non-unital $C^{*}$ -algebras $A$ without maximal ideal and $\operatorname{Ped}(A)\neq A$ , for example, $C^{*}$ -algebras $A$ with $\operatorname{Ped}(A)\neq A$ satisfying Theorem 6.9, or Theorem 6.13, or Corollary 6.15.

Theorem 7.2.

Let

0\longrightarrow I\longrightarrow E\overset{\pi}{\longrightarrow}B\longrightarrow 0,\qquad I\neq 0

be a short exact sequence of $C^{*}$ -algebras. Suppose $I,B$ are in the class $\mathcal{C}$ . Then $E\in\mathcal{C}$ as well. Thus $\mathcal{C}$ is closed under extension.

Proof.

The first statement follows from the definition of $\mathcal{C}$ and Proposition 7.1. For the second statement. It is shown in [21, Corollory 6] that $\pi(\operatorname{Ped}(E))=\operatorname{Ped}(B)$ . Thus, if $\operatorname{Ped}(E)=E$ , then necessarily one has $\operatorname{Ped}(B)=B$ , which is a contradiction. ∎

The same argument applies inductively to finite filtrations whose nonzero subquotients have no maximal ideals and are not algebraically simple.

8. Acknowledgement

The project began in May 2025, during the second author’s visit to Dalian University of Technology, and many of the original ideas are due entirely to humans. But during the development of this work, the authors used GPT 5.6 sol and GPT 6 astra for searching necessary references, examining mathematical arguments, and language editing. The authors have checked and substantially revised all mathematical arguments generated by AI and take full responsibility for the final version.

References

[1] R. Antoine, F. Perera, and H. Thiel, Cuntz semigroups of ultraproduct $C^{*}$ -algebras. J. Lond. Math. Soc. 102 (2020), 994–1029.
[2] B. Blackadar, Traces on simple AF $C^{*}$ -algebras. J. Funct. Anal., 38 (1980), 156–168.
[3] B. Blackadar and D. Handelman, Dimension functions and traces on C*-algebras. J. Funct. Anal., 45 (3) (1982), 297–340.
[4] E. Blanchard and E. Kirchberg, Non-simple purely infinite C*-algebras: the Hausdorff case. J. Funct. Anal., 207 (2004), 461–513.
[5] L. G. Brown, Stable isomorphism of hereditary subalgebras of $C^{*}$ -algebras, Pacific J. Math. 71 (1977), 335–348.
[6] B. Blackadar, L. Robert, A. P. Tikuisis, A. S. Toms, and W. Winter, An algebraic approach to the radius of comparison, Trans. Amer. Math. Soc. 364 (2012), 3657–3674.
[7] N. P. Brown and W. Winter, Quasitraces are traces: a short proof of the finite-nuclear-dimension case. C. R. Math. Acad. Sci. Soc. R. Can., 33 (2) (2011), 44–49.
[8] K. T. Coward, G. A. Elliott and C. Ivanescu, The Cuntz semigroup as an invariant for C*-algebras. J. Reine Angew. Math., 623 (2008), 161–193.
[9] P. J. Cohen, Factorization in group algebras, Duke Math. J. 26 (1959), 199–205.
[10] J. Cuntz, Dimension functions on simple C*-algebras. Math. Ann., 233 (2) (1978), 145–153.
[11] E. G. Effros, D. E. Handelman, and C.-L. Shen, Dimension groups and their affine representations, Amer. J. Math. 102 (1980), 385–407.
[12] G. A. Elliott, L. Robert, and L. Santiago, The cone of lower semicontinuous traces on a $C^{*}$ -algebra, Amer. J. Math. 133 (2011), 969–1005.
[13] T. Fack and H. Kosaki, Generalized $s$ -numbers of $\tau$ -measurable operators, Pacific J. Math. 123 (1986), 269–300.
[14] E. Gardella and H. Thiel, Prime ideals in $C^{*}$ -algebras and applications to Lie theory, Proc. Amer. Math. Soc. 152 (2024), 3647–3656.
[15] E. Gardella, K. Kitamura, and H. Thiel, Semiprime ideals in $C^{*}$ -algebras, J. Eur. Math. Soc., published online first (2025),
[16] U. Haagerup, Quasitraces on exact $C^{*}$ -algebras are traces, C. R. Math. Acad. Sci. Soc. R. Can. 36 (2014), 67–92.
[17] V. Kaftal, P. W. Ng, and S. Zhang, Positive combinations and sums of projections in purely infinite simple $C^{*}$ -algebras and their multiplier algebras, Proc. Amer. Math. Soc. 139 (2011), 2735–2746.
[18] N. Ozawa, Is every maximal ideal in a $C^{*}$ -algebra always closed?, MathOverflow (2017), Question 270654.
[19] G. K. Pedersen, $C^{*}$ -Algebras and Their Automorphism Groups, London Mathematical Society Monographs, vol. 14, Academic Press, London, 1979.
[20] G. K. Pedersen, Measure theory for $C^{*}$ -algebras, Math. Scand. 19 (1966), 131–145.
[21] G. K. Pedersen, A decomposition theorem for $C^{*}$ -algebras, Math. Scand. 22 (1968), 266–268.
[22] G. K. Pedersen and N. H. Petersen, Ideals in a $C^{*}$ -algebra, Math. Scand. 27 (1970), 193–204 (1971).
[23] M. Rørdam, On the structure of simple $\mathrm{C}^{*}$ -algebras tensored with a UHF-algebra, II. J. Funct. Anal., 107 (1992), 255–269.
[24] A. S. Toms, Flat dimension growth for $C^{*}$ -algebras, J. Funct. Anal. 238 (2006), 678–708.
[25] L. G. Brown and G. K. Pedersen, $C^{*}$ -algebras of real rank zero, J. Funct. Anal. 99 (1991), 131–149.
[26] E. Gardella and F. Perera, The modern theory of Cuntz semigroups of $C^{*}$ -algebras, EMS Surv. Math. Sci. 13 (2026), 133–214,
[27] M. Rørdam, The stable and the real rank of $\mathcal{Z}$ -absorbing $C^{*}$ -algebras, Internat. J. Math. 15 (2004), 1065–1084.
[28] M. Rørdam, and W. Winter The Jiang–Su algebra revisited J. Reine Angew. Math., 642 (2010), 129–155.

Maximal Algebraic Ideals in Nonunital C∗C^{*}-Algebras

Abstract.

Key words and phrases:

2020 Mathematics Subject Classification

1. Introduction

Question 1.1.

Corollary 1.2 (Theorem 6.9).

Corollary 1.3 (Corollary 6.14).

Outline of the paper

2. Preliminaries

2.1. Cuntz comparison

Definition 2.1.

Definition 2.2.

Lemma 2.3 ([23]).

Definition 2.4.

Lemma 2.5.

Definition 2.6.

Remark 2.7.

Definition 2.8.

Definition 2.9.

2.2. Algebraic ideals

Definition 2.10.

Theorem 2.11.

Remark 2.12.

Definition 2.13.

Theorem 2.14.

Proof.

Corollary 2.15.

Proposition 2.16.

Proposition 2.17.

Remark 2.18.

3. Generalized singular functions

Lemma 3.1.

Proof.

Lemma 3.2.

Proof.

Definition 3.3.

Proposition 3.4.

Proof.

Proposition 3.5.

Proof.

Corollary 3.6.

Proposition 3.7.

Proof.

4. Admissible quasitracial projection scale

Definition 4.1.

Definition 4.2.

Lemma 4.3.

Proof.

Proposition 4.4.

Proof.

Lemma 4.5.

Proof.

Definition 4.6.

Theorem 4.7.

Proof.

5. The quasitracial-scale theorem

Lemma 5.1.

Proof.

Remark 5.2.

Theorem 5.3.

Proof.

Corollary 5.4.

Proof.

Corollary 5.5 (Nonsimple form).

Proof.

6. Main applications

6.1. Stablization of unital simple algebras

Definition 6.1 ([24]).

Definition 6.2.

Definition 6.3.

Remark 6.4.

Definition 6.5.

Remark 6.6.

Proposition 6.7.

Proof.

Remark 6.8.

Theorem 6.9.

Proof.

Proposition 6.10.

Maximal Algebraic Ideals in Nonunital $C^{*}$ -Algebras

6.2. $\mathcal{Z}$ -stability and comparison

6.4. $C^{*}$ -algebras containing a properly infinite projection

7. Non-simple $C^{*}$ -algebras