A group-algebra trace can remember only finitely many torsion classes

← All digests

Family 207 · Algebra
Importance 5/5 · Provisional editorial rating
Sources consulted 2026-10-21

1. The question

For an ordinary projection matrix, the trace records the dimension of its image. Over a group algebra, a more refined trace can record separate contributions from different conjugacy classes of the group.

Which classes can occur? The Bass trace conjectures predict severe restrictions: infinite-order group elements should contribute nothing. In the analytic algebra $\ell^1(G)$, where coefficients may have infinite but absolutely summable support, there is an additional demand. Only finitely many finite-order conjugacy classes should survive.

Family 207 contains two manuscripts. M2, dated 24 September 2026, claims the complex group-ring Bass theorem and derives the characteristic-zero Kaplansky idempotent conjecture. M1, dated 5 October, claims the stronger $\ell^1$-Bass theorem for every discrete group. Its argument develops the analytic estimates needed to control infinitely supported coefficients.

2. Background and essential definitions

Two algebras built from a group

The complex group ring $\mathbb CG$ consists of sums $\sum_g a_g g$ with only finitely many nonzero complex coefficients. Multiplication extends the group law linearly.

The Banach group algebra $\ell^1(G)$ allows infinitely many nonzero coefficients, provided

$$\|a\|_1=\sum_{g\in G}|a_g|<\infty.$$

Multiplication is convolution:

$$(a*b)(g)=\sum_{h\in G}a(h)b(h^{-1}g).$$

Absolute summability makes this well defined. The distinction between finite support and summable support is important: a summable sequence can have infinitely many nonzero entries.

An element $e$ is idempotent if $e^2=e$. More generally, an idempotent matrix $e\in M_q(A)$ over an algebra $A$ selects a direct summand $eA^q$ of a free module. Such summands are finitely generated projective modules. They generalize vector bundles and provide the objects whose “ranks” are being measured.

A trace indexed by conjugacy classes

Two group elements are conjugate if one has the form $hgh^{-1}$ for the other. Let $C$ be a conjugacy class. For an idempotent matrix with group-algebra entries, define

$$\tau_C(e)=\sum_{g\in C}\sum_{i=1}^q e_{ii}(g).$$

First take the usual matrix trace coefficientwise, then sum the coefficients over the class. These sums converge absolutely in the $\ell^1$ setting. The collection of all $\tau_C(e)$ is the Hattori–Stallings trace.

Grouping by conjugacy is what restores the cyclic trace identity in a noncommutative group algebra. The trace is unchanged by similarity and depends only on the corresponding projective module. It therefore extends to $K_0(A)$, the abelian group formed from projective modules using direct-sum relations and formal differences.

A finite-order element satisfies $g^m=1$ for some positive integer $m$. “Finite-order conjugacy class” refers to the order of its elements, not the number of elements in the class. A torsion element can have an infinite conjugacy class.

3. Development history and previous partial results

Hattori and Stallings introduced the refined rank in the 1960s. Bass connected it with Euler characteristics and formulated trace-support conjectures for group rings. Over the integral group ring $\mathbb ZG$, the conjecture asks for support only at the identity. Over $\mathbb CG$, finite-order classes must be allowed.

The reason is already visible for a finite group $F$. Its averaging element

$$e_F=\frac1{|F|}\sum_{g\in F}g$$

is idempotent. Its trace contribution at a class $C\subseteq F$ is $|C|/|F|$, which need not vanish away from the identity. Complex and integral versions therefore have different permitted supports.

Bass proved important cases, including complex trace vanishing for linear groups. Burghelea’s computation of cyclic homology separated contributions from finite- and infinite-order conjugacy classes. Eckmann and Emmanouil used cyclic-homological methods to establish further cases under structural or homological hypotheses.

Berrick, Chatterji and Mislin formulated the $\ell^1$-Bass conjecture in the finite-support form used here. They proved it for countable groups with rationally surjective degree-zero Bost assembly map. Together with Lafforgue’s work, this includes amenable groups. Their reduction also explains how to pass from suitable countable subgroups to arbitrary groups.

Related work on rapid-decay completions by Ji, Ogle and Ramsey emphasizes a recurring analytic obstacle: moving from finitely supported group-ring elements to completed algebras requires estimates compatible with infinite sums.

The idempotent question has a parallel history. Classical trace arguments show that a scalar idempotent in $\mathbb CG$ has canonical trace between zero and one, and that the endpoint values force triviality. Known rationality of the trace does not suffice, since rational numbers can lie strictly between the endpoints. M2’s torsion-free trace formula claims the missing integrality by identifying the trace with augmentation rank.

4. What the new manuscripts claim

M1: the full ℓ¹ support assertion

For every discrete group $G$ and every idempotent matrix $e\in M_q(\ell^1(G))$, M1 asserts that there is a finite set $F$ of finite-order conjugacy classes such that

$$\tau_C(e)=0\qquad\text{whenever }C\notin F.$$

The surviving set can depend on $e$. No amenability, growth condition, finite generation or assembly-map assumption is imposed on $G$.

Both parts of the conclusion matter: all infinite-order classes vanish, and only finitely many finite-order classes remain. Absolute summability alone would provide neither this finite cutoff nor its uniformity across conjugacy classes.

M2: the algebraic theorem and idempotents

M2 asserts vanishing on infinite-order classes for every projective class over $\mathbb CG$. Finite support is automatic in this algebraic setting. Combining that theorem with the established restriction on integral finite-order coefficients gives the identity-only Bass assertion over $\mathbb ZG$.

For torsion-free $G$, the identity is the only finite-order element. Let $\tau_G$ take the coefficient of the identity, and let the augmentation $\epsilon_G$ sum all coefficients. For an idempotent matrix, M2 obtains

$$\sum_i\tau_G(e_{ii}) =\operatorname{rank}_{\mathbb C}\bigl(\epsilon_G(e)\bigr).$$

The augmentation is applied entrywise. The right side is the ordinary rank of a complex idempotent matrix, hence an integer.

The manuscript then claims that for every torsion-free group and every commutative unital domain $R$ of characteristic zero, a scalar idempotent $e\in RG$ is either $0$ or $1$. This is the characteristic-zero Kaplansky idempotent conjecture. It concerns elements of the ring itself; nontrivial idempotent matrices are not forbidden.

M1 builds on the weighted-configuration ideas of M2, but states that the complex group-ring theorem is not an input to its proof. It develops the necessary ingredients directly for the analytic setting.

5. Why the advance matters

Importance: 5/5, provisional. These are claimed general resolutions of central trace-support problems for group algebras, together with a broad idempotent consequence in characteristic zero.

The trace connects module theory, conjugacy, topology and analysis. Support restrictions limit which group elements can contribute to projective Euler characteristics. The integral theorem also has fixed-point consequences through earlier equivalences involving homotopy-idempotent maps.

The $\ell^1$ statement is a substantial strengthening of the algebraic one. It says that passing to an algebra allowing infinite tails does not produce an infinite tail of torsion-class contributions in this trace.

The score concerns those structural claims. It does not imply that arbitrary group rings have no zero divisors, or that every completed group algebra has no nontrivial projections. Those are different questions with different hypotheses.

6. The main idea

Idempotency makes weighted paths compatible

Fix an element $g$ whose trace coefficient is to be studied. Products of matrix coefficients along lists of group elements assign complex weights to paths whose closing edge is twisted by multiplication by $g$.

The equation $e^2=e$ has a useful combinatorial meaning: summing over an intermediate vertex removes that vertex without changing the total weight. Cyclicity of matrix trace gives another compatibility under rotating the list.

These identities let the manuscript build higher-dimensional cycles whose evaluation by a suitable differential or combinatorial expression recovers $\tau_{[g]}(e)$. The detecting expression involves a Pfaffian, a signed sum over pairings associated with a skew-symmetric matrix. Pfaffians naturally package the repeated two-form contributions arising in even dimensions.

M2 uses finite geometric support

For a group-ring idempotent, only finitely many group steps occur. They define a graph of bounded degree, fixed as the dimension increases. The relevant simplices use connected sets of vertices in that graph.

M2 cuts their supporting complex through a sequence of separating sets. Volume estimates and repeated slicing make the expected number of final zero-dimensional pieces arbitrarily small in high dimension. Equivariant chain comparisons preserve the original cocycle evaluation while allowing it to be bounded by that small quantity.

The construction uses a free probability parameter to separate overlapping translates of local modifications. It does not assume the group is amenable. The resulting high-dimensional vanishing forces the trace coefficient of the chosen infinite-order element to be zero.

M1 first replaces arbitrary tails by exponential decay

An $\ell^1$ idempotent need not have finite support. M1 approximates it by a finite-support matrix, then applies an analytic correction to obtain a nearby exact idempotent $p$. The correction generally restores an infinite tail, but with an exponential moment in a finitely generated subgroup $\Gamma$:

$$\sum_{h\in\Gamma}\|p(h)\|e^{c|h|}<\infty \quad\text{for some }c>0.$$

Here $|h|$ is word length for a finite generating set. The corrected matrix is similar to the original, so every conjugacy-class trace coefficient is preserved exactly. The step is not a claim that every projective class has a finite-support idempotent representative.

Timed paths are then averaged over several time scales. Smooth summaries of their spatial distributions provide differential forms whose derivatives remain controlled even when the path support is large. Normalization produces connection forms, and a cyclic Stokes identity relates their curvature Pfaffians to the trace.

Different scales require different estimates: sparse interactions at the initial scale, Hilbert–Schmidt bounds at intermediate scales, and singular-value bounds for ordered kernels at the final scale. Together they give an exponentially decreasing bound as the simplex dimension grows.

One radius gives finite support

The decisive uniformity is a single spatial radius chosen independently of the conjugacy class to be eliminated. The estimate removes every infinite-order class and every finite-order class with no representative in a fixed finite word ball.

Only finitely many conjugacy classes meet that ball. This gives actual finite support. Passing from $\Gamma$ back to $G$ can merge conjugacy classes but cannot turn a finite surviving list into an infinite one.

For M2’s scalar idempotent consequence, augmentation gives rank zero or one. A classical faithful-trace argument then forces the idempotent to be zero or one. Extension to an arbitrary characteristic-zero domain uses only the finitely generated field containing the finitely many coefficients of the given element, which embeds into $\mathbb C$; no embedding of the whole coefficient domain is required.

7. Scope, limitations, and open questions

The groups are arbitrary discrete groups, possibly uncountable. The analytic statement concerns the complex Banach algebra $\ell^1(G)$; the idempotent corollary is stated for characteristic-zero commutative domains.

The reduced group $C^*$-algebra uses a different completion. M2 identifies the trace on the image of algebraic $K_0(\mathbb CG)$ inside its topological K-theory, but does not assert that every class in that completed K-theory comes from the group ring. It therefore does not claim the full Kadison–Kaplansky conjecture for arbitrary torsion-free groups.

A nontrivial idempotent gives zero divisors through $e(1-e)=0$, but excluding idempotents does not exclude all zero divisors. The positive-characteristic constructions in family 196 and the direct-finiteness questions in family 197 should not be conflated with the characteristic-zero scalar conclusion here.

No explicit universal bound on the number of surviving classes is part of M1’s headline theorem. Its finite cutoff depends on the idempotent and the estimates constructed for it.

8. Sources and further reading

Bass, Euler characteristics and characters of discrete groups, Inventiones Mathematicae 35 (1976), 155–196, gives the foundational setting. Berrick, Chatterji and Mislin, From acyclic groups to the Bass conjecture for amenable groups, Mathematische Annalen 329 (2004), 597–621, formulates the analytic conjecture and proves the assembly-based cases.

Burghelea, The cyclic homology of the group rings, Commentarii Mathematici Helvetici 60 (1985), 354–365, supplies the conjugacy-class decomposition. Eckmann, Cyclic homology of groups and the Bass conjecture, in the same journal 61 (1986), 193–202, develops the higher-degree vanishing approach.

Burger and Valette, Idempotents in complex group rings: theorems of Zalesskii and Bass revisited, Journal of Lie Theory 8 (1998), 219–228, explains the classical trace facts used in the scalar corollary.

Source manuscripts

• M1. OpenAI, The ℓ¹-Bass Conjecture for Discrete Groups (October 5, 2026).
• M2. OpenAI, The Bass trace conjecture and the characteristic-zero Kaplansky idempotent conjecture (September 24, 2026).

Sources consulted on 21 October 2026 at repository revision 1581d4c. Download the manuscript citations.