Torsion Subcomplex Reduction in Group Cohomology

• Torsion subcomplex reduction is a method that isolates cells whose stabilizers contain ℓ-torsion, enabling a smaller, cohomology-equivalent complex for computational ease.
• The approach employs combinatorial operations such as merging adjacent cells and cutting off terminal cells, ensuring the preservation of mod ℓ cohomology under Brown-type conditions.
• Recent advancements use discrete Morse theory to systematically collapse large torsion loci, significantly improving algorithms and runtime for arithmetic groups like PSL₄(ℤ).

Torsion subcomplex reduction is a method in equivariant topology and arithmetic group cohomology that isolates the part of a proper group action responsible for a fixed prime $\ell$, and then replaces that typically large $\ell$-torsion locus by a much smaller equivariantly cohomology-equivalent object. In the standard setting, a discrete group $\Gamma$ acts cellularly on a finite-dimensional CW-complex $X$ with finite stabilizers fixing cells pointwise, and the $\ell$-torsion subcomplex consists of those cells whose stabilizers contain elements of order $\ell$. Under Brown-type hypotheses, the $\ell$-primary part of Farrell–Tate cohomology is already detected on this subcomplex, so the computational problem becomes one of reducing the $\ell$-torsion geometry without changing the relevant equivariant cohomology. This framework was formulated explicitly as torsion subcomplex reduction in Rahm’s work and survey, and in 2025 it was recast for arithmetic groups through a discrete-Morse-theoretic implementation that yields a simpler algorithm and runtime improvements, with a computation of the mod $2$ Farrell–Tate cohomology of $\mathrm{PSL}_4(\mathbb Z)$ as a test case (Rahm, 2021, Bui et al., 21 Jul 2025).

1. Foundational setting and localization to $\ell$0-torsion

The basic input is a group action on a contractible or highly connected model for proper actions. In the arithmetic setting, one considers a group $\ell$1 such as $\ell$2 or $\ell$3 acting properly on a contractible CW-complex $\ell$4, for example a symmetric space, Ash’s well-rounded retract, or a Bruhat–Tits building. Such an $\ell$5 is a model for $\ell$6, and the equivariant topology of $\ell$7 encodes the Farrell–Tate cohomology of $\ell$8 (Bui et al., 21 Jul 2025).

For a fixed prime $\ell$9, the central definition is the $\Gamma$0-torsion subcomplex: $\Gamma$1 Rahm’s survey formulates the corresponding localization statement as a corollary of Brown’s result: under the standard hypotheses on $\Gamma$2, there is an isomorphism

$\Gamma$3

Thus, all $\Gamma$4-primary information in Farrell–Tate cohomology is carried by the $\Gamma$5-torsion subcomplex rather than by the full equivariant cell structure (Rahm, 2021).

This localization is especially significant above the virtual cohomological dimension. For virtual duality groups, Farrell–Tate cohomology agrees with ordinary group cohomology in degrees above $\Gamma$6, so torsion subcomplex reduction becomes a mechanism for computing actual group cohomology in the stable range. The guiding principle is that cells outside $\Gamma$7 have $\Gamma$8-torsion-free stabilizers and therefore do not contribute to the $\Gamma$9-primary part in the relevant range (Rahm, 2021).

2. Classical reduction operations

The original reduction formalism is combinatorial. One starts with the $X$0-torsion subcomplex $X$1 and simplifies it by orbit-wise local operations that preserve equivariant $X$2-primary cohomology. The two fundamental operations are merging adjacent cells and cutting off terminal cells (Rahm, 2021).

The merge operation is formulated for a triple $X$3, where $X$4 is an $X$5-cell in $X$6 contained in the boundaries of two adjacent $X$7-cells $X$8. Condition A requires that no higher-dimensional cells of $X$9 are adjacent to $\ell$0, that the interiors of $\ell$1 contain no two points in the same $\ell$2-orbit, and that $\ell$3. Condition B requires that the inclusion

$\ell$4

induces an isomorphism on mod $\ell$5 cohomology. When both conditions hold, the two $\ell$6-cell orbits can be merged along the orbit of $\ell$7 without changing the equivariant $\ell$8-primary Farrell–Tate cohomology (Rahm, 2021).

The cut-off operation applies to a terminal $\ell$9-cell $\ell$0, meaning one adjacent modulo-$\ell$1 $\ell$2-cell $\ell$3, no further cells in the orbit of $\ell$4 in the boundary of $\ell$5, and no higher-dimensional cells adjacent to $\ell$6. If Condition B holds, the pair $\ell$7 can be removed. Recursively applying merges and cut-offs yields a reduced $\ell$8-torsion subcomplex. Rahm’s theorem states that any such reduced complex computes the same $\ell$9-primary Farrell–Tate cohomology as the original action (Rahm, 2021).

A practical strengthening, Condition $\ell$0, replaces the direct cohomological check by finite-group structure conditions involving normal subgroups with trivial mod $\ell$1 cohomology, $\ell$2-normal finite quotients, and normalizers of centers of Sylow $\ell$3-subgroups. The survey states explicitly that Condition $\ell$4 implies Condition B, making the method implementable in families where stabilizer types are classified (Rahm, 2021).

In low-dimensional cases the reduction becomes especially concrete. For Bianchi groups acting on refined cell complexes in hyperbolic $\ell$5-space, stabilizers of pointwise-fixed $\ell$6-cells and $\ell$7-cells are trivial, edge stabilizers are cyclic of order $\ell$8 or $\ell$9, and only vertices carry the larger finite subgroups. Consequently, the $\ell$0-torsion subcomplex for $\ell$1 is a finite graph, and reduction amounts to a graph simplification by edge fusions at vertices with exactly two adjacent $\ell$2-torsion edges (Rahm, 2011).

3. Cohomological mechanisms and homological variants

The reduced torsion subcomplex is not merely a topological simplification; it is designed to be compatible with equivariant spectral sequences. In the Farrell–Tate setting, the reduced complex enters the standard equivariant spectral sequence

$\ell$3

where the $\ell$4-page is assembled from the cohomology of finite stabilizers. Because $\ell$5 is usually very small, the remaining computation is often symbolic rather than cellularly exhaustive (Rahm, 2021).

For Bianchi groups, Rahm showed a particularly strong form of invariance: for $\ell$6, the $\ell$7-primary part of $\ell$8 in degrees $\ell$9 depends only on the homeomorphism type of the reduced $2$0-torsion subcomplex. The argument passes through the equivariant Leray–Serre spectral sequence for the action on the Flöge complex and shows that edge fusions preserve the $2$1-primary part of the $2$2-page in all rows $2$3 (Rahm, 2011).

A major extension replaces ordinary coefficients by the complex representation ring and thereby moves from group homology to Bredon homology. For Bianchi groups, the finite subgroup structure is sufficiently restricted that the Bredon chain complex splits into three orthogonal summands: a trivial/regular part, a $2$4-torsion part, and a $2$5-torsion part. The key device is representation ring splitting: bases of $2$6 are chosen so that every inclusion-induced map becomes block-diagonal with one block of rank $2$7, one $2$8-torsion block, and one $2$9-torsion block. The resulting theorem states that

$\mathrm{PSL}_4(\mathbb Z)$0

with $\mathrm{PSL}_4(\mathbb Z)$1 and $\mathrm{PSL}_4(\mathbb Z)$2 determined by reduced $\mathrm{PSL}_4(\mathbb Z)$3- and $\mathrm{PSL}_4(\mathbb Z)$4-torsion subcomplexes, respectively (Rahm, 2015).

This adaptation is structurally important. In ordinary Farrell–Tate or group homology, the $\mathrm{PSL}_4(\mathbb Z)$5-torsion subcomplex already isolates the $\mathrm{PSL}_4(\mathbb Z)$6-primary contribution. In Bredon homology, the coefficient system mixes different primes through restriction maps on representation rings, so an additional representation-theoretic splitting is required before geometric torsion subcomplex reduction can be applied prime by prime (Rahm, 2015).

4. Discrete Morse theory and the $\mathrm{PSL}_4(\mathbb Z)$7 computation

The 2025 paper on $\mathrm{PSL}_4(\mathbb Z)$8 introduces a new implementation of torsion subcomplex reduction based on discrete Morse theory. The motivation is that, after enforcing rigidity, the raw torsion subcomplex can be extremely large. In the specific case of $\mathrm{PSL}_4(\mathbb Z)$9 at $\ell$00, the raw $\ell$01-torsion subcomplex after rigid subdivision has many thousands of cells. The objective of torsion subcomplex reduction is therefore to replace this large object by a much smaller equivariantly homotopy equivalent one, with cell stabilizers and incidence data sufficient to preserve the relevant equivariant mod $\ell$02 cohomology (Bui et al., 21 Jul 2025).

The paper states that previous reductions were hand-crafted and combinatorial, whereas the new construction uses discrete Morse theory adapted to complexes of groups and mod $\ell$03 cohomology. In this formulation, the reduction procedure is systematic rather than ad hoc, and the authors explicitly state two consequences: a simpler algorithm and runtime improvements. The method is then demonstrated by computing the mod $\ell$04 Farrell–Tate cohomology of $\ell$05 (Bui et al., 21 Jul 2025).

Conceptually, this discrete-Morse reformulation preserves the defining TSR idea—reduce only the part of the action that sees $\ell$06-torsion in stabilizers—but changes the implementation layer. Instead of relying solely on local merges and cut-offs justified one configuration at a time, the reduction is encoded through a Morse-theoretic collapse adapted to equivariant cell data. This suggests a shift from case-specific graph manipulations toward an algorithmic framework better suited to higher-rank arithmetic groups, where rigid subdivisions can produce combinatorial explosions (Bui et al., 21 Jul 2025).

5. Established applications

TSR has been used across several families of groups, with outputs ranging from explicit Poincaré series to operator $\ell$07-theory. Rahm’s survey emphasizes that TSR has yielded general formulas for the cohomology of tetrahedral Coxeter groups and, at odd torsion, of $\ell$08 groups over arbitrary number rings; that these formulas refine the Quillen conjecture; that TSR has been adapted to Bredon homology for Bianchi groups; and that this leads, via Baum–Connes, to the $\ell$09-theory of reduced group $\ell$10-algebras. The same survey also records applications to Chen–Ruan orbifold cohomology and to Ruan’s crepant resolution conjecture for complexified Bianchi orbifolds (Rahm, 2021).

Setting	Reduced object	Result type
Tetrahedral and triangle Coxeter groups	Reduced $\ell$11-torsion subcomplex in the Davis complex	Formulas for mod $\ell$12 homology
Bianchi groups	Reduced $\ell$13- and $\ell$14-torsion graphs	Integral homology torsion and Poincaré series above $\ell$15
Bianchi groups in Bredon homology	Reduced torsion subcomplexes plus representation ring splitting	Equivariant $\ell$16-homology and $\ell$17
$\ell$18 at odd primes	Reduced torsion data organized by normalizers of cyclic subgroups	Farrell–Tate cohomology and refined Quillen statements

For Bianchi groups, the integral-homological version is particularly explicit. Rahm proved that for $\ell$19, the only torsion primes are $\ell$20 and $\ell$21, that the $\ell$22-torsion subcomplex is a finite graph, and that in degrees $\ell$23 the $\ell$24-primary part of $\ell$25 depends only on the homeomorphism type of the reduced $\ell$26-torsion subgraph. This leads to rational generating functions for the multiplicities of $\ell$27 and $\ell$28 summands in higher homology, and it explains why different Bianchi groups can have identical torsion homology when their reduced torsion graphs are homeomorphic (Rahm, 2011).

In the Bredon-theoretic adaptation, the same geometric reduction feeds directly into equivariant $\ell$29-homology. For Bianchi groups, the reduced $\ell$30- and $\ell$31-torsion subcomplexes determine the torsion summands of the Bredon homology, and a Mislin–Valette theorem for $\ell$32 converts this into explicit formulas for $\ell$33 and $\ell$34. Because the Baum–Connes assembly map is an isomorphism for Bianchi groups, these calculations also determine $\ell$35 (Rahm, 2015).

6. Generalizations, geometric context, and open directions

A broader localization principle appears in the theory of complexes of groups. Suppose a residually finite group $\ell$36 acts cellularly, cocompactly, and rigidly on a contractible complex $\ell$37 with strict fundamental domain $\ell$38, and let

$\ell$39

This $\ell$40 is the subcomplex of nontrivial stabilizers. Under the assumption that all nontrivial stabilizers have normal infinite cyclic subgroups, the homology torsion growth theorem states

$\ell$41

Under weaker $\ell$42-$\ell$43-acyclicity assumptions on stabilizers, one also has

$\ell$44

These formulas realize a torsion-subcomplex reduction at the level of normalized homology growth: asymptotic torsion and mod-$\ell$45 growth are determined by the homology of the singular subcomplex $\ell$46 rather than by the full group action (Okun et al., 2021).

This broader perspective is especially effective for right-angled Artin groups and related graph products, where $\ell$47 is the nerve or link data of the defining complex. In that setting, universal $\ell$48-torsion and ordinary $\ell$49-torsion admit formulas involving torsion in $\ell$50 and Euler characteristics of links, and the paper states that right-angled Artin groups satisfy a torsion analogue of the Lück approximation theorem (Okun et al., 2021).

A related geometric theme arises in buildings. Parker and Tent prove that if $\ell$51 is a convex chamber subcomplex of an irreducible spherical building and every vertex of some fixed type $\ell$52 in $\ell$53 has an opposite in $\ell$54, then $\ell$55 is completely reducible. The building-theoretic discussion tied to this result identifies a dichotomy relevant to torsion subcomplex reduction: fixed point complexes of torsion subgroups can behave either as building-like completely reducible subcomplexes or as cone-like objects with a centre. This distinction guides whether one should seek a structural decomposition into apartments and residues or an equivariant collapse onto a central simplex (Parker et al., 2010).

Open directions stated in the survey include extension to higher-rank arithmetic groups such as $\ell$56 and $\ell$57 over number rings beyond $\ell$58, refinement of TSR in the presence of nontrivial center, and broader use of representation ring splitting and related techniques for other classes of groups, including hyperbolic reflection groups (Rahm, 2021). A plausible implication of the 2025 discrete-Morse implementation is that some of these directions become more tractable once reduction is encoded in a systematic rather than hand-crafted algorithmic framework (Bui et al., 21 Jul 2025).