Ranks over pro-p groups

Henrique Souza

Universidad Autónoma de Madrid

Advisor: Andrei Jaikin-Zapirain

What is this thesis about?

  • How to measure the size of matrices over completed group algebras
  • Two themes:
    • Canonical rank functions
    • Approximations of rank functions
  • Applications to cohomology of subgroups of Lie groups

Two threads

  • Profinite groups
    • inner rank, Sylvester domains, completed group algebras
  • Semisimple groups
    • rational representations, von Neumann rank, Iwasawa algebras
  • Roadmap: Rank functions \to profinite Sylvester domains \to approximations from representations

Rank over a field

Over a field or division ring

\mathrm{rk}(A) = \dim(\mathrm{im}(A))

  • Over division rings:
    • invariant under row and column operations
    • behaves well in exact sequences
  • Over a general ring:
    • no Gaussian elimination
    • “dimension” for the image is not available

Inner rank

Definition: inner rank in a ring R

\mathrm{irk}_R(A) is the smallest k such that A_{m\times n} = B_{m \times k} \cdot C_{k \times n} over R.

  • Agrees with usual rank over fields and division rings
  • Defined over any ring
  • May fail to satisfy the expected formal properties

Pathologies for inner rank

Depends on the ring

Over \mathbb{Q}[x^2,x^3] the matrix \begin{pmatrix} x^2 & x^3 \\ x^3 & x^4\end{pmatrix} = \begin{pmatrix}x^2 \\ x^3\end{pmatrix}\begin{pmatrix}1 & x\end{pmatrix} has inner rank 2

Maximal rank doesn’t imply invertible

Over \mathbb{Q}[x,y,z]: \begin{pmatrix}0 & z & -y\\ -z & 0 & x\\ y & -x & 0\end{pmatrix}\begin{pmatrix} x \\ y \\ z\end{pmatrix} = 0

Sylvester rank functions

Definition: Sylvester rank functions

Is a function \mathrm{rk}\colon \mathrm{Mat}(R) \to \mathbb{R}_{\geq 0} that satisfies the expected formal properties: \mathrm{rk}(0) = 0, \mathrm{rk}(1) = 1, \mathrm{rk}(AB) \leq \min\{\mathrm{rk}(A),\mathrm{rk}(B)\}, \mathrm{rk}\begin{pmatrix}A & 0 \\ 0 & B\end{pmatrix} = \mathrm{rk}(A) + \mathrm{rk}(B) and \mathrm{rk}\begin{pmatrix} A & C\\ 0 & B\end{pmatrix} \geq \mathrm{rk}(A) + \mathrm{rk}(B).

  • The set \mathbb{P}(R) of Sylvester rank functions is usually immense
  • In general, \mathrm{irk}_R \not\in \mathbb{P}(R)

Sylvester rank functions

Examples

  • Over a field/division ring: \mathbb{P}(\mathcal{D}) = \{\mathrm{rk}_{\mathcal{D}}\}
  • Over matrix rings: \mathbb{P}(\operatorname{Mat}_n(\mathcal{D})) consists of the normalization of \mathrm{rk}_{\mathcal{D}}
  • Pullback along a ring map \varphi\colon R \to S: for \mathrm{rk} \in \mathbb{P}(S) and a matrix A over R, \varphi^\#\mathrm{rk}(A) = \mathrm{rk}(\varphi(A))
  • \mathbb{P}(\mathbb{Z}) contains \mathrm{rk}_{\mathbb{Q}}, \mathrm{rk}_{\mathbb{F}_p}, convex combinations…
  • Operator-algebra examples in \mathbb{C}[\Gamma] from group theory:
    • Pullback of ranks through finite-dimensional representations \Gamma \to \mathsf{GL}_n(\mathbb{C})
    • Pullback of the von Neumann rank through \ell^2(\Gamma)

The main questions

  1. When is \mathrm{irk}_R itself a Sylvester rank function?
  2. Is there a canonical / maximal / universal rank?
  3. Can natural rank functions be approximated by simpler ones?

Sylvester domains

Definition: Sylvester domains

Is a ring R such that \operatorname{irk}_R satisfies the Sylvester rank function axioms.

  • Examples:
    • fields and division rings
    • free algebras
    • group algebras of free groups
  • Non-examples:
    • Rings with zero divisors
    • High-dimensional commutative domains (\mathbb{C}[x,y,z])
    • \mathbb{C}[\pi_1(\text{Klein bottle})]

Why are Sylvester domains so rigid?

Theorem. (P. Malcolmson, 1980)

Any integer valued rank function \mathrm{rk} \in \mathbb{P}(R) is the pullback of the usual rank over a unique ring epimorphism R \to \mathcal{D} onto a division ring \mathcal{D}.

  • Sylvester domains are closely tied to localization into division rings
  • If \mathrm{irk}_R \in \mathbb{P}(R), the associated epic division R-ring \mathcal{D} must be universal (Cohn, 1971)
  • Inner-rank preserving homomorphisms are very rare
  • Moreover, Sylvester domains have weak global dimension at most 2; every flat module is locally free and projective modules must be free (Dicks-Sontag, 1978)

Profinite rings

Definition

A profinite ring is a compact Hausdorff ring. Equivalently, it is an inverse limit of finite discrete rings.

  • Key examples:
    • \mathbb{Z}_p
    • R[\![t]\!] when R is profinite
    • \mathbb{F}_p[\![G]\!] for G profinite

Profinite Sylvester domains

  • There are positive examples:
    • completed group \mathbb{F}_p-algebras of free pro-p groups
  • There are strong constraints:
    • \mathbb{F}_p[\![G]\!] is a domain \implies G is pro-p
    • Profinite Sylvester domains are coherent
    • \mathbb{F}_p[\![G]\!] is a Sylvester domain \implies \operatorname{cd}(G) \leq 2
  • Suggests a genuine classification problem

Jaikin’s criterion

Theorem. (Jaikin, 2020)

Let R be a projective-free ring and \varphi\colon R \to \mathcal{D} be a ring homomorphism to a division ring. Assume that:

  1. \mathrm{Tor}_1^R(\mathcal{D},\mathcal{D}) = 0;
  2. Any finitely generated R-submodule of \mathcal{D} has projective dimension \leq 1.

Then, R is a Sylvester domain and \varphi^\# \mathrm{rk}_{\mathcal{D}} = \mathrm{irk}_R.

  • This theorem and its variations are the main technical tool behind recent established examples of Sylvester domains

Commensurability

Theorem. (Jaikin-S., 2025)

If G is a torsion-free finitely generated pro-p group, then \mathbb{F}_p[\![G]\!] is a Sylvester domain if and only if \mathbb{F}_p[\![U]\!] is a Sylvester domain for some/any open subgroup U of G.

  • Neither direction is obvious
  • This is special to the profinite setting

Why this is non-trivial

  • \implies How \mathcal{D}_{\mathbb{F}_p[\![G]\!]} behaves as a \mathbb{F}_p[\![U]\!]-module?
    • We must show \mathrm{Tor}_1^{\mathbb{F}_p[\![G]\!]}(\mathcal{D}_{\mathbb{F}_p[\![G]\!]} \otimes_{\mathbb{F}_p[\![U]\!]} \mathbb{F}_p[\![G]\!],\mathcal{D}_{\mathbb{F}_p[\![G]\!]}) = 0
  • \impliedby How to build \mathcal{D}_{\mathbb{F}_p[\![G]\!]} from \mathcal{D}_{\mathbb{F}_p[\![U]\!]}?
    • A cadidate for \mathcal{D}_{\mathbb{F}_p[\![G]\!]} is \mathcal{D}_{\mathbb{F}_p[\![U]\!]} * G/U.
    • Given M \leq_{f.g.} \mathcal{D}_{\mathbb{F}_p[\![U]\!]} * G/U and 0 \to I \to \mathbb{F}_p[\![G]\!]^n \to M \to 0\,, I is free iff \mathrm{Tor}_1^{\mathbb{F}_p[\![G]\!]}(I,\mathbb{F}_p) = \mathrm{Tor}_2^{\mathbb{F}_p[\![G]\!]}(M,\mathbb{F}_p)=0.

Free-by-\mathbb{Z}_p groups

  • G = F \rtimes \mathbb{Z}_p with F a free pro-p group.
  • Examples:
    • pro-p completions of orientable surface groups
    • PD^2 groups
    • (virtual) pro-p completion of f.g. free-by-cyclic groups
  • Admits “canonical” presentations
  • Rigid filtered structure: virtually mild in the sense of Labute

Flag presentations and mildness

Fact.

G is f.g. free-by-\mathbb{Z}_p iff G admits a flag presentation: G = \langle x_1,\ldots,x_n,t \mid \mathrm{ad}(t)^{k_i}(x_i) = w_i, i=1,\ldots,n\rangle for some k_i \geq 1 and w_i \in F^p[F,F] where F = \langle x_1,\ldots,x_n\rangle.

  • In particular, G is coherent
  • G is virtually mild (k_i = 1 \implies strongly-free presentation)
  • Associated graded ring of \mathbb{F}_p[\![G]\!] can be read off any strongly-free presentation (Labute, 2006)

Main theorem on the profinite side

Theorem. (Jaikin-S., 2025)

If G is a torsion-free finitely generated virtually free-by-\mathbb{Z}_p pro-p group, then \mathbb{F}_p[\![G]\!] is a Sylvester domain.

  • New large family of profinite Sylvester domains
  • Explicit construction of a division ring embedding realizing the inner rank

Proof idea

  • Write the group algebra as a skew power series ring
  • Compute the associated graded algebra with respect to the radical filtration
  • Construct an embedding into a suitable filtered division ring
  • Apply Jaikin’s criterion to identify the induced rank with the inner rank
  • \mathbb{F}_p[\![G]\!] \simeq \mathbb{F}_p[\![F]\!][\![t-1;\sigma,\delta]\!]
  • For R \to \mathcal{D}, we must check:
    • \mathrm{Tor}_1^R(\mathcal{D},\mathcal{D}) = 0
    • Finitely generated R-submodules of \mathcal{D} have projective dimension \leq 1

The graded picture

Proposition.

Assume G has a mild flag presentation. Under the radical filtration, \mathrm{Gr}(\mathbb{F}_p[\![G]\!]) = \mathbb{F}_p\langle y_1,\ldots,y_n,\ldots\rangle [\overline{t-1};\operatorname{id},\overline{\delta}] where \mathbb{F}_p\langle y_1,\ldots,y_n,\ldots\rangle is the free associative algebra.

  • The filtration lets one lift homological information back to the completed ring
  • Using that free and free pro-p algebras are Sylvester domains, we construct \mathbb{F}_p[\![G]\!] \to \mathcal{D} compatible with the radical filtration
  • We are reduced to proving the necessary homological criteria on the graded counterparts

Approximation on the profinite side

Theorem. (Jaikin-S., 2025)

If G = \varprojlim G_i is a torsion-free finitely generated virtually free-by-\mathbb{Z}_p pro-p group, then \mathrm{irk}_{\mathbb{F}_p[\![G]\!]} = \lim \mathrm{rk}_{G_i} = \mathrm{rk}_G\,.

  • The ranks \mathrm{rk}_{G_i} come from the regular representations G \to \mathsf{GL}(\mathbb{F}_p[G_i]).
  • \mathrm{rk}_G is a canonical rank on \mathbb{F}_p[\![G]\!] that does not depend on the decomposition G = \varprojlim G_i.

Topology and localization

  • The upper bound \mathrm{rk}_G \leq \mathrm{irk}_{\mathbb{F}_p[\![G]\!]} trivially holds

Theorem. (Jaikin-S.,2025)

Let \varphi\colon \mathbb{F}_p[\![G]\!] \to \mathcal{D} be a continuous homomorphism to a division ring \mathcal{D} endowed with a discrete valuation. Then, \varphi^\#\mathrm{rk}_{\mathcal{D}} \leq \mathrm{rk}_G as rank functions on \mathbb{P}(\mathbb{F}_p[\![G]\!]).

  • In our construction of \mathcal{D}, we can extend the valuation of \mathbb{F}_p[\![G]\!] to one on \mathcal{D}

Theorem. (Jaikin-S., 2025)

Let R be a profinite Sylvester domain with associated epic universal division R-ring \mathcal{D}. Then, \mathcal{D} admits a Hausdorff ring topology such that the embedding R \to \mathcal{D} is continuous.

Semisimple groups

  • So far: canonical rank functions on completed group algebras and how they interplay
  • Now: approximation of the von Neumann rank function on complex group algebras by ranks induced from rational representations

The asymptotic cohomology problem

Cohomology growth

Let G be a complex semisimple Lie group, \Gamma \leq G be a finitely generated subgroup and V_\lambda be the irreducible rational representation of G with highest weight \lambda. How does the cohomology groups \mathrm{H}^i(\Gamma,V_\lambda) grow as \lambda \to \infty?

  • Geometry of locally symmetric spaces
  • Arithmetic interpretation for lattices as automorphic forms

Highest weights

Fact.

Let G be a semisimple complex Lie group. The irreducible rational representations of G are classified by their highest weight \lambda \in \mathfrak{h}^*, where \mathfrak{h} is a Cartan subalgebra of \operatorname{Lie}(G).

  • For G = \mathsf{SL}_n(\mathbb{C}), \lambda is an (n-1)-tuple of integers \lambda_1\,,\ldots\,,\lambda_{n-1}. V_{(k,0,\ldots,0)} = \{f \in \mathbb{C}[x_1,\ldots,x_n] \mid f\text{ homogeneous of degree }k\} (fA)(x_1,\ldots,x_n) = f\left(A\begin{pmatrix}x_1 \\ \vdots \\ x_n\end{pmatrix}\right)\,.

Cohomology and rank functions

Proposition.

For \Gamma of type FP_\infty(\mathbb{C}) there are fixed matrices A and B over \mathbb{C} and an integer constant C such that for any complex representation \varphi\colon \Gamma \to \mathsf{GL}(V), \dim_{\mathbb{C}} \mathrm{H}^i(\Gamma, V) = \dim_{\mathbb{C}}(V)(C - \mathrm{rk}_{\varphi}(A) - \mathrm{rk}_{\varphi}(B))\,.

  • Cohomology growth becomes a rank approximation problem.

Where the matrices come from

  • Choose a finitely generated free resolution \cdots \to \mathbb{C}[\Gamma]^{n_{i+1}} \overset{\cdot B}{\to} \mathbb{C}[\Gamma]^{n_i} \overset{\cdot A}{\to} \mathbb{C}[\Gamma]^{n_{i-1}} \to \cdots \to \mathbb{C} \to 0
  • Take \Gamma-homomorphisms to V_\lambda
  • Cohomology is: \frac{\ker V_\lambda^{n_i} \overset{\cdot B^T}{\to} V_\lambda^{n_{i+1}}}{\operatorname{im} V_\lambda^{n_{i-1}} \overset{\cdot A^T}{\to} V_\lambda^{n_i}}
  • Computing dimensions yields precisely the previous formula

Approximation for products of \mathsf{SL}_2(\mathbb{C})

Theorem. (Guerrero-S., submitted)

Let G be a quotient of \prod_{i=1}^k \mathsf{SL}_2(\mathbb{C}) by a central subgroup and \Gamma \leq G be a finitely generated torsion-free subgroup. Then, for the rational representations V_\lambda, \mathrm{rk}_{\varphi_\lambda} \to \mathrm{rk}_\Gamma \text{ as }\min_i \lambda_i \to \infty\,.

Consequence for \ell^2-cohomology

Corollary.

Let G be a quotient of \prod_{i=1}^k \mathsf{SL}_2(\mathbb{C}) by a central subgroup and \Gamma \leq G be a torsion-free subgroup of type FP_\infty(\mathbb{C}). Then: \frac{\dim_{\mathbb{C}} \mathrm{H}^i(\Gamma,V_\lambda)}{\dim_{\mathbb{C}} V_\lambda} \to \beta_i^{(2)}(\Gamma) as \min_j \lambda_j \to \infty.

  • If \beta_i^{(2)}(\Gamma) = 0, it gives explicit bounds on \dim_{\mathbb{C}} \mathrm{H}^i(\Gamma, V_\lambda) as a function of \lambda.

Proof Strategy

  • On the complex group algebra:
    • Prove a “commensurability invariance” property for convergence of ranks
    • Reduce to proving it for some finite index subgroup of any f.g. Zariski-dense \Gamma
    • Embed finitely generated subrings of \mathbb{C}[\Gamma] into Iwasawa algebras \mathbb{Z}_p[\![U]\!]
  • On the Iwasawa algebras \mathbb{Z}_p[\![U]\!]
    • Relate the von Neumann rank of \Gamma with the Ore rank of \mathbb{Z}_p[\![U]\!]
    • The Ore rank is the only rank function satisfying \mathrm{rk}(x) = 1 for any non-zero x \in \mathbb{Z}_p[\![U]\!]

Reduction to the finite index

  • Reduce the statement to the case G = \prod_{i=1}^k\mathsf{SL}_2(\mathbb{C}) and \Gamma is Zariski-dense
    • Lifting matrices from central quotients in a way that doesn’t change their rank
  • Reduce further to proving it over a finite index subgroup of \Gamma.
    • The group von Neumann algebra \mathcal{N}(\Gamma) is a factor, so |\mathbb{P}(\mathcal{N}(\Gamma))| = 1
    • The same holds for the *-regular closures of subrings of the form S[\Gamma]
    • Only remains to prove that \lim \mathrm{rk}_{\varphi_\lambda} exists over \Gamma if it exists over a finite index subgroup.

From \mathbb{C} to \mathbb{Q}_p

Theorem. (Cassels, 1976)

If \Gamma is a f.g. subgroup of G, there are infinitely many primes p such that \Gamma embeds in G(\mathbb{Z}_p) = \prod_{i=1}^k \mathsf{SL}_2(\mathbb{Z}_p)\,.

  • Let U_i = \ker(G(\mathbb{Z}_p) \to \prod_{i=1}^k \mathsf{SL}_2(\mathbb{Z}/p^i\mathbb{Z}))\,.
  • Can actually be done in any Cartan type
  • Embeds S[\Gamma \cap U_1] into the Iwasawa algebra \mathbb{Z}_p[\![U_1]\!]

Ranks on \mathbb{Z}_p[\![U]\!]

Facts.

For odd primes p, the Iwasawa algebra \mathbb{Z}_p[\![U_1]\!] is a Noetherian domain. The pullback of the Ore rank function on \mathbb{Z}_p[\![U_1]\!] to S[\Gamma \cap U_1] coincides with the von Neumann rank.

  • First part is well known (Lazard, 1965)
  • Second part follows from a theorem of M. Harris and the Lück approximation theorem
  • On the Iwasawa algebra, the Ore rank is the only rank satisfying \mathrm{rk}(x) = 1 for any non-zero x.

A p-adic reformulation

Conjecture. (Guerrero-S., submitted)

Let U be a compact uniform split semisimple and simply-connected p-adic Lie group, V_\lambda be an irreducible p-adic Lie representation of U and \mathbb{Z}_p[\![U]\!] be its Iwasawa algebra. For any non-zero x \in \mathbb{Z}_p[\![U]\!] we have \frac{\dim_{\mathbb{Q}_p} xV_\lambda}{ \dim_{\mathbb{Q}_p} V_\lambda} = 1 as \min_i \lambda_i \to \infty.

  • The p-adic analogue of the approximation theorem
  • Imples the corresponding complex approximation for any complex Lie group with the same root system.

Rank one vs higher rank

  • Rank one factors: \prod_{i=1}^k \mathsf{SL}_2
    • W. Fu (2022) proves the conjectural picture in this case
    • The proof uses K. Ardakov’s and S. Wadsley’s embedding of \mathbb{Z}_p[\![U]\!] into central quotients of completed universal enveloping algebras, valid for any root system
  • Higher rank factors:
    • Remains open and is part of ongoing work
    • Validity of this conjecture would imply other conjectural results for lattices, such as upper bounds on the growth of the space of automorphic forms

Completed enveloping algebras

  • \mathbb{Z}_p[\![U]\!] embeds in a completion \widehat{U(\mathfrak{g})} of the universal enveloping algebra of \mathfrak{g} = \operatorname{Lie}(U)
  • The action \widehat{U(\mathfrak{g})} \curvearrowright V_\lambda factors through a central quotient \widehat{U(\mathfrak{g})}/I_\lambda

Theorem. (Ardakov-Wadsley, 2014)

For almost all primes p, the composition \mathbb{Z}_p[\![U]\!] \to \widehat{U(\mathfrak{g})} \to \widehat{U(\mathfrak{g})}/I_\lambda is injective.

  • Canonical form for elements of \widehat{U(\mathfrak{g})}/I_\lambda and explicit lower bound for \dim_{\mathbb{Q}_p} xV_\lambda in terms of this form.

Further directions

  1. Classify profinite Sylvester domains more systematically
    • Behaviour under free constructions
    • Coherent non-examples of Sylvester domains among completed group algebras of dimension 2
  2. Extend semisimple results beyond rank one factors
    • New proofs for \mathsf{SL}_2
    • Structure of completed universal enveloping algebras

Thank you!