---
title: Outer Automorphism (Out)
url: https://www.emergentmind.com/topics/outer-automorphism-out
type: topic
---

# Outer Automorphism (Out)

to=arxiv.search  สำนักเลขานุการ  天天送彩票json
{"query":"all:\"outer automorphism\" AND (special groups OR right-angled Coxeter OR II_1 factor OR type III_1 factor)", "max_results": 10, "sort_by": "submittedDate", "sort_order": "descending"}
to=arxiv.search 
to=arxiv.search  ingerlanerjson
{"query":"ti:\"On automorphisms and splittings of special groups\" OR ti:\"A type III_1 factor with the smallest outer automorphism group\" OR ti:\"The Geometry of Outer Automorphism Groups of Universal Right-Angled Coxeter Groups\"", "max_results": 10, "sort_by": "relevance", "sort_order": "descending"}
Outer automorphism, usually written $\mathrm{Out}$, is the quotient of an automorphism group by its inner automorphisms. For a group $G$, one writes
\[
\mathrm{Out}(G)=\mathrm{Aut}(G)/\mathrm{Inn}(G),
\]
where $\mathrm{Inn}(G)$ consists of conjugations by elements of $G$. The same quotient construction appears in operator algebras, where for a von Neumann algebra $M$ one sets $\mathrm{Out}(M)=\mathrm{Aut}(M)/\mathrm{Inn}(M)$. Across geometric group theory, low-dimensional topology, Coxeter and Artin theories, model theory, von Neumann algebras, and even renormalization-group analysis, $\mathrm{Out}$ serves as the residual symmetry object after modding out symmetries already realized internally [1112.0227] [2403.20299].

## 1. Definition, variants, and topological structure

For any group $G$, $\mathrm{Out}(G)$ is defined by quotienting $\mathrm{Aut}(G)$ by $\mathrm{Inn}(G)$. When $G$ is centerless and simple, $\mathrm{Aut}(G)\cong G\rtimes \mathrm{Out}(G)$; this is the form used for sporadic simple groups, where the outer automorphism group is always of order at most $2$ [1106.3760]. In free-group and RAAG settings one also studies relative versions. If
\[
F_n=A_1*\cdots *A_k*B,
\]
then
\[
\mathrm{Out}(F_n;A_1,\dots,A_k)=\mathrm{Aut}(F_n;A_1,\dots,A_k)/\mathrm{Inn}(F_n)
\]
consists of outer automorphisms that restrict on each $A_i$ to conjugation by an element of $F_n$ [1010.4753]. For a right-angled Artin group $A_\Gamma$, one similarly has relative groups
\[
\mathrm{Out}(A_\Gamma;\mathcal{G},\mathcal{H}^t),
\]
defined by preserving each subgroup in $\mathcal G$ and acting trivially on each subgroup in $\mathcal H$ [1712.01583].

Special groups in the Haglund–Wise sense carry another refinement. If $G$ is special and endowed with its canonical coarse median structure, one considers $\mathrm{Aut}_{\rm cmp}(G,[\mu])$ and its image
\[
\mathrm{Out}_{\rm cmp}(G)\le \mathrm{Out}(G),
\]
the subgroup of coarse-median-preserving outer automorphisms [2108.13212]. This distinction is substantive: the paper on special groups identifies precise cases where $\mathrm{Out}(G)$ is infinite but $\mathrm{Out}_{\rm cmp}(G)$ is finite, showing that coarse median preservation isolates a geometrically distinguished part of the full outer automorphism group [2108.13212].

Topological structure becomes essential once $G$ itself is a topological group or $M$ a von Neumann algebra. For a full $\mathrm{II}_1$ factor $M$ with separable predual, $\mathrm{Inn}(M)$ is closed in $\mathrm{Aut}(M)$, and $\mathrm{Out}(M)$ is a Polish group with the quotient topology [2403.20299]. For non-Archimedean Polish groups $G$, $\mathrm{Aut}(G)$ carries a unique Polish topology making its natural action on $G$ continuous under an invariant-countable-basis hypothesis, and $\mathrm{Inn}(G)$ is Polishable; when $\mathrm{Inn}(G)$ is closed, $\mathrm{Out}(G)$ inherits the quotient topology [2512.12589].

## 2. Free groups, relative outer automorphisms, and Outer space

For free groups, $\mathrm{Out}(F_n)$ is the natural object of study because inner automorphisms act trivially on conjugacy classes and on homotopy classes of markings on graphs [1112.0227]. Culler–Vogtmann Outer space $\mathrm{CV}_n$ is a contractible space of marked metric graphs of rank $n$ on which $\mathrm{Out}(F_n)$ acts properly with finite stabilizers [1112.0227]. Relative theories replace graphs by marked graphs carrying prescribed wedge cycles encoding a free factor system. The resulting relative outer space $\mathrm{CV}_n(\mathcal A)$ is contractible and admits an action of $\mathrm{Out}(F_n;\mathcal A)$ [1010.4753] [1112.0227].

The relative spaces come with explicit dimension formulae. If $\mathcal A=\{A_1,\dots,A_k\}$ with $S=\sum_i \mathrm{rank}(A_i)$ and $n-S\ge 1$, then
\[
\dim\big(\mathrm{CV}_n(\mathcal A)\big)=3n+2k-4-3S,
\]
while
\[
\dim\big(\partial\,\mathrm{CV}_n(\mathcal A)\big)=3n+2k-5-3S
=\dim\big(\mathrm{CV}_n(\mathcal A)\big)-1
\]
[1112.0227]. The corresponding relative outer automorphism group has virtual cohomological dimension
\[
\mathrm{vcd}\big(\mathrm{Out}(F_n;A_1,\dots,A_k)\big)
=
2n-2\,s(1)-\cdots-2\,s(k)+2k-2-m,
\]
where $s(i)=\mathrm{rank}(A_i)$ and $m$ is the number of rank-one factors [1010.4753].

A separate algorithmic direction concerns geometricity. An outer automorphism of a free group is geometric if it can be represented by a homeomorphism of a compact surface. Bestvina–Handel solved the irreducible case, and the general case was later completed: there is an algorithm that decides whether a general $\phi\in\mathrm{Out}(F_n)$ is geometric and, if so, constructively produces a realizing surface homeomorphism [2310.04402]. The proof combines CT technology, Guirardel cores of tree actions, and Nielsen–Thurston theory [2310.04402]. This makes the boundary between purely free-group dynamics and mapping-class-type dynamics algorithmically decidable.

## 3. Right-angled Artin groups, Coxeter groups, and special groups

For a finite simplicial graph $\Gamma$, the RAAG $A_\Gamma$ interpolates between free and free abelian groups, and so does $\mathrm{Out}(A_\Gamma)$ [1712.01583]. A fundamental structural result gives a finite subnormal series for $\mathrm{Out}(A_\Gamma)$ whose successive quotients are finite, free-abelian, $\mathrm{GL}(m,\mathbb Z)$, or Fouxe–Rabinovitch groups [1712.01583]. Since the last two act on a symmetric space or a deformation space of trees, respectively, this decomposition supplies a geometric analysis of each piece; in particular, $\mathrm{Out}(A_\Gamma)$ is type $\mathrm{VF}$, and principal congruence subgroups of level $l\ge 3$ are type $\mathrm{F}$ [1712.01583].

For special groups, the central theorem is a splitting criterion for infiniteness. If $G$ is special, then $\mathrm{Out}(G)$ is infinite if and only if $G$ splits over a co-abelian subgroup of a centraliser and admits an infinite-order generalised Dehn twist; similarly, $\mathrm{Out}_{\rm cmp}(G)$ is infinite if and only if $G$ splits over an actual centraliser and admits an infinite-order coarse-median-preserving generalised Dehn twist [2108.13212]. The proof uses non-small, stable $G$-actions on $\mathbb R$-trees whose arc-stabilisers are centralisers or kernels of homomorphisms from centralisers to abelian groups [2108.13212]. This replaces the classical hyperbolic picture of cyclic splittings by a centraliser-based theory adapted to special cube complexes.

For right-angled Coxeter groups, several distinct geometric regimes occur. In the universal case
\[
W_n=\mathbb Z_2*\cdots *\mathbb Z_2,
\]
the McCullough–Miller space $K_n$ is a contractible simplicial model for $\mathrm{Out}(W_n)$, but for $n\ge 4$ it cannot carry an $\mathrm{Out}(W_n)$-equivariant CAT$(0)$ or CAT$(-1)$ piecewise Euclidean or hyperbolic metric [1910.13572]. By contrast, $\mathrm{Out}(W_n)$ is acylindrically hyperbolic for $n\ge 3$, via its relation to $\mathrm{Out}(F_{n-1})$ and fully irreducible elements [1610.08005]. For general RACGs, the outer automorphism group is governed by SIL-type combinatorics: $\mathrm{Out}(W_\Gamma)$ is large if and only if $\Gamma$ contains a STIL or an FSIL, and otherwise it is virtually abelian [1706.07873]. A plausible implication is that the combinatorics of links and separating intersections play, for RACGs, a role analogous to splittings and laminations for free groups and RAAGs.

## 4. Higher outer automorphism groups and algebraic finiteness phenomena

The passage from $\mathrm{Out}(G)$ to $\mathrm{Out}(\mathrm{Aut}(G))$ and $\mathrm{Out}(\mathrm{Out}(G))$ reveals sharp differences between classical groups and RAAGs. For free groups and free abelian groups, rigidity is strong: $\mathrm{Out}(\mathrm{Aut}(F_n))=1$, while $\mathrm{Out}(\mathrm{GL}(n,\mathbb Z))$ has order at most $4$ [1306.6549]. Across all RAAGs, however, there is no uniform upper bound on $\lvert \mathrm{Out}(\mathrm{Aut}(A_\Gamma))\rvert$, and the same unboundedness holds for finite subgroups of $\mathrm{Out}(\mathrm{Out}(A_\Gamma))$ [1306.6549]. Austere graphs yield
\[
\mathrm{Out}(\mathrm{Out}(A_\Gamma))\cong \mathrm{GL}(n,2),
\]
and centerless or centered constructions give explicit lower bounds such as $2^{t-1}$ or $2^{d-1}$ for higher outer automorphism groups [1306.6549].

A complementary construction uses focused graphs. For each $n\ge 2$, there are infinitely many RAAGs $A_\Gamma$ such that
\[
\mathrm{Out}(\mathrm{Out}(A_\Gamma))
\text{ contains } \mathrm{PGL}_n(\mathbb Z),
\]
and, more generally, any $\mathbb Z$-linear or projective $\mathbb Z$-linear group acts faithfully on $\mathrm{Out}(A_\Gamma)$ through automorphisms or outer automorphisms for suitable $\Gamma$ [1507.04359]. This demonstrates a marked departure from the free and free abelian cases, where the second outer automorphism group is trivial or of order at most $4$ [1507.04359].

Outer automorphism groups of hyperbolic and relatively hyperbolic groups exhibit strong residual properties. If $G$ is one-ended and hyperbolic relative to virtually polycyclic groups, then $\mathrm{Out}(G)$ is residually finite; if $G$ is one-ended and toral relatively hyperbolic, then for every prime $p$, $\mathrm{Out}(G)$ is virtually residually $p$-finite [1305.5403]. The same paper proves that finitely generated groups with infinitely many ends have virtually residually $p$-finite outer automorphism groups under a virtually residually $p$-finite hypothesis [1305.5403].

Dimension theory yields a different sort of pathology. For every integer $r\ge 0$, there exists a dimension-rigid CAT$(0)$ Gromov-hyperbolic group $G$ such that $\mathrm{Out}(G)$ is virtually torsion-free, admits a cocompact model for $\underline EG$, and nevertheless satisfies
\[
\mathrm{vcd}(\mathrm{Out}(G))\le \underline{\mathrm{gd}}(\mathrm{Out}(G))-r
\]
[1602.04354]. Thus outer automorphism groups can exhibit arbitrarily large gaps between virtual cohomological dimension and proper geometric dimension even when the base group itself is dimension rigid [1602.04354].

A different algebraic finiteness result comes from $L^2$- and group-ring methods. For $G$ a compact surface group, a finitely generated free group, or a finitely generated RAAG, there exists a torsion-free finite-index subgroup
\[
H\le \mathrm{Out}(G)
\]
satisfying the Strong Atiyah Conjecture, and for every field $\mathbb K$ the group algebra $\mathbb K[H]$ embeds into a division ring [2606.19606]. Consequently, the von Neumann rank function of $\mathrm{Out}(G)$ takes values in a discrete subgroup of $\mathbb Q$ [2606.19606].

## 5. Operator-algebraic outer automorphism groups

In von Neumann algebra theory, one defines
\[
\mathrm{Out}(M)=\mathrm{Aut}(M)/\mathrm{Inn}(M),
\]
with $\mathrm{Inn}(M)=\{\mathrm{Ad}\,u\mid u\in U(M)\}$ [2403.20299]. For full $\mathrm{II}_1$ factors with separable predual, $\mathrm{Inn}(M)$ is closed, so $\mathrm{Out}(M)$ is a Polish group [2403.20299]. A major realization theorem now shows that every locally compact second countable group $G$ occurs as the outer automorphism group of such a factor:
\[
\mathrm{Out}(M)\cong G
\]
for some full $\mathrm{II}_1$ factor $M$ with separable predual [2403.20299]. The proof factors through the notion of a centralizer group in ergodic theory and uses Poisson suspensions and the Maharam extension to realize arbitrary locally compact second countable groups as centralizers, then as outer automorphism groups [2403.20299].

This realization result sharply contrasts with rigidity phenomena in type $\mathrm{III}_1$. For a type $\mathrm{III}_1$ factor $M$, modular theory gives a canonical embedding
\[
\delta:\mathbb R\to \mathrm{Out}(M),\qquad t\mapsto [\sigma_t^\varphi],
\]
independent of the chosen faithful normal state $\varphi$ [2312.04702]. An explicit construction produces a full type $\mathrm{III}_1$ factor with separable predual such that this embedding is an isomorphism:
\[
\mathrm{Out}(M)\cong \mathbb R
\]
[2312.04702]. In that setting, the outer automorphism group is as small as possible subject to the unavoidable modular copy of $\mathbb R$ [2312.04702].

Taken together, these results show that the operator-algebraic notion of outer automorphism admits both maximal flexibility and extreme rigidity. In the $\mathrm{II}_1$ case, every locally compact second countable group appears. In the type $\mathrm{III}_1$ case, one can force the whole outer automorphism group to consist only of modular automorphisms [2403.20299] [2312.04702].

## 6. Model-theoretic, finite-simple, and physical appearances

For oligomorphic groups, outer automorphism groups acquire a distinctly model-theoretic description. If $G$ is oligomorphic and $T$ is the complete theory of a countable structure $M$ with $G=\mathrm{Aut}(M)$, then
\[
\mathrm{Out}(G)\cong B(T),
\]
where $B(T)$ is the group of invertible self-interpretations of $T$ modulo syntactic homotopy [2512.12589]. The group $B(T)$ is totally disconnected and locally compact, so $\mathrm{Out}(G)$ is t.d.l.c.; if the signature of $T$ is finite, then $B(T)$, and hence $\mathrm{Out}(G)$, is discrete [2512.12589]. This gives a topological and model-theoretic interpretation of outer automorphisms unavailable in the purely discrete setting.

At the opposite end of the finite–infinite spectrum, the sporadic finite simple groups exhibit extreme smallness. For the $26$ sporadic simple groups, $\mathrm{Out}(G)$ has order $2$ in exactly $12$ cases and is trivial in the remaining $14$ [1106.3760]. Because these groups are centerless and simple, this completely determines $\mathrm{Aut}(G)$ as either $G$ or $G\rtimes C_2$ [1106.3760]. The same note shows that for a Sylow $2$-subgroup $T$ of a sporadic simple group, $C_{\mathrm{Out}(G)}(T)=1$; equivalently, $Z(T)$ is a Sylow $2$-subgroup of $C_{\mathrm{Aut}(G)}(T)$ [1106.3760].

A formally different, but structurally related, use of outer automorphisms appears in renormalization-group theory. For a symmetry group $G$ of a quantum field theory, an outer automorphism induces a transformation $F$ on the coupling space, and the beta functions satisfy the covariance relation
\[
\beta_{\lambda_n}(F_k(\lambda))
=
\sum_m \frac{\partial F_n(\lambda)}{\partial \lambda_m}\,\beta_{\lambda_m}(\lambda_k).
\]
The fixed locus
\[
H=\{\lambda\mid F(\lambda)=\lambda\}
\]
is then RG invariant, so the existence of an outer automorphism is a sufficient condition for the existence of an RG fixed hyperplane [2603.12318]. In the paper’s terms, the symmetry of the fully coupled system of beta functions can be larger than the symmetry of the action, and “goofy transformations” are essential when the outer automorphism acts nontrivially on kinetic terms [2603.12318].

These examples suggest that $\mathrm{Out}$ is not merely a quotient construction but a recurrent organizing principle. In free-group and RAAG contexts it controls deformation spaces, splittings, and rigidity. In operator algebras it ranges from $\mathbb R$ to arbitrary locally compact second countable groups. In oligomorphic and QFT settings it governs, respectively, self-interpretations of theories and nonperturbative RG constraints [2512.12589] [2603.12318].

Source: https://www.emergentmind.com/topics/outer-automorphism-out