Reversibility in Affine Automorphisms

• Reversibility in affine automorphisms is defined by the existence of a conjugator that transforms an automorphism into its inverse, preserving the affine structure.
• The analysis categorizes cases into semisimple, unipotent, and hyperbolic regimes, with explicit conditions such as generalized Pell equations for hyperbolic maps.
• Strong reversibility is achieved when the reversing element is an involution, providing clear insights into fixed points and the dynamics on structures like tori.

An affine automorphism is a bijective transformation preserving affine structure, typically realized as $x \mapsto A x + v$ for $A$ in $GL(n, D)$ and $v \in D^n$, where $D$ is a division algebra such as $\mathbb{R}$, $\mathbb{C}$, or $\mathbb{H}$. The problem of reversibility of affine automorphisms—i.e., when an element is conjugate to its inverse—has been extensively analyzed, with explicit criteria, structural theorems, and arithmetic and geometric constraints elucidated in the settings of tori, matrix spaces, and general affine groups.

1. Definitions and Basic Notions

Let $G = GL(n, D) \ltimes D^n$ be the affine group, and $g = (A, v) \in G$. The notions of reversibility and strong reversibility are defined as:

• Reversibility: $g$ is reversible if $\exists h \in G$ such that $h g h^{-1} = g^{-1}$.
• Strong Reversibility: $g$ is strongly reversible if $h$ above can be taken as an involution: $h^2 = \mathrm{Id}$.

For affine toral automorphisms, where $D = \mathbb{R}$ and $A \in SL(2, \mathbb{Z})$, the map acts on $\mathbb{T}^2$ via $$f_{A, \mathbf{a}}(\bar{x}) = A \bar{x} + \mathbf{a}\ \mathrm{mod}\ \mathbb{Z}^2$$, and the group is $GL(2, \mathbb{Z}) \ltimes \mathbb{T}^2$ (Banerjee et al., 22 Jan 2026, Gongopadhyay et al., 2022).

2. Algebraic Criteria for Reversibility

The core criterion is that an affine automorphism $(A, v)$ is reversible in $G$ if and only if its linear part $A$ is reversible in $GL(n, D)$ (Gongopadhyay et al., 2022):

• $A$ is reversible if $\exists R \in GL(n, D)$ such that $RAR^{-1} = A^{-1}$.
• Strong reversibility requires that $R^2 = I$ (i.e., $R$ is an involution).

Given $A$, $v$, reversibility further requires the existence of $R$ and $r$ such that:

$(R, r)(A, v)(R, r)^{-1} = (A, v)^{-1}$

which unfolds to the explicit conditions: $$R A R^{-1} = A^{-1}, \qquad (AR + I)v + (A - I) r \equiv 0$$ for toral automorphisms, with congruence in $\mathbb{T}^2$ (Banerjee et al., 22 Jan 2026).

Table: Reversibility Types and Criteria

Type	Linear Criterion	Affine Compatibility
Reversible	$\exists R: RAR^{-1}=A^{-1}$	$(AR+I)v+(A-I)r \equiv 0$
Strongly Reversible	$R^2 = I$ as well	Same as above, with $R$ an involution

If $1$ is not an eigenvalue of $A$, the affine criterion always admits a unique solution for $r$; otherwise, further arithmetic compatibility is needed (Banerjee et al., 22 Jan 2026, Gongopadhyay et al., 2022).

3. Structural Case Analysis

The reversibility problem bifurcates into several algebraic regimes (Banerjee et al., 22 Jan 2026, Bessa et al., 2015, Gongopadhyay et al., 2022):

• Semisimple (No Eigenvalue 1): $A - I$ is invertible, so any $v$ is compatible, provided $A$ is reversible. Every reversible affine map is also strongly reversible when $D=\mathbb{R}$ or $\mathbb{C}$.
• Unipotent ($1 \in \operatorname{spec}(A)$): Nontrivial compatibility—$(AR+I)v \equiv (A-I)r$—requires $(AR+I)v$ to lie in the image of $A-I$, which is at most one-dimensional. Only specific $v$ are compatible for a given $A, R$ (Banerjee et al., 22 Jan 2026).
• Parabolic/Elliptic: Classical results ensure strong reversibility for all affine automorphisms of this type in two dimensions (Banerjee et al., 22 Jan 2026).

For matrix spaces $M_n(K)$ over a general field, all affine preservers of $GL_n(K)$ are in the two-parameter “Frobenius group” (left and right multiplication, possibly with transpose), except in the exceptional case $(n, K) = (2, F_2)$ (Pazzis, 2010).

4. Arithmetic and Algebraic Obstructions

For hyperbolic linear parts in $SL(2, \mathbb{Z})$ (i.e., $|\operatorname{Tr} A| > 2$), reversibility with a linear involution $J$ is equivalent to the existence of integral solutions to a generalized Pell equation (Bessa et al., 2015):

$$x^2 - D y^2 = 4,\quad D=(\operatorname{Tr} A)^2 - 4, \quad y \neq 0$$

along with further divisibility constraints. If such an integral solution exists, one can construct the full family of reducing involutions $J_k$, parametrized by the Pell solutions. If not, $A$ is not reversible by linear involution. For the affine case, a further translation-compatibility constraint $J t = -A^{-1} t$ must be satisfied in $\mathbb{T}^2$ (Bessa et al., 2015).

Classically, the set of reversible $A$ is a “thin” subset of $SL(2,\mathbb{Z})$; for infinitely many values of $D$ there are solutions (hence reversibility), and for infinitely many, there are not. Generically, the centralizer of a hyperbolic ($|\operatorname{Tr} A|>2$) toral automorphism is trivial (Bessa et al., 2015).

5. Fixed Points, Dynamics, and Conjugacy

For $A$ with $\det(A-I)\neq0$, Pick’s theorem gives a geometric criterion for the existence of fixed points: the number of fixed points corresponds to the number of integer points in the parallelogram $(A-I)[0,1)^2$. This is explicitly (Banerjee et al., 22 Jan 2026):

$N = |\det(A-I)| - (g_1 + g_2) + 1$

with $g_1 = \gcd(a-1, c),\, g_2 = \gcd(b, d-1)$. If $N\geq 1$, the affine toral automorphism admits a fixed point.

Dynamically, reversibility implies each periodic orbit is mapped to a time-reversed partner under the action of the reverser. The topological entropy depends only on the eigenvalues of $A$, not the translation part $v$: $$h_{\text{top}}(f_{A, v}) = h_{\text{top}}(f_A) = \sum_{|\lambda_i| > 1} \log |\lambda_i|$$ Conjugacy classes in $GL(2,\mathbb{Q})$-similarity classes are finite if $1 \notin \operatorname{spec}(A)$, but uncountable if $1$ is an eigenvalue, since the translation component can vary along a nontrivial sublattice (Banerjee et al., 22 Jan 2026).

6. Exceptional and Higher-Dimensional Cases

For affine automorphisms of $M_n(K)$, reversibility is governed by linear structure except when $n=2$ and $K = F_2$. In this case, the group of affine invertibility-preservers corresponds to $S_6$ via the symplectic group $Sp_4(F_2)$ acting on quadratic forms; many such maps are genuinely non-linear (Pazzis, 2010).

In all classical settings $D=\mathbb{R}, \mathbb{C}$, and even $D = \mathbb{H}$ (with caveats), the reversibility and strong reversibility of affine automorphisms reduce to that of their linear parts, with at most four involutions in the decomposition for $\det(A)=\pm 1$ (Gongopadhyay et al., 2022). Over $\mathbb{H}$, strongly real and real elements need not coincide, but in the affine group, the classification remains uniform.

7. Summary and Open Directions

The reversibility problem for affine automorphisms admits a complete algebraic and arithmetic classification in $\mathbb{T}^2$ and higher dimensions, controlled by the reversibility of the linear part and specific compatibility with the translation. Arithmetic obstructions (generalized Pell equations) determine reversibility in the hyperbolic case, with unipotent and parabolic cases exhibiting distinct phenomena.

Current directions include extension to higher-dimensional tori, classification with non-involutive reversers, and understanding in settings with non-linear reversing symmetries (Bessa et al., 2015, Gongopadhyay et al., 2022, Banerjee et al., 22 Jan 2026, Pazzis, 2010).