---
title: Bowditch's Q-Conditions in Rank-Two Representations
url: https://www.emergentmind.com/topics/bowditch-s-q-conditions
type: topic
---

# Bowditch's Q-Conditions in Rank-Two Representations

Bowditch’s Q-conditions are trace-theoretic conditions on representations of the rank-two free group \(F_2\) that single out a large \(\mathrm{Out}(F_2)\)-invariant domain of discontinuity in the corresponding character variety. In the classical setting, they are imposed on primitive elements of a representation \(\rho:F_2\to \mathrm{SL}(2,\mathbb C)\) or \(\rho:F_2\to \mathrm{PSL}(2,\mathbb C)\), and are encoded by Markoff maps on the Farey graph or Farey tree. Bowditch introduced them in the punctured-torus, type-preserving setting, and Tan–Wong–Zhang generalized them to relative character varieties with arbitrary commutator trace. Independent proofs by Lee–Xu and by Series established that, in rank two, Bowditch’s domain \(BQ\) coincides with Minsky’s primitive stable domain \(PS\), with no discreteness hypothesis and for general \(\kappa=\mathrm{tr}([A,B])\) [1812.04237] [1901.01396].

## 1. Rank-two setting and Markoff formalism

Let \(F_2=\langle a,b\rangle\) be the free group of rank \(2\). An element \(u\in F_2\) is primitive if it belongs to some generating pair \((u,v)\) of \(F_2\). Up to inverse and conjugacy, primitive elements are enumerated by extended conjugacy classes, which correspond bijectively to rational slopes \(p/q\in \mathbb Q\cup\{\infty\}\) via the Farey tree. For a fixed initial triple \((a,b,ab)\), representatives \(w_{p/q}\) are obtained by Farey addition and concatenation.

For a representation \(\rho:F_2\to \mathrm{SL}(2,\mathbb C)\), the character variety \(\chi(F_2)\) is the GIT quotient of \(\mathrm{Hom}(F_2,\mathrm{SL}(2,\mathbb C))\) by conjugation, and for irreducible representations it is identified with \(\mathbb C^3\) through
\[
(\mathrm{tr}(U),\mathrm{tr}(V),\mathrm{tr}(UV)).
\]
In the \(\mathrm{PSL}(2,\mathbb C)\) formulation, traces are defined up to sign by choosing lifts to \(\mathrm{SL}(2,\mathbb C)\), so the invariant quantities are \(|\mathrm{tr}_\rho(g)|\) and trace identities.

The trace geometry is governed by two basic identities:
\[
\mathrm{tr}(AB^{-1})=\mathrm{tr}(A)\mathrm{tr}(B)-\mathrm{tr}(AB),
\]
and
\[
\mathrm{tr}(A)^2+\mathrm{tr}(B)^2+\mathrm{tr}(AB)^2-\mathrm{tr}(A)\mathrm{tr}(B)\mathrm{tr}(AB)=\mathrm{tr}([A,B])+2.
\]
Writing \(x=\mathrm{tr}(A)\), \(y=\mathrm{tr}(B)\), \(z=\mathrm{tr}(AB)\), and \(\mu=\mathrm{tr}([A,B])+2\), this becomes
\[
x^2+y^2+z^2-xyz=\mu.
\]
In the type-preserving punctured torus case, \(\mathrm{tr}([A,B])=-2\), hence \(\mu=0\), and the classical Markoff equation is recovered.

These identities define the Markoff-map framework. In one description, the complementary regions \(\Omega\) of the Farey tree are identified with extended conjugacy classes of primitive elements, and a representation \(\rho\) defines
\[
\phi_\rho:\Omega\to \mathbb C,\qquad \phi_\rho(\mathfrak u)=\mathrm{tr}(U),
\]
where \(\mathfrak u\) corresponds to the primitive class of \(u\). In the dual Farey-graph description, primitive classes are the vertices. At each Farey triangle the three associated values satisfy the Markoff relation, and across an edge they satisfy the basic trace recursion. This is the combinatorial substrate of Bowditch’s Q-conditions [1812.04237].

## 2. Definition of the Q-conditions

Bowditch’s Q-conditions are conditions on primitive elements only. In the \(\mathrm{PSL}(2,\mathbb C)\) formulation used by Lee–Xu, if \(V\) denotes the set of unoriented primitive conjugacy classes of \(F_2\), then \(\rho\) satisfies the Q-conditions when

\[
(Q1)\qquad \forall [\gamma]\in V,\quad \rho(\gamma)\text{ is loxodromic},
\]
equivalently, after choosing lifts, \(\mathrm{tr}(\widetilde{\rho(\gamma)})\notin [-2,2]\), and

\[
(Q2)\qquad \exists K>2\text{ such that }\#\{[\gamma]\in V: |\mathrm{tr}_\rho(\gamma)|\le K\}<\infty.
\]

In the \(\mathrm{SL}(2,\mathbb C)\) formulation used by Series, an irreducible representation satisfies \(BQ\) if

\[
(Q1)\qquad \mathrm{tr}(\rho(g))\notin [-2,2]\quad \text{for every primitive }g\in F_2,
\]
and
\[
(Q2)\qquad \{g\in \mathrm{Prim}(F_2): |\mathrm{tr}(\rho(g))|\le 2\}\text{ is finite.}
\]

In the strict formulation with \(Q1\), this finite set may be empty; \(Q2\) is retained to match generalized settings and to emphasize openness. In the rank-two literature these conditions are routinely expressed in terms of the Markoff map: \(Q1\) means the map avoids \([-2,2]\) on primitive classes, while \(Q2\) is the finiteness of bounded-trace primitive classes. Under \(Q1\), Bowditch and Tan–Wong–Zhang proved that \(Q2\) is equivalent to a uniform linear lower bound
\[
\log |\mathrm{tr}_\rho(\gamma)| \ge m\,\ell_e(\gamma)-c
\]
for constants \(m,c>0\), where \(\ell_e(\gamma)\) is the cyclically reduced length in a fixed basis \(e=(A,B)\); in this rank-two setting \(\ell_e(\gamma)\) coincides with the Christoffel or Fibonacci length [1901.01396] [1812.04237].

## 3. Farey-tree dynamics, sink subtrees, and effective criteria

A central feature of Bowditch’s theory is that trace bounds propagate combinatorially along the Farey tree. Along an edge with adjacent regions \(\mathfrak u,\mathfrak v\) and third regions \(\mathfrak w,\mathfrak z\) at the endpoints, the recursion
\[
\phi_\rho(\mathfrak z)=\phi_\rho(\mathfrak u)\phi_\rho(\mathfrak v)-\phi_\rho(\mathfrak w)
\]
follows from \(\mathrm{tr}(AB^{-1})=\mathrm{tr}(A)\mathrm{tr}(B)-\mathrm{tr}(AB)\). This permits an orientation of Farey edges by comparing the moduli of the two third-region traces.

For \(m\ge 2\), define
\[
\Omega_\rho(m)=\{\mathfrak u\in \Omega: |\phi_\rho(\mathfrak u)|\le m\}.
\]
Then \(\Omega_\rho(m)\) is connected, and it is finite if and only if \(\rho\in BQ\). Infinite rays oriented away from their initial vertex meet regions with \(|\phi_\rho|<2\), infinitely many unless the ray stays on the boundary of a single region. A finite sink subtree \(T_F\) exists with the property that any path of strictly decreasing arrows eventually lands on an edge of \(T_F\); all sink vertices and all edges abutting a sink lie in \(T_F\). Moreover, there is \(M_0\ge 2\) such that any region adjacent to an edge of \(T_F\) has \(|\mathrm{tr}|\le M_0\).

Tan–Wong–Zhang’s generalized Markoff-map theory also isolates an exceptional set along the boundary of a single region,
\[
E=[-2,2]\cup \{\pm \sqrt{\mu}\},
\]
for \(\mu=\mathrm{tr}([A,B])+2\). Values outside \(E\) grow exponentially along the boundary; values in \(E\) obstruct the simplest monotone-growth picture. This dependence on \(\kappa=\mathrm{tr}([A,B])\) is one of the main distinctions between the classical type-preserving case and the relative-character-variety setting.

The practical verification scheme follows the same combinatorics. One fixes \(\kappa=\mathrm{tr}([A,B])\), computes \(\mu=\kappa+2\), builds the Markoff map from initial trace data, propagates it with the two trace identities, orients edges by third-region moduli, identifies the finite sink subtree, verifies that no primitive trace lies in \([-2,2]\), and checks that only finitely many primitive classes have bounded trace. The wake estimates give Fibonacci growth: for a directed edge whose wake is oriented toward it, there exist \(c>0\) and \(n_0\) such that
\[
\log |\phi_\rho(\mathfrak u)|\ge c\,F_{\to e}(\mathfrak u)
\]
for all but at most \(n_0\) regions in the wake. This is the effective combinatorial core of \(BQ\) [1901.01396].

## 4. Equivalence with primitive stability

Minsky’s primitive stability is a geometric condition on the orbit map into \(\mathbb H^3\). Fix generators \((u,v)\) and a basepoint \(O\in \mathbb H^3\). For a cyclically shortest word \(w=e_1\cdots e_n\), the broken geodesic \(\Gamma_\rho(w;(u,v))\) is obtained from the orbit points
\[
\dots,E_{n-1}^{-1}E_n^{-1}O,\ E_n^{-1}O,\ O,\ E_1O,\ E_1E_2O,\dots
\]
with \(E_i=\rho(e_i)\). Primitive stability means that there are constants \(K,\epsilon>0\) such that every broken geodesic arising from a primitive cyclic word is a \((K,\epsilon)\)-quasi-geodesic; equivalently, for the relevant cyclic subwords,
\[
K^{-1}\|w\|-\epsilon \le d(O,\rho(w)O)\le K\|w\|+\epsilon.
\]

The main theorem in rank two is
\[
PS=BQ.
\]
Series proves that Bowditch’s BQ-condition is equivalent to Minsky’s primitive stability for \(\rho:F_2\to \mathrm{SL}(2,\mathbb C)\), and Lee–Xu prove the analogous equality for \(\mathrm{PSL}(2,\mathbb C)\). Both proofs hold without any discreteness hypothesis; in Series’ formulation the equivalence also holds on relative character varieties with fixed \(\kappa=\mathrm{tr}([A,B])\) [1812.04237] [1901.01396].

The implication \(PS\Rightarrow BQ\) is comparatively direct. Uniform quasi-geodesicity forces every primitive image to be loxodromic and yields a uniform linear lower bound on hyperbolic displacement and hence on trace. In Lee–Xu’s formulation,
\[
\log |\mathrm{tr}_\rho(\gamma)| \ge \tfrac12\,\ell_{\mathbb H}(\rho(\gamma))-1 \ge m'\,\ell_e(\gamma)-c'.
\]
This gives \(Q1\) and \(Q2\).

The converse \(BQ\Rightarrow PS\) is the deeper direction. In Series’ proof, \(BQ\) first implies the bounded intersection property, then Tan–Wong–Zhang’s Fibonacci growth on wakes is combined with estimates on bending and segment lengths in broken geodesics to obtain uniform quasi-geodesicity. In Lee–Xu’s proof, one passes through a Coxeter extension and an oriented right-angled hexagon, uses the law of cosines
\[
\frac{2\,\mathrm{tr}(\tilde Z)}{\mathrm{tr}(\tilde X)\mathrm{tr}(\tilde Y)}
=
1+\tanh \eta_X \tanh \eta_Y \cosh \eta_Q,
\]
derives an angle lemma forcing \(\theta(\rho,f)=|\mathrm{Im}\,\eta_Q|\) to be small at high Farey level, and then exploits Christoffel bases so that high-level primitive words are positive in suitable generators. The resulting lower bounds on orbit distances along primitive axes establish primitive stability.

## 5. Bounded intersection property, examples, and failure modes

The bounded intersection property (BIP) was introduced to capture a geometric remnant of the Q-conditions. In Lee–Xu’s formulation, after fixing a basic triple \((a,b,c)\) with \(abc=1\) and the corresponding half-turn data \((P,Q,R)\), one considers the sets
\[
J_P,\quad J_Q,\quad J_R
\]
of intersections between the axes of palindromic primitive elements and the special half-turn axes \(\mathrm{Axis}_P,\mathrm{Axis}_Q,\mathrm{Axis}_R\). A representation has BIP if it is irreducible, every primitive element is loxodromic, and these three sets are bounded subsets of \(\mathbb H^3\). Series formulates the same phenomenon in terms of axes of palindromic primitives intersecting fixed special hyperelliptic axes within uniformly bounded distance of a basepoint.

In rank two,
\[
BQ\subset BI,
\]
and both \(BQ\) and \(PS\) imply BIP. The converse fails in general. A standard counterexample is an \(\mathrm{SO}(3)\)-valued representation for which all images are elliptic with a common fixed point: all axes pass through that point, so BIP holds, but \(Q1\) fails and primitive stability fails. In the discrete faithful setting, however, Lee–Xu prove that BIP forces the Q-conditions.

The classical positive examples are Schottky groups and many quasi-Fuchsian once-punctured torus groups. In these cases primitive words are loxodromic and trace growth along the Farey tree is exponential or at least Fibonacci-linear on the logarithmic scale, so only finitely many primitive classes can have small trace. The standard failure modes lie on the boundary of \(BQ\): either some primitive element becomes elliptic or parabolic, violating \(Q1\), or infinitely many primitive classes retain bounded trace, violating \(Q2\). Geometrically this destroys the uniform quasi-geodesic control required by primitive stability; combinatorially it destroys attraction to a finite sink subtree [1812.04237] [1901.01396].

## 6. Generalizations, dynamics, and open directions

The dynamical significance of Bowditch’s Q-conditions is that \(BQ\) is an open \(\mathrm{Out}(F_2)\)-invariant subset of the character variety on which \(\mathrm{Out}(F_2)\cong \mathrm{GL}(2,\mathbb Z)\) acts properly discontinuously. This domain strictly contains the Schottky or convex-cocompact locus, and it can contain non-discrete representations. Thus the dynamical decomposition of the character variety does not coincide with the discrete-versus-dense dichotomy.

A higher-dimensional generalization replaces trace by translation length. For a proper geodesic Gromov-hyperbolic space \(X\), a representation \(\rho:F_2\to \mathrm{Isom}(X)\) satisfies the \(Q_\lambda\)-conditions if every primitive element is loxodromic and only finitely many primitive conjugacy classes have translation length at most \(\lambda\). Fléchelles proves that for \(\rho:F_2\to \mathrm{Isom}(\mathbb H^d)\), \(d\ge 3\), and under the half-length property (HLP), the following are equivalent:
\[
\text{primitive stable}
\iff
\text{primitive displacing}
\iff
\text{strong }Q\text{-conditions}
\iff
Q_{\lambda_d\circ \mathrm{PrimSys}},
\]
where \(\lambda_d:\mathbb R_{>0}\to \mathbb R_{>0}\) is non-increasing and
\[
\mathrm{PrimSys}(\rho)=\inf\{\mathrm{trans}(\rho(x)):\ x\text{ primitive}\}.
\]
All Coxeter- or \(W_3\)-extensible representations satisfy HLP, so this recovers the classical \(\mathrm{PSL}(2,\mathbb C)\) equivalence and extends it to a substantial class of higher-dimensional representations. The bounded intersection property is also generalized to \(\mathrm{Isom}(\mathbb H^d)\), with primitive stability implying BIP and, in the discrete faithful case, BIP implying the strong Q-conditions [2507.09771].

Two nuances remain central. First, Bowditch’s Q-conditions do not require discreteness; nondiscrete \(BQ\) representations exist when \(\kappa\) varies. Second, in the punctured torus case \(\kappa=-2\), Bowditch conjectured that all \(BQ\) representations are quasi-Fuchsian; this remains open. These facts delimit both the reach and the unresolved boundary of the theory [1901.01396].

Source: https://www.emergentmind.com/topics/bowditch-s-q-conditions