Fibonacci Trace Map Dynamics

• Fibonacci trace map is a polynomial automorphism in R3 that preserves the Fricke–Vogt cubic invariant, organizing the dynamics into invariant surfaces.
• It exhibits two regimes: uniformly hyperbolic bounded orbits for V>0 and mixed conservative dynamics with homoclinic tangencies for V<0.
• The map underpins the spectral analysis of the Fibonacci Hamiltonian, linking its dynamical properties to fractal spectral measures.

The Fibonacci trace map is the polynomial automorphism of $\mathbb R^3$

$T(x,y,z)=(2xy-z,\;x,\;y),$

often written in shift-coordinate form as $$(x_{n+1},x_n,x_{n-1})\mapsto(2x_nx_{n-1}-x_{n-2},x_{n-1},x_n)$$. Its defining feature is preservation of the Fricke–Vogt cubic invariant, which organizes the dynamics into invariant cubic surfaces and makes the map a common object of study in geometry, algebra, analysis, mathematical physics, and number theory. In one regime the restriction to invariant surfaces supports uniformly hyperbolic bounded-orbit dynamics; in another it exhibits conservative mixed behavior with homoclinic tangencies, stochastic seas, and elliptic islands. The same map also underlies the spectral analysis of the Fibonacci Hamiltonian, where bounded trace-map orbits encode spectral membership and thermodynamic quantities of the dynamics become spectral exponents and fractal dimensions (Yessen, 2015, Damanik et al., 2014).

1. Definition, invariant, and invariant surfaces

The basic invariant is written in two equivalent normalizations in the literature. One form is

$I(x,y,z)=x^2+y^2+z^2-2xyz-1,$

with level sets

$S_V=\{(x,y,z)\in\mathbb R^3:I(x,y,z)=V\}.$

Another writes

$x^2+y^2+z^2-2xyz=1+\lambda^2/4,$

or, equivalently, $G(x,y,z)=x^2+y^2+z^2-2xyz-1=V^2/4$, with the associated invariant surface denoted $S_\lambda$ or $S_V$ according to context. In each notation, $T$ preserves the cubic quantity and therefore preserves each corresponding surface (Yessen, 2015, Damanik et al., 2012).

This elementary algebraic fact governs the entire subject. For real $T(x,y,z)=(2xy-z,\;x,\;y),$0, the surface $T(x,y,z)=(2xy-z,\;x,\;y),$1 carries a uniformly hyperbolic limit set $T(x,y,z)=(2xy-z,\;x,\;y),$2 on which $T(x,y,z)=(2xy-z,\;x,\;y),$3 is topologically mixing. By contrast, when $T(x,y,z)=(2xy-z,\;x,\;y),$4, the compact component of $T(x,y,z)=(2xy-z,\;x,\;y),$5, denoted $T(x,y,z)=(2xy-z,\;x,\;y),$6, is a two-sphere, and on that sphere the map exhibits mixed dynamics analogous to the Taylor–Chirikov standard map. The distinction between these regimes is central: the same explicit polynomial map furnishes both a rigid hyperbolic model and a conservative system with highly non-trivial nonhyperbolic phenomena (Yessen, 2015).

A recurrent source of confusion is the parameter $T(x,y,z)=(2xy-z,\;x,\;y),$7. In the Newhouse-phenomena setting, $T(x,y,z)=(2xy-z,\;x,\;y),$8 denotes the value of the Fricke–Vogt invariant and the interval $T(x,y,z)=(2xy-z,\;x,\;y),$9 is the relevant mixed-dynamics regime. In the Fibonacci-Hamiltonian literature, $$(x_{n+1},x_n,x_{n-1})\mapsto(2x_nx_{n-1}-x_{n-2},x_{n-1},x_n)$$0 or $$(x_{n+1},x_n,x_{n-1})\mapsto(2x_nx_{n-1}-x_{n-2},x_{n-1},x_n)$$1 usually denotes the coupling constant, so that the surface relation is $$(x_{n+1},x_n,x_{n-1})\mapsto(2x_nx_{n-1}-x_{n-2},x_{n-1},x_n)$$2. The formulas define the same family of cubic surfaces under different parameterizations.

2. Hyperbolic bounded-orbit dynamics

For the Hamiltonian parameter regime $$(x_{n+1},x_n,x_{n-1})\mapsto(2x_nx_{n-1}-x_{n-2},x_{n-1},x_n)$$3 or $$(x_{n+1},x_n,x_{n-1})\mapsto(2x_nx_{n-1}-x_{n-2},x_{n-1},x_n)$$4, the bounded-orbit set

$$(x_{n+1},x_n,x_{n-1})\mapsto(2x_nx_{n-1}-x_{n-2},x_{n-1},x_n)$$5

or, in alternative notation, $$(x_{n+1},x_n,x_{n-1})\mapsto(2x_nx_{n-1}-x_{n-2},x_{n-1},x_n)$$6, is a compact, invariant, locally maximal, transitive hyperbolic set. The restriction $$(x_{n+1},x_n,x_{n-1})\mapsto(2x_nx_{n-1}-x_{n-2},x_{n-1},x_n)$$7 is topologically conjugate to a finite-type subshift; in one description it is a horseshoe obtained from a Markov partition and symbolic coding by a $$(x_{n+1},x_n,x_{n-1})\mapsto(2x_nx_{n-1}-x_{n-2},x_{n-1},x_n)$$8 matrix of ones, and in another it is topologically conjugate to the full shift on two symbols. The stable and unstable laminations are $$(x_{n+1},x_n,x_{n-1})\mapsto(2x_nx_{n-1}-x_{n-2},x_{n-1},x_n)$$9 or $I(x,y,z)=x^2+y^2+z^2-2xyz-1,$0, depending on the formulation used, and the system admits the standard Axiom A apparatus: local product structure, dense periodic points, and symbolic coding by a topological Markov chain (Damanik et al., 2014, Leclerc, 31 Jul 2025).

The topological entropy on the non-wandering hyperbolic set is

$I(x,y,z)=x^2+y^2+z^2-2xyz-1,$1

where $I(x,y,z)=x^2+y^2+z^2-2xyz-1,$2 is the golden mean. This value reflects the substitution structure encoded by the map and serves as the entropy term in the dimension and transport formulas associated with the Fibonacci Hamiltonian (Damanik et al., 2014).

At weak coupling, the geometry approaches a singular limit. As $I(x,y,z)=x^2+y^2+z^2-2xyz-1,$3, the invariant surfaces collapse to the Cayley cubic $I(x,y,z)=x^2+y^2+z^2-2xyz-1,$4; in the small-coupling analysis the dynamics on $I(x,y,z)=x^2+y^2+z^2-2xyz-1,$5 converges to a system conjugate to the linear toral automorphism

$I(x,y,z)=x^2+y^2+z^2-2xyz-1,$6

and the measure of maximal entropy converges to Lebesgue measure. This limiting picture explains why several dimensional quantities tend to $I(x,y,z)=x^2+y^2+z^2-2xyz-1,$7 in the weak-coupling regime (Damanik et al., 2012).

3. Mixed dynamics for negative invariant values

For $I(x,y,z)=x^2+y^2+z^2-2xyz-1,$8, the restriction

$I(x,y,z)=x^2+y^2+z^2-2xyz-1,$9

to the compact component $S_V=\{(x,y,z)\in\mathbb R^3:I(x,y,z)=V\}.$0 displays the full conservative Newhouse phenomenology. Yessen proves that there exists $S_V=\{(x,y,z)\in\mathbb R^3:I(x,y,z)=V\}.$1 such that for every $S_V=\{(x,y,z)\in\mathbb R^3:I(x,y,z)=V\}.$2, $S_V=\{(x,y,z)\in\mathbb R^3:I(x,y,z)=V\}.$3 admits a locally maximal hyperbolic set $S_V=\{(x,y,z)\in\mathbb R^3:I(x,y,z)=V\}.$4 with four key properties: if $S_V=\{(x,y,z)\in\mathbb R^3:I(x,y,z)=V\}.$5, then $S_V=\{(x,y,z)\in\mathbb R^3:I(x,y,z)=V\}.$6 contains the continuation of $S_V=\{(x,y,z)\in\mathbb R^3:I(x,y,z)=V\}.$7; as $S_V=\{(x,y,z)\in\mathbb R^3:I(x,y,z)=V\}.$8, $S_V=\{(x,y,z)\in\mathbb R^3:I(x,y,z)=V\}.$9 converges in the Hausdorff metric to the Cayley-cubic limit $x^2+y^2+z^2-2xyz=1+\lambda^2/4,$0; $x^2+y^2+z^2-2xyz=1+\lambda^2/4,$1 as $x^2+y^2+z^2-2xyz=1+\lambda^2/4,$2; and $x^2+y^2+z^2-2xyz=1+\lambda^2/4,$3 exhibits quadratic homoclinic tangencies that unfold generically in the parameter $x^2+y^2+z^2-2xyz=1+\lambda^2/4,$4 (Yessen, 2015).

The same theorem yields a residual parameter set $x^2+y^2+z^2-2xyz=1+\lambda^2/4,$5 such that for every $x^2+y^2+z^2-2xyz=1+\lambda^2/4,$6 there exists an increasing family of hyperbolic sets $x^2+y^2+z^2-2xyz=1+\lambda^2/4,$7 with $x^2+y^2+z^2-2xyz=1+\lambda^2/4,$8, a transitive stochastic sea

$x^2+y^2+z^2-2xyz=1+\lambda^2/4,$9

with $G(x,y,z)=x^2+y^2+z^2-2xyz-1=V^2/4$0, and infinitely many elliptic islands accumulating on every point of $G(x,y,z)=x^2+y^2+z^2-2xyz-1=V^2/4$1. In this regime the map combines full-dimensional hyperbolic behavior with KAM-type elliptic stability, which is precisely the mixed conservative behavior associated with the Newhouse paradigm (Yessen, 2015).

The proof proceeds through several geometric mechanisms. On the singular surface $G(x,y,z)=x^2+y^2+z^2-2xyz-1=V^2/4$2, one constructs locally maximal Cantor horseshoes $G(x,y,z)=x^2+y^2+z^2-2xyz-1=V^2/4$3 with thickness $G(x,y,z)=x^2+y^2+z^2-2xyz-1=V^2/4$4, hence Hausdorff dimension tending to $G(x,y,z)=x^2+y^2+z^2-2xyz-1=V^2/4$5. These persist to nearby $G(x,y,z)=x^2+y^2+z^2-2xyz-1=V^2/4$6. Near the conic singularity at $G(x,y,z)=x^2+y^2+z^2-2xyz-1=V^2/4$7, the invariant surfaces are straightened so that local stable and unstable manifolds lie in coordinate planes. A cone-field argument and a $G(x,y,z)=x^2+y^2+z^2-2xyz-1=V^2/4$8-estimate then force quadratic tangencies of stable and unstable laminations for $G(x,y,z)=x^2+y^2+z^2-2xyz-1=V^2/4$9 close to $S_\lambda$0, and generic unfolding follows from a one-parameter transversality argument. To obtain persistent tangencies, one constructs transverse foliations and a smooth curve of tangencies $S_\lambda$1, shows that its intersections with the two laminations are Cantor sets of thickness $S_\lambda$2, and invokes the Newhouse Gap Lemma. Finally, a general result of Gorodetski for area-preserving unfoldings of a quadratic homoclinic tangency yields transitivity of the homoclinic class, full Hausdorff dimension, and infinitely many elliptic islands in the residual parameter set (Yessen, 2015).

4. Trace-map coding of the Fibonacci Hamiltonian

For the Fibonacci Hamiltonian with coupling constant $S_\lambda$3, one works on the invariant surface

$S_\lambda$4

The spectral parameter $S_\lambda$5 is encoded by the line of initial conditions

$S_\lambda$6

Sütő’s theorem identifies spectral membership with bounded forward trace-map orbit: $S_\lambda$7 Equivalently, the spectrum is the pullback of the intersection of $S_\lambda$8 with the stable lamination of the bounded-orbit hyperbolic set (Damanik et al., 2014).

This description implies that the spectrum $S_\lambda$9 is a dynamically defined Cantor set. In the formulation of Damanik–Gorodetski–Yessen, $S_V$0 meets the stable lamination transversally for all $S_V$1. In earlier weak-coupling work, transversality was established for all sufficiently small $S_V$2 and for all $S_V$3, with the intermediate openness problem then still unresolved. The later global transversality statement removes the possibility of tangency breakdown along the spectral line and makes the Cantor geometry uniform across all couplings (Damanik et al., 2014, Damanik et al., 2012).

The density of states measure is realized dynamically. The measure of maximal entropy on the bounded-orbit set is projected along stable leaves onto the line of initial conditions, yielding the density of states measure $S_V$4 or $S_V$5. In particular, the support of the projected measure is the spectrum, so the spectral set and the canonical spectral measure are both encoded by the stable foliation of the trace-map hyperbolic set (Damanik et al., 2012, Leclerc, 31 Jul 2025).

5. Thermodynamic formalism, dimensions, and exponents

On $S_V$6, the trace map admits the full thermodynamic formalism of a mixing hyperbolic system. There is a unique invariant probability measure $S_V$7 of maximal entropy, and for each Hölder potential $S_V$8 there is a unique equilibrium measure. For the unstable Jacobian potential

$S_V$9

the pressure function

$T$0

is real-analytic, strictly convex, and strictly decreasing, with the usual variational characterization. These properties provide the dynamical bridge from symbolic coding to quantitative spectral data (Damanik et al., 2014).

Several exact identities follow. The Hausdorff dimension of the spectrum is the unique zero $T$1 of the pressure: $T$2 The density of states measure is exact-dimensional and satisfies

$T$3

where

$T$4

More generally, equilibrium measures satisfy

$T$5

For the transport and regularity exponents one has

$T$6

and these identities imply the strict inequalities

$T$7

because $T$8 is not cohomologous to a constant on $T$9 (Damanik et al., 2014).

In the weak-coupling regime, the density of states measure has almost-everywhere local scaling exponent

$T(x,y,z)=(2xy-z,\;x,\;y),$00

It is exact-dimensional, satisfies $T(x,y,z)=(2xy-z,\;x,\;y),$01 for sufficiently small $T(x,y,z)=(2xy-z,\;x,\;y),$02, and obeys $T(x,y,z)=(2xy-z,\;x,\;y),$03. In the large-coupling regime, the asymptotics are explicit: $T(x,y,z)=(2xy-z,\;x,\;y),$04

$T(x,y,z)=(2xy-z,\;x,\;y),$05

These formulas make the trace map a complete dictionary between hyperbolic dynamics, pressure, equilibrium states, and spectral fractal data (Damanik et al., 2012, Damanik et al., 2014).

6. Extensions and later developments

The trace-map framework extends beyond the one-dimensional Fibonacci Hamiltonian. For the square Fibonacci Hamiltonian, Yessen proved that for all but countably many $T(x,y,z)=(2xy-z,\;x,\;y),$06,

$T(x,y,z)=(2xy-z,\;x,\;y),$07

where $T(x,y,z)=(2xy-z,\;x,\;y),$08. The argument uses trace-map periodic points, unstable multipliers $T(x,y,z)=(2xy-z,\;x,\;y),$09, and the fact that for two explicit analytic families of periodic points $T(x,y,z)=(2xy-z,\;x,\;y),$10 and $T(x,y,z)=(2xy-z,\;x,\;y),$11, the ratio

$T(x,y,z)=(2xy-z,\;x,\;y),$12

is irrational for all but countably many $T(x,y,z)=(2xy-z,\;x,\;y),$13. Hochman–Shmerkin’s theorem on sums of regular Cantor sets then yields the dimension formula for the spectral sumset (Yessen, 2014).

A more recent direction concerns Fourier decay. For small $T(x,y,z)=(2xy-z,\;x,\;y),$14, the trace-map restriction $T(x,y,z)=(2xy-z,\;x,\;y),$15 on $T(x,y,z)=(2xy-z,\;x,\;y),$16 is treated as an Axiom-A map whose nonwandering set is a Cantor-type hyperbolic set with $T(x,y,z)=(2xy-z,\;x,\;y),$17 stable and unstable line fields. The measure of maximal entropy $T(x,y,z)=(2xy-z,\;x,\;y),$18 and its stable-holonomy projection $T(x,y,z)=(2xy-z,\;x,\;y),$19 both have positive lower Fourier dimension, and in fact satisfy power-law Fourier decay: $T(x,y,z)=(2xy-z,\;x,\;y),$20 The proof combines a Bourgain–Dyatlov sum-product reduction, a temporal distance function for a suspension flow with roof $T(x,y,z)=(2xy-z,\;x,\;y),$21, Tsujii–Zhang-type exponential mixing for the corresponding three-dimensional flow, and a nonlinearity check at a nearby $T(x,y,z)=(2xy-z,\;x,\;y),$22-periodic point that yields the required uniform non-integrability estimate on the temporal distance function (Leclerc, 31 Jul 2025).

Taken together, these developments situate the Fibonacci trace map at the intersection of smooth ergodic theory, conservative dynamics, spectral theory of quasicrystals, and fractal geometry. It serves simultaneously as a hyperbolic coding device, a source of exact dimension formulas, and a concrete polynomial conservative system with fully mixed behavior.