Discrete Lorenz Attractors

• Discrete Lorenz attractors are chaotic attractors in 3D maps characterized by pseudohyperbolicity, a split of contracting and expanding directions, and unique bifurcation pathways.
• They emerge via codimension-3 bifurcations including non-simple homoclinic or heteroclinic tangencies, resonance, and symmetry-induced degeneracies in Hénon-like maps.
• Exhibiting positive maximal Lyapunov exponents and robust invariant splittings, DLAs model complex dynamics in physics, engineering, and secure communication systems.

A discrete Lorenz attractor (DLA) is a robust, strange pseudohyperbolic attractor that arises in three-dimensional maps and diffeomorphisms, exhibiting chaotic dynamics structurally analogous to the classical Lorenz attractor, but realized in discrete time. DLAs emerge in parameter regimes of three-dimensional Hénon-like maps and other polynomial or transcendental diffeomorphisms as a result of codimension-3 bifurcations involving strong-saddle fixed points or periodic orbits, accompanied by mechanisms such as non-simple homoclinic or heteroclinic tangencies, resonance, and symmetry-induced degeneracies. These attractors are characterized by a splitting of the tangent bundle into a strongly contracting direction and a two-dimensional central-unstable bundle exhibiting strict volume expansion, together with a positive maximal Lyapunov exponent for every orbit. DLAs are observed in both orientation-preserving and reversing maps, Z₄-symmetric flows, and high-dimensional parameter families, with parameter windows accumulating near degeneracy.

1. Formal Definition, Normal Forms, and Dynamical Criteria

A map $f:\mathbb{R}^3 \to \mathbb{R}^3$, often of the generalized Hénon type,

$$\bar x = y,\quad \bar y = z,\quad \bar z = Bx + Az + Cy + g(y,z),$$

with $g(0,0) = Dg(0,0) = 0$, admits a DLA if it has an invariant compact set $\mathcal{A}$ containing a unique fixed point $O$ whose multipliers $(\lambda_1, \lambda_2, \lambda_3)$ satisfy strict inequalities (for $B>0$) $$\lambda_1 < -1, \quad 0 < \lambda_2 < 1, \quad -1 < \lambda_3 < 0, \quad \lambda_2 > |\lambda_3|, \quad |\lambda_1 \lambda_2| > 1.$$ This ensures one-dimensional instability (real $\lambda_1 < -1$) and two-dimensional contraction, giving rise to a "butterfly" separatrix configuration analogous to the flow setting. The pseudo-hyperbolicity condition requires an invariant splitting $E^{ss}\oplus E^{cu}$ on $$\bar x = y,\quad \bar y = z,\quad \bar z = Bx + Az + Cy + g(y,z),$$0 with uniform strong contraction on $$\bar x = y,\quad \bar y = z,\quad \bar z = Bx + Az + Cy + g(y,z),$$1 and strict two-dimensional volume expansion on $$\bar x = y,\quad \bar y = z,\quad \bar z = Bx + Az + Cy + g(y,z),$$2, verified via Lyapunov exponents $$\bar x = y,\quad \bar y = z,\quad \bar z = Bx + Az + Cy + g(y,z),$$3 fulfilling $$\bar x = y,\quad \bar y = z,\quad \bar z = Bx + Az + Cy + g(y,z),$$4, $$\bar x = y,\quad \bar y = z,\quad \bar z = Bx + Az + Cy + g(y,z),$$5, and $$\bar x = y,\quad \bar y = z,\quad \bar z = Bx + Az + Cy + g(y,z),$$6 (Gonchenko et al., 2015, Ovsyannikov, 2021).

The canonical discrete normal form for the bifurcation is the 3D Hénon map: $$\bar x = y,\quad \bar y = z,\quad \bar z = Bx + Az + Cy + g(y,z),$$7 with the DLA existing in explicit open regions in $$\bar x = y,\quad \bar y = z,\quad \bar z = Bx + Az + Cy + g(y,z),$$8-space (e.g., $$\bar x = y,\quad \bar y = z,\quad \bar z = Bx + Az + Cy + g(y,z),$$9, $g(0,0) = Dg(0,0) = 0$0–$g(0,0) = Dg(0,0) = 0$1, $g(0,0) = Dg(0,0) = 0$2 small), as established theoretically and numerically (Ovsyannikov, 2021, Gonchenko et al., 2015, Capinski et al., 2017). The map supports global chain transitivity and sensitive dependence on initial data, as in the Lorenz flow case.

2. Bifurcation Mechanisms and Geometric Origins

DLAs are born in homoclinic or heteroclinic bifurcations involving codimension-3 degeneracies. Key scenarios are:

• Homoclinic tangency to a saddle or saddle-focus with Jacobian $g(0,0) = Dg(0,0) = 0$3 (conservative type), with the tangency being non-simple so that no two-dimensional invariant manifold captures the full return dynamics. Local degeneracy includes resonance ($g(0,0) = Dg(0,0) = 0$4), Belyakov transitions (real to complex multipliers), and flip-fold points (Gonchenko et al., 2014, Ovsyannikov, 2023, Ovsyannikov, 2021).
• Heteroclinic cycles, especially between saddles of type $g(0,0) = Dg(0,0) = 0$5, with one connection being a transverse intersection and one a quadratic tangency, again typically non-simple or involving a resonance to avoid dimensional reduction (Ovsyannikov, 2017, Ovsyannikov, 2023).

For both cases, parameter unfoldings involve three parameters $g(0,0) = Dg(0,0) = 0$6: $g(0,0) = Dg(0,0) = 0$7 splits the main tangency, $g(0,0) = Dg(0,0) = 0$8 unfolds the degeneracy (resonance, non-simple geometry), and $g(0,0) = Dg(0,0) = 0$9 controls the Jacobian/product of multipliers. Through a sequence of affine and nonlinear rescalings (Rescaling Lemma), the $\mathcal{A}$0-th return map is shown to be $\mathcal{A}$1-close to the 3D Hénon normal form (Gonchenko et al., 2014, Ovsyannikov, 2021). Each large $\mathcal{A}$2 generates small parameter windows accumulating at the degenerate bifurcation point, resulting in a cascade of domains with DLAs.

Multi-winged and period-doubled discrete Lorenz attractors (including "Simo angels") arise when the underlying periodic orbit has multipliers $\mathcal{A}$3 and the system possesses Z₄ symmetry (Karatetskaia et al., 2024, Karatetskaia et al., 2024).

3. Pseudohyperbolicity, Lyapunov Spectra, and Robustness

The signature of DLAs is pseudohyperbolicity: one direction is contracted strongly and area is expanded on a two-dimensional bundle. This is seen via:

• Invariant splitting $\mathcal{A}$4: the former is the strong-stable direction, and the latter (central-unstable plane) is everywhere transverse, with the angle bounded away from zero.
• Numerical Lyapunov exponents: DLAs exhibit spectra with $\mathcal{A}$5, $\mathcal{A}$6, $\mathcal{A}$7 and the sum $\mathcal{A}$8. For example: (LA) $\mathcal{A}$9, (SA) $O$0 (Karatetskaia et al., 2024, Gonchenko et al., 2015).
• Persistence: The pseudohyperbolic structure is robust under $O$1 perturbations (Gonchenko et al., 2015). This implies the attractor cannot bifurcate to a stable periodic sink under arbitrarily small perturbations, unlike quasi-attractors in Newhouse domains.

Pseudohyperbolicity is verified using cone-field techniques, Lyapunov diagram analysis, and continuity diagrams of the splitting (Gonchenko et al., 2015, Karatetskaia et al., 2024).

4. Phenomenological Scenarios and Routes to Discrete Lorenz Attractors

Discrete Lorenz attractors emerge in multiple bifurcation scenarios:

1. Period-doubling (flip) bifurcation: A stable fixed point becomes a saddle, creating a stable period-2 orbit. Upon further parameter change, either an Andronov–Hopf (Neimark–Sacker) bifurcation or a global homoclinic “butterfly” occurs, leading to a DLA (Gonchenko et al., 2020, Muni, 12 Jun 2025).
2. Successive period-doublings: Especially in non-orientable settings, a sequence of period-doublings and global homoclinic/heteroclinic events yields period-2 or higher-period generalized Lorenz attractors with richer topology (Gonchenko et al., 2020).
3. Global geometric events: Homoclinic/heteroclinic tangencies of invariant manifolds (“butterfly” configuration), mediated by the presence of a one-dimensional unstable manifold and two-dimensional stable manifold (Gonchenko et al., 2014, Gonchenko et al., 2015).

Each scenario is characterized by the creation of a “butterfly” geometry—a twin-lobed structure—observable in projections of the dynamics and guaranteed by the intersection pattern of local (stable/strong-stable) and global (unstable) manifolds (Gonchenko et al., 2015, Gonchenko et al., 2020).

5. Multi-winged, Symmetric, and Complex Discrete Lorenz Attractors

Beyond the classical two-wing discrete Lorenz attractors, recent work rigorously demonstrates the existence of other multi-wing robust pseudohyperbolic attractors:

Attractor Type	Normal Form Context	Parameter Example ($O$2)	Number of Wings
Discrete Lorenz	3D Hénon, Z₄-symmetric	$O$3	1
Two-wing "Simo angel"	Z₄-symmetric unfolding	$O$4	2
Four-wing "Simo angel"	Z₄-symmetric unfolding	$O$5	4

These attractors arise via codimension-3 bifurcations of periodic orbits with multipliers $O$6 and through refined symmetry conditions (Karatetskaia et al., 2024, Karatetskaia et al., 2024). The mathematical structure supporting their robustness is the same: area expansion in a center-unstable bundle and strong-stable contraction; Lyapunov spectra indicate strict pseudohyperbolicity in all cases.

6. Applications, Modeling, and Numerical and Experimental Evidence

DLAs are observed and modeled in diverse settings:

• Periodically-forced flows: The Poincaré map of flows such as the Lorenz or Shimizu–Morioka systems under periodic forcing exhibits DLA behavior in suitable parameter regions, underpinned by analytic and computer-assisted proofs (Capinski et al., 2017, Ovsyannikov, 2021).
• Sinusoidal and transcendental maps: Three-dimensional sinusoidal maps demonstrate DLA formation via period-doubling, Neimark–Sacker bifurcation, and homoclinic butterfly, with precise Lyapunov spectrum measurements and topological characterizations (Muni, 12 Jun 2025).
• Physical and engineering models: Nonholonomic systems (e.g., the Celtic stone or rattleback), laser dynamics with saturable absorber, and thermosolutal convection exhibit parameter regimes with robust DLAs.
• Applications in security and computation: Hyperchaotic DLA regimes have been used to design differential-attack-resistant video encryption schemes, exploiting the high-dimensional, unpredictable dynamics for cryptography (Muni, 12 Jun 2025).

Extensive numerical evidence confirms the phase-space morphology (two-winged, multi-winged, twisted/non-twisted), Lyapunov exponents, fractal dimensions, and robustness under variations (Gonchenko et al., 2015, Muni, 12 Jun 2025, Karatetskaia et al., 2024).

7. Open Problems and Directions

While codimension-3 scenarios leading to DLAs are fully classified, several questions remain:

• Non-orientable pseudohyperbolic Lorenz attractors: The full bifurcation structure and robustness in non-orientable settings and for attractors of higher periodicity remain open.
• Existence of Lorenz attractors with positive multipliers: Most explicit criteria are for negative $O$7; positive cases are under investigation.
• Further rigorous computer-assisted proofs: Extending analytic criteria and validated numerics to a broader class of maps, including transcendental cases as in high-dimensional networks and neural field models.
• Extension to systems with symmetry: Z₄ and other symmetry classes open new families of multi-winged DLAs, relevant for modeling in symmetric physical systems (Karatetskaia et al., 2024, Karatetskaia et al., 2024).

References:

(Gonchenko et al., 2015, Ovsyannikov, 2021, Gonchenko et al., 2014, Gonchenko et al., 2015, Capinski et al., 2017, Ovsyannikov, 2023, Gonchenko et al., 2020, Ovsyannikov, 2017, Muni, 12 Jun 2025, Karatetskaia et al., 2024, Karatetskaia et al., 2024, Ovsyannikov et al., 2015)