Sequential Chaotic Oscillations (SCOs)

• Sequential Chaotic Oscillations (SCOs) are graph-ordered chaotic dynamics in E-I TLNs, characterized by sequential state visitation with irregular oscillatory activity.
• The mechanism uses threshold-linear network models where strong inhibition and unstable singleton fixed points lead to metastable chaotic itinerancy governed by graph structure.
• This framework unifies diverse systems, linking structured chaotic pulse trains, switching attractors, and parameter-driven routes to chaos for practical insights into nonperiodic dynamics.

Searching arXiv for the core SCO paper and closely related dynamical-systems analogues. Sequential Chaotic Oscillations (SCOs) are a class of graph-organized chaotic dynamics proposed in excitatory-inhibitory threshold-linear networks (E-I TLNs) as a candidate mechanism for sequential metastability under constant input (Zang et al., 29 May 2026). In this setting, trajectories spend extended times near a succession of metastable states, switch among them in an order predicted by the underlying directed graph, and exhibit irregular or chaotic dwell times together with oscillatory excitatory-inhibitory activity (Zang et al., 29 May 2026). More broadly, the term also admits a wider dynamical-systems interpretation: temporally ordered or parameter-ordered oscillatory regimes in which chaos organizes repeated pulse events, switching patterns, or oscillatory episodes, as in bounded-phase opto-radiofrequency pulse trains (Romanelli et al., 2015), chaotic switching in symmetric Kerr cavities (Bitha et al., 2024), graph-ordered chaotic itinerancy in E-I TLNs (Zang et al., 29 May 2026), and sequentially structured routes to chaos in nano-oscillators (Wolba et al., 2021), magnonic combs (Sun et al., 29 May 2025), and related systems. By contrast, the acronym “SCO” in “spatially confined oscillations” denotes a different concept in wave mechanics and is unrelated to sequential chaotic dynamics (Stadtler et al., 2021).

1. Definition and conceptual scope

In E-I TLNs, SCOs are introduced as a simple form of chaotic itinerancy whose transition order is predicted by the graph underlying the network (Zang et al., 29 May 2026). The defining features are threefold. First, the dynamics are sequential: on an $n$-path, the observed order is $1\to 2\to \cdots \to n$, whereas on an $n$-cycle the order is $1\to 2\to \cdots \to n\to 1$ (Zang et al., 29 May 2026). Second, they are chaotic: the observed attractors are numerically chaotic, with irregular dwell times and sensitivity to initial conditions (Zang et al., 29 May 2026). Third, they are oscillatory: the local dynamics near each metastable state are built from excitatory-inhibitory oscillations rather than static switching alone (Zang et al., 29 May 2026).

The paper that explicitly names SCOs frames them as a candidate dynamical mechanism for sequential metastability, with trajectories lingering near attractor ruins associated with unstable singleton fixed points and nearby chaotic attractors before moving on to the next graph-predicted state (Zang et al., 29 May 2026). This places SCOs within the broader family of chaotic itinerancy, but in a particularly structured form: graph architecture determines the order of visitation, while local instability and nonlinear oscillations determine the residence-time variability (Zang et al., 29 May 2026).

A broader usage is also supported by related literature. In an opto-radiofrequency oscillator, a chaotic attractor can generate a near-periodic train of bounded-phase pulses with excitable-like triggering and refractory times, yielding a temporally ordered but nonperiodic chaotic pulse sequence (Romanelli et al., 2015). In a symmetric Kerr cavity, trajectories can switch chaotically between two symmetry-related localization states, with the switching order encoded by kneading sequences and organized by global bifurcations (Bitha et al., 2024). These are not SCOs by name, but they exhibit the same core motif: recurrent, ordered oscillatory or switching episodes governed by deterministic chaos.

The same caution applies to acronym overlap. “The dynamics of spatially confined oscillations” uses “SCO” to mean a localized superposition of plane waves with emergent geodesic-like motion in inhomogeneous media, not sequential or chaotic oscillations (Stadtler et al., 2021). Any encyclopedia treatment must therefore distinguish the nonlinear-dynamical concept from this unrelated wave-mechanical usage.

2. Mathematical realization in excitatory-inhibitory threshold-linear networks

The explicit SCO framework is formulated for E-I TLNs with $n$ excitatory units $x_1,\dots,x_n$ and one inhibitory unit $x_I$ (Zang et al., 29 May 2026). The governing equations are

$$\tau_E \frac{dx_i}{dt} = -x_i + \left[\sum_{j=1}^n W_{ij}x_j + W_{iI}x_I + b_i\right]_+, \qquad i=1,\dots,n,$$

$$\tau_I \frac{dx_I}{dt} = -x_I + \left[\sum_{j=1}^n W_{Ij}x_j + W_{II}x_I + b_I\right]_+,$$

with threshold nonlinearity $[y]_+=\max\{y,0\}$ and $1\to 2\to \cdots \to n$0 by convention (Zang et al., 29 May 2026). The network obeys Dale’s law: $1\to 2\to \cdots \to n$1 (Zang et al., 29 May 2026).

For graph-based families, excitatory coupling is specified by a directed graph $1\to 2\to \cdots \to n$2, with

$1\to 2\to \cdots \to n$3

(Zang et al., 29 May 2026). Thus an E-I TLN is parameterized by $1\to 2\to \cdots \to n$4, where $1\to 2\to \cdots \to n$5 and $1\to 2\to \cdots \to n$6 (Zang et al., 29 May 2026).

The paper distinguishes support and excitatory support via

$1\to 2\to \cdots \to n$7

with $1\to 2\to \cdots \to n$8 because the inhibitory node is active at every fixed point (Zang et al., 29 May 2026). Piecewise-linearity is organized by chambers $1\to 2\to \cdots \to n$9, determined by which excitatory threshold variables are on or off; within each chamber the dynamics are linear (Zang et al., 29 May 2026). This chamber structure is central because SCO trajectories move through multiple chambers while preserving a graph-constrained visitation order.

The simplest local oscillatory building block already appears in the singleton E-I TLN. With one excitatory variable, the fixed point

$n$0

has Jacobian

$n$1

which is stable for

$n$2

and unstable for

$n$3

(Zang et al., 29 May 2026). When unstable, boundedness and planar geometry yield a stable limit cycle, providing the elementary E-I oscillation from which more complex attractors are assembled (Zang et al., 29 May 2026).

3. Graph rules, fixed-point structure, and conditions for SCOs

The emergence of SCOs in E-I TLNs is tied to two conditions emphasized throughout the paper: unstable singleton fixed points and sufficiently strong inhibition (Zang et al., 29 May 2026). The strong-inhibition regime is

$n$4

while singleton instability requires

$n$5

(Zang et al., 29 May 2026). Together, these conditions create a fixed-point architecture rich enough to support graph-ordered chaotic itinerancy.

The fixed-point support theorems for paths and cycles are especially important. For the $n$6-path $n$7, one has

$n$8

$n$9

$1\to 2\to \cdots \to n\to 1$0

(Zang et al., 29 May 2026). For the $1\to 2\to \cdots \to n\to 1$1-cycle $1\to 2\to \cdots \to n\to 1$2, $1\to 2\to \cdots \to n\to 1$3,

$1\to 2\to \cdots \to n\to 1$4

$1\to 2\to \cdots \to n\to 1$5

$1\to 2\to \cdots \to n\to 1$6

(Zang et al., 29 May 2026).

These theorems show why strong inhibition is decisive. In that regime, every nonempty subset is a fixed-point support on paths and cycles, so all singleton supports are present. Once those singleton fixed points are unstable, they can organize local oscillatory or chaotic attractors that later turn into attractor ruins, producing SCOs (Zang et al., 29 May 2026).

The graph-theoretic analysis is refined by domination and uniform in-degree rules. If $1\to 2\to \cdots \to n\to 1$7 has uniform in-degree $1\to 2\to \cdots \to n\to 1$8, the on-neuron condition is

$1\to 2\to \cdots \to n\to 1$9

and the corresponding fixed point is

$n$0

(Zang et al., 29 May 2026). For proper subgraphs, the off-neuron condition becomes

$n$1

where $n$2 counts incoming edges from $n$3 to $n$4 (Zang et al., 29 May 2026). These rules are used to classify all fixed-point supports on paths and cycles, thereby predicting which metastable structures can participate in SCOs.

On paths, SCOs are transient in the sense that the first $n$5 singleton-associated chaotic attractors become attractor ruins and the trajectory eventually settles near the terminal node’s attractor (Zang et al., 29 May 2026). On cycles, all singleton-associated attractors can become attractor ruins and the switching continues indefinitely in the cyclic graph order (Zang et al., 29 May 2026). This distinction between terminating and sustained sequential chaos is one of the defining structural properties of the theory.

4. Oscillatory modes, chaotic itinerancy, and mode decomposition

A notable feature of the E-I TLN framework is that oscillations need not be synchronized across excitatory units (Zang et al., 29 May 2026). To characterize this, the paper introduces a decomposition into the $n$6-mode and the mean mode. For $n$7 excitatory nodes,

$n$8

$n$9

(Zang et al., 29 May 2026). The $x_1,\dots,x_n$0-mode captures excitatory differences and therefore encodes sequential or asymmetric structure, whereas the mean mode captures overall E-I population activity (Zang et al., 29 May 2026).

On cycles, in the full-support chamber $x_1,\dots,x_n$1, the $x_1,\dots,x_n$2-mode is governed by

$x_1,\dots,x_n$3

$x_1,\dots,x_n$4

or $x_1,\dots,x_n$5, with eigenvalues

$x_1,\dots,x_n$6

(Zang et al., 29 May 2026). Hence $x_1,\dots,x_n$7 is stable iff

$x_1,\dots,x_n$8

(Zang et al., 29 May 2026). The mean mode obeys

$x_1,\dots,x_n$9

$x_I$0

with stability condition

$x_I$1

(Zang et al., 29 May 2026).

This yields a four-way classification of full-support cycle attractors. If both modes are stable, the full-support fixed point is stable. If the mean mode is unstable but the $x_I$2-mode is stable, synchronized E-I oscillations occur. If the mean mode is stable but the $x_I$3-mode is unstable, CTLN-like oscillations arise. If both are unstable, more complex regimes appear, including synchronized E-I oscillations or flower-like quasi-periodic attractors depending on chamber switching (Zang et al., 29 May 2026). This decomposition clarifies that SCOs are not merely a byproduct of global E-I periodicity; they rely on unstable excitatory-difference structure as well as oscillatory mean-mode activity.

The broader literature supports similar distinctions between observable regularity and underlying geometrical complexity. In the opto-radiofrequency oscillator of (Romanelli et al., 2015), pulses that look nearly identical in intensity can traverse well-separated regions of the chaotic attractor, showing that temporal order at the observable level need not imply a unique state-space route. In the Kerr cavity of (Bitha et al., 2024), symmetric and asymmetric switching patterns correspond to different itineraries around symmetry-related states $x_I$4 and $x_I$5, and these itineraries are organized by symbolic dynamics rather than by a single periodic orbit. This suggests that SCOs should be understood as structured but nonrigid recurrent chaos: what repeats is a family of related excursions, not a single exact loop.

5. Related mechanisms in optics, magnetism, Josephson systems, and slow-fast chaos

Several arXiv papers provide closely related mechanisms even though they do not use the SCO label.

In optics, an opto-radiofrequency oscillator based on a self-injected dual-frequency laser exhibits “chaotic pulses with excitable-like properties” in a bounded-phase regime (Romanelli et al., 2015). The field equations involve complex amplitudes $x_I$6, population inversions $x_I$7, detuning $x_I$8, injection strength $x_I$9, cross saturation $$\tau_E \frac{dx_i}{dt} = -x_i + \left[\sum_{j=1}^n W_{ij}x_j + W_{iI}x_I + b_i\right]_+, \qquad i=1,\dots,n,$$0, pump $$\tau_E \frac{dx_i}{dt} = -x_i + \left[\sum_{j=1}^n W_{ij}x_j + W_{iI}x_I + b_i\right]_+, \qquad i=1,\dots,n,$$1, and inversion lifetime parameter $$\tau_E \frac{dx_i}{dt} = -x_i + \left[\sum_{j=1}^n W_{ij}x_j + W_{iI}x_I + b_i\right]_+, \qquad i=1,\dots,n,$$2. The chaotic pulse train emerges when the locked state destabilizes through a subcritical Hopf bifurcation in the window

$$\tau_E \frac{dx_i}{dt} = -x_i + \left[\sum_{j=1}^n W_{ij}x_j + W_{iI}x_I + b_i\right]_+, \qquad i=1,\dots,n,$$3

(Romanelli et al., 2015). At $$\tau_E \frac{dx_i}{dt} = -x_i + \left[\sum_{j=1}^n W_{ij}x_j + W_{iI}x_I + b_i\right]_+, \qquad i=1,\dots,n,$$4, the dynamics transitions from a phase-locked fixed point to a self-pulsating chaotic state, with pulse amplitudes nearly constant and interspike times quite regular, yet with a positive Lyapunov exponent (Romanelli et al., 2015). Refractory-like scales are measured as

$$\tau_E \frac{dx_i}{dt} = -x_i + \left[\sum_{j=1}^n W_{ij}x_j + W_{iI}x_I + b_i\right]_+, \qquad i=1,\dots,n,$$5

in normalized time, and pulses occur without $$\tau_E \frac{dx_i}{dt} = -x_i + \left[\sum_{j=1}^n W_{ij}x_j + W_{iI}x_I + b_i\right]_+, \qquad i=1,\dots,n,$$6 phase slips, distinguishing the mechanism from Adler-type excitability (Romanelli et al., 2015). This provides a concrete example of near-periodic chaotic pulse sequencing.

In a symmetric Kerr cavity with two counter-propagating fields,

$$\tau_E \frac{dx_i}{dt} = -x_i + \left[\sum_{j=1}^n W_{ij}x_j + W_{iI}x_I + b_i\right]_+, \qquad i=1,\dots,n,$$7

$$\tau_E \frac{dx_i}{dt} = -x_i + \left[\sum_{j=1}^n W_{ij}x_j + W_{iI}x_I + b_i\right]_+, \qquad i=1,\dots,n,$$8

with $$\tau_E \frac{dx_i}{dt} = -x_i + \left[\sum_{j=1}^n W_{ij}x_j + W_{iI}x_I + b_i\right]_+, \qquad i=1,\dots,n,$$9, the dynamics exhibits chaotic switching oscillations and self-switching oscillations between symmetry-related states (Bitha et al., 2024). The symbolic coding uses $$\tau_I \frac{dx_I}{dt} = -x_I + \left[\sum_{j=1}^n W_{Ij}x_j + W_{II}x_I + b_I\right]_+,$$0 and a kneading sequence $$\tau_I \frac{dx_I}{dt} = -x_I + \left[\sum_{j=1}^n W_{Ij}x_j + W_{II}x_I + b_I\right]_+,$$1, with kneading invariant

$$\tau_I \frac{dx_I}{dt} = -x_I + \left[\sum_{j=1}^n W_{Ij}x_j + W_{II}x_I + b_I\right]_+,$$2

(Bitha et al., 2024). Global bifurcations of Shilnikov homoclinic type, organized by a $$\tau_I \frac{dx_I}{dt} = -x_I + \left[\sum_{j=1}^n W_{Ij}x_j + W_{II}x_I + b_I\right]_+,$$3-equivariant Belyakov transition, generate infinitely many switching patterns $$\tau_I \frac{dx_I}{dt} = -x_I + \left[\sum_{j=1}^n W_{Ij}x_j + W_{II}x_I + b_I\right]_+,$$4, providing a symbolic-dynamics realization of sequential chaotic switching (Bitha et al., 2024).

In magnetism, a spin-torque-driven antiferromagnetic nano-oscillator exhibits the parameter sequence

$$\tau_I \frac{dx_I}{dt} = -x_I + \left[\sum_{j=1}^n W_{Ij}x_j + W_{II}x_I + b_I\right]_+,$$5

along $$\tau_I \frac{dx_I}{dt} = -x_I + \left[\sum_{j=1}^n W_{Ij}x_j + W_{II}x_I + b_I\right]_+,$$6 as current $$\tau_I \frac{dx_I}{dt} = -x_I + \left[\sum_{j=1}^n W_{Ij}x_j + W_{II}x_I + b_I\right]_+,$$7 increases (Wolba et al., 2021). The reduced four-dimensional phase space is $$\tau_I \frac{dx_I}{dt} = -x_I + \left[\sum_{j=1}^n W_{Ij}x_j + W_{II}x_I + b_I\right]_+,$$8, and chaos is diagnosed by Lyapunov signatures such as $$\tau_I \frac{dx_I}{dt} = -x_I + \left[\sum_{j=1}^n W_{Ij}x_j + W_{II}x_I + b_I\right]_+,$$9, while hyperchaos has $[y]_+=\max\{y,0\}$0 (Wolba et al., 2021). This is SCO-like in the parameter-driven sense: an ordered succession of oscillatory attractors culminating in chaos.

A related magnonic example is the “magnonic chaotic comb” (Sun et al., 29 May 2025). In a synthetic antiferromagnet with ultra-strong magnon-magnon coupling, three-wave mixing generates frequency combs with lines

$[y]_+=\max\{y,0\}$1

and comb spacing $[y]_+=\max\{y,0\}$2 (Sun et al., 29 May 2025). Depending on detuning $[y]_+=\max\{y,0\}$3, the system transitions to chaos through a subcritical Hopf bifurcation, torus-doubling bifurcation, or torus breakdown (Sun et al., 29 May 2025). The clearest sequence is

$[y]_+=\max\{y,0\}$4

near resonance, as seen in Poincaré maps, bifurcation diagrams, and positive largest Lyapunov exponents (Sun et al., 29 May 2025). This is a parameter-sweep analogue of sequential chaotic oscillation formation.

Josephson dynamics supplies another structured example. In an irradiated underdamped Josephson junction,

$[y]_+=\max\{y,0\}$5

structured chaotic windows alternate with subharmonic Shapiro steps in a devil’s staircase, especially in the sequence

$[y]_+=\max\{y,0\}$6

approaching $[y]_+=\max\{y,0\}$7 (Shukrinov et al., 2014). The onset of chaos on a subharmonic step follows a Feigenbaum period-doubling scenario, with estimates converging to $[y]_+=\max\{y,0\}$8 and $[y]_+=\max\{y,0\}$9, and the structured set has fractal dimension

$1\to 2\to \cdots \to n$00

(Shukrinov et al., 2014). Here the “sequence” lies in control-parameter space rather than in graph-ordered state visitation, but the idea of a correlated family of chaotic oscillatory windows is closely related.

A slow-fast perspective is given by “fast chaos” in relaxation systems (Jaquette et al., 2022). In the chaotic Rulkov map,

$1\to 2\to \cdots \to n$01

fast-scale chaos can coexist with nearly periodic slow burst timing, quantified by small coefficient of variation

$1\to 2\to \cdots \to n$02

for interburst intervals (Jaquette et al., 2022). Passage of the slow cycle through a crisis of the fast chaotic attractor near $1\to 2\to \cdots \to n$03 opens shortcut routes, producing irregular long and short cycles (Jaquette et al., 2022). This refines SCO-like behavior into a distinction between chaotic episode content and chaotic episode sequencing.

6. Terminology, boundaries of the concept, and non-equivalent usages

The literature supports both narrow and broad senses of SCOs. In the narrow sense established by (Zang et al., 29 May 2026), SCOs are graph-ordered chaotic E-I oscillations in threshold-linear networks with strong inhibition and unstable singleton fixed points. Under this definition, the graph predicts the sequence, and the attractor structure is explicitly metastable and itinerant.

In a broader sense, SCOs can refer to recurrent chaotic pulse, switching, or oscillatory episodes whose order is sufficiently structured to admit symbolic or statistical description. This broader interpretation encompasses bounded-phase chaotic pulse trains (Romanelli et al., 2015), symmetric chaotic switching organized by kneading invariants (Bitha et al., 2024), sequential parameter routes through fixed points, tori, chaos, and hyperchaos (Wolba et al., 2021), or relaxation cycles carrying fast chaotic content (Jaquette et al., 2022).

Several distinctions remain important. SCOs are not identical to generic chaos; they require recurrent organization. They are not identical to ordinary periodic self-oscillation; the recurrence is nonperiodic and attractor-based. They are not necessarily noise-driven; in several cases, including the opto-radiofrequency system and the E-I TLN framework, the deterministic skeleton is chaotic, although noise may affect triggering statistics (Romanelli et al., 2015). They are also not identical to mode switching among a finite list of stable periodic attractors, because the relevant structures are often attractor ruins or chaotic attractors rather than stable limit cycles (Zang et al., 29 May 2026).

Finally, the unrelated wave-mechanical use of “SCO” for “spatially confined oscillation” must be excluded from this concept (Stadtler et al., 2021). That paper studies localized superpositions of plane waves in inhomogeneous media, with emergent geodesic-like motion and no chaos, no strange attractors, and no sequential oscillatory switching in the nonlinear-dynamical sense (Stadtler et al., 2021).

Taken together, the literature suggests that Sequential Chaotic Oscillations are best understood as recurrent chaotic oscillatory dynamics with nontrivial temporal organization. In the strongest current formulation, they are graph-ordered, metastable, and itinerant E-I oscillations under constant input (Zang et al., 29 May 2026). In the broader dynamical-systems landscape, they include ordered chaotic pulse trains, switching attractors, and structured routes through oscillatory states into and through chaos (Romanelli et al., 2015, Bitha et al., 2024, Wolba et al., 2021, Sun et al., 29 May 2025). This suggests a unifying viewpoint: SCOs occupy the middle ground between rigid periodic oscillation and unstructured chaos, with recurrence preserved but exact repetition lost.