WorksPhysics and Astronomy
Unstable Periodic Orbit Localization in a Nonlinear System-Environment Model
Abstract
The role of unstable periodic orbits (UPOs) is investigated in the chaotic dynamics of a nonlinear system-environment model composed of a central two-level system coupled to an environment of two-level systems. Families of UPOs embedded in chaotic regions are identified and their local instability is characterized. Their behavior is examined under variations of the environment size, energy, and system symmetry, allowing the model to serve as a controlled testbed in which the dimensionality and dynamical conditions can be systematically modified. Characteristic UPO-based dynamical quantities are used to assess the influence of these periodic orbits on nearby chaotic trajectories, including occupation of the UPO tube, its persistence at late times, return fidelity enhancement, and recurrence timing. Across the different model variations considered, trajectories initialized close to the UPOs retain statistically detectable spatial and temporal signatures associated with the periodic orbits. These results indicate that unstable periodic orbits can retain a measurable organizing influence on surrounding chaotic trajectories across changes in dimensionality, energy, and system symmetry, giving rise to scar-like dynamical signatures in a simple nonlinear system–environment model.
Introduction
Chaotic dynamical systems are characterized by a strong sensitivity to initial conditions: trajectories starting arbitrarily close to one another may eventually evolve in very different ways. Despite this apparent irregularity, chaotic dynamics contains recurrent structures that can provide a geometrical description of the underlying phase-space organization [1, 2, 3, 4]. Among these structures, unstable periodic orbits (UPOs) are trajectories that return exactly to their initial state after a finite period , while remaining unstable against perturbations. An initial condition placed exactly on a UPO repeats the same motion indefinitely, whereas a nearby trajectory generally follows a similar evolution only temporarily before separating from it. This local influence suggests that UPOs may affect the surrounding chaotic dynamics even after nearby trajectories begin to separate from the periodic path. Collections of UPOs can therefore provide a useful skeleton for describing chaotic dynamics and, in suitable systems, their statistical properties [2, 3, 4].
UPOs are particularly relevant in quantum physics through their connection with quantum scars. In its original single-particle formulation, quantum scarring refers to quantum states that become unusually concentrated around unstable periodic orbits of their classical counterparts. Heller first identified this behavior in eigenfunctions of chaotic Hamiltonian systems, and subsequent work established connections between scar strength, recurrence, and the instability of the underlying periodic orbit [5, 6, 7, 8]. The concept has since been extended to many-body systems, where atypical nonthermal states and persistent dynamical revivals can appear within otherwise ergodic spectra [9, 10, 11, 12]. Closely related phenomena have also been studied in the Dicke model, coupled-spin systems, chaotic mean-field dynamics, and interacting bosonic systems, where unstable or recurrent classical or mean-field motion has been connected with localization, dynamical revivals and persistent memory [13, 14, 15, 16, 17, 18, 19, 20].
The periodic-orbit viewpoint also has applications well beyond quantum physics. A particularly developed example is fluid turbulence, where the flow can be represented as a trajectory through a very high-dimensional state space that repeatedly approaches unstable invariant solutions. Periodic and relative periodic solutions have been identified in shear flows and two-dimensional turbulence, and increasingly large catalogues of recurrent solutions have been used to characterize both the geometry and the statistics of turbulent motion [21, 22, 23, 24, 25, 26, 27, 28, 29]. Related UPO-based descriptions have also been applied to other nonlinear problems, including neuronal dynamics, nonlinear fluid systems, atmospheric regime transitions, and the stabilization and control of chaotic systems [30, 31, 32, 33, 34, 35]. These examples illustrate the broader role of unstable recurrent structures in the analysis of complex nonlinear dynamics.
Periodic-orbit methods form part of a broader set of tools for characterizing deterministic chaos. Alternative approaches based on symbolic dynamics, ordinal patterns, permutation entropy, complexity measures, and mathematical analyses of nonlinear evolution equations provide complementary ways of identifying structure in complex dynamical systems [36, 37, 38, 39, 40, 41].
The model studied here shares several dynamical ingredients with these UPO-related scar studies. It possesses unstable periodic trajectories embedded in chaotic regions, and it is examined whether trajectories initialized close to them retain enhanced spatial localization and recurrent dynamics associated with their characteristic periods. There is, however, an important conceptual distinction. The present system is governed by nonlinear, state-dependent system–environment equations rather than by the eigenstates of a fixed linear quantum Hamiltonian. Consequently, the behavior investigated here does not satisfy the conventional definition of Heller-type quantum scarring. Therefore, the phenomenon studied below is referred to as UPO-associated dynamical localization. Nevertheless, the similarities in periodic-orbit localization, recurrence, and the competition between orbital instability and recurrence timescale provide a direct point of contact with the scar-related literature [8, 13, 15, 18, 19].
The model originates from the geometry-based formulation of interacting TLS dynamics introduced in Ref. [42] and subsequently applied to a chiral system–environment setting in Ref. [43]. Each TLS is represented through variables associated with its Bloch-sphere geometry, while the interaction depends on the population imbalance of the coupled subsystems. The resulting nonlinear equations have a mathematical form reminiscent of Gross–Pitaevskii-type dynamics, although their physical origin and interpretation are different. Previous work focused primarily on the resulting system–environment evolution and the emergence of oscillatory, tunneling-suppressed, and dissipative-like regimes [42]. Here, the same comparatively simple model is used as a controlled setting for studying how unstable periodic motion can organize nonlinear chaotic dynamics.
Homogeneous environments containing and TLSs at a common energy density are first investigated. The chaotic background is characterized using finite-time Lyapunov exponents (FTLEs) [44], while the local instability of the identified periodic orbits is quantified through a fixed-sector Floquet analysis that respects the invariant structure of the homogeneous dynamics. As described in Sec. IV, complementary spatial and temporal metrics are then used to quantify the influence of these UPOs on the surrounding chaotic dynamics. The representative case is also considered at five different energy densities, for which the complete analysis is repeated in order to examine how the observed behavior changes across different dynamical regimes.
Finally, the analysis is repeated for two additional cases. In the first, a finite energetic bias is introduced in the central TLS while the environment remains homogeneous. In the second, different energetic biases are assigned to the environmental TLSs, breaking the environmental homogeneity. These calculations are again performed for and at the reference energy density and test whether the observed UPO-associated dynamical signatures survive as the symmetry and collective structure of the reference model are progressively removed.
The purpose of the present work is therefore broader than the particular TLS realization. The model provides a compact nonlinear testbed in which the number of environmental degrees of freedom, energy, symmetry, and collective structure can be varied independently while retaining direct access to unstable periodic orbits and the surrounding chaotic dynamics. Because the formulation is expressed in terms of generic Bloch-vector degrees of freedom and state-dependent couplings, the results apply directly to systems governed by the same nonlinear equations rather than to a single specific physical implementation. More generally, the methodology developed here can be transferred to other nonlinear dynamical systems in which unstable recurrent structures coexist with chaotic motion. The established role of such structures in areas ranging from turbulence and geophysical dynamics to neuronal systems and chaos control further illustrates the broader relevance of this dynamical-systems viewpoint [23, 26, 28, 30, 33, 35].
This manuscript is structured as follows. § 1.1 introduces the nonlinear system–environment model and its dynamical formulation. Section III characterizes the chaotic dynamics, identifies the unstable periodic orbits considered in the subsequent analysis, and quantifies their local instability through Floquet analysis. Section IV examines the influence of these UPOs on nearby chaotic trajectories through spatial and temporal dynamical signatures. The dependence on energy is analyzed in § 2.1, while Section VI investigates the robustness of the observed behavior under symmetry-breaking modifications of the model. Finally, the main conclusions are summarized in Section VII.
Nonlinear system–environment model
The nonlinear system–environment model considered here is based on the geometry-based representation of two-level systems introduced in Ref. [42]. In this formulation, the quantum state of each TLS is mapped onto the Bloch sphere after removing its physically irrelevant global phase. The resulting Bloch coordinates can be expressed as expectation values of the Pauli operators, providing a classical-like geometrical representation of the TLS dynamics. In particular, the population difference and relative phase form a pair of canonically conjugate variables. A detailed derivation of this mapping is given in Ref. [42].
A central TLS coupled to an environment formed by additional TLSs is considered. Each environmental TLS interacts with the central TLS, while no direct interaction is introduced between different environmental TLSs.
The central TLS is represented by the Bloch vector
and the th environmental TLS by
For pure states,
Here, and denote the corresponding population differences, while and denote the relative phases of the populations. The Bloch-sphere coordinates can therefore be written as
Throughout the manuscript, the subscript 0 denotes the central TLS, whereas the index labels the TLSs forming the environment.
In units with , the Hamiltonian function is written as
with
and
representing the system, the environment, and the interaction Hamiltonian, respectively. Here, and control the tunneling dynamics of the central and environmental TLSs, respectively, and describe their energetic asymmetries, and is the coupling strength between the central TLS and the th environmental TLS.
In this representation, the population difference plays the role of a generalized momentum. The interaction in Eq. (9) can therefore be interpreted as a momentum–momentum-type coupling. A detailed discussion of this construction and of alternative bilinear couplings is given in [42].
An important feature of the present model is that the interaction differs from the conventional spin–spin coupling of a linear many-spin Hamiltonian. In the quantum representation,
so that the interaction is constructed from products of expectation values rather than from an operator product such as . The resulting evolution is therefore nonlinear and explicitly dependent on the instantaneous state of the interacting subsystems [42].
A homogeneous environment is first considered,
and use the parameters
with setting the energy scale. In this homogeneous case, the identical environmental TLSs also allow an exact collective description of the dynamics. The derivation and the associated conserved quantities are given in Appendix § 1.2.1; only their consequences when constructing dynamically comparable trajectories will be needed below.
To compare systems containing different numbers of environmental TLSs, the energy per TLS is kept fixed,
where is the conserved total energy, i.e. the value of the Hamiltonian for a system formed by the central TLS and environmental TLSs. The denominator is therefore the total number of TLSs in the system. The following environment sizes are considered:
Keeping fixed, rather than the total energy , allows systems with different numbers of TLSs to be compared at the same energy per TLS.
The homogeneous system defined above is used as the reference model. In the following section, it is first determined whether the corresponding energy regions display chaotic dynamics, after which unstable periodic orbits embedded in them are searched for.
Chaotic dynamics and unstable periodic orbits
The chaotic dynamics of the homogeneous model are next characterized, and the periodic orbits subsequently used to test UPO-associated dynamical localization are identified.
Chaotic dynamics at fixed energy density
As defined in § 1.1, environments with
are compared at the same energy per TLS,
The purpose of varying is to determine how the chaotic dynamics changes as the finite environment becomes larger. It is necessary to remark at this point that these values are not intended to represent a thermodynamic limit. Instead, they provide a controlled finite-size sequence in which the number of environmental degrees of freedom is progressively increased while keeping the remaining model parameters fixed.
A direct evaluation of the accessible energy ranges confirms that the chosen energy corresponds to comparable regions of the available energy range for the five environment sizes. The accessible energy intervals and the relative position of the selected energy within them are discussed in Appendix § 1.2.2.
Before searching for periodic orbits, it is first verified that the selected energy regions exhibit sensitivity to initial conditions. This is quantified using finite-time Lyapunov exponents (FTLEs) [44].
The basic idea is to compare two trajectories whose initial conditions differ by a very small amount. If their separation is denoted by , in a chaotic region it grows approximately as
where is the Lyapunov exponent. A positive value therefore means that initially close trajectories separate exponentially with time and provides a direct measure of sensitivity to initial conditions.
In the present calculations, the exponent is evaluated over a finite evolution time, giving the finite-time Lyapunov exponent . The calculation starts from a reference trajectory and a nearby perturbed trajectory separated by a small distance . Both trajectories are then evolved for a short time interval . At the end of the first interval, their separation has changed from to , and the corresponding growth factor is recorded.
Before starting the next interval, the perturbed trajectory is redefined so that its distance from the reference trajectory is again . Its direction relative to the reference trajectory is kept equal to the direction of the separation reached at the end of the previous interval. The two trajectories are then evolved for another interval , at the end of which a new separation is obtained. Repeating this procedure produces a sequence of separations , where labels the successive evolution intervals.
The finite-time Lyapunov exponent is then calculated from the accumulated growth of these separations,
A positive value indicates that nearby trajectories tend to separate exponentially over the considered time interval. All trajectories considered in the same-energy ensembles yield positive values of . The median values are
for , respectively. The median is used rather than the mean because the FTLE distributions are not necessarily symmetric and may contain tails toward larger values. The median therefore provides a representative measure of the typical chaotic sensitivity that is less affected by a small number of unusually large FTLE values.
As shown in Figure 1, the distributions for and are concentrated at relatively small FTLE values. From onward, the distributions progressively shift toward larger values, with the strongest chaotic sensitivity observed for and . The selected energy regions therefore provide a chaotic background in which to search for unstable periodic orbits.
Figure 1. Distributions of the finite-time Lyapunov exponent for , and 10 at the common energy density used in the analysis. The horizontal axis gives the FTLE obtained for each sampled trajectory, while the vertical axis gives the fraction of sampled trajectories whose FTLE lies within each histogram interval. The orange vertical line marks the median value, , for each distribution.
Search for unstable periodic orbits
Having established chaotic behavior at the selected energies, periodic trajectories embedded in these chaotic regions are subsequently searched for. A periodic trajectory must satisfy a much more restrictive condition than an arbitrary trajectory: after a finite time , the complete dynamical state must return to its initial value.
The complete state is denoted by
A periodic orbit therefore satisfies
This condition requires not only the central TLS but also every individual environmental TLS to return to its initial state after one period. It is therefore the complete system–environment configuration that must close periodically, rather than only a reduced set of collective coordinates or a low-dimensional projection of the trajectory.
A randomly chosen chaotic trajectory will generally not satisfy this exact condition. The numerical procedure used to locate periodic-orbit candidates, refine them into closed trajectories, and validate their periodicity is described in Appendix § 1.2.3.
For each environment size four distinct periodic orbits are retained, giving a total of
selected trajectories. Using the same number of orbits for every ensures that each environment size contributes equally to the subsequent analysis.
Considering the 20 selected periodic orbits, the fundamental period spans the range
The analysis therefore includes periodic trajectories with substantially different recurrence times.
For visualization, each periodic orbit is represented through the Bloch trajectory of the central TLS, using the coordinates
This representation provides a direct and physically transparent visualization of the recurrent motion of the central subsystem.
It should be emphasized that the periodicity criterion is not established from this visualization alone. The actual UPO condition is always verified for the complete dynamical state through Eq. (21). The central Bloch representation is therefore used only as a convenient way to display representative periodic trajectories.
Figure 2 shows three representative UPOs with different periods and instability values. The figure illustrates the range of recurrent motions found in the final catalogue, while the periodicity of each orbit is validated in the full phase space.
Figure 2. Three representative unstable periodic orbits for , shown through the Bloch trajectory of the central TLS in the coordinates . The value indicated in each panel is the period of the complete dynamical state, i.e., the time required for the full orbit to return to its initial condition.
Local instability of the periodic orbits
Finding a periodic trajectory does not yet characterize its local stability. The selected orbits are therefore analyzed using Floquet theory [13, 45]. The basic idea is to determine how an infinitesimal perturbation of a point on the periodic orbit evolves after one complete orbital period.
Let
be a point on a UPO and let
be a nearby initial condition. After one complete orbital period , the UPO returns to its initial state,
whereas the small perturbation has evolved to . Within the linear approximation, the relation between the initial and final perturbations can be written as
where is the monodromy matrix of the periodic orbit. This matrix describes how infinitesimal changes in the initial state are transformed after one complete period.
The eigenvalues of the monodromy matrix, , are the Floquet multipliers [13, 45]. Several multipliers are obtained because the initial state can be perturbed along different independent directions. A multiplier satisfying
corresponds to a direction in which the perturbation grows after one period and therefore indicates local instability. Conversely, corresponds to a contracting direction, while describes a neutral direction.
An additional point must be considered for the homogeneous model studied here. The environmental TLSs possess conserved geometrical quantities given by the pairwise scalar products
which are the elements of the environmental Gram matrix [46]. Because all environmental TLSs experience the same instantaneous effective field in the homogeneous model, these quantities remain constant during the evolution. Their conservation and the resulting collective structure are derived in Appendix § 1.2.1.
This conservation must be considered when defining the perturbation . A nearby state
is a valid initial condition. However, if the environmental Bloch vectors are perturbed independently, their relative angles generally change. Since , such a perturbation will in general modify the values of . The perturbed state would then belong to a different invariant sector from the original UPO. To construct a fixed-sector perturbation, must therefore be restricted so that all remain unchanged.
For this reason, the Floquet calculation considers perturbations that preserve the Gram invariants of the UPO. This is referred to as the fixed-sector Floquet analysis. It measures the local stability of the periodic orbit with respect to nearby trajectories belonging to the same invariant sector.
Some Floquet multipliers correspond to neutral directions and are not used to quantify the instability. The largest modulus among the remaining fixed-sector multipliers is denoted by
The corresponding local instability exponent is
A larger value of therefore means that a small perturbation separates more rapidly from the periodic orbit. Because the selected UPOs have different periods, comparing alone does not provide the complete picture. The instability is therefore normalized by the orbital angular frequency,
and define the dimensionless instability parameter
This quantity compares the local instability rate of the periodic orbit with its characteristic orbital frequency. Since the selected periodic orbits are already identified as unstable through their Floquet multipliers, is not used to determine whether an orbit is stable or unstable. Instead, it provides a dimensionless measure of the strength of that instability relative to the orbital timescale. Smaller values indicate weaker divergence of nearby trajectories over one orbital cycle, whereas larger values correspond to stronger relative instability. In UPO-based studies, the condition
is commonly associated with conditions favorable for the persistence of periodic-orbit signatures [16, 19, 47]. Accordingly, is used here to compare the relative instability of the identified UPOs on a common dimensionless scale.
As summarized in Figure 3, the selected periodic orbits display a broad range of instability values. The mean does not vary monotonically with , so no simple dependence of the average UPO instability on the number of environmental TLSs can be established from the present set of periodic orbits.
Figure 3. Local instability of the final collective UPO representatives as a function of the environment size . Colored circles represent the four individual UPO families obtained for each , with the colors identified in the legend. Red squares connected by a line show the mean value for each environment size. The instability is quantified by the dimensionless parameter .
Both the chaoticity of the surrounding dynamics, through the FTLE, and the local instability of the periodic orbits, through the Floquet analysis, have therefore been characterized. The next step is to determine whether these periodic structures also organize the dynamics of nearby chaotic trajectories. In the following section, trajectories initialized close to each UPO are compared with corresponding reference trajectories to test whether they retain a preferential spatial association with the periodic path and temporal recurrences related to its fundamental period.
UPO-associated dynamical localization
The previous section characterized the chaotic background of the system and the local instability of the UPOs embedded within it. A different question is now considered in order to extend this analysis: can these periodic orbits also influence the evolution of nearby chaotic trajectories? In particular, it is investigated whether trajectories initialized close to a UPO retain measurable spatial or temporal properties associated with that orbit after evolving away from their initial conditions.
The term UPO memory is used to describe this behavior. A trajectory displays UPO memory when, after the immediate effect of its initial proximity to the orbit has been excluded, its subsequent dynamics retains measurable properties associated with that particular periodic orbit. Such behavior would indicate that the UPO contributes to organizing the surrounding chaotic dynamics.
This question is closely related to the periodic-orbit picture of quantum scarring. In conventional scar studies, the influence of an unstable periodic orbit is investigated through quantities such as enhanced phase-space weight around the orbit, recurrent returns associated with its period, and frequency components related to the orbital motion [5, 8, 13, 18, 19]. Here, analogous dynamical quantities adapted to the nonlinear system–environment model are used.
For each of the 20 UPO families identified in Sec. III, two ensembles of trajectories are constructed. The first, referred to as the near-UPO ensemble, consists of trajectories initialized in the immediate vicinity of the periodic orbit. The second, referred to as the matched-reference ensemble, consists of trajectories initialized away from the UPO while belonging to the same fixed-energy invariant sector.
Consequently,
and
The relevant difference between the two ensembles is therefore their initial relation to the UPO rather than their energy or invariant sector. The specific procedure used to construct these initial conditions is described below, after introducing Eq. (42).
For every UPO,
independent trajectories are propagated for a total time of , where is the fundamental period of the corresponding UPO. Once the initial conditions have been generated, both ensembles evolve freely according to the same nonlinear equations of motion. In particular, the near-UPO trajectories are not forced to remain close to the periodic orbit at any later time.
To compare complete system–environment configurations, the product-state overlap is defined. Let
and
denote two different complete system–environment configurations. Their overlap is defined as
where corresponds to the central TLS and to the environmental TLSs [19, 48]. This quantity measures the similarity between two complete system–environment configurations. In particular, corresponds to identical configurations, whereas decreasing values of indicate progressively smaller similarity.
The near-UPO and matched-reference initial conditions can now be specified more precisely. For the near-UPO ensemble, a random point along the periodic orbit is selected and slightly perturbed. The perturbation corresponds to a characteristic angular displacement of rad of the Bloch vectors and is constructed so that the total energy and Gram invariants of the UPO remain unchanged.
For the matched-reference ensemble, initial conditions are sampled from the same fixed-energy invariant sector but are required to remain sufficiently far from the complete UPO. Using the product-state overlap defined in Eq. (42), a candidate reference state is accepted only if its maximum overlap with the periodic orbit does not exceed 0.20. The reference trajectories therefore share the same energy and Gram invariants as the near-UPO trajectories but do not initially lie in the immediate neighborhood of the UPO.
Several complementary quantities are used to determine whether trajectories initialized close to a UPO retain information about its spatial path and characteristic period. Phase-space localization and recurrence-based quantities are commonly used to characterize periodic-orbit signatures [8, 13, 14, 18, 19]. Other quantities, including spectral measures of the orbital frequency, can provide additional information, but the analysis here focuses on the spatial localization and recurrence observables introduced below.
Full-window tube localization.
The first quantity measures the association between a trajectory and the complete periodic path rather than with one particular point of the UPO. A given periodic orbit is labeled by and represented using uniformly distributed points,
where specifies the position along the periodic orbit.
For any trajectory , either from the near-UPO or reference ensemble, the quantity
is defined. Each term in the sum compares the state of the trajectory at time with one point of the UPO. Averaging over the 360 points provides a single measure of how strongly that trajectory is associated with the complete periodic path. A larger value of therefore means that the trajectory at time has a stronger association with UPO .
This provides an orbit-specific measure analogous to the enhanced phase-space weight used in periodic-orbit scar analyses [8, 13, 19].
For a time interval , the time-averaged value
is defined, and the two ensembles are compared through
Thus,
indicates that trajectories initialized near the UPO remain more strongly associated with its periodic path than the matched-reference trajectories.
The main interval considered here is –. The interval from 0 to is excluded because the near-UPO ensemble is deliberately initialized very close to the periodic orbit. A large value of at very early times could therefore simply reflect how the initial conditions were constructed. By starting the quantitative analysis at , it is tested whether the stronger association with the UPO remains visible after the trajectories have evolved freely for two complete orbital periods.
Late-time tube localization.
To determine whether this spatial association remains detectable much later in the evolution, the same analysis is repeated using only the final interval –. The corresponding quantity is
This provides a more restrictive test of persistent UPO-associated localization because it measures the difference between the two ensembles long after their initialization.
Return fidelity.
Tube localization measures association with the complete periodic path. It is next examined whether a trajectory also returns close to its own initial state. This is quantified through the return fidelity
By construction, . Values close to unity at later times indicate that the complete system–environment configuration has returned close to its initial state. Return fidelities or survival probabilities around integer multiples of a UPO period are standard dynamical signatures of periodic-orbit memory [13, 18].
Because the exact UPO returns after every integer multiple of its period, the first five expected recurrences are analyzed,
For each recurrence , is defined as the largest value of the return fidelity within a small interval around ,
This finite time window is used because a trajectory that is not exactly periodic is not expected to reach its maximum return fidelity precisely at an integer multiple of the UPO period. The maximum can instead occur slightly before or after , while still remaining dynamically associated with the recurrence imposed by the periodic orbit.
The average amplitude of the first five returns is then
The relative enhancement between the two ensembles is defined as
Therefore,
means that trajectories initialized close to the UPO exhibit stronger returns toward their initial configurations than dynamically comparable reference trajectories. This provides a temporal signature that the periodic orbit is influencing the surrounding chaotic dynamics.
Return timing
The amplitude of a return does not determine how accurately that return follows the period of the UPO. The positions of the first five return fidelity maxima are therefore also compared with the recurrence times predicted by the periodic orbit.
If denotes the time at which occurs, the normalized timing error is defined as
A smaller value of means that the observed recurrences occur more closely to the times
defined by the periodic orbit. UPO-associated temporal memory therefore corresponds to
This quantity provides a direct test of whether the fundamental period of the UPO continues to organize the timing of nearby chaotic trajectories [13, 18].
For each UPO family, the near-UPO and matched-reference ensembles contain several independent trajectories. Comparing only their mean values is not enough, because one ensemble may have a larger mean while the individual values within both ensembles are widely dispersed. A Mann–Whitney test [49] is therefore used to assess whether the two ensembles differ consistently.
A UPO family is classified as significant for a given diagnostic when the corresponding Mann–Whitney test gives
where is the parameter provided by the test. In practical terms, this indicates that the criterion associated with that diagnostic is statistically supported by the comparison between the near-UPO and matched-reference ensembles. In summary, significant means that the near-UPO ensemble shows a statistically supported difference in the expected direction: higher full-window and late-time tube localization, higher return fidelity, or lower return timing than the matched-reference ensemble.
The main results are summarized in Table I.
| Diagnostic | UPO-memory criterion | Significant families |
| Tube localization – | 18/20 | |
| Tube localization – | 14/20 | |
| Return fidelity – | 19/20 | |
| Return timing – | 20/20 |
Table I. Summary of the main diagnostics used to quantify UPO-associated dynamical localization. The last column gives the number of UPO families showing a statistically significant difference between the near-UPO and matched-reference ensembles.
As shown in Table I, 18 of the 20 UPO families show an enhanced occupation of the complete periodic path over the interval –. This effect remains statistically detectable during the final – interval for 14 families, showing that the spatial association with the UPO often persists long after initialization.
The return fidelity analysis provides a complementary temporal result. Enhanced return amplitudes over the first five periods are found for 19 of the 20 UPO families. Furthermore, the recurrence timing remains associated with the UPO period for all 20 families: the return maxima of the near-UPO trajectories occur significantly closer to , , , than those of the matched-reference trajectories.
The results therefore indicate a specific form of dynamical organization: trajectories initialized near a UPO retain both a preferential association with its periodic path and recurrent dynamics related to its fundamental period.
To illustrate these quantities directly, Figure 4 shows one representative UPO. In panel (a), the near-UPO trajectories retain a larger tube occupation than the matched-reference trajectories after the initial – interval excluded from the quantitative analysis. In panel (b), their return fidelity maxima remain associated with the recurrence times . The figure therefore provides a representative example of the spatial and temporal effects quantified statistically across the complete set of UPO families.
Figure 4. Representative visualization of UPO-associated dynamical memory for an periodic orbit with and . (a) Phase-averaged tube occupation for the near-UPO and matched-reference ensembles. The shaded interval indicates the initial interval excluded from the quantitative tube-localization analysis. (b) Return fidelity for the same two ensembles. The vertical dashed lines indicate the recurrence times , with .
Taken together, these results show that the periodic orbits identified in Sec. III leave statistically reproducible spatial and temporal signatures in nearby chaotic trajectories. In most cases, trajectories initialized close to a UPO retain an enhanced association with its complete periodic path, while their recurrences remain related to the fundamental period of the orbit.
This behavior is referred to as UPO-associated dynamical localization. The term is used here to describe persistent orbit-specific spatial and temporal memory in the nonlinear dynamics, rather than conventional eigenstate scarring of a fixed linear Hamiltonian.
Having established this phenomenon at the reference energy density, the following section examines how these signatures change when the energy is varied.
Energy dependence of chaos and UPO-associated localization
The previous section established that unstable periodic orbits leave measurable spatial and temporal signatures in nearby chaotic trajectories at the reference energy density. It is now examined whether these signatures change when the energy of the system is varied.
To study the effect of energy independently of the environment size, the number of environmental TLSs is fixed at
Five energy shells centered on the reference energy density are considered, including two lower- and two higher-energy cases. The selection of these energies from the accessible energy interval is described in Appendix § 1.2.2. The corresponding energy densities are
At each energy density, four independent UPOs are selected and the same near-UPO and matched-reference analysis introduced in Sec. IV is repeated. The objective is not to introduce new localization measures, but to determine how the previously defined dynamical quantities change with energy. The results are summarized in Table II. For each energy density, the table reports the mean values obtained from the four selected UPOs, allowing the evolution of chaoticity, local UPO instability, and UPO-associated localization to be compared over the energy scan.
| Timing | |||||
| 0.2152 | 0.0517 | 0.3796 | 1.668 | 3.169 | 4/4 |
| 0.5262 | 0.1012 | 0.4975 | 1.695 | 3.151 | 4/4 |
| 0.8371 | 0.2092 | 0.6733 | 1.065 | 1.779 | 4/4 |
| 1.1481 | 0.3240 | 0.8625 | 1.033 | 1.523 | 4/4 |
| 1.4590 | 0.4625 | 0.5619 | 1.045 | 1.777 | 4/4 |
Table 2. Energy dependence of the main dynamical and localization diagnostics for . Four independent UPOs are considered at each energy density. denotes the median finite-time Lyapunov exponent evaluated within the invariant sector, while quantifies the dimensionless local instability of the periodic orbit. The angle brackets denote the mean over the four UPOs selected at each energy. is the tube-localization enhancement over the interval –, and is the enhancement of the first five return fidelity peaks. The final column gives the number of UPO families, out of four, satisfying the statistically significant recurrence-timing criterion introduced in Sec. IV.
As can be seen in Table II, the chaoticity of the system shows a clear dependence on energy. For each energy density, the sector FTLE is calculated for the four selected UPO families and their median values are averaged. The resulting increases monotonically from 0.0517 at the lowest energy to 0.4625 at the highest one. The dynamics therefore becomes progressively more chaotic over the energy interval considered.
The remaining quantities do not vary monotonically with energy. The mean local UPO instability, , increases from 0.3796 to 0.8625 over the first four energy densities, before decreasing to 0.5619 at the highest energy. The selected UPOs therefore tend to become more unstable as the energy is increased, although this tendency does not extend monotonically over the complete scan.
The two localization measures display the opposite general behavior. The mean return fidelity enhancement decreases from values above 3 at the two lowest energies to values between approximately 1.5 and 1.8 at the three higher energies. Similarly, takes values of 1.668 and 1.695 at the two lowest energies and remains much closer to unity at the three higher energies. Neither dependence is strictly monotonic, since both quantities increase again at the highest energy, but the overall trend is a reduction of the UPO-associated localization strength as the energy is increased.
These trends are shown directly in Figure 5, which compares the energy dependence of the chaoticity, the local Floquet instability of the UPOs, and the two localization enhancements.
Figure 5. Energy dependence of chaoticity, local UPO instability, and UPO-associated dynamical localization for . (a) Mean invariant-sector finite-time Lyapunov exponent as a function of the energy density . (b) Mean dimensionless local UPO instability , with . (c) Mean tube-localization enhancement . (d) Mean enhancement of the first five return fidelity peaks, . In all panels, each point represents the mean over the four collectively inequivalent UPO families obtained at the corresponding energy density.
The recurrence timing shows a different behavior. At every energy considered, all four UPO families satisfy the timing criterion introduced in Sec. IV. Thus, for all 20 periodic orbits in the energy scan, the first five return fidelity maxima of the near-UPO trajectories occur significantly closer to
than those of the corresponding matched-reference trajectories.
The energy scan therefore shows that increasing energy is accompanied by a strong increase in chaoticity and, in general, by weaker tube-localization and return fidelity enhancements. The local Floquet instability also tends to increase over most of the scan, although its dependence is not monotonic. Despite these changes in magnitude, the recurrence timing remains statistically significant for every UPO considered, showing that the characteristic period of the UPO continues to influence nearby chaotic trajectories throughout the energy interval studied.
Robustness to central bias and environmental inhomogeneity
The previous sections established UPO-associated dynamical localization in the homogeneous model and examined how its main signatures vary with the energy density. It is now examined whether these signatures persist when the bias parameters of the model are changed.
This question is addressed in two steps. First, a finite bias is introduced in the central TLS while all environmental TLSs are kept identical. Second, different biases are also introduced among the environmental TLSs, thereby breaking the homogeneity of the environment.
For the parameter values considered throughout this work, the complete Hamiltonian can be written as
where is the energetic bias of the central TLS and is the corresponding bias of the th environmental TLS. The homogeneous model considered in the previous sections corresponds to
Because all environmental TLSs have identical dynamical parameters, their pairwise Gram quantities
remain constant during the evolution. As discussed previously, this invariant structure must be respected when constructing perturbations for the Floquet analysis and when generating dynamically comparable reference trajectories.
In the first modified model, a bias is introduced only in the central TLS,
Although the central dynamics is modified, all environmental TLSs remain identical and continue to experience the same instantaneous effective field. Consequently, the Gram invariants remain conserved and the same collective description used in the homogeneous reference model continues to apply.
In the second modified model, the central bias is retained and the environment is made inhomogeneous by assigning a different bias to each environmental TLS,
where the deterministic set is chosen to have zero mean and unit root-mean-square amplitude,
The environmental TLSs therefore no longer have identical dynamical parameters. In this case they do not, in general, experience the same instantaneous effective field, and the pairwise quantities are no longer constants of motion.
For both modified models, the calculations are performed at the reference energy density used in the main analysis,
and for
For each value of , four independent UPO families are considered, giving 20 UPO families for each model. The same dynamical observables introduced in Sec. IV are then evaluated.
One difference must be considered when constructing the reference ensembles. In the two homogeneous models, the near-UPO and reference trajectories are matched in both energy and Gram invariants. Once the environmental homogeneity is broken, the Gram quantities are no longer conserved, and the reference trajectories are therefore matched only in energy. The comparison below focuses on whether the same dynamical signatures remain statistically detectable in each model.
Following the same analysis used in the previous sections, the main results are summarized in Table III.
| Homogeneous | Central bias | Inhomogeneous environment | |
| 0 | 1 | 1 | |
| 0 | 0 | ||
| 18/20 | 20/20 | 20/20 | |
| 14/20 | 17/20 | 20/20 | |
| 19/20 | 17/20 | 19/20 | |
| 20/20 | 20/20 | 20/20 |
Table III. Comparison of the main UPO-associated dynamical signatures in the homogeneous reference model, the centrally biased homogeneous model, and the environmentally inhomogeneous model. Each value indicates the number of statistically significant UPO families out of the 20 analyzed.
The effect of the two modifications on the chaoticity and on the main UPO-associated signatures is displayed in Figure 6. In panel (a), the median finite-time Lyapunov exponent is shown as a function of for the three models. The original homogeneous model with displays the largest median FTLE for every environment size considered. Introducing the central bias substantially reduces the FTLE. When environmental inhomogeneity is subsequently introduced, the FTLE increases again relative to the centrally biased homogeneous model for , and 10, although it remains below the values of the original homogeneous model.
Figure 6. Robustness of the chaotic background, local UPO instability, and UPO-associated dynamical signatures under central bias and environmental inhomogeneity. (a) Median finite-time Lyapunov exponent of the same-energy trajectory ensemble as a function of the environment size for the homogeneous reference model, the centrally biased homogeneous model, and the environmentally inhomogeneous model. (b) Mean normalized local UPO instability of the four representative UPO families obtained at each . For the homogeneous reference model and the centrally biased homogeneous model, the quantity shown is , evaluated within the corresponding invariant sector. For the environmentally inhomogeneous model, the corresponding full-space normalized Floquet exponent is shown. (c) Percentage of UPO families satisfying the statistical criterion (Mann–Whitney ) for the main dynamical diagnostics: tube localization over –, late-time tube localization over –, return fidelity enhancement over –, and recurrence timing over –.
Panel (b) shows the normalized Floquet instability of the selected UPOs. Unlike the FTLE, no simple dependence on the bias configuration is observed. The centrally biased model gives the largest mean Floquet instability for , whereas the original homogeneous model gives the largest values for and . The effect of the bias on the local instability of the UPOs therefore depends on the environment size and does not follow a monotonic trend.
Panel (c) compares the number of UPO families showing statistically significant localization and recurrence signatures. For the full-window tube localization, significant behavior is found for 18/20 families in the homogeneous model and for 20/20 families in both modified models. At late times, the corresponding numbers are 14/20, 17/20, and 20/20. For the return fidelity enhancement over the first five periods, significant results are obtained for 19/20, 17/20, and 19/20 families, respectively. Finally, the recurrence-timing criterion is satisfied by all 20 UPO families in the three models.
The central bias therefore does not produce a systematic increase or decrease in the number of UPO families displaying these signatures. The environmentally inhomogeneous model, however, gives significant results for almost all UPO families for each of the four diagnostics considered here. The UPO-associated signatures therefore remain highly robust after the environmental homogeneity and the corresponding Gram conservation are lost.
Taken together, these results show that the UPO-associated dynamical signatures identified in the homogeneous reference model persist after a finite central bias is introduced and also when different biases are assigned to the environmental TLSs. In the latter case, the Gram quantities are no longer conserved, yet the spatial and temporal signatures remain highly robust across the complete UPO catalogue. The observed UPO-associated memory is therefore not restricted to the exact homogeneous structure of the original model.
Conclusions
In this work, it is studied whether unstable periodic orbits can organize the dynamics of a nonlinear system–environment model composed of a central two-level system coupled to a finite environment of additional TLSs. Unlike conventional quantum-scar studies based on the eigenstates of a fixed linear Hamiltonian, the present system evolves according to nonlinear, state-dependent equations. The analysis therefore focuses on dynamical signatures associated with UPOs rather than on eigenstate localization.
The study is organized around three complementary analyses. First, the chaotic dynamics and UPO-associated localization of the homogeneous reference model are characterized while varying the number of environmental TLSs, , and , at fixed energy density. Second, the dependence on energy is examined by keeping fixed and repeating the analysis over different energy shells. Finally, the robustness of the observed behavior against changes in the symmetry of the model is tested by first introducing an energetic asymmetry in the central TLS and then an inhomogeneous environment in which the environmental TLSs have different energetic biases.
For the homogeneous reference model, finite-time Lyapunov exponents confirm that the selected energy regions are chaotic, while the periodic-orbit search identifies unstable recurrent trajectories embedded in this background. Their local stability is characterized through fixed-sector Floquet analysis, and their influence on nearby chaotic trajectories is examined through spatial and temporal observables. In particular, the occupation of the complete UPO tube, its persistence at late times, the enhancement of return fidelity, and the timing of recurrent returns are evaluated. Taken together, these quantities show that trajectories initialized close to a UPO retain a measurable association with both its spatial path and its characteristic recurrence time after the immediate effect of the initial condition has been excluded. This behavior is referred to as UPO-associated dynamical localization.
In addition, the energy dependence of this behavior is studied for . Increasing the energy produces a pronounced increase in the chaoticity of the surrounding dynamics, while the strength of the UPO-associated spatial and return fidelity signatures generally decreases. The local Floquet instability does not follow a simple monotonic dependence on energy. By contrast, the timing of the recurrent returns remains closely linked to the UPO period throughout the energy range considered. These results indicate that the strength of UPO-associated localization depends on the dynamical regime, while the characteristic recurrence times of the UPOs remain detectable even as the surrounding dynamics becomes more chaotic.
Finally, the robustness analysis shows that the observed UPO-associated signatures are not restricted to the exact symmetry of the reference model. Introducing a finite bias in the central TLS reduces the chaoticity of the system while preserving the localization and recurrence signatures. When the environmental TLSs are subsequently made nonidentical, the chaoticity increases again for some environment sizes, although it remains below that of the original homogeneous model. The UPO-associated signatures also persist in this case, even though the Gram invariants of the homogeneous environment are no longer conserved. The UPO-associated organization of nearby trajectories therefore survives both changes in the central system and the loss of environmental homogeneity, even when these modifications substantially alter the surrounding chaotic dynamics.
The three analyses lead to a common picture. UPOs remain dynamically relevant across changes in environment size, energy, and model symmetry, although the magnitude of the individual localization measures depends on the surrounding chaotic regime. The results do not suggest a simple one-to-one relation between chaoticity, Floquet instability, and localization strength. Instead, spatial localization, recurrence amplitude, and recurrence timing provide complementary information about how unstable periodic motion influences nearby chaotic trajectories.
More generally, the model considered here provides a convenient nonlinear chaotic framework in which UPOs can be studied while independently varying system size, energy, symmetry, and system–environment parameters. Its system–environment structure also makes it possible to examine how recurrent dynamics changes when a central subsystem interacts with an increasingly complex or inhomogeneous environment. In this setting, a scar-like dynamical analysis can be performed by testing whether trajectories initialized close to UPOs retain measurable spatial and temporal signatures of the underlying periodic motion. Beyond the specific TLS realization, the dynamical-systems framework employed here is relevant to a broad class of nonlinear problems in applied mathematics and physics in which unstable recurrent structures organize chaotic motion. The present analysis therefore connects nonlinear system–environment dynamics with the broader use of unstable periodic orbits as organizing structures of complex chaotic dynamics.
Collective structure and Gram invariants
For a homogeneous environment, all environmental TLSs have the same dynamical parameters,
This does not imply that the environmental TLSs have identical instantaneous states. Instead, it means that all of them follow the same equations of motion and interact in the same way with the central TLS. The collective environmental Bloch vector is defined as
Using this quantity, the Hamiltonian of the homogeneous model becomes
The interaction between the central TLS and the environment therefore depends only on the collective environmental coordinate . Consequently, the evolution of the central TLS is determined by collective environmental variables rather than by the individual environmental Bloch vectors separately.
The same homogeneity also imposes an additional invariant structure on the environmental dynamics. Since every environmental Bloch vector evolves under the same instantaneous effective field generated by the central TLS, all environmental vectors undergo the same time-dependent rotation. Their relative geometry is therefore preserved during the evolution.
For any two environmental TLSs, the pairwise Gram quantity is defined as
Because all environmental TLSs obey the same equations of motion with the same time-dependent central coordinate , differentiation of gives
Therefore,
and the relative geometry of the environmental Bloch vectors remains unchanged during the evolution. The conserved values of therefore define invariant sectors of the dynamics.
This invariant structure is used in the main analysis when constructing fixed-sector perturbations for the Floquet calculation and when comparing near-UPO trajectories with reference trajectories belonging to the same dynamically accessible sector.
Accessible energy ranges
For the homogeneous model considered in the main analysis, the parameters are
The Hamiltonian can therefore be written as
The accessible energy bounds can be obtained by maximizing and minimizing this Hamiltonian over all allowed Bloch-sphere coordinates. Since all population coordinates are restricted to the interval , maximizing the Hamiltonian over the allowed Bloch-sphere coordinates gives
On the other hand, the symmetry of the Hamiltonian gives the corresponding lower bound,
Therefore, for the five environment sizes considered in this work, the minimum and maximum energies are listed in Table IV.
| 2 | -6.0000 | 6.0000 |
| 4 | -10.0000 | 10.0000 |
| 6 | -14.1421 | 14.1421 |
| 8 | -18.4391 | 18.4391 |
| 10 | -22.8035 | 22.8035 |
Table IV. Minimum and maximum energies accessible to the homogeneous model for the five numbers of environmental TLSs considered in this work.
To compare energies across systems with different numbers of environmental TLSs, the normalized energy coordinate is defined as
With this definition, corresponds to the minimum accessible energy and to the maximum. The reference energy originally selected for the system is
Its normalized energy coordinate is
Since the system contains three TLSs in total, this reference energy corresponds to the energy per TLS
This value of is kept fixed in the main -dependent analysis. The total energy used for each environment size is therefore
The corresponding total energies and their normalized positions are
| 2 | 2.5114 | 0.7093 |
| 4 | 4.1856 | 0.7093 |
| 6 | 5.8598 | 0.7072 |
| 8 | 7.5341 | 0.7043 |
| 10 | 9.2083 | 0.7019 |
Table V. Total energies used in the main -dependent analysis and their corresponding normalized energy coordinates. All values are obtained by keeping the energy per TLS fixed at .
The values of vary only from 0.7093 to 0.7019. Thus, keeping the energy per TLS fixed at
selects comparable regions of the accessible energy interval for all five environment sizes considered in the main -dependent analysis.
The energy-dependent analysis of Sec. V instead keeps the environment size fixed at . For this system, the reference energy
corresponds to the normalized energy coordinate
This value is taken as the central point of the energy scan. Four additional energy shells are selected using two values below and two above the reference, with a constant spacing
The five normalized energy coordinates are therefore
Using Eq. (80), these values correspond to the energy densities
These are the five energy densities used in Sec. V.
Search and numerical validation of unstable periodic orbits
Trajectories that return close to a previous state are therefore searched for first, and these approximate recurrences are subsequently refined numerically. The search is based on a Poincaré section [3].
A Poincaré section reduces a continuous trajectory to the sequence of points at which it crosses a chosen surface in phase space. This makes close returns considerably easier to identify. The section
is used, with crossings taken in a fixed direction. Since
the conditions and select the central phase and remove the second crossing corresponding to .
Long trajectories are generated from different initial conditions at the target energy. Each time a trajectory crosses the Poincaré section, the complete state of the central and environmental TLSs is recorded. The condition in Eq. (91) is therefore imposed only on the central TLS; the environmental variables are not fixed and take the values generated dynamically at each crossing.
Pairs of section crossings that are very close to one another but separated by a finite time provide approximate periodic-orbit candidates. For each candidate, both the initial state and the return time are then adjusted numerically until the periodicity condition
is satisfied while keeping the total energy fixed. In other words, the close recurrence provides an initial estimate of the periodic orbit, and the numerical refinement progressively modifies that estimate until the complete state closes after one period.
Once a periodic solution has been found, it is integrated again using stricter numerical tolerances. It is then independently verified that the final state reproduces the initial state and that the energy remains constant throughout one complete period. This additional integration ensures that the periodic solution is not an artifact of the numerical optimization. The Python implementation used for the recurrence search, numerical refinement, and validation is provided.
Acknowledgements
D. M.-G. gratefully acknowledges Pedro Bargueño and Salvador Miret-Artés for their valuable discussions, guidance, and support during the development of the initial model on which this work is based.
Data availability
All Python code used in this work, together with instructions to reproduce the reported results, are available at https://github.com/daniel-martinez-gil/UPOs.
References
References
- [1]Jean-Pierre Eckmann and David Ruelle. Ergodic theory of chaos and strange attractors. Reviews of Modern Physics, 57(3):617–656, 1985.DOI
- [2]Ditza Auerbach, Predrag Cvitanović, Jean-Pierre Eckmann, Gemunu Gunaratne, and Itamar Procaccia. Exploring chaotic motion through periodic orbits. Physical Review Letters, 58(23):2387–2389, 1987.pubmed.ncbi.nlm.nih.gov/10034735
- [3]Predrag Cvitanović. Periodic orbits as the skeleton of classical and quantum chaos. Physica D: Nonlinear Phenomena, 51(1–3):138–151, 1991.DOI
- [4]Predrag Cvitanović. Periodic orbit theory in classical and quantum mechanics. Chaos, 2(1):1–4, 1992.pubmed.ncbi.nlm.nih.gov/12779944
- [5]Eric J. Heller. Bound-state eigenfunctions of classically chaotic hamiltonian systems: Scars of periodic orbits. Physical Review Letters, 53(16):1515–1518, 1984.DOI
- [6]E. B. Bogomolny. Smoothed wave functions of chaotic quantum systems. Physica D: Nonlinear Phenomena, 31(2):169–189, 1988.DOI
- [7]M. V. Berry. Quantum scars of classical closed orbits in phase space. Proceedings of the Royal Society of London A, 423(1864):219–231, 1989.DOI
- [8]L. Kaplan and E. J. Heller. Measuring scars of periodic orbits. Physical Review E, 59(6):6609–6628, 1999.pubmed.ncbi.nlm.nih.gov/11969649
- [9]C. J. Turner, A. A. Michailidis, D. A. Abanin, M. Serbyn, and Z. Papić. Weak ergodicity breaking from quantum many-body scars. Nature Physics, 14:745–749, 2018.DOI
- [10]Maksym Serbyn, Dmitry A. Abanin, and Zlatko Papić. Quantum many-body scars and weak breaking of ergodicity. Nature Physics, 17:675–685, 2021.DOI
- [11]Sanjay Moudgalya, B. Andrei Bernevig, and Nicolas Regnault. Quantum many-body scars and hilbert space fragmentation: A review of exact results. Reports on Progress in Physics, 85(8):086501, 2022.DOI
- [12]Anushya Chandran, Thomas Iadecola, Vedika Khemani, and Roderich Moessner. Quantum many-body scars: A quasiparticle perspective. Annual Review of Condensed Matter Physics, 14:443–469, 2023.DOI
- [13]Saul Pilatowsky-Cameo, David Villasenor, Miguel A. Bastarrachea-Magnani, Sergio Lerma-Hernandez, Lea F. Santos, and Jorge G. Hirsch. Quantum scarring in a spin-boson system: fundamental families of periodic orbits. New Journal of Physics, 23:033045, 2021.DOI
- [14]Saúl Pilatowsky-Cameo, David Villaseñor, Miguel A. Bastarrachea-Magnani, Sergio Lerma-Hernández, Lea F. Santos, and Jorge G. Hirsch. Identification of quantum scars via phase-space localization measures. Quantum, 6:644, 2022.DOI
- [15]Debabrata Mondal, Sudip Sinha, and S. Sinha. Chaos and quantum scars in a coupled top model. Physical Review E, 102:020101, 2020.DOI
- [16]Quirin Hummel, Klaus Richter, and Peter Schlagheck. Genuine many-body quantum scars along unstable modes in bose-hubbard systems. Physical Review Letters, 130:250402, 2023.DOI
- [17]Bertrand Evrard, Andrea Pizzi, Simeon I. Mistakidis, and Ceren B. Dag. Quantum scars and regular eigenstates in a chaotic spinor condensate. Physical Review Letters, 132:020401, 2024.DOI
- [18]Bertrand Evrard, Andrea Pizzi, Simeon I. Mistakidis, and Ceren B. Dag. Quantum many-body scars from unstable periodic orbits. Physical Review B, 110:144302, 2024.DOI
- [19]Andrea Pizzi, Long-Hei Kwan, Bertrand Evrard, Ceren B. Dag, and Johannes Knolle. Genuine quantum scars in many-body spin systems. Nature Communications, 16:6722, 2025.DOI
- [20]Igor Ermakov, Oleg Lychkovskiy, and Boris V. Fine. Classical periodic trajectories and quantum scars in many-spin systems. Physical Review E, 112:064109, 2025.DOI
- [21]Genta Kawahara and Shigeo Kida. Periodic motion embedded in plane couette turbulence: Regeneration cycle and burst. Journal of Fluid Mechanics, 449:291–300, 2001.DOI
- [22]Divakar Viswanath. Recurrent motions within plane couette turbulence. Journal of Fluid Mechanics, 580:339–358, 2007.DOI
- [23]Genta Kawahara, Markus Uhlmann, and Lennart van Veen. The significance of simple invariant solutions in turbulent flows. Annual Review of Fluid Mechanics, 44:203–225, 2012.DOI
- [24]Gary J. Chandler and Rich R. Kerswell. Invariant recurrent solutions embedded in a turbulent two-dimensional kolmogorov flow. Journal of Fluid Mechanics, 722:554–595, 2013.DOI
- [25]Lucas, D., & Kerswell, R. R. (2015). Recurrent flow analysis in spatiotemporally chaotic 2-dimensional kolmogorov flow. Physics of Fluids, 27(4), 045106.
- [26]Michael D. Graham and Daniel Floryan. Exact coherent states and the nonlinear dynamics of wall-bounded turbulent flows. Annual Review of Fluid Mechanics, 53:227–253, 2021.DOI
- [27]Gökhan Yalniz, Björn Hof, and Nazmi Burak Budanur. Coarse graining the state space of a turbulent flow using periodic orbits. Physical Review Letters, 126(24):244502, 2021.DOI
- [28]Jacob Page, Peter Norgaard, Michael P. Brenner, and Rich R. Kerswell. Recurrent flow patterns as a basis for two-dimensional turbulence: Predicting statistics from structures. Proceedings of the National Academy of Sciences, 121(23):e2320007121, 2024.DOI
- [29]Omid Ashtari, Laurette S. Tuckerman, and Tobias M. Schneider. Computing simple invariant solutions in chaotic fluid flows. Annual Review of Fluid Mechanics, 59, 2026. Review in Advance, in press.DOI
- [30]Paul So, Joseph T. Francis, Theoden I. Netoff, Bruce J. Gluckman, and Steven J. Schiff. Periodic orbits: A new language for neuronal dynamics. Biophysical Journal, 74(6):2776–2785, 1998.pubmed.ncbi.nlm.nih.gov/9635732
- [31]Alberto Enciso, Alejandro Luque, and Daniel Peralta-Salas. Beltrami fields with hyperbolic periodic orbits enclosed by knotted invariant tori. Advances in Mathematics, 373:107328, 2020.arxiv.org/abs/1909.07448
- [32]Alberto Enciso, Daniel Peralta-Salas, and Álvaro Romániega. Beltrami fields exhibit knots and chaos almost surely. Forum of Mathematics, Sigma, 11:e56, 2023.arxiv.org/abs/2006.15033
- [33]Ott, E., Grebogi, C., & Yorke, J. A. (1990). Controlling chaos. Physical Review Letters, 64(11), 1196–1199.
- [34]Alexandre R. Nieto, Thomas Lilienkamp, Jesús M. Seoane, Miguel A. F. Sanjuán, and Ulrich Parlitz. Control of escapes in two-degree-of-freedom open hamiltonian systems. Chaos, 32(6):063118, 2022.DOI
- [35]Oisín Hamilton, Jonathan Demaeyer, Michel Crucifix, and Stéphane Vannitsem. Using unstable periodic orbits to understand blocking behavior in a low order land–atmosphere model. Chaos, 35(8):083126, 2025.DOI
- [36]Christoph Bandt and Bernd Pompe. Permutation entropy: A natural complexity measure for time series. Physical Review Letters, 88(17):174102, 2002.pubmed.ncbi.nlm.nih.gov/12005759
- [37]José M. Amigó, Matthew B. Kennel, and Ljupco Kocarev. The permutation entropy rate equals the metric entropy rate for ergodic information sources and ergodic dynamical systems. Physica D: Nonlinear Phenomena, 210(1–2):77–95, 2005.DOI
- [38]José M. Amigó, Ljupco Kocarev, and Janusz Szczepański. Order patterns and chaos. Physics Letters A, 355(1):27–31, 2006.DOI
- [39]José M. Amigó. Permutation Complexity in Dynamical Systems: Ordinal Patterns, Permutation Entropy and All That. Springer Series in Synergetics. Springer, Berlin, Heidelberg, 2010.DOI
- [40]Valeri A. Makarov, Ricardo Muñoz, Oscar Herreras, and Julia Makarova. Correlation dimension of high-dimensional and high-definition experimental time series. Chaos, 33(12):123114, 2023.DOI
- [41]J. Alberto Conejero, Félix Martínez-Giménez, Alfredo Peris, and Francisco Ródenas. Chaotic asymptotic behaviour of the solutions of the Lighthill–Whitham–Richards equation. Nonlinear Dynamics, 84(1):127–133, 2016.DOI
- [42]Daniel Martinez-Gil, Pedro Bargueno, and Salvador Miret-Artes. Dissipative evolution of a two-level system through a geometry-based classical mapping. New Journal of Physics, 27:124512, 2025.arxiv.org/abs/2501.03760
- [43]Daniel Martínez-Gil, Pedro Bargueño, and Salvador Miret-Artés. Chiral environment effects on the dynamics of a central chiral molecule. Physical Review A, 112(2):022810, 2025.arxiv.org/abs/2504.03344
- [44]Giancarlo Benettin, Luigi Galgani, Antonio Giorgilli, and Jean-Marie Strelcyn. Lyapunov characteristic exponents for smooth dynamical systems and for hamiltonian systems: A method for computing all of them. part 2: Numerical application. Meccanica, 15(1):21–30, 1980.DOI
- [45]Cvitanović, P., Artuso, R., Mainieri, R., Tanner, G., & Vattay, G. (2020). Chaos: Classical and Quantum.
- [46]Horn, R. A., & Johnson, C. R. (2013). Matrix Analysis.
- [47]L. Kaplan. Scars in quantum chaotic wavefunctions. Nonlinearity, 12:R1–R40, 1999.DOI
- [48]Arrechi, F. T., Courtens, E., Gilmore, R., & Thomas, H. (1972). Atomic coherent states in quantum optics. Physical Review A, 6(6), 2211–2237.
- [49]H. B. Mann and D. R. Whitney. On a test of whether one of two random variables is stochastically larger than the other. The Annals of Mathematical Statistics, 18(1):50–60, 1947.DOI





