Identification of quantum scars via phase-space localization measures

Sa´ulPilatowsky-Cameo,1 David Villase˜nor,1 Miguel A. Bastarrachea-Magnani,2 Sergio Lerma-Hern´andez,3 Lea F. Santos,4 and Jorge G. Hirsch1 1Instituto de Ciencias Nucleares, Universidad Nacional Aut´onomade M´exico, Apdo. Postal 70-543, C.P. 04510 CDMX, Mexico 2Departamento de F´ısica, Universidad Aut´onomaMetropolitana-Iztapalapa, San Rafael Atlixco 186, C.P. 09340 CDMX, Mexico 3Facultad de F´ısica, Universidad Veracruzana, Circuito Aguirre Beltr´ans/n, C.P. 91000 Xalapa, Veracruz, Mexico 4Department of Physics, Yeshiva University, New York, New York 10016, USA There is no unique way to quantify the degree of delocalization of quantum states in unbounded continuous spaces. In this work, we explore a recently introduced localization measure that quantifies the portion of the classical phase space occupied by a quantum state. The measure is based on the α-moments of the Husimi function and is known as the R´enyi occupation of order α. With this quantity and random pure states, we find a general expression to identify states that are maximally delocalized in phase space. Using this expression and the Dicke model, which is an interacting spin- boson model with an unbounded four-dimensional phase space, we show that the R´enyi occupations with large α are highly effective at revealing quantum scars. Furthermore, by analyzing the large moments of the Husimi function, we are able to find the unstable periodic orbits that scar some of the eigenstates of the model.

I. INTRODUCTION in the chaotic regime and distinguish the eigenstates that are maximally delocalized from those that are highly lo- The notion of localization in quantum systems presup- calized in phase space. The Dicke model has a rich and poses a basis representation. Anderson localization [1, 2], unbounded phase space. This spin-boson model [34] was for example, refers to the suppression of the classical dif- initially introduced to explain the phenomenon of super- fusion of particles in real space due to quantum inter- radiance [35–38] and has since fostered a wide variety of ferences. This phenomenon has a dynamical counterpart theoretical studies, including the behavior of out-of-time- originally studied in the context of the kicked rotor [3–6]. ordered correlators [39–41], manifestations of quantum Localization is an extremely broad subject that extends scarring [42–46], non-equilibrium dynamics [38, 47–51], to the hydrogen atom in a monochromatic field [7, 8], and measures of quantum localization with respect to Rydberg atoms [9], quantum billiards [10–16], banded phase space [45, 52]. The model is also of great inter- random matrices [17], interacting systems [18–22], and est to experiments with trapped ions [53, 54], supercon- kicked interacting systems [23–26], among others. ducting circuits [55], and cavity assisted Raman transi- In finite spaces, the degree of delocalization of a quan- tions [56, 57]. tum state is based on how much it spreads on a chosen By comparing the R´enyi occupations of the high- finite basis representation, as quantified by participation energy eigenstates of the Dicke model with the R´enyi oc- ratios [5, 27] and R´enyi entropies [28, 29]. In unbounded cupations of random states, as a function of α, we range continuous spaces, such measures can reach arbitrarily the eigenstates according to their degree of delocaliza- large values. If one wishes to investigate maximally de- tion. While all eigenstates are scarred [45], the most localized states in these systems, one must define a finite localized states are the ones scarred by unstable periodic region as a reference volume. There are multiple ways orbits with short periods. The analysis of the fourth- to select this bounded region, and each choice produces moment of the Husimi functions of these highly localized a different localization measure [30]. A natural volume states allow us to identify the classical unstable periodic orbits that cause their scarring. These orbits are different arXiv:2107.06894v1 [quant-ph] 14 Jul 2021 of reference is obtained by measuring the degree of lo- calization of a quantum state over the classical phase- from the ones found in [46] and their uncovering repre- space energy shell, as proposed in [30]. This measure of sents a step forward in the difficult task of identifying localization has been named R´enyi occupation, because the unstable periodic orbits underlying quantum scars. it is related to the exponential of the R´enyi-Wehrl en- In addition, we study the dynamical properties of the tropies [31, 32]. Quantum states can be represented in unveiled orbits and use them to explain the structures of the phase space through quasi-probability distributions, the eigenstates and their degree of localization. such as the Husimi function [33]. The R´enyi occupations The paper is organized as follows. The Dicke model of order α refer to averages of normalized α-moments of and its phase space are presented in Sec. II. In Sec. III, the Husimi functions. we introduce a measure of localization of quantum states In this work, we use maximally delocalized random based on the R´enyi occupations of order α and provide pure states and obtain a general analytical expression an analytical expression for maximally delocalized states. for the R´enyi occupations of order α. We employ this In Sec. IV we discuss quantum scarring in the light of the expression to analyze the eigenstates of the Dicke model R´enyi occupations and identify the unstable periodic or- 2 bits that scar the most localized eigenstates. We also system size, interpret our localization measure in terms of the geo- hx|HˆD|xi metric and dynamical properties of the classical periodic hcl(x) = = (2) orbits. Our conclusions are presented in Sec. V. j r ω ω P 2 + Q2 p2 + q2
+ 0 P 2 + Q2
+ 2γqQ 1 − − ω . 2 2 4 0 II. DICKE MODEL The bosonic Glauber and atomic Bloch coherent states are, respectively, h√ i 2 2 † The Dicke Hamiltonian is a fundamental model in |q, pi = e−(j/4)(q +p )e j/2(q+ip) aˆ |0i, (3) quantum optics that describes a single mode of the elec- 2 2 h √ i  P + Q j 2 2 ˆ tromagnetic field interacting with a set of N two-level |Q, P i = 1 − e (Q+iP )/ 4−P −Q J+ |j, −ji, atoms [34]. Setting ~ = 1, the Hamiltonian is given by 4 with |0i being the photon vacuum and |j, −ji the state † γ † with all the atoms in their ground state. HˆD = ωaˆ aˆ + ω0Jˆz + √ (ˆa +a ˆ)(Jˆ+ + Jˆ−), (1) N

B. Phase space wherea ˆ† (ˆa) is the bosonic creation (annihilation) oper- ator of the field mode and Jˆ+ (Jˆ−) is the raising (lower- The classical Hamiltonian hcl(x) has a four- ing) operator defined by Jˆ± = Jˆx ± iJˆy, where Jˆx,y,z = N dimensional phase space M in the coordinates x = P k 2 2 (1/2) k=1 σˆx,y,z are the collective pseudo-spin opera- (q, p; Q, P ), where P + Q ≤ 4, because the pseudo- tors satisfying the SU(2) algebra andσ ˆx,y,z are the Pauli spin degree of freedom is bounded. This space can be matrices. The Hamiltonian parameters are the radiation partitioned into a family of classical energy shells frequency of the electromagnetic field ω, the transition M() = {x ∈ M | h (x) = } frequency of the two-level atoms ω0, and the atom-field cl coupling strength γ. Since the Hamiltonian commutes in terms of the rescaled classical energy  = E/j. Due ˆ2 ˆ2 ˆ2 ˆ2 with the total pseudo-spin operator J = Jx + Jy + Jz , to energy conservation, for a given initial condition x ∈ the Hilbert space is separated into different invariant sub- M(), the classical evolution given by hcl remains in 2 spaces for each eigenvalue j(j +1) of Jˆ . We work within M(), that is, x(t) ∈ M() for all times t. the totally symmetric subspace that includes the ground- The classical energy shells are bounded with respect to state of the system and is defined by the maximum value the three-dimensional surface measure dx δ(hcl(x) − ). of the pseudo-spin length j = N /2. The model also The finite volume of M() is given by † ˆ ˆ iπ(ˆa aˆ+Jz +j) Z possesses a discrete parity symmetry, Π = e , 2 which leads to an additional subspace separation accord- dx δ(hcl(x) − ) = (2π~eff) ν(), (4) M ing to the eigenvalues of the parity operator Π± = ±1. where ~eff = 1/j is the effective Planck constant [65], The model has been used in studies of the quantum 2 which determines the volume of the Plank cell (2π~eff) , phase transition from a normal (γ < γc) to a superradiant and ν() is the semiclassical density of states obtained by (γ > γ ) phase, which arises when the coupling strength c √ taking only the first term in the Gutzwiller trace formula reaches the critical value γc = ωω0/2 [35, 36, 58, 59]. [62, 66, 67] (see App. A for more details on this quantity). The model also exhibits regular or chaotic behavior de- To calculate the average value, hfi, of a phase-space pending on the Hamiltonian parameters and excitation function f(x) over the classical energy shell M(), we in- energies [60]. In this paper, we choose the resonant fre- tegrate with respect to the surface measure dx δ(hcl(x)− quency case ω = ω0 = 1, system size j = N /2 = 100, ) and divide the result by the total volume (2π )2ν(), and γ = 2γ to work in the superradiant phase. We focus ~eff c on high energies, where the system displays hard-chaos hfi = f(x) x∈M() (5) behavior. 1 Z ≡ 2 dx δ(hcl(x) − )f(x). (2π~eff) ν() M

A. Classical Limit III. RENYI´ OCCUPATIONS

The classical Hamiltonian of the Dicke model is ob- The R´enyi entropy of order α [68], with α ≥ 0, that is, tained by taking the expectation value of the quantum   Hamiltonian HˆD under the tensor product of bosonic ∞ 1 X α Glauber and atomic Bloch coherent states, that is |xi = S (B, ρˆ) = log hφ |ρˆ|φ i , (6) α 1 − α  i i  |q, pi ⊗ |Q, P i [42, 51, 60–64], and dividing it by the i=1 3 is a common tool to measure the degree of delocalization by any realistic pure state and not even by random pure of a quantum stateρ ˆ over a given basis B = {|φii | i ∈ N} states, as explained in the subsection below. that forms a countable set of states. The quantity In this work, we explore how the R´enyi occupation de- Lα = exp(Sα) is the generalized participation ratio to pends on α and which information about the degree of the power 1/(α − 1), and it counts the effective number localization of the quantum state can be extracted from of states |φii that compose stateρ ˆ. The fact that B is different α’s. In general, α < 1 makes the Husimi distri- P∞ an orthonormal basis, ensures that i=1 hφi|ρˆ|φii = 1. bution look more homogeneous over the classical energy If B is an arbitrary countable set of states, the measure shell. In the limit α = 0, one has Lα can be generalized by performing an additional nor- malization as follows: D 0E Qρˆ h1i L (, ρˆ) =  =  = 1 (13) !1/(1−α) 0 0 P∞ α Q 1 i=1 hφi|ρˆ|φii ρˆ  Lα(B, ρˆ) = α . (7) P∞ hφ |ρˆ|φ i
i=1 i i for any state, because the Husimi function can be zero The same idea can be extended to the case of non- only on a zero-measure set of points. In contrast, for countable sets of states. Consider the set α > 1, the R´enyi occupation becomes less sensitive to the regions of the classical energy shell where the Husimi

C = {|xi | x ∈ M()} (8) function is relatively small. As α increases, these regions cease to contribute to Lα(, ρˆ), leaving only those regions of coherent states over a classical energy shell M() at where the Husimi function attains relatively large values, energy . In this case, a measure similar to that in Eq. (7) and thus decreasing the occupation value. We return to can be defined, but written in terms of the Husimi func- this discussion at the end of Sec. III C and on how to tion Qρˆ. This function is a quasi-distribution used to make use of the α-moments of the Husimi function in represent a quantum state in the phase space M, and is the analysis of scarring in Sec. IV. given by To benchmark the R´enyi occupations as a measure of localization, we describe below the behavior of Lα(, ρˆR) Qρˆ(x) = hx|ρˆ|xi ≥ 0 (9) for random pure statesρ ˆR = |ψRi hψR|, which are the most delocalized states over the phase space. For this, we for each point x ∈ M. need to investigate the Husimi moments hQα i , which ρˆR  For a non-countable set of states we can no longer per- we do first for systems with a finite Hilbert space, before form a discrete sum over the states of C, but we can in- proceeding with the Dicke model. stead use the three-dimensional surface measure of M() to perform an average over M(), as defined in Eq. (5), A. Maximally delocalized states in finite Hilbert hQαi = h hx|ρˆ|xiαi . (10) ρˆ  x∈M() spaces P∞ α Substituting the sums i=1 hφi|ρˆ|φii in Eq. (7) by these averages, we obtain the R´enyi occupations of or- It has been shown that for random pure states |ψRi der α ≥ 0 [30], completely spread in a Hilbert space of dimension N, the statistical averages h · i of the projections of |ψ i ψR R α !1/(1−α) into an arbitrary state |φi gives [32, 69, 70] Qρˆ  Lα(, ρˆ) = α . (11) Qρˆ  D 2αE Γ(N)Γ(1 + α) hψR|φi = (14) ψR Γ(N + α) In the limit α → 1, Eq. (11) gives where Γ is the gamma function. Applying this result to ! Qρˆ log Qρˆ coherent states |φi = |xi and performing an additional L (, ρˆ) = Q exp −  . (12) 1 ρˆ  average over all points x of the phase space M, we obtain Qρˆ    The name “R´enyi occupations” is inspired by the rela- D 2αE Γ(N)Γ(1 + α) hψR|xi = . (15) x∈M Γ(N + α) tionship of these measures with the exponential of the ψR R´enyi entropies. The values of Lα(, ρˆ) range from 0 to 1, and they indicate the percentage of the classical en- As N increases, one expects that that the variance of the ergy shell that is occupied by the quantum state. The averages h · i will decrease [45]. Thus, for sufficiently ψR R´enyi occupation of a quantum state equals 1 if the cor- large N, a single, but typical, random state |ψRi will responding Husimi function is constant in the classical satisfy energy shell, which means that the state is uniformly delocalized, covering the classical energy shell homoge- D 2αE Γ(N)Γ(1 + α) hψR|xi = . (16) neously. Notice, however, that this value is not reached x∈M Γ(N + α) 4

Moreover, in the limit of large N, we can do the approx- A state that is as delocalized as a random pure state α imation Γ(N + α)/Γ(N) ≈ N , which leads to the result gives Λα ≈ 1, which is the lower bound for this measure. As the stateρ ˆ becomes more localized in the classical 1/(1−α) D 2αE  energy shell at ,Λ (, ρˆ) increases. For a given α, the hψ |xi α R value Λ (, ρˆ) = n indicates that the stateρ ˆ is n times  x∈M  ≈ Γ(1 + α)1/(1−α), (17) α  D 2Eα  more localized than a random state. hψR|xi x∈M 5 that is independent of the dimension N.

B. Maximally delocalized states in the Dicke model 4 In analogy to Eq. (17), we say that an arbitrary state ρˆ = |ψihψ| in the unbounded space of the Dicke model is maximally delocalized if

max 1/(1−α) Lα(, ρˆ) ≈ Lα ≡ Γ(1 + α) . (18)

This maximum level of delocalization is attained, on aver- age, by random pure states, although several eigenstates 2 of the Dicke model are maximally delocalized as well. The reason why we can extrapolate the result in Eq. (17), valid for large finite-dimensional Hilbert spaces, to the infinite-dimensional Hilbert space of the Dicke model is because we perform averages over individual 1 classical energy shells, which are bounded. Their finite volume in phase space induces a large but finite effective 0 1 3 4 dimension that allows us to treat the Hilbert space of the

Dicke model as if it were finite dimensional [71]. FIG. 1. Localization measure Λα(k, ρˆk) for the 2437 eigen- In our studies of the Dicke model in Ref. [45], we found statesρ ˆk = |EkihEk| with eigenenergies k inside the energy max numerically that L2 = 1/2 for pure random states. interval [−0.6, −0.4]; j = 100. The 8 states labeled A-H are In Ref. [72], an approximate maximum value of 0.7 was further analyzed in Fig. 2. found numerically in billiards for a measure similar to L1. These values are obtained directly from Eq. (18), which max −1 max In Fig. 1, we study Λα(k, ρˆk) as a function of α for gives L2 = Γ(3) = 1/2 and L1 = limα→1 Γ(1 + all the eigenstatesρ ˆ = |E ihE | of the Dicke model α)1/(1−α) ≈ 0.66. These results reinforce the validity of k k k with eigenenergies k = Ek/j inside the energy interval the expression (18) and suggest that it may be applicable [−0.6, −0.4], where chaos is predominant. Each eigen- to other models. state is represented by a thin gray line. One sees that Notice that the upper bound for the degree of delocal- max most eigenstates cluster slightly above Λα = 1, indicat- ization of pure states given by Lα is not equal to one, ing that they are nearly as delocalized as random pure instead Lmax < 1 for α > 0. This happens because quan- α states. Values of Λα < 1 are possible, but disappear as j tum interference effects prevent a quantum pure state increases [45]. A portion of the eigenstates has values of from homogeneously covering the phase space and cause Λα that are much higher than 1, some reaching Λ4 = 5. its Husimi function to be zero in some points of the phase As we show next, these high values of localization are space [73]. Only mixed states, which can be obtained caused by strong quantum scarring. The thick colored by performing infinite-time averages, can homogeneously lines in Fig. 1 mark 8 eigenstates, labeled A-H, that we cover the classical energy shells [45]. select for further analysis. Visualizing the moments of the four-dimensional Husimi function comes to our aid for explaining the dif- C. Eigenstates ferent degrees of localization of the eigenstates shown in Fig. 1. We consider the projection into the atomic (Q, P ) To quantify how close the degree of delocalization of plane of the α-moments of the Husimi function Qk ofρ ˆk a quantum state is to the maximal value attained by intersected with the classical energy shell at k, so that random states, we define the following measure: we obtain two-dimensional pictures, that is,

Lmax ZZ α α α Λα(, ρˆ) = . (19) Qfk (Q, P ) = dq dp δ(hcl(x) − k) Qk(x) . (20) Lα(, ρˆ) 5

1 In Refs. [45, 46], we studied Qfk(Q, P ) and verified that OA OB OC OD, OD it allows for the visual identification of quantum scars. In Ref. [45], we concluded that all the eigenstates of the λ 0.143 0.191 0.253 0.153 Dicke model are scarred, even though most of them are almost as delocalized as random states. This is corrob- orated again by the representative eigenstates shown in T λ 1.86 2.97 3.43 2.21 Fig. 2. α In Fig. 2, we compare Qfk (Q, P ) for the 8 eigenstates A- Pk 33.2 31.1 28.3 22.5 H from Fig. 1 and also Qα (Q, P, ) = RR dq dp δ(h (x)− gρˆR cl α TABLE I. Lyapunov exponent λ, its product with the period ) QρˆR (x) for a random stateρ ˆR centered at energy . This state is obtained by weighting the positive par- T , and the scarring measure Pk(O) (22) for the orbits A-D. ity eigenstates inside a rectangular energy window with real coefficients normally distributed as in the Gaus- sian orthogonal ensembles of random matrix theory. For ordinates xi ∈ M(k) where the Husimi function attains each state, we show the distributions for five values of a maximum, thus yielding a set of possible initial condi- α ∈ {0.5, 1, 2, 3, 4}. tions for a periodic orbit on all variables xi. We evolve In contrast to the random state [Fig. 2 (R0)-(R4)], these initial conditions and select those that roughly fol- the projected moments of the Husimi functions of the low the green lines in Figs. 2 (A4), (B4), (C4), and (D4). eigenstates [Fig. 2 (A0)-(H4)] always display structures The evolution is halted when the evolved point xi(Ti) ap- that are related with underlying unstable periodic orbits, proximately returns to the initial condition xi(0). This and therefore imply that the eigenstates are scarred [45]. gives an approximate period Ti and an initial condition When examining the figure, one should keep in mind xi, which can be converged to a truly periodic condi- that values of α < 1 tend to homogenize the distribution tion by means of an algorithm known as the monodromy α 0 method [46, 74–76]. This procedure leads to the five un- Qfk (Q, P ), up to the limiting case where Qfk(Q, P ) = 1 in- dependently of Q, P . On the other hand, values of α > 1 stable periodic orbits labeled OA, OB, OC, OD, and OD tend to erase the small contributions from the Husimi in Figs. 2 (A2), (B2), (C2), and (D2). The mirrored α orbit OD is obtained by changing the signs of the coor- function, so that regions where Qfk (Q, P ) is large get en- hanced. The high moments of the Husimi function are dinates Q and q of OD. We only mirror OD, because the therefore useful tools in the analysis of quantum scarring, other orbits are symmetric under the parity transforma- as discussed next. tion (Q, q) 7→ (−Q, −q). We stress that these unstable periodic orbits are different from those that we found in Ref. [46], whose origin could be traced down to the fun- IV. QUANTUM SCARRING AND UNSTABLE damental excitations around the ground state configura- PERIODIC ORBITS tion. Further studies of these new orbits may eventually reveal the properties and origin of these new families of Quantum scarring corresponds to the concentration of unstable periodic orbits for the Dicke model. a quantum state around the phase-space region occupied The Lyapunov times 1/λ of the unstable periodic or- by a periodic orbit bits identified here are smaller but of the same order of their respective periods T . Table I gives the values of the O = {x(t) | t ∈ [0,T ]} ⊆ M() (21) Lyapunov exponents, λ, and the periods divided by the Lyapunov times, T λ, for each of the five unstable peri- with a periodic initial condition x(T ) = x(0) that is odic orbits that we found. The competition between the unstable, that is, has a positive Lyapunov exponent λ. period of the orbit and the Lyapunov time is important These orbits are always present in chaotic systems and for scarring. As originally discussed in Ref. [77], a non- are deeply connected with the quantum spectrum of such stationary state will be affected by a neighboring periodic systems [66, 67]. orbit if the Lyapunov time of the orbit is larger than its period. If instead the Lyapunov time is shorter than the period, the periodic orbit does not have time to develop A. Identifying Unstable Periodic Orbits and scar the non-stationary state. Here, we observe that even if this condition is slightly broken, T λ & 1, the pe- By tracking the peaks of the projected fourth-moment riodic orbits still cause a significant localization of the of the Husimi function displayed in Figs. 2 (A4), (B4), eigenstates. (C4), and (D4), we can uncover the unstable periodic or- To verify that the states A-D are indeed scarred by the bits that significantly scar the states A-D. The procedure unstable periodic orbits OA-OD, we employ the following goes as follows. Those peaks give a set of possible initial scarring measure [46]: conditions for periodic orbits on the variables Qi,Pi. By varying the bosonic coordinate p and selecting q accord- tr(ˆρkρˆO) Pk(O) = , (22) ing to the fixed energy, we identify the whole set of co- tr(ˆρρˆO) 6

Eigenstates

0.8

-0.8

0.8

-0.8

0.8

-0.8

0.8

-0.8

0.8

-0.8

0.8

-0.8

0.8

-0.8

0.8

-0.8

Random State

0.8

-0.8

-1 1 -1 1 -1 1 -1 1 -1 1

FIG. 2. Projected moments of the Husimi functions, Qα(Q, P ), for the states (A)-(H) from Fig. 1 and Qα (Q, P, ) for a pure fk gρˆR random state (R) from a Gaussian orthogonal ensemble obtained by weighting the positive parity eigenstates inside an energy window of width 1.5 centered at the energy  = −0.5. Each column corresponds to a different value of α ∈ {0.5, 1, 2, 3, 4}. The solid red line in (A2)-(D2) represent the projections in Q, P of the unstable periodic orbits OA, OB, OC, OD, and the dashed red line in (D2) is for the mirrored orbit OD. 7 where shapes seen in Figs. 3 (A1)-(D1) are precisely the orbits in Figs. 3 (A2)-(D2) projected into the (Q, P )-plane. Z T 1 Figure 3 contains only semiclassical information, which ρˆO = dt x(t) x(t) (23) T 0 helps us understand specific features and the degree of localization of the eigenstates A-D. For example, when is a tubular state composed of all coherent states |x(t)i the classical dynamics is fast, the scars are less intense. In whose centers lie in the periodic orbit O = {x(t) | Figs. 3 (A1)-(D1), we place red arrows at constant time t ∈ [0,T )} andρ ˆ = h|xihx|ix∈M() is a mixed state intervals along the corresponding periodic orbits. The completely delocalized in the classical energy shell at closer those arrows are, the slower the classical dynamics  = hcl(O), that is, a mixture of all coherent states whose are. One sees that the brighter green regions correspond centers lie within this classical energy shell (see details to the regions with a high density of red arrows. These in Ref. [46]). A value Pk(O) = n indicates that the state are regions of longer permanence of the classical orbit, ρˆk = |EkihEk| is n times more likely to be found near the which produce deeper imprinted scars. orbit O than a completely delocalized state, for which The dynamical properties of the orbits also explain P (O) = 1. The results of the scarring measure obtained k why Λα increase with α for some eigenstates. The faster with Eq. (22) for the states A-D are shown in Table I and the dynamics are in a given region of the phase space, confirm that these states are significantly scarred by the the less probable it is to find the orbit there. This gets unstable periodic orbits OA-OD. reflected in the high-moments of the Husimi function, One observes secondary structures in Figs. 2 (A0), where small contributions tend to be erased, increasing (B0), (C0), and (D0) that are not directly generated by Λα. the unstable periodic orbits delineated in Figs. 2 (A2), The general features discussed in the two previous (B2), (C2), and (D2). These structures may be related to paragraphs are described below in more detail for the the self-interferences of the unstable periodic orbits [78] eigenstates A-D. or to other secondary scars associated with other unsta- ble periodic orbits that we have not identified. We leave • The unstable periodic orbit OA [Fig. 3 (A1)], which this question for future investigations. heavily scars state A, has slower dynamics near the small oval region in the center of the orbit, where most red arrows are concentrated. Compar- B. Highly localized eigenstates and scarring ing OA with OB, one sees that OA is larger in phase space, so state A is less localized than state B and The dynamical properties of the unstable periodic or- Λα(A) < Λα(B) for α < 3, as shown in Fig. 1. Nev- bits that scar states A-D allow us to explain the struc- ertheless, α = 4 is sufficiently large to erase the less tures of the distributions in Figs. 2 (A0)-(D4). To do this, bright regions of OA, where the classical dynamics we consider the Husimi function of the tubular stateρ ˆO, is fast, leaving only the contributions from the cen- tral oval region and thus leading to a higher value 1 Z T 2
of localization and to Λα(A) > Λα(B) for α > 3. QO(x) = hx|ρˆO|xi = dt x y(t) y(t) ∈ O , T 0 (24) • The unstable periodic orbit OB covers only a small and compute the projection of this function into the portion of the classical energy shell and has a rela- (Q, P ) plane, tively constant speed in the phase space, as seen by the constant density of red arrows in Fig. 3 (B1), so ZZ the state B maintains a high value of localization Q (Q, P ) = dq dp δ(h (x) − ) Q (x), (25) eO cl O in Fig. 1 for all values of α ∈ [0, 4]. at energy  = hcl(O). In Figs. 3 (A1), (B1), and (C1) • The unstable periodic orbit OC presents two bright we plot in green the projections QeO for the orbits OA, regions in Fig. 3 (C1), at Q = 0 and P ≈ ±0.8, OB, and OC, respectively, and in Fig. 3 (D1) we plot where the dynamics is slow. As seen in Fig. 1, state the added projections (Q + Q )/2 for orbit O and C has a lower value of Λα than state B, because eOD eOD D OC is larger in phase-space than OB. However, the the mirrored orbit OD. One sees that the projections slope of Λα at α = 4 is larger for state C than for QeO(Q, P ) of the Husimi function of the tubular state state B, because the less bright loops of OC, where in Figs. 3 (A1), (B1), (C1), and (D1) are similar to the the classical dynamics is fast, disappear for α = 4 α projections Qfk (Q, P ) of the Husimi function of the eigen- [Fig. 2 (C4)], leaving only the two bright spots at states A-D shown in Figs. 2 (A2)-(D2), Figs. 2 (A3)-(D3), Q = 0. Figs. 2 (A4)-(D4). In Figs. 3 (A2)-(D2), we show three-dimensional plots • The behavior of Λα for state D in Fig. 1 is peculiar of the four-dimensional periodic orbits OA, OB, OC, and in that it initially grows for α < 2, but flattens both OD and OD. The plots display the variables Q, P , after α > 2. The localization measure increases as and p, and the color represents the fourth variable q. The α grows to 2, because the contribution of the fast 8

0.8

-0.8

-1 1 -1 1 -1 1 -1 1

3

2

1

0

-1

-2

-3

FIG. 3. Top panels: in green, the Husimi projection QeO for the orbits OA (A1), OB (B1), OC (C1), and added projections (QO + Q )/2 for orbit OD and the mirrored orbit OD (D1). The red arrows are placed at constant time intervals along the e D eOD unstable periodic orbits OA (A1), OB (B1), OC (C1), and OD, OD (D1). Bottom panels: Three-dimensional plots of the POs OA (A2), OB (B2), OC (C2), and OD (opaque) OD (translucent) (D2). In these panels, the variables Q, P , and p correspond to each of the three axis, as marked, and the color represents the fourth variable q.

outer loops strongly diminish. But then, for α > 2, patterns, that is, they exhibit closed loops. These pat- since the dynamics in the loops of intermediate sizes terns must belong to a set of periodic orbits that scar the has a relatively constant speed, Λα flattens. In fact, eigenstates [45], but we have not yet been able to identify Fig. 1 suggests that the flattening for large α holds these orbits with the numerical method that we have im- also for state B, while Λα for states A and C keeps plemented. They must have periods that are much larger growing at least up to α = 4. than their corresponding Lyapunov times, which makes them more difficult to find. Since these states are scarred We close this subsection by discussing the state E, by unstable periodic orbits of larger periods, they are which in terms of localization is midway between the more delocalized than the states A-D, which explains why highly localized eigenstates A-D and the maximally delo- Λα(F ), Λα(G), Λα(H) < Λα(A), Λα(B), Λα(C), Λα(D). calized eigenstates F-H described in the next subsection. The results in this work support our claims in Ref. [45] The projected Husimi moment distributions depicted in that all eigenstates are scarred by unstable periodic or- Figs. 2 (E0)-(E4) show localization around some regions bits, yet some of them are as delocalized in phase space as that must be associated with a classical unstable peri- random states. Identifying the unstable periodic orbits odic orbit with slow dynamics at Q ≈ 0 and P ≈ ±0.9. that scar the states E-H would bring further confirmation However, the numerical method used to identify the or- to this conjecture. bits that scar the eigenstates A-D fails for state E. This suggests that the ratio between the period and the Lya- punov time, T λ, of the unstable periodic orbit that scars this eigenstate should be much larger than one. V. CONCLUSIONS

R´enyi entropies of order α quantify the degree of delo- C. Maximally delocalized eigenstates calization of quantum states in Hilbert space. Our inter- est in this work was instead in the structure of quantum As seen in Fig. 1, the eigenstates F, G and H have states in phase space. For this analysis, we employed a values of Λα ≈ 1, similar to those of random states. measure of localization called R´enyi occupations of or- However, contrary to Figs. 2 (R0)-(R4), the panels (F0)- der α, which is defined over classical energy shells and is (H4) show that these eigenstates still display orbit-like based on the α-moments of the Husimi functions. The 9 latter are distributions that represent a quantum state D´ıaz,and Eduardo Murrieta. SP-C, DV, and JGH ac- in phase space. We showed that the R´enyi occupations knowledge financial support from the DGAPA- UNAM of random states have a simple analytical form that is a project IN104020, and SL-H from the Mexican CONA- function of α only. We used this result to define a lo- CyT project CB2015-01/255702. LFS was supported by calization measure given by the ratio of the energy-shell the NSF Grant No. DMR-1936006. occupation of random states to that of the state under investigation. This measure was then used to analyze the eigenstates of the Dicke model. Appendix A: Semiclassical Density of States As α increases beyond 1, the smaller contributions from the Husimi function are erased and only the larger The semiclassical approximation of the density of peaks survive. By examining these peaks in the fourth- states of a quantum system is obtained by taking the first moment of the Husimi function of highly localized eigen- term of the well-known Gutzwiller trace formula [66, 67]. states, we were able to find the classical unstable periodic This semiclassical density of states is the ratio of the orbits that scar those states. three dimensional volume of the classical energy shell to Our study of the classical dynamics of the identified the four dimensional volume of the Planck cell, orbits assisted our understanding of the structure and the degree of localization of the eigenstates. In regions 1 Z of the phase space where the classical dynamics is slow, ν() = 2 dx δ(hcl(x) − ). (A1) (2π~eff) M the scars are deeper imprinted, so the values of the α- The explicit expression of ν() for the Dicke model was moments of the Husimi functions are larger and persist derived in Ref. [62]. Here, this expression is slightly mod- as α increases. ified to be consistent with the notation used in this article Highly localized states are scarred by orbits of rela- and it reads tively short period and the peaks in the higher moments  of their Husimi distributions provide a powerful tool to 1 R y+  dyf(y, ) if 0 ≤  ≤ −ω0, identify classical unstable periodic orbits. This is a great 2j2  π y− 1+/ω0 1 R y+ ν() = + dyf(y, ) if || < ω0, accomplishment given the difficulty in identifying the un- ω 2 π /ω0  stable periodic orbits underlying scarred states. This  1 if  ≥ ω0, shows that quantum states supply valuable information (A2) about the non-linear dynamics of the classical limit of the model. where The localization measure introduced in this work ap- s  plies to other models with unbounded phase space, such 2 2γc (y − /ω0) as quantum billiards, and our technique to find classi- f(y, ) = arccos   , (A3) cal unstable periodic orbits could be extended to those γ2(1 − y2) systems as well. and

ACKNOWLEDGMENTS  s  ! γ γ 2( −  ) 1 γ2 γ2 y = − c c ∓ 0 ,  = − c + . ± γ  γ ω  0 2 γ2 γ2 We acknowledge the support of the Computation Cen- 0 c ter - ICN, in particular to Enrique Palacios, Luciano (A4)

[1] P. W. Anderson, “Absence of diffusion in certain random [7] Giulio Casati, B. V. Chirikov, and D. L. Shepelyansky, lattices,” Phys. Rev. 109, 1492–1505 (1958). “Quantum limitations for chaotic excitation of the Hy- [2] P. A. Lee and T. V. Ramakhrishnan, “Disordered elec- drogen atom in a monochromatic field,” Phys. Rev. Lett. tronic systems,” Rev. Mod. Phys. 57, 287 (1985). 53, 2525–2528 (1984). [3] B.V. Chirikov, F. M. Izrailev, and D. L. Shepelyansky, [8] G. Casati, B.V. Chirikov, D. L. Shepelyansky, and “Dynamical stochasticity in classical and quantum me- I. Guarnesi, “Relevance of classical chaos in quantum me- chanics,” Sov. Scient. Rev. C 2, 209–267 (1981). chanics: The Hydrogen atom in a monochromatic field,” [4] Shmuel Fishman, D. R. Grempel, and R. E. Prange, Phys. Rep. 154, 77–123 (1987). “Chaos, quantum recurrences, and Anderson localiza- [9] R. Bl¨umeland U. Smilansky, “Localization of Floquet tion,” Phys. Rev. Lett. 49, 509–512 (1982). states in the rf excitation of Rydberg atoms,” Phys. Rev. [5] F. M. Izrailev, “Simple models of quantum chaos: Spec- Lett. 58, 2531–2534 (1987). trum and eigenfunctions,” Phys. Rep. 196, 299–392 [10] Fausto Borgonovi, Giulio Casati, and Baowen Li, “Dif- (1990). fusion and localization in chaotic billiards,” Phys. Rev. [6] S. Fishman, “Anderson localization and quantum chaos Lett. 77, 4744–4747 (1996). maps,” Scholarpedia 5, 9816 (2010), revision #186577. 10

[11] Benjamin Batisti´c and Marko Robnik, “Semiempiri- [29] Y. Y. Atas and E. Bogomolny, “Calculation of multi- cal theory of level spacing distribution beyond the fractal dimensions in spin chains,” Phil. Trans. R. Soc. Berry–Robnik regime: modeling the localization and the A 372 (2014), 10.1098/rsta.2012.0520. tunneling effects,” Jour. Phys. A 43, 215101 (2010). [30] D. Villase˜nor,S. Pilatowsky-Cameo, M. A. Bastarrachea- [12] Benjamin Batisti´cand Marko Robnik, “Dynamical local- Magnani, S. Lerma-Hern´andez, and J. G. Hirsch, “Quan- ization of chaotic eigenstates in the mixed-type systems: tum localization measures in phase space,” Phys. Rev. E spectral statistics in a billiard system after separation of 103, 052214 (2021). regular and chaotic eigenstates,” J. Phys. A 46, 315102 [31] Alfred Wehrl, “General properties of entropy,” Rev. Mod. (2013). Phys. 50, 221–260 (1978). [13] Benjamin Batisti´cand Marko Robnik, “Quantum local- [32] Sven Gnutzmann and Karol Zyczkowski, “R´enyi-Wehrl ization of chaotic eigenstates and the level spacing distri- entropies as measures of localization in phase space,” J. bution,” Phys. Rev. E 88, 052913 (2013). Phys. A 34, 10123–10139 (2001). [14] Max D. Porter, Aaron Barr, Ariel Barr, and L. E. Re- [33] KˆodiHusimi, “Some Formal Properties of the Density ichl, “Chaos in the band structure of a soft Sinai lattice,” Matrix,” Proc. Phys.-Math. Soc. of Jap. 22, 264–314 Phys. Rev. E 95, 052213 (2017). (1940). [15] Benjamin Batisti´c, Crtˇ Lozej, and Marko Robnik, “Sta- [34] R. H. Dicke, “Coherence in spontaneous radiation pro- tistical properties of the localization measure of chaotic cesses,” Phys. Rev. 93, 99 (1954). eigenstates and the spectral statistics in a mixed-type [35] Klaus Hepp and Elliott H Lieb, “On the superradiant billiard,” Phys. Rev. E 100, 062208 (2019). phase transition for molecules in a quantized radiation [16] Marko Robnik, “Recent advances in quantum chaos of field: the Dicke maser model,” Ann. Phys. (N.Y.) 76, generic systems,” in Synergetics, edited by Axel Hutt and 360 – 404 (1973). Hermann Haken (Springer US, 2020) pp. 133–148. [36] Y. K. Wang and F. T. Hioe, “Phase transition in the [17] G. Casati, B. V. Chirikov, I. Guarneri, and F. M. Dicke model of superradiance,” Phys. Rev. A 7, 831–836 Izrailev, “Band-random-matrix model for quantum lo- (1973). calization in conservative systems,” Phys. Rev. E 48, [37] Barry M. Garraway, “The Dicke model in quantum op- R1613–R1616 (1993). tics: Dicke model revisited,” Philos. Trans. Royal Soc. A [18] L. F. Santos, G. Rigolin, and C. O. Escobar, “Entan- 369, 1137 (2011). glement versus chaos in disordered spin systems,” Phys. [38] Peter Kirton, Mor M. Roses, Jonathan Keeling, and Rev. A 69, 042304 (2004). Emanuele G. Dalla Torre, “Introduction to the Dicke [19] L. F. Santos, M. I. Dykman, M. Shapiro, and F. M. model: From equilibrium to nonequilibrium, and vice Izrailev, “Strong many-particle localization and quantum versa,” Adv. Quantum Technol. 2, 1800043 (2019). computing with perpetually coupled qubits,” Phys. Rev. [39] Jorge Ch´avez-Carlos, B. L´opez-del-Carpio, Miguel A. A 71, 012317 (2005). Bastarrachea-Magnani, Pavel Str´ansk´y, Sergio Lerma- [20] D. M. Basko, I. L. Aleiner, and B. L. Altshuler, “Metal- Hern´andez,Lea F. Santos, and Jorge G. Hirsch, “Quan- insulator transition in a weakly interacting many-electron tum and classical Lyapunov exponents in atom-field in- system with localized single-particle states,” Ann. Phys. teraction systems,” Phys. Rev. Lett. 122, 024101 (2019). 321, 1126 (2006). [40] R. J. Lewis-Swan, A. Safavi-Naini, J. J. Bollinger, and [21] Vadim Oganesyan and David A. Huse, “Localization of A. M. Rey, “Unifying , thermalization and entanglement interacting fermions at high temperature,” Phys. Rev. B through measurement of fidelity out-of-time-order corre- 75, 155111 (2007). lators in the Dicke model,” Nat. Comm. 10, 1581 (2019). [22] F. Dukesz, M. Zilbergerts, and L. F. Santos, “Interplay [41] Sa´ulPilatowsky-Cameo, Jorge Ch´avez-Carlos, Miguel A. between interaction and (un)correlated disorder in one- Bastarrachea-Magnani, Pavel Str´ansk´y, Sergio Lerma- dimensional many-particle systems: delocalization and Hern´andez,Lea F. Santos, and Jorge G. Hirsch, “Posi- global entanglement,” New J. Phys. 11, 043026 (2009). tive quantum Lyapunov exponents in experimental sys- [23] Achilleas Lazarides, Arnab Das, and Roderich Moessner, tems with a regular classical limit,” Phys. Rev. E 101, “Fate of many-body localization under periodic driving,” 010202(R) (2020). Phys. Rev. Lett. 115, 030402 (2015). [42] M.A.M de Aguiar, K Furuya, C.H Lewenkopf, and M.C [24] Pedro Ponte, Z. Papi´c, Fran¸cois Huveneers, and Nemes, “Chaos in a spin-boson system: Classical analy- Dmitry A. Abanin, “Many-body localization in period- sis,” Ann. Phys. 216, 291 – 312 (1992). ically driven systems,” Phys. Rev. Lett. 114, 140401 [43] K Furuya, M.A.M de Aguiar, C.H Lewenkopf, and M.C (2015). Nemes, “Husimi distributions of a spin-boson system and [25] Michele Fava, Rosario Fazio, and Angelo Russomanno, the signatures of its classical dynamics,” Ann. of Phys. “Many-body dynamical localization in the kicked Bose- 216, 313–322 (1992). Hubbard chain,” Phys. Rev. B 101, 064302 (2020). [44] L. Bakemeier, A. Alvermann, and H. Fehske, “Dynamics [26] Colin Rylands, Efim B. Rozenbaum, Victor Galitski, and of the Dicke model close to the classical limit,” Phys. Rev. Robert Konik, “Many-body dynamical localization in a A 88, 043835 (2013). kicked Lieb-Liniger gas,” Phys. Rev. Lett. 124, 155302 [45] Sa´ul Pilatowsky-Cameo, David Villase˜nor, Miguel A. (2020). Bastarrachea-Magnani, Sergio Lerma-Hern´andez,Lea F. [27] J T Edwards and D J Thouless, “Numerical studies of Santos, and Jorge G. Hirsch, “Ubiquitous quantum scar- localization in disordered systems,” J. Phys. C 5, 807– ring does not prevent ergodicity,” Nat. Commun. 12 820 (1972). (2021), 10.1038/s41467-021-21123-5. [28] Y. Y. Atas and E. Bogomolny, “Multifractality of eigen- [46] Sa´ul Pilatowsky-Cameo, David Villase˜nor, Miguel A. functions in spin chains,” Phys. Rev. E 86, 021104 Bastarrachea-Magnani, Sergio Lerma-Hern´andez,Lea F. (2012). Santos, and Jorge G. Hirsch, “Quantum scarring in a 11

spin-boson system: fundamental families of periodic or- Localization-delocalization in the Husimi distributions,” bits,” New J. of Phys. 23, 033045 (2021). EPL (Europhys. Lett.) 15, 125 (1991). [47] Alexander Altland and Fritz Haake, “Equilibration and [62] M. A. Bastarrachea-Magnani, S. Lerma-Hern´andez, and macroscopic quantum fluctuations in the Dicke model,” J. G. Hirsch, “Comparative quantum and semiclassical New J. Phys. 14, 073011 (2012). analysis of atom-field systems. I. Density of states and [48] Michal Kloc, Pavel Str´ansk´y, and Pavel Cejnar, “Quan- excited-state quantum phase transitions,” Phys. Rev. A tum quench dynamics in Dicke superradiance models,” 89, 032101 (2014). Phys. Rev. A 98, 013836 (2018). [63] M. A. Bastarrachea-Magnani, S. Lerma-Hern´andez, and [49] Sergio Lerma-Hern´andez, Jorge Ch´avez-Carlos, J. G. Hirsch, “Comparative quantum and semiclassical Miguel A. Bastarrachea-Magnani, Lea F. Santos, analysis of atom-field systems. II. Chaos and regularity,” and Jorge G. Hirsch, “Analytical description of the sur- Phys. Rev. A 89, 032102 (2014). vival probability of coherent states in regular regimes,” [64] Miguel Angel Bastarrachea-Magnani, Baldemar L´opez- J. Phys. A 51, 475302 (2018). del-Carpio, Sergio Lerma-Hern´andez, and Jorge G [50] S. Lerma-Hern´andez,D. Villase˜nor,M. A. Bastarrachea- Hirsch, “Chaos in the Dicke model: quantum and semi- Magnani, E. J. Torres-Herrera, L. F. Santos, and J. G. classical analysis,” Phys. Scripta 90, 068015 (2015). Hirsch, “Dynamical signatures of quantum chaos and re- [65] A. D. Ribeiro, M. A. M. de Aguiar, and A. F. R. laxation time scales in a spin-boson system,” Phys. Rev. de Toledo Piza, “The semiclassical coherent state propa- E 100, 012218 (2019). gator for systems with spin,” J. Phys. A 39, 3085 (2006). [51] David Villase˜nor, Sa´ul Pilatowsky-Cameo, Miguel A [66] Martin C. Gutzwiller, “Periodic orbits and classical quan- Bastarrachea-Magnani, Sergio Lerma, Lea F Santos, and tization conditions,” J. Math. Phys. 12, 343–358 (1971). Jorge G Hirsch, “Quantum vs classical dynamics in a [67] M. C. Gutzwiller, Chaos in classical and quantum me- spin-boson system: manifestations of spectral correla- chanics (Springer-Verlag, New York, 1990). tions and scarring,” New J. Phys. 22, 063036 (2020). [68] Alfr´ed R´enyi, “On measures of entropy and informa- [52] Qian Wang and Marko Robnik, “Statistical properties tion,” in Proceedings of the Fourth Berkeley Symposium of the localization measure of chaotic eigenstates in the on Mathematical Statistics and Probability, Volume 1: Dicke model,” Phys. Rev. E 102, 032212 (2020). Contributions to the Theory of Statistics (University of [53] J Cohn, A Safavi-Naini, R J Lewis-Swan, J G Bohnet, California Press, 1961) pp. 547–561. M G¨arttner,K A Gilmore, J E Jordan, A M Rey, J J [69] M Ku´s, J Mostowski, and F Haake, “Universality of Bollinger, and J K Freericks, “Bang-bang shortcut to eigenvector statistics of kicked tops of different symme- adiabaticity in the Dicke model as realized in a penning tries,” J. of Phys. A 21, L1073–L1077 (1988). trap experiment,” New J. Phys. 20, 055013 (2018). [70] K R W Jones, “Entropy of random quantum states,” J. [54] A. Safavi-Naini, R. J. Lewis-Swan, J. G. Bohnet, Phys. A 23, L1247–L1251 (1990). M. G¨arttner,K. A. Gilmore, J. E. Jordan, J. Cohn, J. K. [71] This is further discussed in a paper in preparation. Freericks, A. M. Rey, and J. J. Bollinger, “Verification [72] B. Batistic, Cˇ Lozej, and M. Robnik, “The distribution of a many-ion simulator of the Dicke model through slow of localization measures of chaotic eigenstates in the sta- quenches across a phase transition,” Phys. Rev. Lett. dium billiard,” Nonlinear Phenomena in Complex Sys- 121, 040503 (2018). tems 23, 17–32 (2020). [55] Tuomas Jaako, Ze-Liang Xiang, Juan Jos´eGarcia-Ripoll, [73] H J Korsch, C M¨uller, and H Wiescher, “On the zeros and Peter Rabl, “Ultrastrong-coupling phenomena be- of the Husimi distribution,” J. Phys. A 30, L677–L684 yond the Dicke model,” Phys. Rev. A 94, 033850 (2016). (1997). [56] Markus P. Baden, Kyle J. Arnold, Arne L. Grimsmo, [74] M. Baranger, K.T.R. Davies, and J.H. Mahoney, “The Scott Parkins, and Murray D. Barrett, “Realization calculation of periodic trajectories,” Ann. of Phys. 186, of the Dicke model using cavity-assisted Raman transi- 95–110 (1988). tions,” Phys. Rev. Lett. 113, 020408 (2014). [75] M.A.M. de Aguiar and C.P. Malta, “Isochronous and pe- [57] Zhiqiang Zhang, Chern Hui Lee, Ravi Kumar, K. J. riod doubling bifurcations of periodic solutions of non- Arnold, Stuart J. Masson, A. L. Grimsmo, A. S. Parkins, integrable Hamiltonian systems with reflexion symme- and M. D. Barrett, “Dicke-model simulation via cavity- tries,” Physica D 30, 413–424 (1988). assisted Raman transitions,” Phys. Rev. A 97, 043858 [76] N. S. Simonovi´c,“Calculations of periodic orbits: The (2018). monodromy method and application to regularized sys- [58] Klaus Hepp and Elliott H. Lieb, “Equilibrium statisti- tems,” Chaos 9, 854–864 (1999). cal mechanics of matter interacting with the quantized [77] Eric J. Heller, “Bound-state eigenfunctions of classically radiation field,” Phys. Rev. A 8, 2517–2525 (1973). chaotic Hamiltonian systems: Scars of periodic orbits,” [59] Clive Emary and Tobias Brandes, “Chaos and the quan- Phys. Rev. Lett. 53, 1515–1518 (1984). tum phase transition in the Dicke model,” Phys. Rev. E [78] Michael Victor Berry, “Quantum scars of classical closed 67, 066203 (2003). orbits in phase space,” Proc. of the R. Soc. of London. [60] J. Ch´avez-Carlos, M. A. Bastarrachea-Magnani, A. 423, 219–231 (1989). S. Lerma-Hern´andez, and J. G. Hirsch, “Classical chaos in atom-field systems,” Phys. Rev. E 94, 022209 (2016). [61] M. A. M. de Aguiar, K. Furuya, C. H. Lewenkopf, and M. C. Nemes, “Particle-spin coupling in a chaotic system: