Order, chaos and quasi symmetries in a first-order quantum phase transition A Leviatan and M Macek 4 Racah Institute of Physics, The Hebrew University, Jerusalem 91904, Israel 1 0 E-mail: [email protected], [email protected] 2 n Abstract. Westudythecompetingorderandchaosinafirst-orderquantumphasetransition a with a high barrier. The boson model Hamiltonian employed, interpolates between its U(5) J (spherical)andSU(3)(deformed)limits. Aclassicalanalysisrevealsregular(chaotic)dynamics 8 atlow(higher)energyinthesphericalregion,coexistingwitharobustlyregulardynamicsinthe deformed region. A quantum analysis discloses, amidst a complicated environment, persisting ] regularmultipletsofstatesassociatedwithpartialU(5)andquasiSU(3)dynamicalsymmetries. h t - l c u n 1. Introduction [ Quantum phase transitions (QPTs) are qualitative changes in the ground state properties of a 1 physical system induced by a variation of parameters λ in the quantum Hamiltonian Hˆ(λ) [1,2]. v 7 Such ground-state transformations have received considerable attention in recent years and 9 have found a variety of applications in many areas of physics and chemistry [3]. The particular 0 type of QPT is reflected in the topology of the underlying mean-field (Landau) potential V(λ). 2 . Most studies have focused on second-order (continuous) QPTs [4], where V(λ) has a single 1 minimum which evolves continuously into another minimum. The situation is more complex 0 4 for discontinuous (first-order) QPTs, where V(λ) develops multiple minima that coexist in a 1 range of λ values and cross at the critical point, λ=λ . The interest in such QPTs stems from c : v their key role in phase-coexistence phenomena at zero temperature. Examples are offered by the i metal-insulatorMotttransition[5],heavy-fermionsuperconductors[6],quantumHallbilayers[7] X and shape-coexistence in mesoscopic systems, such as atomic nuclei [8]. r a ThecompetinginteractionsintheHamiltonianthatdrivetheseground-statephasetransitions can affect dramatically the nature of the dynamics and, in some cases, lead to the emergence of quantum chaos. This effect has been observed in quantum optics models of N two-level atoms interacting with a single-mode radiation field [9], where the onset of chaos is triggered by continuous QPTs. In the present contribution, we address the mixed regular and chaotic dynamics associated with a first order QPT [10–12]. For that purpose, we employ an interacting boson model which describes such QPTs between spherical and axially-deformed nuclei. The model is amenable to both classical and quantum treatments, has a rich group structure and inherent geometry, which makes it an ideal framework for studying the intricate interplay of order and chaos and the role of symmetries in such shape-phase transitions. 2. Symmetries, geometry and quantum phase transitions in the IBM Theinteractingbosonmodel(IBM)[13]describeslow-lyingquadrupolecollectivestatesinnuclei in terms of N interacting monopole (s) and quadrupole (d) bosons representing valence nucleon pairs. The bilinear combinations G ≡ b†b = {s†s, s†d , d† s, d† d } span a U(6) algebra, ij i j m m m m(cid:48) which serves as the spectrum generating algebra. The IBM Hamiltonian is expanded in terms of these generators, Hˆ = (cid:80) (cid:15) G +(cid:80) u G G , and consists of Hermitian, rotational- ij ij ij ijk(cid:96) ijk(cid:96) ij k(cid:96) invariant interactions which conserve the total number of s- and d- bosons, Nˆ = nˆ + nˆ = s d s†s+(cid:80) d† d . A dynamical symmetry (DS) occurs if the Hamiltonian can be written in terms m m m of the Casimir operators of a chain of nested sub-algebras of U(6). The Hamiltonian is then completely solvable in the basis associated with each chain. The three dynamical symmetries of the IBM and corresponding bases are U(6) ⊃ U(5) ⊃ O(5) ⊃ O(3) |N,n ,τ,n ,L(cid:105) spherical vibrator (1a) d ∆ U(6) ⊃ SU(3) ⊃ O(3) |N,(λ,µ),K,L(cid:105) axially−deformed rotor (1b) U(6) ⊃ O(6) ⊃ O(5) ⊃ O(3) |N,σ,τ,n ,L(cid:105) γ−unstable deformed rotor (1c) ∆ The associated analytic solutions resemble known limits of the geometric model of nuclei [14], as indicated above. The basis members are classified by the irreducible representations (irreps) of the corresponding algebras. Specifically, the quantum numbers N,n ,(λ,µ),σ,τ and L label the d relevantirrepsofU(6),U(5),SU(3),O(6),O(5)andO(3),respectively. n andK aremultiplicity ∆ labels needed for complete classification in the reductions O(5) ⊃ O(3) and SU(3) ⊃ O(3), respectively. Each basis is complete and can be used for a numerical diagonalization of the Hamiltonianinthegeneralcase. Ageometricvisualizationofthemodelisobtainedbyapotential surface, V(β,γ)=(cid:104)β,γ;N|Hˆ|β,γ;N(cid:105), defined by the expectation value of the Hamiltonian in the intrinsic condensate state [15,16] |β,γ;N(cid:105) = (N!)−1/2[Γ†(β,γ)]N|0(cid:105) , (2a) c Γ†(β,γ) = (cid:104)βcosγd† +βsinγ(d† +d† )/√2+(cid:112)2−β2s†(cid:105)/√2 . (2b) c 0 2 −2 Here(β,γ)arequadrupoleshapeparametersanalogoustothevariablesofthecollectivemodelof nuclei. Their values (β ,γ ) at the global minimum of V(β,γ) define the equilibrium shape for eq eq a given Hamiltonian. For one- and two-body interactions, the shape can be spherical (β = 0) eq or deformed (β > 0) with γ = 0 (prolate), γ = π/3 (oblate), or γ-independent. eq eq eq The dynamical symmetries of Eq. (1), correspond to phases of the system, and provide analytic benchmarks for the dynamics of stable nuclear shapes. Quantum phase transitions (QPTs) between such stable shapes have been studied extensively in the IBM framework [16,17] and are manifested empirically in nuclei [8]. The Hamiltonians employed mix interaction terms from different DS chains, e.g., Hˆ(λ) = λHˆ +(1−λ)Hˆ . The coupling coefficient (λ) responsible a b for the mixing, serves as the control parameter and the surface, V(λ) ≡ V(λ;β,γ), serves as the Landau potential. In general, under such circumstances, solvability is lost, there are no remaining non-trivial conserved quantum numbers and all eigenstates are expected to be mixed. However, for particular symmetry breaking, some intermediate symmetry structure can survive. The latter include partial dynamical symmetry (PDS) [18] and quasi-dynamical symmetry (QDS) [19]. In a PDS, the conditions of an exact dynamical symmetry (solvability of the complete spectrum and existence of exact quantum numbers for all eigenstates) are relaxed andapplytoonlypartoftheeigenstates. InaQDS,particularstatescontinuetoexhibitselected characteristic properties (e.g., energy and B(E2) ratios) of the closest dynamical symmetry, in the face of strong-symmetry breaking interactions. This “apparent” symmetry is due to the coherent nature of the mixing. As discussed below, both PDS and QDS are relevant to quantum phase transitions [19,20]. In view of the central role of the Landau potential, V(β,γ), for QPTs, it is convenient to resolvetheHamiltonianintotwoparts,Hˆ = Hˆ +Hˆ [21]. Theintrinsicpart(Hˆ )determines int col int the potential surface while the collective part (Hˆ ) is composed of kinetic terms which do not col affect the shape of V(β,γ). For first-order QPTs, the resolution allows the construction of an intrinsic Hamiltonian with a high-barrier [22], and by treating it separately, one avoids the complication of rotation-vibration couplings that can obscure the simple pattern of mixed dynamics, reportedbelow. Henceforth, weconfinethediscussiontothedynamicsoftheintrinsic Hamiltonian. 3. Intrinsic Hamiltonian for a first-order QPT Focusing on first-order QPTs between stable spherical (β = 0) and prolate-deformed (β > 0, eq eq γ = 0) shapes, the intrinsic Hamiltonian reads eq Hˆ (ρ)/h¯ = 2(1−2ρ2)nˆ (nˆ −1)+2R†(ρ)·R˜ (ρ) , (3a) 1 2 d d 2 2 Hˆ (ξ)/h¯ = ξP†P +P†·P˜ , (3b) 2 2 0 0 2 2 √ √ where nˆ is the d-boson number operator, R† (ρ)= 2s†d† +ρ 7(d†d†)(2), P†=d†·d†−2(s†)2 d √ 2µ µ µ 0 and P† =2s†d† + 7(d†d†)(2). Here R˜ =(−1)µR , P˜ =(−1)µP and the dot implies a 2µ µ µ 2µ 2,−µ 2µ 2,−µ scalar product. Scaling by h ≡ h /N(N −1) is used throughout, to facilitate the comparison 2 2 withtheclassicallimit. ThecontrolparametersthatdrivetheQPTareρandξ,with0 ≤ ρ ≤ √1 2 and 0 ≤ ξ ≤ 1. The intrinsic Hamiltonian in the spherical phase, Hˆ (ρ), has by construction the intrinsic 1 state of Eq. (2) with β = 0 as zero energy eigenstate. For large N, its normal modes eq involve quadrupole vibrations about the spherical global minimum of the potential surface, with frequency (cid:15)=4h¯ N. For ρ = 0 it reduces to 2 Hˆ (ρ = 0)/h¯ = 2nˆ (nˆ −1)+4(Nˆ −nˆ )nˆ . (4) 1 2 d d d d Since nˆ is the linear Casimir operator of U(5), Hˆ (ρ = 0) has U(5) dynamical symmetry d 1 (DS). The spectrum is completely solvable E = [2n (n − 1) + 4(N − n )n ]h¯ , and the DS d d d d 2 eigenstates, |N,n ,τ,n ,L(cid:105), are those of the U(5) chain, Eq. (1a). The spectrum resembles that d ∆ ofananharmonicsphericalvibrator,describingquadrupoleexcitationsofasphericalequilibrium shape. The lowest U(5) multiplets involve states with quantum numbers (n =0,τ =0,L=0), d (n =1,τ=1,L=2), (n =2,τ=2,L=2,4;τ = 0,L = 0), (n =3,τ=3,L=6,4,3,0;τ=1,L=2). d d d For ρ > 0, Hˆ (ρ) has an additional ρ[(d†d†)(2) ·d˜s+s†d† ·(d˜d˜)(2)] term, which breaks the 1 U(5) DS, and induces U(5) and O(5) mixing subject to ∆n = ±1 and ∆τ = ±1,±3. The d explicitbreakingofO(5)symmetryleadstonon-integrabilityand,aswillbeshowninsubsequent discussions, is the main cause for the onset of chaos in the spherical region. Although Hˆ (ρ > 0) 1 is not diagonal in the U(5) chain, it retains the following selected solvable U(5) basis states |N,n = τ = L = 0(cid:105) E = 0 , (5a) d PDS |N,n = τ = L = 3(cid:105) E = 12(cid:0)N −2+3ρ2(cid:1)h¯ , (5b) d PDS 2 while other eigenstates are mixed with respect to U(5). As such, it exhibits U(5) partial dynamical symmetry [U(5)-PDS]. The intrinsic Hamiltonian in the deformed phase, Hˆ (ξ), has by construction the intrinsic 2 state of Eq. (2) |βeq = √2 ,γeq = 0;N(cid:105) as zero energy eigenstate. For large N, its normal modes 3 involvebothβ andγ vibrationsaboutthedeformedglobalminimumofV(β,γ),withfrequencies (cid:15) =4h¯ N(2ξ+1) and (cid:15) =12h¯ N. For ξ = 1, the Hamiltonian reduces to β 2 γ 2 Hˆ (ξ = 1)/h¯ = −Cˆ +2Nˆ(2Nˆ +3) , (6) 2 2 SU(3) It involves the quadratic Casimir of SU(3) and hence has SU(3) DS. The spectrum is completely solvable, E /h¯ = [−(λ2 + µ2 + λµ + 3λ + 3µ) + 2N(2N + 3)]h¯ , and the eigenstates, DS 2 2 |N,(λ,µ),K,L(cid:105), are those of the SU(3) chain , Eq. (1b). The spectrum resembles that of an axially-deformed rotor with degenerate K-bands arranged in SU(3) (λ,µ) multiplets, K being the angular momentum projection on the symmetry axis. The rotational states in each band have angular momenta L = 0, 2, 4..., for K = 0 and L = K,K +1,K +2,..., for K > 0. The lowest SU(3) multiplets are (2N,0) which contains the ground band g(K = 0), and (2N −4,2) which contains the β(K = 0) and γ(K = 2) bands. For ξ < 1, Hˆ (ξ) has an additional term, (ξ−1)P†P , which breaks the SU(3) DS and most 2 0 0 eigenstates are mixed with respect to SU(3). However, the following states |N,(2N,0)K = 0,L(cid:105) L = 0,2,4,...,2N E = 0 (7a) PDS |N,(2N −4k,2k)K = 2k,L(cid:105) L = K,K +1,...,(2N −2k) k > 0 , E = 6k(2N −2k+1)h¯ (7b) PDS 2 remainsolvablewithgoodSU3)symmetry. Assuch,Hˆ (ξ < 1)exhibitsSU(3)partialdynamical 2 symmetry [SU(3)-PDS]. The selected states of Eq. (7) span the ground band g(K = 0) and γk(K = 2k) bands. The intrinsic Hamiltonians, Hˆ1(ρ) and Hˆ2(ξ) of Eq. (3), with 0 ≤ ρ ≤ √1 and 0 ≤ ξ ≤ 1, 2 interpolatebetweentheU(5)(ρ=0)andSU(3)(ξ=1)DSlimits. ThetwoHamiltonianscoincide at the critical point ρc=√1 and ξc=0: Hˆ1(ρc) = Hˆ2(ξc). Both Hamiltonians support subsets of 2 solvable PDS states, Eqs. (5) and (7), whose analytic properties provide unique signatures for their identification in the quantum spectrum. 4. Classical limit and topology of the Landau potential The classical limit of the IBM is obtained through the use of Glauber coherent states. This √ amounts to replacing (s†, d†) by six c-numbers (α∗, α∗) rescaled by N and taking N → ∞, µ s µ with 1/N playing the role of (cid:126) [23]. Number conservation ensures that phase space is 10- dimensional and can be phrased in terms of two shape (deformation) variables, three orientation (Euler) angles and their conjugate momenta. The shape variables can be identified with the β,γ variables introduced through Eq. (2). Setting all momenta to zero, yields the classical potential which is identical to V(β,γ) mentioned above. In the classical analysis presented below we consider, for simplicity, the dynamics of L = 0 vibrations, for which only two degrees of freedom areactive. TherotationaldynamicswithL > 0isexaminedinthesubsequentquantumanalysis. For the intrinsic Hamiltonian of Eq. (3), constrained to L = 0, the above procedure yields the following classical Hamiltonian H (ρ)/h = H2 +2(1−H )H +2ρ2p2 1 2 d,0 d,0 d,0 γ (cid:113) +ρ 2(1−H )(cid:2)(p2/β −βp2 −β3)cos3γ +2p p sin3γ(cid:3) , (8a) d,0 γ β β γ H (ξ)/h = H2 +2(1−H )H +p2 2 2 d,0 d,0 d,0 γ +(cid:112)1−H (cid:2)(p2/β −βp2 −β3)cos3γ +2p p sin3γ(cid:3) d,0 γ β β γ +ξ(cid:2)β2p2 + 1(β2−T)2−2(1−H )(β2−T)+4(1−H )2(cid:3) . (8b) β 4 d,0 d,0 √ Here the coordinates β ∈ [0, 2], γ ∈ [0,2π) and their canonically conjugate momenta √ p ∈ [0, 2] and p ∈ [0,1] span a compact classical phase space. The term, H = (T +β2)/2 β γ d,0 with T = p2 + p2/β2, denotes the classical limit of nˆ (restricted to L = 0) and forms an β γ d isotropic harmonic oscillator Hamiltonian in the β and γ variables. Notice that the classical Hamiltonian of Eq. (8) contains complicated momentum-dependent terms originating from the two-body interactions in the Hamiltonian (3), not just the usual quadratic kinetic energy T. Setting p = p = 0 in Eq. (8) leads to the following classical potential β γ (cid:112) V (ρ)/h = 2β2−2ρ 2−β2β3cos3γ − 1β4 , (9a) 1 2 2 (cid:112) V (ξ)/h = 2(1−3ξ)β2− 2(2−β2)β3cos3γ + 1(9ξ−2)β4+4ξ . (9b) 2 2 4 The variables β and γ can be interpreted as polar coordinates in an abstract plane parametrized by Cartesian coordinates x = βcosγ and y = βsinγ. Using these relations together with p =p cosγ−(p /β)sinγ andp =(p /β)cosγ+p sinγ,onecancasttheclassicalHamiltonian, x β γ y γ β Eq. (8) and potential, Eq. (9), in Cartesian form. Thus, H = (p2 +p2 +x2 +y2)/2 and the d,0 x y potentials V(β,γ) = V(x,y) depend on the combinations β2 =x2 + y2, β4 = (x2 + y2)2 and β3cos3γ=x3−3xy2. Thecontrolparametersρandξ determinethelandscapeandextremalpointsofthepotentials V (ρ;β,γ) and V (ξ;β,γ), Eq. (9). Important values of these parameters at which a pronounced 1 2 changeinstructureisobserved, arethespinodalpoint(ρ∗)whereasecond(deformed)minimum occurs, an anti-spinodal point (ξ∗∗) where the first (spherical) minimum disappears and a critical point (ρ ,ξ ) in-between, where the two minima are degenerate. For the potentials c c under discussion, the values of the control parameters at these points are ρ∗ = 12 , (ρc = √12 , ξc = 0) , ξ∗∗ = 13 . (10) The critical point separates the spherical and deformed phases. The spinodal and anti-spinodal points embrace it and mark the boundary of the phase coexistence region. (cid:112) In general, the only γ-dependence in the potentials (9) is due to the 2−β2β3cos3γ term. This induces a three-fold symmetry about the origin β =0. As a consequence, the deformed extremal points are obtained for γ=0, 2π, 4π (prolate shapes), or γ= π,π, 5π (oblate shapes). 3 3 3 3 It is therefore possible to restrict the analysis to γ = 0 and allow for both positive and negative values of β, corresponding to prolate and oblate deformations, respectively. The potentials V(β,γ = 0)=V(x,y = 0) for several values of ξ,ρ, are shown at the bottom rows of Figs. 2-4. The relevant potential in the spherical phase is V (ρ;β,γ), Eq. (9a), with 0 ≤ ρ ≤ ρ . In 1 c this case, β = 0 is a global minimum of the potential at an energy V = 0, representing the eq sph spherical equilibrium shape. The limiting value at the domain boundary is V = V (ρ;β = √ lim 1 2,γ) = 2h . For ρ = 0, (the U(5) limit), the potential is independent of γ and has β = 0 as 2 eq a single minimum. For ρ > 0, the potential depends on γ, and β = 0 remains a single minimum for 0 ≤ ρ < ρ∗, At the spinodal point ρ∗, V (ρ) acquires an additional deformed local minimum 1 at an energy V > 0, and a barrier develops between the two minima. The spherical and def deformed minima cross and become degenerate at the critical point (ρ ,ξ ). c c The relevant potential in the deformed phase is V (ξ;β,γ), Eq. (9b), with ξ ≥ ξ . In 2 c this case, [βeq = √2 ,γeq = 0] is a global minimum of the potential at an energy Vdef = 0, 3 representing the deformed equilibrium shape. The limiting value of the domain boundary is √ V = V (ξ;β = 2,γ) = (2+ξ)h . The two potentials coincide at the critical point (ρ ,ξ ), lim 2 2 √ c c V (ξ ;β,γ) = V (ρ ;β,γ), andthebarrierheightisV =1h (1− 3)2 = 0.268h . Thespherical 2 c 1 c bar 2 2 2 minimum turns local in V (ξ) for ξ > ξ with energy V = 4h ξ > 0, and disappears at the 2 c sph 2 anti-spinodal point ξ∗∗. For ξ > ξ∗∗, β = 0 turns into a maximum and the potential remains with a single deformed minimum, reaching the SU(3) limit at ξ = 1. The indicated changes in the topology of the potential surfaces upon variation of the control parameters (ρ,ξ), identify three regions with distinct structure. Figure 1. Behavior of the order parameter, β , as a function of the control parameters (ρ,ξ) eq of the intrinsic Hamiltonian, Eq. (3). Here ρ∗, (ρ , ξ ), ξ∗∗, are the spinodal, critical and anti- c c spinodalpoints,respectively,withvaluesgiveninEq.(10). Thedeformationattheglobal(local) minimum of the Landau potential (9) is marked by solid (dashed) lines. βeq=0 (βeq= √2 ) on 3 the spherical (deformed) side of the QPT. Region I (III) involves a single spherical (deformed) shape, while region II involves shape-coexistence. I. The region of a stable spherical phase, ρ ∈ [0,ρ∗], where the potential has a single spherical minimum. II. The region of phase coexistence, ρ ∈ (ρ∗,ρ ] and ξ ∈ [ξ ,ξ∗∗), where the potential has both c c spherical and deformed minima which cross and become degenerate at the critical point. III. The region of a stable deformed phase, ξ > ξ∗∗, where the potential has a single deformed minimum. The potential surface in each region serves as the Landau potential of the QPT, with the equilibrium deformations as order parameters. The latter evolve as a function of the control parameters (ρ,ξ) and exhibit a discontinuity typical of a first order transition. As depicted in Fig 1, the order parameter β is a double-valued function in the coexistence region (in-between eq ρ∗ andξ∗∗)andastep-functionoutsideit. Inwhatfollows,weexaminethenatureoftheclassical and quantum dynamics in each region. 5. Classical analysis Chaotic properties of the IBM have been studied extensively [24], albeit, with a simplified Hamiltonian, giving rise to an extremely low barrier. A new element in the present study is the presence of a high barrier at the critical-point, V /h =0.268, compared to V /h =0.0018 bar 2 bar 2 in previous works. This allows the uncovering of a rich pattern of regularity and chaos across a generic first-order QPT in a wide coexistence region. The classical dynamics constraint to L = 0, can be depicted conveniently via Poincar´e surfaces of sections in the plane y = 0, plotting the values of x and the momentum p each x time a trajectory intersects the plane [25]. Regular trajectories are bound to toroidal manifolds within the phase space and their intersections with the plane of section lie on 1D curves (ovals). In contrast, chaotic trajectories randomly cover kinematically accessible areas of the section. Figure 2. Poincar´e sections in the stable spherical phase (region I). Upper five rows depict the classical dynamics of H (ρ) (8a) with h = 1, for several values of ρ ≤ ρ∗∗. The bottom 1 2 row displays the Peres lattices {x ,E }, portraying the quantum dynamics for (N = 80,L = 0) i i eigenstates of Hˆ (ρ) (3a), overlayed on the classical potentials V (ρ;x,y = 0) (9a). The five 1 1 energies at which the sections were calculated consecutively, are indicated by horizontal lines. Figure 3. Poincar´e sections in the region of phase-coexistence (region II). The panels are as in Fig. 2, but for the classical Hamiltonians H (ρ) (8a) with ρ∗∗ < ρ ≤ ρ , and H (ξ) (8b) with 1 c 2 ξ ≤ ξ < ξ∗∗. The classical potentials are V (ρ;x,y = 0) (9a) and V (ξ;x,y = 0) (9b). The Peres c 1 2 lattices involve the quantum Hamiltonians Hˆ (ρ) (3a) and Hˆ (ξ) (3b). 1 2 Figure 4. Poincar´e sections in the stable deformed phase (region III). The panels are as in Fig. 2, but for the classical Hamiltonian H (ξ) (8b) and potential V (ξ;x,y = 0) (9b), with 2 2 ξ ≥ ξ∗∗. The Peres lattices involve the quantum Hamiltonian Hˆ (ξ) (3b). 2 The Poincar´e sections associated with the classical Hamiltonian of Eq. (8) are shown in Figs. 2-3-4 for representative energies, below the domain boundary, and control parameters (ρ,ξ) in regions I-II-III, respectively. The bottom row in each figure displays the corresponding classical potential V(β,γ = 0) = V(x,y = 0), Eq. (9). In region I (0 ≤ ρ ≤ ρ∗), for ρ = 0, H (ρ = 0) = H (2 − H ), involves the 2D harmonic oscillator Hamiltonian and 1 d,0 d,0 V (ρ = 0) ∝ 2β2 − β4/2. The system is in the U(5) DS limit and is completely integrable. 1 The orbits are periodic and, as shown in Fig. 2, appear in the surface of section as a finite collection of points. The sections for ρ=0.03 in Fig. 2, show the phase space portrait typical of an anharmonic (quartic) oscillator with two major regular islands, weakly perturbed by the small ρcos3γ term. The orbits are quasi-periodic and appear as smooth one-dimensional √ invariant curves. For small β, V (ρ)≈β2−ρ 2β3cos3γ and resembles the well-known H´enon- 1 Heiles potential (HH) [26]. Correspondingly, as shown for ρ=0.2 in Fig. 2, at low energy, the dynamics remains regular and two additional islands show up. The four major islands surround stable fixed points and unstable (hyperbolic) fixed points occur in-between. At higher energy, one observes a marked onset of chaos and an ergodic domain. The chaotic component of the dynamics increases with ρ and maximizes at the spinodal point ρ∗ =0.5. The chaotic orbits densely fill two-dimensional regions of the surface of section. The dynamics changes profoundly in region II of phase coexistence (ρ∗ < ρ ≤ ρ and c ξ ≤ ξ < ξ∗∗). The Poincar´e sections before at and after the critical point, (ρ=0.6, ξ =0, c c ξ=0.1) are shown in Fig. 3. In general, the motion is predominantly regular at low energies and gradually turning chaotic as the energy increases. However, the classical dynamics evolves differently in the vicinity of the two wells. As the local deformed minimum develops, robustly regulardynamicsattachedtoitappears. Thetrajectoriesformasingleislandandremainregular even at energies far exceeding the barrier height V . This behavior is in marked contrast to the bar HH-type of dynamics in the vicinity of the sphericalminimum, where a change with energy from regularity to chaos is observed, until complete chaoticity is reached near the barrier top. The clearseparationbetweenregularandchaoticdynamics, associatedwiththetwominima, persists all the way to the barrier energy, E = V , where the two regions just touch. At E > V , bar bar the chaotic trajectories from the spherical region can penetrate into the deformed region and a layer of chaos develops, and gradually dominates the surviving regular island for E (cid:29) V . As bar ξ increases, the spherical minimum becomes shallower, and the HH-like dynamics diminishes. Fig.4displaystheclassicaldynamicsinregionIII(ξ∗∗ ≤ ξ ≤ 1). Thelocalsphericalminimum andtheassociatedHH-likedynamicsdisappearattheanti-spinodalpointξ∗∗ = 1/3. Theregular motion, associated with the single deformed minimum, prevails for ξ≥ξ∗∗. Here a single stable fixedpoint, surroundedbyafamilyofellipticorbits, continuestodominatethePoincar´esection. In certain regions of the control parameter ξ and energy, the section landscape changes from a single to several regular islands, reflecting the sensitivity of the dynamics to local degeneracies of normal modes. A notable exception to such variation is the SU(3) DS limit (ξ = 1), for which the system is integrable and the phase space portrait is the same for any energy. 6. Quantum analysis Quantum manifestations of classical chaos are often detected by statistical analyses of energy spectra [25]. In a quantum system with mixed regular and irregular states, the statistical properties of the spectrum are usually intermediate between the Poisson and the Gaussian orthogonal ensemble (GOE) statistics. Such global measures of quantum chaos are, however, insufficient to reflect the rich dynamics of an inhomogeneous phase space structure encountered in Figs. 2-4, with mixed but well-separated regular and chaotic domains. To do so, one needs to distinguish between regular and irregular subsets of eigenstates in the same energy intervals. For that purpose, we employ the spectral lattice method of Peres [27], which provides additional properties of individual energy eigenstates. The Peres lattices are constructed by plotting