* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download --Fundamental Problems and Application to Material Science-
Spin (physics) wikipedia , lookup
Molecular Hamiltonian wikipedia , lookup
Perturbation theory (quantum mechanics) wikipedia , lookup
Ferromagnetism wikipedia , lookup
Quantum decoherence wikipedia , lookup
Aharonov–Bohm effect wikipedia , lookup
Quantum dot wikipedia , lookup
Wave–particle duality wikipedia , lookup
Copenhagen interpretation wikipedia , lookup
Probability amplitude wikipedia , lookup
Measurement in quantum mechanics wikipedia , lookup
Particle in a box wikipedia , lookup
Renormalization wikipedia , lookup
Quantum fiction wikipedia , lookup
Quantum electrodynamics wikipedia , lookup
Many-worlds interpretation wikipedia , lookup
Quantum entanglement wikipedia , lookup
Density matrix wikipedia , lookup
Orchestrated objective reduction wikipedia , lookup
Coherent states wikipedia , lookup
Quantum field theory wikipedia , lookup
Topological quantum field theory wikipedia , lookup
Bell's theorem wikipedia , lookup
Quantum computing wikipedia , lookup
Renormalization group wikipedia , lookup
Path integral formulation wikipedia , lookup
Hydrogen atom wikipedia , lookup
Quantum teleportation wikipedia , lookup
Scalar field theory wikipedia , lookup
Theoretical and experimental justification for the Schrödinger equation wikipedia , lookup
Interpretations of quantum mechanics wikipedia , lookup
Quantum key distribution wikipedia , lookup
Relativistic quantum mechanics wikipedia , lookup
EPR paradox wikipedia , lookup
Quantum machine learning wikipedia , lookup
Quantum group wikipedia , lookup
Symmetry in quantum mechanics wikipedia , lookup
Hidden variable theory wikipedia , lookup
Quantum state wikipedia , lookup
Progress of Theoretical Physics Supplement No. 98, 1989 383 Quantum Chaos --Fundamental Problems and Application to Material ScienceKatsuhiro NAKAMURA*> The fames Franck Institute, The University of Chicago Chicago, Illinois 60637, U.S.A. (Received October 28, 1988) We investigate quantum mechanics of nonintegrable and chaotic systems. Two realistic examples of quantum chaos in magnetic phenomena are given: (1) Quantum billiard in a magnetic field; (2) quantum dynamics of a pulsed spin system. In these examples, we discuss salient aspects of irregular energy spectra and complicated quantum diffusion. Then, funda· mental problems of quantum chaos are examined from a viewpoint of integrable dynamical systems. Singularities at degenerate points, which abound in the generic case with several nonintegrability parameters being included, are incorporated into a high·dimensional field theory. Contents §1. Introduction §2. Quantum oilliard in a magnetic field--avoided level-crossings and diamag- netism §3. Quantum dynamics of a pulsed spin system--anoma1ous behavior of semiclassical wavefunctions §4. Universal dynamical system behind the quantum chaos--a single parameter case §5. Reduction to a field-theoretical complex-Grassmannian sigma model---several parameters case §6. Summary and discussion § 1. Introduction The subject of quantum chaos has received a growing interest in contemporary physics. 1> The terminology of quantum chaos is currently used for the quantum mechanics of systems which exhibit dynamical chaos in the classical limit. Strictly speaking, the linear nature of Schrodinger equation suppresses a chaotic diffusion and eventually leads to the vanishing of Kolmogorov-Simii entropy and of other characteristic exponents. Nevertheless, many fundamental problems remain unsolved, e.g., those of irregular energy spectra and of complicated quantum diffusion even in the recurrent time region. Theoretical tools borrowed from a field of random systems *> Present and permanent address: Fukuoka Institute of Technology, Fukuoka 811-02, Japan. 384 K. Nakamura are not always effective here. The Wigner-Dyson level-spacing distribution in the random matrix theory2l explains only a very limited aspect of energy spectra. The concept of localization of wavefunctions3l loses its significance when the localization length ( ocn- 2 ) is larger than the dimensionality of Hilbert space. Recent challengers4 l have derived coupled dynamical equations for energy eigenvalues and matrix elements following a philosophy due to Dyson. 5 l Most of their efforts, however, have been directed towards confirming the effectiveness of the random matrix theory in the subject of quantum chaos. But, this new subject should contain much more fruitful informations than those due to the random matrix theory. On the other hand, the spirit of natural science requires to find realization of quantum chaos and thereby to establish the connection between chaotic dynamics and experimentally-accessible macroscopic quantum observables. In this paper, we provide two realistic examples of quantum chaos in magnetic phenomena. In § 2, we discuss single-electron energy spectra for the free electron gas cofined in a billiard in the presence of magnetic field. In § 3, the ergodic behavior of semiclassical wavefunctions of a pulsed spin system is examined in the long-time regime. In §§ 4 and 5, we seek for a universal theory behind the quantum chaos. In §4, we derive a universal dynamical system from generic quantum systems whose Hamiltonians possess only a singl~ nonintegrability parameter. In § 5, a high-dimensional field theory is constructed, corresponding to adiabatic-ansatz induced eigenvalue problems which are characterized by several nonintegrability parameters. We emphasize there the role of topological singularities at degenerate points.6 l The final section will be devoted to a summary and discussion. § 2. Quantum billiard in a magnetic field --avoided level-crossings and diamagnetism We choose here a typical autonomous system found in condensed matter physics. Let us consider the quantum mechanics of noninteracting electrons in a planar billiard (e.g., a thin conducting disk) in a uniform magnetic field normal to the plane.1l The shape of the boundary is taken as elliptic. Unless the boundary effects are taken into consideration, quantum-mechanical treatment merely yields the Landau diamagnetic susceptibility. 8l Recent analysis of single-electron classical dynamics, 9l however, elucidated the onset of chaos in the case when the Larmor radius is comparable to the linear dimension of the billiard, indicating the crucial role of the convex boundary. Quantum aspects of chaos will be captured by incorporating Dirichlet-type boundary conditions. In the following, we shall first solve the Dirichlet eigenvalue problem for a single-electron system. Then, global aspects of the spectra, i.e., the sensitivities of energies and their average over "occupied" levels- the diamagnetic susceptibility at absolute zero- will be investigated by changing the magnetic field. Let us consider an ellipse with area 1rV=1rab where a and bare semi-major and semi-minor axes in x- and y-directions, respectively. The eigenvalue problem is given by H¢=E¢ with ¢=0 at the boundary. Here H=(l/2m)((n/i)f7+(e/c)A) 2• We take a symmetric gauge: A=( -(1/2)yB, (1/2)xB) with B being the magnetic field. The present system has C2-symmetry. Using the map (x, y)~(r,8) via x=arcos8, y =brsin8, the eigenvalue problem is reduced to fi(r, 8)(f(r,8)=E¢(r,8) with ¢=0 Quantum Chaos 385 at the boundary of the unit disk. Basis functions are now constructed in terms of integer-order Bessel functions. For 40 convenience, we introduce the dimensionless parameters a=b/a, a=a/L, li =b/L and B=B(ctt/(eV))- 1• Using computed integrals in the matrix elements, we have solved the eigenvalue problem. We present the results for a nonintegrable case a=0.5 (elliptic bil0. 20.8 40. liard) and the integrable case (circular billiard). Comparison of the two cases Fig. 1. .8-dependent energy spectra for the evenparity manifold; (a) circle billiard (a=l); (b) helps to elucidate the effects of noninteellipse billiard (a=O_S). grability. In Fig. 1, the even-parity part of the energy spectra is shown. In both of Figs. 1(a) and (b), most of the levels are found not to be well bunched into Landau levels. Level repulsion leading to avoided crossings is widely seen in the case a=0.5 (Fig. 1(b)), while true crossings between levels with different k-values predominate in the case a=1 (Fig. 1(a)). The presence of many avoided crossings corresponds to chaos in the underlying classical dynamics. 9> For a=F1, chaos around the unstable diametral orbit and/ or flyaway chaos are reported to dominate phase space, provided the Larmor radius rc=mvc/eB satisfies rc/L';::; Pmln for the smallest curvature radius Pm1n= b2 /a. ity, we find (2·1) Noting E=1/2mv 2 for electron veloc(2·2) In consistence with the classical findings, we clearly observe in Fig. 1(b) that avoided crossings dominate the spectra in the region E/B 2 ';::;0.35 for a=0.5. We consider a sample containing 2N electrons, and neglect the Zeeman splitting of the spin states. Then, in the free-electron ground state at a given value of the applied field, the N lowest levels Ej(B) (j = 1, · · ·, N) are filled with two electrons each, the Fermi level cF lying between EN(B) and EN+I(B). The isothermal susceptibility per electron at absolute zero is given by the 2nd order derivative of the total energy as (2·3) The contributions to the sum in Eq. (2·3) are computed separately for each manifold of different symmetry (parity for a=F 1, k-values for a= 1). Singularities of .tP Ej/L1B 2 at true crossings are thus removed. In Fig. 2, the negative of x is shown as a function of B for the case N = 100 occupied levels, where approximately 50 levels are of even parity and the remaining ones are of odd parity. For a= 1, x is found to retain the essential features of Landau diamagnetism in 386 K. Nakamura 2D systems : - x takes the largest value aa~------------~ in the vicinity of B=O, and decreases monotonically with increasing B. Let us trace back to Fig. 1(a): All crossings appearing in this figure are true crossings between levels with different k. Therefore, each energy level shows a 0.1 smooth variation with B with positive curvature L12 l{/L1B 2 > 0, which decreases with increasing B. The characteristics ao~4H-~:H--------L-~ of the diamagnetic susceptibility for 11 = 1 in Fig. 2 are thus well explained by the regular behavior of the spectrum -0.1 which is a direct consequence of the integrability of this case. For 11=0.5, on the contrary, x shows -0.2l-------'-'r--~---~~ remarkably different features: The value 20. 40. of - x is greatly reduced at B=O as compared to the Landau value. It Fig.2. -x as a function of B for N=lOO. Circles and squares indicate a=l and 0.5 cases, reincreases on the average with B, spectively. Heavy symbols and lines denote recovering the value for 11= 1 only for combined contributions from both even- and E10o/B 2 ~ Pmln, i.e., for B ~50. This odd-parity manifolds, while fine cc;mnterparts increase is accompanied by large fluctuadenote the contribution from the even-parity manifold alone with energies below cF. x is tions and anomalous dips. These feascaled by (2mL"/h 2)J.lB2• tures can be traced back to the behavior of the spectrum, which shows a multitude of avoided crossings (AC) (see Fig.1 (b)): The rapid variation of the two levels with JJ near an AC gives rise to anomalous contributions of L1 2Ei/L1iJ2 of opposite signs. If the AC is narrow and lies below cF, the two contributions cancel in Eq. (2·3). But most ACs have width~their mutual distance. This leads to a rather flat variation of each level with B, with a greatly reduced average curvature and large fluctuations due to the nonuniform distribution of ACs. With increasing B most ACs become extremely narrow, and the Bdependence of the levels approaches that for 11= 1. Thus, the anomalous features of the diamagnetic susceptibility for !1=0.5 shown in Fig. 2 reflect the effects of level repulsion and avoided crossings typical for a nonintegrable system. In the case of ellipse billiards, in general, there exists a critical field Be: For B<Be the anomalous feature of diamagnetic susceptibility is found and for B >Be Landau-like behaviors are recovered. Be is determined by using Eqs. (2 ·1) and (2 · 2) as: o. e (2·4) Our findings cannot simply be interpreted in terms of the concept of bulk states and edge states, since such a distinction of states is not possible for the case re/L = 0(1). Figure 3 shows a typical pair of wavefunctions at an avoided crossing indicated in Fig. 1(b). Indeed, they can be attributed neither to bulk nor to edge Quantum Chaos 387 states. Our prediction will be tested in thin conducting disks containing a lowdensity 2D electron gas. Let us consider, for example, the interface layer in semiconductor heterojunctions, where current works concentrate on the quantum Hall effect. In these systems, the electron concentration n is very low -typically, n~l0 12cm- 2 • Noting that the Fermi wave number kF ~ n 112 in 2D, we have kF~l0 6 cm- 1 • For the above n, N=l0 2 corresponds to L~lo-s em. Then EN~ 8.0 X 102 for a= 0.5 and the interesting aspect of quantum chaos is seen for O<B~50, i.e., for O<B~2.5T. The de Broglie wavelength ~ kF - 1 is large enough for the results to be insensitive to the lattice discreteness both within and along the boundary of the interface layer. In ordinary metal Fig. 3. Wavefunctions 1¢1 at the avoided crossing indicated by arrow in Fig. l(b). o-=0.5 and B disks, on the other hand, n amounts to =10: (a) Esz=284.0230; (b) Eaa=285.1459. 1016cm-I, which corresponds to kF~l0 8 cm- 1 • Then the de Broglie wavelength is comparable to the lattice constant so that the results will be more or less modified. So our predictions due to the deterministic chaos will be verified in real experiments for small 2D devices so long as both the temperatures and the concentrations of atomic-scale defects are low enough. In this way, chaotic dynamics in the nonintegrable elliptic billiard in a uniform magnetic field is found to result in a multitude of avoided level-crossings, thereby inducing a remarkable reduction and large fluctuations of diamagnetic susceptibility. On the other hand, the integrable circular billiard yields results close to Landau diamagnetism in 2D. § 3. Quantum dynamics of a pulsed spin system -anomalous behavior of semiclassical wavefunctions We now move to another example of quantum chaos in a driven nonautonomous system. Chaos in driven spin system has recently become realized and its study constitutes a very active research field, 10' though experiments to date have been limited to dissipative spin-wave dynamics.11l Fully nonlinear dynamics for noninteracting spins, both classical and quantum, is also an attractive candidate by which to study classical chaos and its quantum counterpart. The advantage of spin systems is that, as a result of the finite-dimensional Hilbert space, we can readily tune the value of n without artificial truncation procedures in quantum-mechanical treatments. We consider wavefunctions for a periodically kicked spin system and demonstrate their fractal structures in semiclassically large-spin regions. 12 '' 13 ' Let us con- 388 K. Nakamura sider the single-spin Hamiltonian, common to both classical and quantum spin variables S, co H=A(Sz)2-pBSx 1: 8(t-2;rn), (3·1) n==-oo where A(>O) and B(>O) are an easy-plane anisotropy and magnetic field along the x axis, respectively. In Eq. (3 ·1), we have chosen a tentative model Hamiltonian. One may replace A(Sz)2 and pB by ASz and pBcos(mt), respectively. This kind of replacement will not alter the essential part of the outcome described below. Further, in the experiment of spin echoes, for instance, an assembly of spin 1/2 systems behaves coherently and constitutes actually a single large quantum-spin system. Before proceeding to the quantum-mechanical treatment, we shall present brief results for classical dynamics. Then Sis a three-component vector S=Sxex+Syey + Szez ( e is a unit vector), which obeys the equation of motion co dS/dt=SX(-8H/8S)=Sx(-2ASzez+Ji.Bex 1: 8(t-2;rn)). n=-oo (3·2) The magnitude 8 2 is conserved in Eq. (3 · 2), and is now normalized to unity. Sis then described in polar coordinates, i.e., S=(Sx, Sy, Sz)=(sin8cos¢, sin8sin¢, cosO). The discrete map can be constructed for successive values {Sn}, where Sn is the value of Sat t=2;rn+O, i.e., just after the n-th pulse. In the interval2;rn+O:::;; t:;:;;2;r(n+ 1)-0, the magnetic field is not operative in Eq. (3·2) and thus Sn is transformed into r =Rz(a)Sn at t=2;r(n+ 1)-0, where the operator Rz(a) denotes a rotation by angle a =4;rASnz around the z axis. Then, the (n+1)-th pulse during 2;r(n+1)-0~t:::;;2;r x(n+1)+0 rotates r by angle [3=-pB around the x axis, yielding Sn+I=Rx([J)r. Eventually, the combined map Sn+I = Rx([J)Rz(a)Sn is obtained, which we have solved with A=l.O and 0.0::5:pB~l.O. Investigation of our extensive data as a function of pB indicates the presence of two characteristic fields pB1 ~ 0.1 and pB2 ~ 0.5, where the fraction of chaotic trajectories increases strongly and the last KAM torus disappears, respectively. 12J The corresponding quantum dynamics is governed by the time-dependent Schrodinger equation for a wavefunction with the Hamiltonian being the quantum version of Eq. (3 ·1). We solve this equation after rewriting it in matrix form at the outset: A set of eigenstates of Sz is used as basis kets. A coefficient vector C for the wavefunction satisfies the matrix equation in dC/dt =fie . (3·3) fi is a (2S+l)X(2S+l) real-symmetric matrix ii=fio+V ~ 8(t-2;rn). Noting n=-oo Floquet's theorem, the solution of Eq. (3·3) just after the n-th pulse is C(2;rn+O)= 1:[exp( -2;rinEa/n)][Xa t • C( +O)]Xa, a (3·4) where the Ea's and Xa's are the quasienergies and eigenstate vectors, respectively, obtained by solving the eigenvalue problem UXa=exp( -2;riEa/n)Xa. Here U is a unitary matrix defined in terms of the time-ordering operator T as follows: Quantum Chaos (a) <I> (b) 389 ja) (c 0 <I> 0 EJ EJ te) (f) Fig. 4. Contour maps of Pn(B, ¢) in the case JJ.B =0.01 for very early stages of time evolution: (a) n=O; (b) n=l; (c) n=2; (d) n=3; (e) n=4; (f) n=8. Fig. 5. The same as in Fig. 4, but in the case JJ.B=l.O. U = Texp[l:no(- i/n)H(t')dt'] =exp[(- i/n) V]exp[(- i/n)2n-ilo] . (3·5) The probability density function is given in terms of SU(2S+ 1) coherent state representations as Pn(B, ¢)=1<8, ¢127Z'n+O>I 2 • In the following, we employ S=128 and choose n=(S(S+ 1))- 112 so that the observable spin magnitude maintains the scaled value for the classical spin vector, i.e., S 2 =S(S+ 1)n2 =1. In Figs. 4 and 5, very early stages of temporal evolution of initially (n=O) localized wavepackets are shown. For a weak pulse (,uB=0.01), Pn(8,¢) shows a unidirectional diffusion (see Fig. 4) corresponding to regular behaviors in classical dynamics. However, for a strong pulse (.uB=l.O), remarkable isotropic and irregular diffusions begin after the period of "classical" stretching and folding-type diffusion (see Fig. 5). Figures 4 and 5 also indicate the crossover time tc= 0(1) at which the classical and quantum correspondence breaks down. We proceed to examine Pn(B, ¢)in large n(~1) regions beyond tc. Figures 6(a) "-'(c), (a')""'(c') and (a")""'(c"), while they have no exact classical counterparts in these time regimes, retain some signatures of periodic orbits at ,uB=0.01, coexisting KAM orbits and localized chaos at ,uB=0.2 and global chaos at ,uB=l.O, respectively. We have attempted to characterize these structures in terms of the singularity spectra /(a)/ 4 > which have proven effective recently in quantifying multifractal aspects of chaotic systems. While details will be described elsewhere, 13 > we summarize here the major issue: Widths of /(a), i.e., distribution of local dimensions of Pn(B, ¢), for ,uB =0.01 and 1.0 fall into a narrow range during the time evolution. On the other hand, 390 K. Nakamura (C) Fig. 6. Time evolution of three-dimensional pictures of Pn(B, ¢)for n:>l. (a)-(c) JlB=O.Ol. From the left, n=70, 90 and 110; (a')-(c') tlB=0.2. From the left, n=60, 110 and 120; (a")-(c") tlB=l.O. From the left, n=90, 110 and 120. those for J.tB=0.2 extend over a much broader range. The large fluctuation of local dimensions for J.tB=0.2 signifies the inhomogeneous distribution of measures Pn(B, ¢),as seen in Figs. 6(a'}·"(c'). It clearly reflects the coexistence of KAM orbits and chaos characteristic of a transitional region leading to a global chaos. The small fluctuations for J.tB=O.Ol and 1.0 signify the uniform distribution of measures in Figs. 6(a)--.....(c) and 6(a")--.....(c"). The small fluctuation for J.tB=l.O reflects highly organized ergodicity in the corresponding classical dynamics. Thus semiclassical wavefunctions exhibit quite new phenomena which can be seen in neither the classical nor quantum limit. The outcome of this section is summarized as: (1) Despite the complete absence of classical and quantum correspondence, the long-time .behavior of wavefunctions maintains the ergodic or nonergodic features possessed by the underlying classical dynamics; (2) the enhanced fluctuation of local dimensions of wavefunctions, which is found at a transitional region leading to a global chaos, persists throughout the time evolution. §4. Universal dynamical system behind the quantum chaos --a single parameter case We have so far observed salient aspects of quantum chaos by taking its typical examples experimentally-accessible in magnetic phenomena. We have solved eigenvalue problems for Hermitian operator H(t) and Unitary one U(t) where t represents hereafter a nonintegrability parameter (e.g., strength of magnetic fields). In the analyses of quantum chaos, in general, we encounter the eigenvalue problems Quantum Chaos 391 of the forms H(t)ln>=xnln> and U(t)ln>=e-.:rnln> corresponding to autonomous and driven nonautonomous systems, respectively. Here we can naturally ask a question: If tis regarded as quasi-time and both of eigenfunctions {In>} and eigenvalues {xn} as dynamical variables, what will be the resultant dynamical system? While Pechukas derived some interesting equations along this line, 4 > his major concern was lying in the derivation of level-spacing distributions such as Wigner-Dyson2 > or Poisson distributions. Therefore, he chose, besides the variables of {xn}, matrix elements for a nonintegrable part of H(t) as dynamical variables. But, this methodology inevitably suppresses the essential part of information possessed by eigenfunctions: In the case of the finite (N-) dimensional Hilbert space, for instance, a set of eigenfunctions of H(t) has N 2 complex variables, while nondiagonal matrix elements employed by Pechukas have only N(N -1)/2 complex variables. Since the ingredients of quantum chaos are involved not only in energy spectra but also in wavefunctions, it is highly desirable to take {In>} and {xn} on the same footing. We now concentrate on autonomous nonintegrable systems. Let H(t)=Ho+ tV be the Hamiltonian for a quantum bound system which is classically nonintegrable. Ho and V are a classically integrable part and a nonintegrable perturbation, respectively. Both of them are Hermitian, i.e., Hot=Ho and Vt= V. A single parameter t denotes the strength of nonintegrability. A manifold with a definite symmetry is considered. Then we can assume discrete eigenvalues {xn(t)} to be nondegenerate, suppressing a negligible possibility of accidental degeneracies. (When the Hamiltonian includes several nonintegrability parameters, this assumption is not valid which will be described in § 5.) Eigenfunctions {ln(t)>} are complex orthonormal and form a complete set. If we take t as quasi-time, equations of motion for Xn(t), ln(t)> and <n(t)l can be obtained from the time (real time!)-independent Schrodinger equation H(t)ln(t)>=xnln(t)> as 15 > dxn/dt=Pn[ = Vnn=<n(t)l Vln(t)>], (4·1a) (4 ·1b) (4·1c) (4·1d) Equations (4·1c) and (4·1d) have been derived as follows: Taking the t derivative of the SchrOdinger equation, we have Vln>+Hdln>/dt=Pnln>+xndln>/dt. Its lefthand side equals ~lm> Vmn+ ~lm>xm<ml(dln>/dt) m m where <ml(dln>/dt)= Vmn(xn-xm)- 1 for m=l=n and <nl(dln>/dt)=(d<nl/dt)ln>=O are used. Noting xn(t)=l=const (=0), we arrive at Eqs. (4·1c) and (4·1d). Here, the equality <nl(dln>/dt)=O and its Hermitian conjugate prove to be valid so long as a single parameter is changed (see § 5). For the study below, it is convenient to rewrite 392 K. Nakamura Eqs. (4 ·1) in a perfectly canonical formalism. Let us define Anm as (4·2) which reduces to Anm=<nli- 1 [Ho, V]lm>=<niAim>. The operator A is /-independent and proves to be Hermitian. A can thereby be decomposed as A= L t L. Here L is an appropriate /-independent operator which has its unique inverse L - 1• Anm thus becomes (4·3) Anm=<niLt Lim>. By using Eqs. (4·2) and (4·3), Eqs. (4·1b~d) become dPn/dt=2 m(*n) ~ AnmAmn(Xn-Xm)- 3 , (4·1b') (4 ·1c') (4·1d') Equation (4·1c') and (4·1d'), though substantially the same as Eqs. (4·1c) and (4·1d), take advantageous forms for our study on the complete integrability below. Lin> and <niL t are now read as complex dual vectors. In contrast with previous results, 4' Eqs. (4 ·1a) and (4 ·1b' ~ d') describe the dynamics for both eigenfunctions and eigenvalues and suggest a perfectly canonical formalism. We now show that these equations can also be attained from an effective 'classical' Hamiltonian. Let us introduce a classical N-particle Hamiltonian with internal complex dual space in N dimensions for each particle: (4·4) with Anm defined as a scalar product of complex vectors <niLt and Lim>, consistent with Eqs. (4·2) and (4·3). Applying Poisson brackets given by aA aB ) aiLin> a<niLt ' (4·5) one gets a set of canonical equations from Eq. (4·4) as 15 ' dxn/dt={fi, Xn}=ailjapn, (4·6a) dPn/dt={fi, Pn}=-afi/axn, (4·6b) d(iLin>)/dt={il, iLin>}=ail;a<niLt, (4·6c) d(<niLt)/dt={il, <niLt}=-ail/aiLin>. (4·6d) Equations (4·6) together with Eq. (4·4) are exactly the same with Eqs. (4·1a) and (4 ·1b' ~d'), if the number N in the latter is finite. Clearly, this equivalence holds even in the limit N=oo. Equations (4·6) with (4·4) prove to be a Calogero-Moser N-particle system in 1 + 1 dimensions with internal complex vector space, which Quantum Chaos 393 happened to be proposed in a completely different context of nonlinear dynamics,I 6 > though in the present case Amn, Lin> and <niLt are defined in a much more intricate way. The number of degree of freedom of the system, Eqs. (4·6), is N(N+1), consisting of N for positions and momenta of N particles (levels) and N 2 for internal freedom (complex eigenstates) possessed by particles (levels). The complete integrability of Eqs. (4·6) will be elucidated, merely by rewriting the novel findings of Gibbons and Hermsen 16> in our notations. Let us define N X N matrices: Pnm = OnmPn + (1- Onm)iAnm(Xn- Xm)-l , Xnm=OnmXn, L=(LID, LIZ>,···, Lin>,···, LIN>), where Ll n> are N-component column vectors. Then, Lax representation of Eqs. (4 · 6) can be obtained. Here we show its complete algebraic solution: X(t)= U(t)[X(O)+ P(O)t] u- 1(t), P(t)= U(t)P(O)U- 1(t), L(t)=L(O)U- 1(t), Lt(t)=U(t)Lt(o). (4·7) U(t) is a unitary matrix defined in terms of the time-ordering operator T, i.e., Thus, if we would have a knowledge of eigenfunctions and eigenvalues at t = + 0 (i.e., in the integrable or weak-coupling limit), those at t >O (i.e., in the nonintegrable or strong-coupling regime) will be provided by the solution in Eqs. (4·7). While we have been confining ourselves to the autonomous system, the same argument as above proves to hold even for nonintegrable driven systems in the case when the driving field consists of periodic pulse trains.17l The problem of the time (real time!)-dependent Schrodinger equation is now reduced to an eigenvalue problem for the one-period propagator: exp(- itV)·exp( -2m"Ho)ln(t))=exp(- ixn(t))in(t)>, (4·8) where t is a nonintegrability parameter, representing the pulse strength. By taking the t derivative of Eq. (4·8), we obtain: 17> dxn/dt=Pn, dPn/dt= 41 m-¢n ~ <niLt Llm><miLt Lln>cos[(xn-xm)/2]sin- 3[(xn-xm)/2], d(Lin>)/dt=( -i/4) ~ m*n Llm><miLt Lln>sin- 2 [(xn-xm)/2], 394 K. Nakamura (4 ·9) L and L t are !-independent operators, though their definition here is a bit different from Eq. (4·3). We find that Eqs. (4·9) constitute a generalized Sutherland system in 1 + 1 dimensions whose complete integrability was proved by Nakamura and Mikeska. 17> In this way, we arrive at the issue: There exists a universal 'classical' dynamical system which underlies the quantum chaos. It is a completely-integrable manyparticle system with internal complex vector space, whose effective Hamiltonian is given by N 1 1 N H= "'£12Pi+---{~1 ~1 <niLt ·Lim><miLt ·Lin>.P(xn-xmlw, w'). (m*n) (4·10) Poisson brackets and equations of motion also take universal forms described by Eqs. (4·5) and (4·6), respectively. In Eq. (4·10), .P(zlw, w') is the doubly-periodic Weierstrass function. .P(zloo, oo)=z- 2 and .P(ziO, i7r/2)= fsin- 2 (z/2) correspond to the nonintegrable autonomous and driven (pulsed) nonautonomous systems, respectively. § 5. Reduction to a field-theoretical complex-Grassmannian sigma model -several parameters case So long as a single parameter is to be changed in the Hamiltonian, the desymmetrized spectra exhibit no degeneracy expect for accidental ones. We can see only avoided crossings. The presence of a multitude of avoided crossings is a quantummechanical manifestation of the classical chaos. In§ 4, this situation was examined from a viewpoint of nonlinear dynamics. When several parameters are to be changed, however, we see degeneracies in general. 18 > For example, crossings of eigenvalues at degenerate points have codimensions 2, 3 and 5 depending on the real, complex and quaternionic Hermitian nature of Hamiltonians, respectively. The codimension here implies the number of independent parameters. Let us consider the eigenvalue problem: d H(t)in(t)>=(Ho+ L! tp. Vp.)in(t)>=xn(t)in(t)>, P.=l (5·1) where t=(ft, t2, ···, td) denotes a set of nonintegrability parameters. Here the adiabatic ansatz is imposed: Throughout the change of coordinates in the highdimensional parameter space (t), no transition between eigenstates is assumed to occur. Just as in the previous section, the Hamiltonian H(t) in Eq. (5·1) is assumed to have bound spectra. Hot =Ho and VP. t = VP. for ,u=1, 2, ···,d. Let us consider the desymmetrized spectra of Eq. (5 ·1). The spectra are now constructed in the space t, in which each eigenvalue constitutes a hyper-energy surface. In general, we can see the contacts of adjacent energy surfaces at degenerate points. In the neighbourhood of each degeneracy, a pair of energy surfaces constitute two sheets of a double-cone (diabolo). 19 > Figure 7 illustrates a diabolical degenerate point and an avoided crossing in the parameter space. The diabolical point is the source of topological singularity: Let us consider a closed circuit around a degenerate point in t space (see Fig. 7). Quantum Chaos Then, the eigenfunction ln(t)> joining the degeneracy acquires the topological phase6 > given by (we take Einstein's contraction for Greek letters) 395 energy (5·2) This phase is a manifestation of the fact that eigenfunctions are multi-valued in a parameter space, though of course I single-valued in the configuration space. I I -i<nlatpln> in Eq. (5·2) plays the role of a gauge potential, which will be shown below. We may expect that the presc ence of many degenerate points and avoided crossings in energy spectra Fig. 7. Diabolical degenerate point and avoided crossing in 2D parameter space. would be an indicator of the quantum chaos characterized by several nonintegrability parameters. We take {t"}, {xn(t)} and {ln(t)>} as d-dimensional 'space-time coordinates', components of a vector field and elements of a matrix field, respectively. By making the t" derivative of Eq. (5 ·1), we obtain (hereafter, atp is abbreviated as a") Q:j (5·3) with A;:'n=- i<mla"ln>. (5·4) We have also a"xn=<nl V"ln>. (5·5) Equations (5·3) and (5·4) form a closed system, which governs {In>}. On the other hand, {xn} can be solved, separately by using Eq. (5·5) whose right-hand side is completely determined by the solution {In>} in Eqs. (5·3) and (5·4). So, we focus our attention on Eqs. (5·3) and (5·4). Let us define the rna trix field X=(ID, 12>, ·--,IN>). (5·6) Equations (5·3) and (5·4) are then rewritten as a~=iXA", (5. 3') Ap=-iXta~. (5 ·4') These forms remind us of a complex-Grassmannian sigma model. 20> In fact, for a transformation 396 K. Nakamura X'(t)=X(t) U(t) with (5·7) ut U=1, (5·4') transforms as (5·8) which is characteristic of a non-Abelian gauge field. Under the transformation in Eqs. (5·7) and (5·8), Eq. (5·3') remains invariant. Equations like (5·3') prove to be obtained from the gauge-invariant action: (5·9) The topological invariants are given by (5·10) Diagonal elements Qnn in Eq. (5·10), which were recently introduced by Mead in the case d=2 (for an infinitesimal rectangle), 21 > are generalizations of the adiabatic phase in nondegenerate systems. 6> (So long as one treats a single level, the 3rd term in the integrand in Eq. (5·10) does not appear except for in the degenerate case. 2 ll' 22 l) The system in Eqs. (5·3') and (5·4') is completely integrable and admits instanton solutions in the case d=2. Further details will be described elsewhere.23> Thus, the problem of quantum chaos characterized by several nonintegrability parameters proves to be reduced to a simplified version of the d -dimensional complex-Grassmannian sigma model. Singularities at degenerate points are described by gauge fields. We can extend the present theory to a driven system as well. § 6. Summary and discussion In the initial half of the present paper, we have shown two realistic examples of quantum chaos in condensed matter physics. The first one is concerned with the chaotic cyclotron motion of electrons in a conducting disk of mesoscopic size. The irregularity in a single-electron energy spectrum is found to be induced by a multitude of avoided level-crossings. Resultant homogeneous level structures lead to a reduction of diamagnetism in 2D dilute electron-gas systems. We find also the critical magnetic field below which the anomalous feature of diamagnetism can be observed. These findings will be a vehicle for our future study on open billiards: If we make small holes on the billiard boundary and connect them with current-flowing thin channels, we have a new possibility of observing the quantum chaos by means of quantum transports such as magneto-conductance or Hall conductance. We may see there an interesting interplay between a deterministic quantum chaos and disorder (impurity)-induced quantum Hall effect in 2D. The other example is a semiclassical spin system radiated by periodic pulse trains. Such a spin system is conceivable, e.g., in an assembly of spin 1/2 systems in electron spin resonances. We find: Despite the complete absence of the classical and quantum correspondence, semiclassical wavefunctions exhibit turbulent diffusions in Quantum Chaos 397 the long-time regime. The wavefunctions show also a quantum analog of the coexistence of KAM tori and chaos in a transitional region leading to a global chaos. Its invariant probability measure is characterized by anomalous fluctuations of singularities.14> Our spin system, when suitable dissipations will be incorporated, will yield remarkable phenomena such as chaotic trains of spin-echo bursts. This system will also be a nice model to simulate Bloch oscillations in driven Josephson junctions24 > because Cooper-pair operators in a small junction can be described by quantum spins. 25 > In the last half of the present paper, we have found a universal dynamical system lying behind the quantum chaos. In case when a single nonintegrability parameter characterizes the Hamiltonian, the eigenvalue problem is reduced to a (1 +I)dimensional many-particle system, where the parameter and both of eigenfunctions and eigenvalues are regarded as quasi-time and dynamical variables, respectively. The above system has inter-particle interactions of Weierstrass Sf function type and proves to be completely integrable. On the basis of our findings, many conjectures become possible: An avoided crossing and its motion for a change of the quasi-time may be interpreted as a single soliton motion. 26 > Excitation of small or large number of solitons will correspond to the near-integrable or chaotic nature of the underlying classical dynamical system, respectively. The statistical mechanics approach attempted by Dyson5>and Pechukas4>is of course interesting, though it is not a unique way of analysing the quantum chaos. Regarding this approach, however, the proof for the ergodicity hypothesis on our completely-integrable system remains unprovided. Extention of our analyses has been made to more generic quantum systems characterized by several nonintegrability parameters. Since we see many degeneracies in this case, a drastic change in the methodology is required. A set of nonintegrability parameters and both of eigenfunctions and eigenvalues are taken as 'spacetime coordinates' and field variables, respectively. Matrix gauge fields in Eq. (5·4) are introduced, whose diagonal elements give rise to the adiabatic phase of current interest. 6> The resultant field theory proves to be described by a simplified version of a d-dimensional complex-Grassmannian sigma model. The gauge fields capture the topological singularities at degenerate points. In a particular case d=2, this model admits instanton solutions with various topological charges. Excitation of small or large number of instantons will be related to the near-integrability and chaos of underlying classical systems, respectively. The mixing and ergodic features of classical chaos have helped to establish its relationship with the formalism of equilibrium statistical mechanics. 27 > In the field of quantum chaos, however, neither the nonzero Kolmogorov-Sinai entropy nor positive Lyapunov exponent can be expected, except for in systems with infinite degrees of freedom and/ or in some of open quantum systems. Nonetheless, we see many complicated issues in the quantum-mechanical treatment of chaotic systems. Both the universal dynamical system and complex-Grassmannian sigma model that we have obtained will be a guiding vehicle for understanding the above complexities. 398 K. Nakamura Acknowledgements The main part of the present paper is based on a series of our latest works. The author would like to thank collaborators, in particular, C. Jeffries, M. Lakshmanan, H. J. Mikeska, A. Shuda and H. Thomas. He acknowledges valuable conversations with P. Gaspard and S. A. Rice and their nice hospitalities during his sabbatical leave at the University of Chicago. References 1) 2) 3) 4) 5) 6) 7) 8) 9) 10) 11) 12) 13) 14) 15) 16) 17) See, e.g., Chaotic Behavior in Quantum Mechanics, ed. G. Casati (Plenum, New York, 1985); Quantum Chaos and Statistical Nuclear Physics, ed. T. H. Seligman and H. Nishioka (Springer·Verlag, Berlin, 1986). M. L. Mehta, Random Matrices and Statistical Theory of Energy Levels (Academic, New York, 1967). M. Carmeli, Statistical Theory and Random Matrices (Dekker, New York, 1983). E. Abrahams, P. W. Anderson, D. C. Licciardello and T.V. Ramakrishnan, Phys. Rev. Lett. 42 (1979), 67...3. For an application of this concept to a kicked rotator, see : S. Fishman, D. R. Grempel and R. E. Prange, Phys. Rev. Lett. 49 (1982), 509; D. L. Shepelyansky, Phys. Rev. Lett. 56 (1986), 677. P. Pechukas, Phys. Rev. Lett. 51 (1983), 943; ]. Phys. Chern. 88 (1984), 4823. See also, T. Yukawa, Phys. Rev. Lett. 54 (1985), 1883; Phys. Lett. A116 (1986), 227; H. Hasegawa, H. ]. Mikeska and H. Frahm, Phys. Rev. A38 (1988), 395. F. ]. Dyson, ]. Math. Phys. 3 (1962), 1191. M. V. Berry, Proc. R. Soc. London A392 (1984), 45. B. Simon, Phys. Rev. Lett. 51 (1983), 2167. K. Nakamura and H. Thomas, Phys. Rev. Lett. 61 (1988), 247. K. Nakamura, in Proceedings of Symposium on Magnetism, Washington, D. C., August 15·16, 1988 (Springer·Verlag, Berlin), in press. R. Peierls, Surprises in Theoretical Physics (Princeton Univ. Press, Princeton, 1979). M. Robnik and M. V. Berry, ]. of Phys. A18 (1985), 1361. K. Nakamura, S. Ohta and K. Kawasaki,]. of Phys. C15 (1982), L143. S. Ohta and K. Nakamura,]. of Phys. C16 (1983), L605. H. Suhl and X. Y. Zhang, Phys. Rev. Lett. 57 (1986), 1480. S. P. Lim and D. L. Huber, Phys. Rev. B37 (1988), 5426. T. L. Gill and W. W. Zachary, Phys. Lett. 128A (1988), 419. G. Gibson and C. Jeffries, Phys. Rev. A29 (1984), 811. · M. Mino and H. Yamazaki, J. Phys. Soc. Jpn. 55 (1986), 4168. F. M. de Aguiar and S. M. Rezende, Phys. Rev. Lett. 56 (1986), 1070. T. L. Carroll, L. M. Pecora and F. ]. Ratchford, Phys. Rev. Lett. 59 (1987), 2891. P. Bryant, C. Jeffries and K. Nakamura, Phys. Rev. Lett. 60 (1988), 1185. For further references, see H. Yamazaki and M. Mino, this issue., p. 400. K. Nakamura, Y. Okazaki and A. R. Bishop, Phys. Rev. Lett. 57 (1986), 5. K. Nakamura, A. R. Bishop and A. Shudo, Phys. Rev. B39 (1989), 12422. T. Halsey, M. H. Jensen, L. P. Kadanoff, ]. Procaccia and B. I. Shraiman, Phys. Rev. A33 (1986), 1141. K. Nakamura and M. Lakshmanan, Phys. Rev. Lett. 57 (1986), 1661. K. Nakamura, in Proceedings of Workshop on Nonlinear Evolution Equations : Integrability and Spectral Methods, Como, july 4·15, 1988 (Manchester Univ. Press), in press. J. Gibbons and T. Hermsen, Physica Dll (1984), 337. See also, S. Wojciechowski, Physica Scripta 34 (1986), 304. K. Nakamura and H.]. Mikeska, Phys. Rev. A35 (1987), 5294. Quantum Chaos 18) 19) 20) 21) 22) 23) 24) 25) 26) 27) 399 J. von Neumann and E. P. Wigner, Z. Phys. 30 (1929), 467. E. Teller, J. Phys. Chern. 41 (1937), 109. M. V. Berry and H. Wilkinson, Proc. R. Soc. London A392 (1984), 15. A. D' Adda, M. LUscher and P. Di Vecchia, Nucl. Phys. B146 (1978), 63. H. Eichenherr, Nucl. Phys. B146 (1978), 215. C. A. Mead, Phys. Rev. Lett. 59 (1987), 161. F. Wilczek and A. Zee, Phys. Rev. Lett. 52 (1984), 2111. ]. E. Avron, L. Sadun, J. Segert and B. Simon, Phys. Rev. Lett. 61 (1988), 1329. K. Nakamura et al., in preparation. K. K. Likharev and A. B. Zorin, J. Low Temp. Phys. 59 (1985), 347. D. J. Thouless, Phys. Rev. 117 (1960), 1256. A vivid realization of this idea has successfully been made in P. Gaspard, S. A. Rice and K. Nakamura, preprint. R. Bowen, Equilibrium States and the Ergodic Theory of Anosov Diffeomorphisms (Springer-Verlag, Berlin, 1975). D. Ruelle, Thermodynamic Formalism (Addison-Wesley, Reading, Massachusetts, 1978).