• No results found

Ehrenfest times for classically chaotic systems

N/A
N/A
Protected

Academic year: 2021

Share "Ehrenfest times for classically chaotic systems"

Copied!
4
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

Ehrenfest times for classically chaotic systems

Beenakker, C.W.J.; Silvestrov, P.G.

Citation

Beenakker, C. W. J., & Silvestrov, P. G. (2002). Ehrenfest times for classically chaotic

systems. Retrieved from https://hdl.handle.net/1887/1219

Version:

Not Applicable (or Unknown)

License:

Leiden University Non-exclusive license

Downloaded from:

https://hdl.handle.net/1887/1219

(2)

RAPID COMMUNICATIONS

PHYSICAL REVIEW E, VOLUME 65, 035208(R)

Ehrenfest times for classically chaotic Systems

P G Silvestiov

12

and C W J Beenakker

1

llnstituut Lorentz Umversiteit Leiden, PO Box 9506 2300 RA Leiden The Netheilands

Budker Institute of Nucleai Physics, 630090 Novosibirsk, Russia

(Received 26 November 2001, pubhshed 7 March 2002)

We descnbe the quantum-mechamcal spreadmg of a Gaussian wave packet by means of the semiclassical WKB approximation of Berry and Balazs [J Phys A 2, 625 (1979)] We find that the time scale τ οη which this approximation breaks down m a chaotic System is larger than the Ehrenfest times considered previously In one dimension τ— |λ~Ίη(/4/Α), with λ the Lyapunov exponent and A a typical classical action

DOI 10 1103/PhysRevE 65 035208 PACS number(s) 05 45 Mt, 03 65 Sq, 03 65 Yz, 73 63 Kv

Accordmg to Ehrenfest's theorem [1], the propagation of

a quantum-mechamcal wave packet is described for short

times by classical equations of motion The time scale at

which this conespondence between quantum and classical

dynamics breaks down is called the Ehrenfest time If the

classical dynamics is chaotic with Lyapunov exponent λ,

then the Ehienfest time τ is of order λ"

1

ln(A/ft) (with A a

typical classical action of the dynamical System) [2] There is

actually more than a smgle Ehrenfest time, correspondmg to

different types of semiclassical approximations Although

they differ only by a numencal coefficient, τ,

= c,\~

1

ln(A/h), the structure of the wave function changes

quahtatively from one time scale to the next

Up to a time τ

{

, with c

l

= i/6, the initial coherent state

will retam its Gaussian form with vamshing erroi in the hmit

h—>0 [3,4] Foi longer times up to τ

2

, with 02= i/2, the

uncertainty m the position and momentum of the paiticle

remams small but the phase-space stiucture of the wave

packet deviates strongly from a Gaussian For times greatei

than τ

2

the wave function no longer has the form of a wave

packet (this is the "mixing regime" of Refs [5,6]), but up to

a time τ

3

it can still be described semiclassically by the

time-dependent WKB approximation of Beny and Balazs

[7] As we will show in this papei, the WKB icpresentation

imphes 03 = 7/6 for a smgle degree of freedom (with simple

geneiahzations for higher dimensions) This is larger than the

value c

3

= 2/3 obtamed by Bouzouma and Robeit [6] from a

diffeient semiclassical approximation

Let us start with the Gaussian one-dimensional wave

packet

1/4

expl p0x

2h (D

Imtially ß(t = 0)=0 and a(t = Q)=p

F

/L, wheie p

F

and L

aie the typical classical momentum and length The typical

classical action is A-p

F

L The paiameters x

0

(t),p

Q

(t)

fol-low the classical tiajectoiy for h<A We will measuie the

momentum and cooidmate in units of p

F

and L, lespectively,

so that a ( 0 ) = l and A = l Foi chaotic dynamics with

Lyapunov exponent λ one has a(t)<*e.\p(—2\t), hence α

«i foi t>l/\

To descnbe the time evolution in phase space we consider

the Wignei function

. η|φ*| v-^L-w'ft—^_

2/ 2irh

a(x-x

0

)

2

[p-p

0

-ß(x-x

0

)]

2

h

ah

(2)

The wave packet is centered at x

0

(t),p

0

(t) and for a(t)

<ä l becomes highly elongated and tilted with slope Αρ/Δχ

^ß It has length /||= ^h(l + β

2

)!α and width Zj_

= ^jhal(l + ß

2

), so that the area in phase space is conserved

exactly, l^l

L

= h The Gaussian quantum wave packet

satis-fies the classical Liouville theorem

The Gaussian form (1) takes into account the elongation

of the wave packet, but not the curvature that develops m

time and lesults in a bendmg of the packet To descnbe the

curvatuie we add an imagmary cubic teim m the exponent in

Eq (i),

πη

1/4

exp

p

0

x (ιβ-α)χ

2 γχ (3)

(Foi simphcity we have put XQ = 0 ) The cubic term leads to

an appieciable phase stuft over a length l\\ — (h/a)

112

when

(y/£)(ft/a)

3 / 2

>l, hence when a(t)<h

y

For a<^h

y

the Wignei function takes agam a simple

form, m terms of the Airy function AI

W(x,p) =

a

1/2

exp( — ax

2

lh) ρ

-AI

l 0

+ βχ+γχ

2

—ρ

\ il/3

(4)

One can check that W(x,p)^S(x)S(p-p

0

) when h—>0 (at

fixed a), by means of the identity lim

e

^

0

Αι(ζ/ε)/ε

= Λ/ττδ(ζ) At finite h the wave packet is extended along the

curved line p = p

0

+ ßx+ γχ

2

Since p,p

0

,x aie of oider

umty, the two paiameteis β and γ aie of oider unity äs well

(in contiast to a, which is <§1) The tiansverse width is of

oidei

(5)

The length of the packet lemams at l^ %/ft(l + ß

2

)/a Since

now /||/

±

S>Ä, the Liouville theoiem no longer holds

(3)

RAPID COMMUNICATIONS

P G SILVESTROV AND C W J BEENAKKER PHYSICAL REVIEW E 65 035208(R)

To obtain the Ehrenfest time, we paiametnze time äs

c

l

Wkk(x,p)= exp

·+-iy^'"\\f(x)\2dy

24h

2-n-h

(6)

The classical hmit for a chaotic System means Ä—>0, f—>co at fixed c Diffeient coefficients c follow from different semi-classical approximations If we use the Gaussian wave packet (1), without the cubic term to account foi the cuiva-tuie, then we need a(t)9>hlßy2ß Smce a^e~2^'^h2c we need c<l/6 The uppei hmit of c gives the first Ehienfest time TI = ^λ~ι ln(l/ft)

The classical hmit can be reached for longer times if we use the wave packet (3), mcluding the cubic teim The di-mensions of the packet for t>r{ scale with h äs

(7)

For c < 1/2 the length of the packet approaches zero m the classical hmit This upper hmit of c gives the second Ehien-fest time τ2= j X "1 ln(l/Ä)

For t>r2 the length of the wave packet exceeds the size of the System and is no longer small compared to the ladius of curvature For these large times we may adopt the semi-classical WKB appioximation of Berry and Balazs [7] Con-sidei a curve m phase space p(x) and a phase-space distn-bution p(p(x),x) Both p and p evolve in accordance with classical equations of motion Foi t> τ2 the function p ( x ) is

multivalued with an exponentially large numbei of branches ~ exp[X(f— r2)] The quantum wave function in this

"mix-mg" regime has the foi m

(8)

The summation over k accounts for the different bianches of the multivalued function p (x) The two functions/and σ are related for Ä^O to p and p by the correspondence pnnciple

(9)

An explicit descnption of the evolution of the wave function (8) for quantum maps can be found in Ref [8]

Near the pomt xb at which p ( x ) bifuicates into two branches, one has p — pb±a\jx—xb, p = b/\lx — xb The wave function theie is

(10)

up to an overall phase The phase diffeience between the bifuication pomts can be deteimmed from Eqs (8) and (9) Because the cuive p ( x ) is not closed, there is no analog of the Bohi-Sommeifeld quantization mle

The Wignei function conesponding to the wave function

(8), bemg quadiatic m Ψ, contams both diagonal (WLk

<x\fk 2) and oscillatmg nondiagonal (V/km^f[fm)

contnbu-tions Fai fiom bifuicacontnbu-tions, the diagonal contubucontnbu-tions to the Wigner function lead

!A i

2(cr'-p)

(£V")1/3 (H)

We have made a Taylor expansion of a(x±y/2) and ne-glected the difference between/(jc±y/2) and/(*)

If we parametnze time äs in Eq (6) we have foi both l^

and /± the same scalmg with fi, äs in Eq (7) The ränge of

vahdity of Eq (8) is limited by the condition that the diffei-ent branches should be distinguishable This lequires that the diffeient parts of the cuive p ( x ) m phase space should not get closer than /_,_ Their spacmg is of oidei 1//| (assummg a umfoim fillmg of phase space), hence

The uppei hmit of 7/6 foi c leads to the thnd Ehienfest time

7 l

r3 =—In- (13)

The third derivative σ'" m Eq (11) vamshes at the pomts

of inflecüon of the curve p ( x ) In oidei to find the Wigner

function there, one should expand σ(χ±γ/2) up to terms of

01 der y5 This leads to a different scalmg I±x f l4 / 5 of the

width of the Wigner function near the inflection pomts Be-cause these are isolated pomts, they will not contiibute to the matnx elements of nonsmgular opeiators (contammg only smooth functions of χ and p) This different scalmg should therefore not affect the Ehienfest time (13)

The nondiagonal contnbutions Wkm to the Wignei

func-tion lead to the "ghost curves" discussed in Ref [9] (Ghost curves are legions of large values of the Wigner function which do not conespond to classical trajectones) The Wigner function near these curves is given by the same Any function äs m Eq (11), but m addition acquires a strongly

oscillatmg factor Due to these oscillations the nondiagonal teims do not contribute to the maüix elements of nonsingulai opeiators (They may play a lole in the decoheience by the envnonment [10]) At r a τ3 the ghost cuives meige with the

(multivalued) curve p (x) and become indistmguishable The time scale (13) for the bieakdown of the WKB ap-proximation is gieatei than the Ehrenfest time j\~lln(l/h)

m the mixed regime obtamed m Ref [6] That shortei time scale may Signal the bieakdown of the senes expansion

ak(x)^i,J = 0akj(x)1iJ Then Eq (9) would no longei hold,

but for t< τ3 the lepiesentation (8) with a lenoimahzed

func-tion ak(x) would still be valid

So fai we have discussed a one-dimensional (1D) chaotic System, which m geneial can be lepiesented by an aiea pie-seivmg map [8] A famihai example is the kicked lotatoi [11] Foi mesoscopic quantum dots, howevei, a moie lel-evant model is the rf-dimensional (cf = 2,3) Schiodmgei equation with a smooth potential V (r) The Gaussian wave packet then takes the foi m

(4)

RAPID COMMUNICATIONS

EHRENFEST TIMES FOR CLASSICALLY CHAOTIC SYSTEMS PHYSICAL REVIEW E 65 035208(R)

x +—x,xn (14)

1=1,2, (16)

Heie S is the action for the classical üajectoiy r0(t) and we

have defined p0 = mr0, x = r—r0, ζίη = β1η + ια1η As be-fore, we rescale the momentum and coordmate such that the typical classical action A = l Imtially, ζιη — ιδ1η Similai to the one dimensional case, aln defines the foim of the packet in coordmate space and β1η = Δρι/Δχη give the angles m phase space Substitutmg the wave function (14) into the Sclirodinger equation one finds Newton's equation of motion foi r0 The spreading of the wave packet m phase space is descnbed by

(15) This is the equation descnbmg the spieadmg m phase space of a small Gaussian bunch of classical paiticles

The Wigner function conespondmg to the wave function (14) has the Gaussian foim W°cexp(-Q;M;„2„/Ä), where

Q = (r—ro,p—po) is a vectoi m 2ui-dimensional phase space The d Lyapunov exponents λ, (ι = 1,2, ,d) govern

the large-time behavior of the eigenvalues /«,= l/m2ii-( + i <*εχρ(2λ,ί) of the leal Symmetrie matnx M Because of en-eigy conservation one Lyapunov exponent vamshes We or-dei the λ 's from large to small, so that λι is the laigest and Xr f=0

The wave packet lemains Gaussian (preseiving the vol-ume °cftrf in phase space) until the cuivature Starts to play a i öle (via a cubic teim in the action) The corresponding

Ehienfest time T1 = ^ X f1l n ( l / Ä ) is the same äs m 1D, only

now it is defined through the laigest Lyapunov exponent X j The second Ehrenfest time, when the length of the packet

exceeds the size of the System, also has the same form τ2

The thiid time τ3 is diffeient foi d = 2,3 from the 1D case Instead of Eq (7), one now has

The longitudmal dimensions l^ correspond to eigenvalues

ml with l ^i=Si/— l , and the tiansveise dimensions /^ to m, with d+2^i^2d The two umt eigenvalues mr f=mr f + 1 = l contnbute anothei factor \fh each to the total volume V m phase space coveied by the wave packet

rf-l rf-l

(17) The available aiea Vmax is lestncted to a shell of constant energy with thickness \fK, hence Vm^\lfi We requue V sVm a x foi the semiclassical appioximation, which leads to the Ehrenfest time

Td-4 A

(18) In conclusion, we exammed different time scales r,

= cIX ~1 ln(l/Ä) foi the bieakdown of diffeient types of

semiclassical appioximations These Ehrenfest ümes differ

only by a numencal coefficient cl, which may seem

msig-mficant However, this difference is actually a Signal of a different power law scahng with h of the volume V in phase space coveied by the wave packet For short times Liou-ville's theoiem dictates V^h Foi long times [parametenzed äs f = (c/X)ln(l/%)] the WKB approximation gives V

α^ν/6-Γ f01 a one_dimensional quantum map (such äs the

kicked lotatoi) and V<xfildl6~1/6~c for a c?-dimensional con-seivative System These diffeient powei laws reflect the fun-damental change in the stiuctuie of the wave function with mcieasmg time and should, therefore, have obseivable con-sequences Two possible applications are the Loschmidt echo [12] and the quantum shot noise [13], where the Ehrenfest time plays a key role

This work was supported by the Dutch Science Founda-tion NWO/FOM and by the NaFounda-tional Science FoundaFounda-tion undei Grant No PHY99-07949

[1] P Ehienfest, Z Phys 45, 455 (1927)

[2] G P Berman and G M Zaslavsky, Physica A 91, 450 (1978), G M Zaslavsky, Phys Rep 80, 157 (1981)

[3] M Combescure and D Robert, Asymptotic Anal 14, 377 (1997)

[4] G A Hagedorn and A Joye, Ann Henri Pomcare l, 837 (2000)

[5] F Bonechi and S De Bievre, Commun Math Phys 211, 659 (2000)

[6] A Bouzouma and D Robert, Duke Math J (to be pubhshed) [7] M V Berry and N L Balazs J Phys A 12, 625 (1979) [8] M V Berry, N L Balazs, M Tabor, and A Voros, Ann Phys

( N Y ) 122,26(1979)

[9] M A Alonso and G W Forbes, J Opt Soc Am A 17, 2288 (2000)

[10] WH Zurek, Nature (London) 412, 712 (2001)

[11] G Casati, B V Chmkov, FM Izrailev, and J Ford, Lect Notes Phys 93, 334 (1979)

[12] R A Jalabert and H M Pastawski, Phys Rev Lett 86, 2490 (2001), Ph Jacquod, PG Silvestrov, and CWJ Beenakker, Phys Rev E 64, 055203(R) (2001)

[13] O Agam, I Alemer, and A Larkm, Phys Rev Lett 85, 3153 (2000), I L Alemer and A I Larkm, Phys Rev B 54, 14 423 (1996)

Referenties

GERELATEERDE DOCUMENTEN

Although in the emerging historicity of Western societies the feasible stories cannot facilitate action due to the lack of an equally feasible political vision, and although

Schrijf je naam op elke bladzijde en start een nieuwe pagina bij elke vraag.. Kladwerk dien je ook in,

The task is now to design a DOE, here a grating that, if illuminated by laser light, transmits only three orders, each containing approximately the same energy. In practice this is

We mvestigate the low-frequency dynamics for transmission or reflection of a wave by a cavity with chaotic scattermg We compute the probabihty distn- bution of the phase derivative φ

The density of states m a noimal metal is suppressed near the Feimi eneigy when it is brought into contact with a superconductoi The history of this pioximity effect goes back to

single exponential with a different coefficient λ ] The volume of phase space that contnbutes to M vanishes in the classical hmit Ä—»0 for times less than the Ehienfest time T E = j

\selectVersion, eqexam will perform modular arithmetic on the number of available versions of a problem, in this way each problem will be properly posed; consequently, when we

Notwithstanding the relative indifference toward it, intel- lectual history and what I will suggest is its necessary complement, compara- tive intellectual history, constitute an