20.11.2012 Views

Physics@Technion

Physics@Technion

Physics@Technion

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Anderson localization<br />

in some nonlinear<br />

systems<br />

Yevgeny Bar-Lev (né Krivolapov),<br />

Shmuel Fishman and Avy Soffer,<br />

Technion, May 2010


Outline<br />

� Anderson localization<br />

� Some Nonlinear phenomena<br />

� Nonlinear optics<br />

� Nonlinearity in many-body physics<br />

� Nonlinearity and Disorder: Motivation<br />

� Heuristics and numerics<br />

� Perturbation theory<br />

� Results<br />

� Summary


Anderson localization<br />

� Anderson model (1958)<br />

where are independent random variables<br />

� All eigenstates are exponentially localized<br />

° = » ¡1<br />

un (x) » exp ¡° (x ¡ xn)<br />

were is the inverse localization length<br />

� Dynamical localization: Transport is<br />

exponentially suppressed


Some nonlinear phenomena<br />

BEC<br />

He 4<br />

Many-Body<br />

Nonlinearity due to<br />

approximations<br />

Dynamics<br />

Gross-Pitaevskii or<br />

NL Schrodinger eq.<br />

Optics/Hydrodynamics<br />

Nonlinearity due to<br />

media<br />

Waves<br />

NL materials<br />

Solitons


Nonlinear Schrödinger Equation<br />

(NLSE)<br />

� Transverse modes of light propagating in<br />

nonlinear media<br />

i @A<br />

@z<br />

= ¡ 1<br />

2k0<br />

@2A k0n2<br />

¡<br />

@x2 n0<br />

jAj 2 A<br />

� Bose-Einstein Condensate in a mean field<br />

approximation<br />

i~@tà = ¡ ~2 @<br />

2m<br />

2Ã @x2 + ¯ jÃj2 Ã


Properties of NLSE<br />

� Without external potential and in the<br />

continuum the equation is integrable<br />

Nonlinearity<br />

Attractive<br />

Repulsive<br />

Bright solitons<br />

Dark solitons,<br />

Spreading<br />

wavepacket


Nonlinearity and Disorder: Motivation<br />

Disorder<br />

Localization of<br />

all eigenfunctions.<br />

transport is<br />

suppressed<br />

?<br />

Nonlinearity<br />

Enhancement of<br />

transport<br />

by soliton<br />

formation


Does dynamical localization<br />

survives the nonlinearity?<br />

The model: motivated by experimental relevance<br />

NLSE with a random potential<br />

i@tà = ¡J [Ã(x +1) +Ã(x ¡1)] +"xà +¯ jÃj 2 Ã


Simple heuristic arguments<br />

� Yes, if there is spreading the magnitude of the nonlinear term decreases<br />

and localization takes over.<br />

� Depends, for a localization length of the relevant energy spacing is ,<br />

and the perturbation because of the nonlinear term is<br />

therefore for small β there is no spreading. (Shepelyansky)<br />

� Depends, there will be no spreading for large enough β (theorem), but there<br />

is subdiffusion even for very small β (Flach)<br />

� Yes, because quasiperiodic localized perturbation does not destroy<br />

localization (Soffer, Wang-Bourgain)<br />

»<br />

» ¡1<br />

¯ jÃj 2 ¼ ¯» ¡1


Numerics – second moment<br />

Pikovsky, Shepelyansky<br />

YK<br />

Ð<br />

2<br />

log10 x ® Ð x 2®


Numerics - wavefunction<br />

Pikovsky, Shepelyansky


Problem of all numerics<br />

� “Asymptotics curse”: It is impossible to decide<br />

whether there is a saturation in the expansion of<br />

the wavefunction. In any case it looks like the<br />

expansion is very slow, at most sub-diffusional.<br />

� Convergence: All long time numerics are done<br />

without convergence to true solution.<br />

� Time scale: The time scale of the problem is not<br />

clear


Perturbation theory: Change basis<br />

and expand in<br />

� Expand in eigenstates of the linear Anderson model<br />

� The equation for the expansion coefficients is<br />

where<br />

� Expand cn<br />

in powers of


Example: 1 st order<br />

(0)<br />

n n0<br />

c �<br />

�<br />

� �1 � 000 i( En �E0<br />

) t<br />

c0 ��iVn<br />

e<br />

�t n � 0<br />

n �<br />

n �<br />

0<br />

0<br />

i( E E ) t<br />

(1) 000 �1�e� n � n � �<br />

En�E0 c V<br />

� �<br />

The first problem: Secular terms<br />

c � �iV �t<br />

1 000<br />

0 0<br />

� �<br />

The second problem: Small denominator problem


Elimination of secular terms<br />

Change<br />

Expand<br />

For example:<br />

à (x; t) = X<br />

à (x; t) = X<br />

i@tcn = ¡ E (0)<br />

n ¡ E0 ¢<br />

n cn + ¯ X<br />

cn (t) e<br />

n<br />

¡iE0 nt<br />

cn (t) e un (x)<br />

n<br />

¡iEnt<br />

un (x)<br />

m1m2m3<br />

V m1m2m3<br />

n c ¤ m1cm2cm3e i(E0 n+E0 m ¡E<br />

1 0 m ¡E<br />

2 0 m3)t i@tc (1)<br />

n = ¡E (1)<br />

n ±n0 + V 000<br />

n e i(E 0 n ¡E0 0)t<br />

We set<br />

E (1)<br />

0 = V 000<br />

0


Bounding the general term<br />

The general term is a product of terms of the form<br />

³ m1m2m3<br />

n ´<br />

Using Cauchy-Shwartz inequality<br />

hj³ m1m2m3<br />

n j s i ·<br />

We find<br />

*<br />

¯<br />

1<br />

V m1m2m3<br />

n<br />

E 0 n ¡ fE0 g mi<br />

¯E0 n ¡ fE0g ¯<br />

mi<br />

2s<br />

¯<br />

+ 1=2<br />

Ð m1m2m3 jVn j 2s®1=2 ;<br />

hj³ m1m2m3<br />

n j s ¡1<br />

i · D" 0e 3 (°¡"0 )s ijxn¡xm ij


Structure of general term


Bounding the general term<br />

Finally, we can show using generalized Hölder inequality<br />

* ¯ ¯¯¯¯ X<br />

³ m1m2m3<br />

n ³ m4m5m6<br />

m1<br />

fmig<br />

or for the expansion coefficient<br />

D¯¯c ¯<br />

(k) ¯<br />

n<br />

sE<br />

±<br />

¢ ¢ ¢ ³ 000<br />

mN¡1<br />

¯<br />

¯s+<br />

¯<br />

±<br />

· F (k)<br />

± e¡(°¡"¡"0 )sjxnj<br />

· e 2k F (k)<br />

± e¡(°¡"¡"0 )sjxnj<br />

All the expansion coefficients are exponentially localized uniformly in time !<br />

Does it prove dynamical localization ?


The remainder<br />

Equation of motion<br />

i@tQn = Wn (t) + X<br />

Mnm (t) Qm + X<br />

Qn (t = 0) = 0<br />

m<br />

m<br />

¹Mnm (t) Q ¤ m + F (Q)<br />

For t · t one can neglect the nonlinear term (Bootstrap argument)<br />

¤<br />

jQn (t)j · A¯ N+1 t ¢ e ¡°jnj<br />

t ¤ » O ¡ ¯ ¡1¢<br />

?<br />

t ¤ » O ¡ ¯ ¡N¢


Main result – Logarithmic front<br />

X<br />

jQnj 2 < " 2<br />

If we require that it is satisfied for some<br />

n>¹n<br />

µ ¶ N A¯ t<br />

¹n (t) = » ln<br />

"<br />

The front after which the solution is exponentially bounded and quasi-<br />

periodic in time propagates at most logarithmically !*<br />

* Similar to a conjecture of W.-M. Wang<br />

¹n


¡¹n (t)<br />

µ ¶ N A¯ t<br />

¹n (t) = » ln<br />

"<br />

jª (n; t)j<br />

?<br />

¹n (t) n<br />

Exponential tail Exponential tail


Numerical implementation


lim<br />

jQnj<br />

= 0<br />

¯N Is ?<br />

¯!0


A new time scale<br />

If the series is asymptotic it has an optimal order<br />

Therefore<br />

N! ¢ ¯ N » 1<br />

N = ¯ ¡1<br />

t¤ = O ¡ ¯ ¡N¢ = O<br />

³ 1<br />

¡<br />

¯ ¯<br />

Is this time scale connected to the Lyapunov exponent of the system ?<br />

´


Summary<br />

� Renormalized perturbation theory was developed, including secular<br />

terms removal in all orders, bounding of the general term and the<br />

remainder (for finite time).<br />

� Symbolic and numerical implementation of the theory was done.<br />

Good agreement with numerical solutions and fast calculation.<br />

� The wavepacket spreads at most logarithmically up to large<br />

times.<br />

� The asymptotic nature of the series was supported numerically.<br />

� A new time scale was proposed.


Open questions<br />

� Does the logarithmic front persists to<br />

arbitrarily long time ?<br />

� Are the series asymptotic ?<br />

� Is the new time proposed time scale<br />

connected to the Lyapunov exponent of<br />

the system ?<br />

Thank you !

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!