Professional Documents
Culture Documents
1310.8116v1 ScalingChaosStronglyNonlinearLattices Pikovsky
1310.8116v1 ScalingChaosStronglyNonlinearLattices Pikovsky
Mario Mulansky1, 2
1
Department of Physics and Astronomy, Potsdam University, Karl-Liebknecht-Str 24, D-14476, Potsdam-Golm, Germany
2
Max-Planck-Institut für Physik komplexer Systeme, Nöthnitzer Str. 38, D-01159 Dresden, Germany
(Dated: October 31, 2013)
Although it is now understood that chaos in complex classical systems is the foundation of thermo-
dynamic behavior, the detailed relations between the microscopic properties of the chaotic dynamics
and the macroscopic thermodynamic observations still remain mostly in the dark. In this work, we
numerically analyze the probability of chaos in strongly nonlinear Hamiltonian systems and find
different scaling properties depending on the nonlinear structure of the model. We argue that these
different scaling laws of chaos have definite consequences for the macroscopic diffusive behavior,
as chaos is the microscopic mechanism of diffusion. This is compared with previous results on
chaotic diffusion [New J. Phys. 15, 053015 (2013)], and a relation between microscopic chaos and
macroscopic diffusion is established.
Bridging the gap between chaotic trajectories in lier works [12]. Specifically, we look at the probability
the microscopic perspective and the macroscopic phe- of chaos in chains of nonlinear oscillators, which is the
nomenology of thermodynamic behavior is one of the fun- probability to have a chaotic trajectory when starting
damental problems of statistical physics and still remains from some random initial condition, thus it is also a mea-
an open challenge for theoreticians. In recent years, sure for the fraction of the phase space belonging to the
substantial progress has been made in this direction by chaotic component. We analyze the scaling of this proba-
studying the diffusive behavior in nonlinear disordered bility around the transition point from chaotic to regular
models. Starting from a localized initial state, e.g. an phase space in dependence of the system size. We find
Anderson mode, one finds spreading of energy due to the in numerical studies that this scaling depends crucially
nonlinear interactions. on the choice of nonlinearity. By connecting the scaling
This phenomenon was first understood as the “destruc- behavior of chaos with the phenomenon of chaotic diffu-
tion of Anderson localization by weak nonlinearities” [1] sion in such systems we are able to deduce the existence
and studied extensively in numerical experiments using of fundamentally different spreading laws. Indeed, such
the Discrete Nonlinear Schrödinger Equation with local a difference in the spreading behavior has been observed
disorder, e.g. [2–5], and [6] for a recent review. Later, it in nonlinear oscillator chains [7, 8] indicating that the
was realized that this effect can be viewed in the broader detailed properties of chaos play a fundamental role for
context of “chaotic diffusion” which led to an increased thermodynamic effects.
understanding of the spreading phenomenon [7, 8]. The This article is structured as follows: first we introduce
most prominent observation is a subdiffusive spreading the model of Hamiltonian oscillator chains, followed by
of the width of an initially localized wave packet L ∼ tν , a description of the numerical procedure to measure the
ν < 1/2 [1–3]. In contrast to these numerical results, probability of chaos. Then we present our results and
rigorous arguments showed that on very long time scales analyze the scaling of chaos and finally the implications
spreading has to be slower than any power law [9], while for nonlinear spreading are discussed.
a perturbation analysis of the nonlinearity suggested a The subject of study in this work are Hamiltonian
propagating front moving as ln t beyond which localiza- chains of harmonic or nonlinear oscillators with nonlin-
tion occurs [10]. It is, however, unclear if those ana- ear nearest-neighbor coupling. Such models are called
lytical results are applicable for the parameter regimes strongly nonlinear because of the absence of linear cou-
and time scales considered in the numerical experiments. pling terms and hence the absence of linear waves. They
Nevertheless, there also exist numerical results indicating belong to the class of systems considered in [14]. The
a slowing down of spreading away from the subdiffusive Hamilton function for this model is written in terms of
power-law, based on the scaling of chaos [11] but also on position qk and momentum pk of the oscillator at site k:
direct numerical observations of spreading in nonlinear N 2 N −1
ωk2 κ
X pk 1 X
Hamiltonian oscillator chains [7, 12]. A possible mecha- H= + q + (qk+1 − qk )λ . (1)
nism for the devation from power-law spreading was pre- 2 κ k λ
k=1 k=1
sented in [13]. The on-site potential has power κ ≥ 2, so we consider
Here, we want to draw further connection between harmonic or nonlinear local oscillators, while the cou-
the microscopic chaos and the diffusion properties in pling is strictly nonlinear λ > 2 and of higher power
strongly nonlinear Hamiltonian chains introduced in ear- λ > κ. In the following, we will consider two choices
2
of the nonlinear powers: first a fully nonlinear model is changed. Hence this parametrization is better suited
where κ = 4 and λ = 6, called model A. The second for numerical simulations, while the original model (1)
choice are harmonic oscillators with nonlinear coupling: allows to easier connect the results with the spreading
κ = 2, λ = 4 which will be called model B. Further- behavior of trajectories.
more, we consider regular lattices with ωk = 1, but also We will now examine the emergence of regular regions
the case of disorder where ωk ∈ [1/2, 3/2] chosen uni- in phase space when approaching the integrable limit of
formly and independently identically distributed. For zero density. Therefore, a measure of the fraction of
the fully nonlinear case in Model A, the introduction of the chaotic phase space is introduced which we call the
disorder makes no fundamental difference for the scaling “probability of chaos” Pch . This quantity has already
of chaos as will be seen later. For harmonic oscillators, been followed for the Discrete Nonlinear Schrödinger
on the other hand, disorder in terms of random oscil- Equation [11] and Hamiltonian oscillator chains [15, 16].
lator frequencies is crucial as otherwise the whole chain Pch is measured numerically by categorizing trajectories
would always be in resonance which strongly favors the as being chaotic or regular based on the largest Lyapunov
existence of chaos. Note, that for studying chaotic dif- exponent λ. Therefore, we simulate a trajectory accord-
fusion one usually considers disordered systems to elim- ing to (1) starting from a random initial condition in
inate nonlinear waves [7]. For fixed nonlinearities κ and equilibrium (uniform energy distribution) and calculate λ
λ and disorder ωk , the only parameter in model (1) is by standard methods. As explained above, we use the
the conserved total energy E = H(q, p). An important parametrization (3) for our numerical simulations. The
observable in this system is the local energy at site k: initial conditions are chosen in the following way: first
the positions qk and momenta pk of the oscillators are
p2k ωk2 κ 1 initialized as random numbers chosen from a normal dis-
Ek = + q + [(qk+1 − qk )λ + (qk − qk−1 )λ ]. (2)
2 κ k 2λ tribution with zeros mean and variance σ 2 : N (0, σ). The
From λ > κ it then follows that in the limit of small variance is calculated such that the expectation value of
local energies Ek → 0 the system (1) becomes an in- the energy density has the desired value hEk′ i = 1, i.e.
tegrable chain of uncoupled oscillators. Assuming a E ′ = N , where the average here is taken over many initial
roughly uniform energy distribution, this corresponds to conditions. Then, in a correction step, the momenta are
the limit of vanishing energy density W → 0, where scaled such that the actual energy density of this initial
W = hEk i = E/N . Consequently, for W → 0 one ex- condition has exactly the desired value E ′ (p, q)/N = 1.
pects a completely regular phase space without chaotic The numerical time evolution of the trajectories is com-
regions. puted using a fourth order symplectic algorithm [17] and
Note, that system (1) can also be parametrized differ- we performed the simulations on GPU devices using the
ently by introducing a coupling parameter β: Boost.odeint C++ library [18, 19].
We compute the Lyapunov exponents for M = 2000
N ′2 N −1 such initial conditions for several lattice sizes N
ωk2 ′κ
X p β X ′
H′ = k
+ q + (qk+1 − qk′ )λ . (3) and coupling parameter β. Then we perform
2 κ k λ
k=1 k=1 a back-transformation to parametrization (1) with
W = β κ/(λ−κ) according to (4). The result is a distribu-
This involves rescaled amplitudes qk′ , p′k and a rescaled
tion of λ for each parameter value N and W . Figure 1a
time t′ and gives the coupling strength β as the only
shows histograms of these Lyapunov exponents for Model
parameter while the dynamics become independent from
A of length N = 32 with a regular local potential ωk = 1
the total energy E ′ [8]. Fixing the rescaled total energy to
and energy densities W = 1.4 · 10−13 , 3.8 · 10−9 , 3 · 10−7 .
E ′ = N , one finds that changing the nonlinear strength
One clearly sees that for small densities almost all Lya-
corresponds to a change of the original energy according
punov exponents are close to λ ≈ 10−5 . For regular
to [8]:
dynamics the maximum Lyapunov exponent should be
E = N β κ/(λ−κ) . (4) exactly zero, but numerically, with finite integration time
T = 106 , such small values are below the numerical accu-
For the numerical simulations presented below, we used racy and essentially indicate regular motion. For larger
this second parametrization (3). This is especially valu- energy densities one finds Lyapunov exponents of the or-
able for simulations of model A, where κ = 4 and λ = 6, der of λ ≈ 0.01, clearly indicating chaotic dynamics.
as there for formulation (1) the uncoupled oscillator fre- Based on this observation we introduce a threshold
quencies and thus the fundamental time-scales depend on value λc = 20/T = 2 · 10−5 and consider all trajecto-
the nonlinearity parameter E. This would imply that the ries with smaller Lyapunov exponents λ < λc as regular,
simulation times need to be increased for smaller densi- and those with larger values chaotic λ > λc . Note, that
ties W = E/N . In the rescaled formulation however, the this threshold is solely defined by the finite integration
energy density and thus the typical frequencies are kept time. Longer integrations would allow for smaller λc .
constant and only the nonlinear coupling parameter β Using this distinction of regular and chaotic trajectories
3
N =8
Pch
600 0.4
0.2
N =16
400 N =32
0.0
200 1.0 N =8 N 1.5 W
0.8 N =16 Spreading
0 0.6 N =32
Pch
−5 −4 −3
log10λ
−2 −1 0.4 N =64
0.2
1.0 (b) 0.0 (b)
0.8 10-12 10-10 10-8 10-6 10-4 10-2
N W
1.5
0.6
Pch
0.2
N =16 terpreted as an effective increase of the dimensionality
N =32 of phase space accessed by the trajectory. So assuming
0.0 N =64 W = E/L and N ∼ L, the scaled variable in Fig. 2 in-
creases for spreading states as N 1.5 W ∼ L1.5 W ∼ L0.5 E.
10-4 10-3 10-2 10-1 100
N 0.35 W That means in the course of spreading the trajectory is
driven away from the regular parts of the phase space!
This is a quite surprising, even counter-intuitive result as
FIG. 3: (color online) Probability of chaos for Model B with one would naturally expect that for small energy densi-
local disorder ωk ∈ [1/2, 3/2] in dependence of the scaled
ties eventually a KAM-regime will be reached. But this
density N 0.35 W .
is only true for a fixed system size. For spreading states,
however, together with decreasing W , also the effective
dimensionality is increased and it turns out numerically
is not addressed here, but we briefly note that the W 2 that for model A the latter effect is stronger (∼ N 1.5 vs.
behavior derived for harmonic oscillators does not hold ∼ 1/N ). This is also indicated by the arrows in Figure 2.
for nonlinear oscillators, because due to the amplitude Hence, the KAM regime will not be reached for spread-
dependent frequencies the resonance structure of nonlin- ing states in this case. This is true for both regular and
ear oscillators is completely different from the harmonic disordered potential in model A with κ = 4, λ = 6.
case. For the harmonic oscillators in model B, the scaling of
For the case of harmonic oscillators with random fre- chaos was found above as N 0.35 . With the same assump-
quencies ωk ∈ [1/2, 3/2] and nonlinear coupling, κ = 2, tions as before this means that the probability of chaos
λ = 4, the scaling exponent was found to be α24 = 0.35. for spreading states decreases as N 0.35 W ∼ E/L0.65 with
Notably, this value is in perfect agreement with the nu- time. The trajectory thus moves towards the regular
merical results reported by Basko, who found for the in- KAM-regime while spreading, where more and more reg-
termediate regime that g(W ) ∼ W 2.85 [16], which trans- ular islands appear in the phase space. Hence the typical
lates to α = 1/2.85 = 0.35. The numerical results with phase space environment for spreading trajectories is fun-
a rescaled density according to this parameter value are damentally different for nonlinear (model A, κ = 4) and
shown in Figure 3. Again, the rescaled curves show a harmonic (model B, κ = 2) oscillators. Therefore, one
convincing overlap indicating also here that changing the can expect a qualitative difference in the spreading be-
system size only shifts the transition point. So we find havior for these two cases. It should be noted, however,
that for both cases, nonlinear and harmonic local oscilla- that a quantitative implication from the different scaling
tors, the size dependence of the transition from chaotic to observation for the spreading can not be deduced. More-
regular phase space can be modeled with a simple power- over, the fact that the spreading trajectories are driven
law Wtr ∼ N −α . The numerical value for the exponent, towards the regular regime does not necessarily mean
however, is significantly different: for the fully nonlinear that the spreading process has to stop at some point.
case we find α46 = 1.5, while for harmonic oscillators we Although the phase space becomes more and more reg-
get α24 = 0.35. This different scaling has its origin in the ular there always exist thin layers of chaos where the
different resonance structure of harmonic and nonlinear trajectory can travel along, a mechanism called Arnol’d
oscillators, but a rigorous derivation of the exact scaling Diffusion [21].
exponents α is beyond the scope of this work. Nevertheless, a qualitative difference for spreading in
The scaling law identified above has some remark- fully nonlinear and harmonic oscillators has indeed been
able consequences for the spreading process in such se- reported recently [7, 12]. There, the spreading process
tups. This spreading is typically observed in the follow- was quantified in terms of the average time required to
ing sense: One starts with only a few excited oscilla- excite one new oscillator ∆T as function of the current
tors in the center of a large chain of model (1) while all excitation length L. For details on this observable and its
other oscillators are at rest. Then, as time evolves, more averaging we refer to [7]. Here, we only want to empha-
and more oscillators become excited due to the nearest- size the fact that for model A a pure power-law spreading
neighbor interactions. The number of excited oscillators, was found, while model B showed a clear deviation from
usually called excitation width L, grows but consequently the power-law behavior and thus a qualitatively differ-
the energy density of the excited oscillators decreases be- ent spreading. This is illustrated in Figure 4 showing
5
7
the past, we were able to explain the deviation from the
6.5
pure power-law spreading reported recently. Analyzing
6
E=0.001
the properties of the microscopic chaos we were thus able
5.5
E=0.002
E=0.003
to provide the first explanation for this slowing down of
5
4.5
E=0.005
E=0.01
spreading, which has been controversially discussed in
4
E=0.02
E=0.05
the recent past. Although the comparison of spreading
3.5
2.6*x-3.4 states with the properties of chaos for fixed sized chains
2.8 3 3.2 3.4 3.6 3.8 4 4.2 4.4 4.6 4.8
seems physically reasonable, it should be noted that this
log10 L/E
is merely an assumption and requires further fortifica-
10
tion. Nevertheless, we believe that the results reported
8 here provide new and valuable insight on how the proper-
8 6 ties of microscopic chaos influences macroscopic processes
a(W )
4 such as diffusion.
6
log10 ∆T /L
2
I thank A. Pikovsky for numerous fruitful discussions.
0
0 1 2
4 − log10 W