betaFeedback

WorksPhysics and Astronomy

Nonlinear dynamics and chaos in multidimensional disordered Hamiltonian systems

Abstract

We study the chaotic behavior of multidimensional Hamiltonian systems in the presence of nonlinearity and disorder. It is known that any localized initial excitation in a large enough linear disordered system spreads for a finite amount of time and then halts forever. This phenomenon is called Anderson localization (AL). What happens to AL when nonlinearity is introduced is an interesting question which has been considered in several studies over the past decades. However, the characteristics and the asymptotic fate of such evolutions still remain an issue of intense debate due to their computational difficulty, especially in systems of more than one spatial dimension. As the spreading of initially localized wave packets is a non-equilibrium thermalization process related to the ergodic and chaotic properties of the system, in our work we investigate the properties of chaos studying the behavior of observables related to the system's tangent dynamics. In particular, we consider the disordered discrete nonlinear Schrödinger (DDNLS) equation of one (1D) and two (2D) spatial dimensions. We present detailed computations of the time evolution of the system's maximum Lyapunov exponent (MLE–ΛΛ), and the related deviation vector distribution (DVD). We find that although the systems' MLE decreases in time following a power law tαΛt^{α_Λ} with αΛ<0α_Λ<0 for both the weak and strong chaos regimes, no crossover to the behavior Λ∝t−1Λ\propto t^{-1} (which is indicative of regular motion) is observed. In addition, the analysis of the DVDs reveals the existence of random fluctuations of chaotic hotspots with increasing amplitudes inside the excited part of the wave packet, which assist in homogenizing chaos and contribute to the thermalization of more lattice sites.

Phd Dissertation

Supervisor: A/Prof. Charalampos Skokos

Plagiarism declaration

I, the undersigned, hereby declare that the work contained in this thesis is my original work, and that any work done by others or by myself previously has been acknowledged and referenced accordingly.

Bertin MANY MANDA, 19 October 2020.

Acknowledgments

The completion of my doctoral project was only made possible due to the outstanding support and guidance of my supervisor A/Prof. Haris Skokos. The several months he spent proof-reading this thesis is a glance into how engaged and supportive he has been during these last 4 years.

I am grateful to my colleagues Bob Senyange, Malcolm Hillebrand and Henok Moges for proof-reading this work and for the innumerable fruitful discussions I had with them. We spent hours arguing on analytical and numerical methods, which contributed enormously to improving several aspects of my doctoral work.

During my PhD work, I had the opportunity to interact with many other people whose feedback increased my understanding of nonlinear lattice dynamics and deeply shaped the content of this work. More specifically I mention Prof. Anastasios Bountis (Nazarbayev University, Kazakhstan), Dr. George Kalosakas (University of Patras, Greece), Dr. Carlo Danieli (Max Planck Institute for the Physics of Complex Systems, Germany), Dr. Thudiyangal Mithun (University of Massachusetts, USA), Dr. Konstantinos Kaloudis (Nazarbayev University, Kazakhstan), Dr. Thomas Oikonomou (Nazarbayev University, Kazakhstan) and Mr. Malcolm Hillebrand (University of Cape Town, South Africa) with whom I had the privilege to co-author research papers. In addition, the insights I obtained from discussions with Prof. Sergej Flach (Institute for Basic Science, Korea), Prof. Panayotis Kevrekidis (University of Massachusetts, USA) and Dr. Yagmur Kati (Institute for Basic Science, Korea) are also encompassed in this work. Furthermore, I thank Prof. Giorgos Tsironis (University of Crete, Greece) who as an organizer financially assisted me to participate in the International Conference ‘Nonlinear Localization in Lattices’ which was held in Spetses, Greece in June 2018 and where I met most of the researchers I mentioned above.

Finally, I would like to thank my whole family. My brothers and sisters for their encouragement and my mother Marie Bernadette Biloa Onana who made travelling to South Africa possible through great financial and personal sacrifices.

This work was funded by the National Research Foundation of South Africa under the Grant Number: 98974. I also received additional support from the University of Cape Town. I thank the Centre for High Performance Computing of South Africa and the University of Cape Town’s ICTS High Performance Computing Centre for providing the computational resources and the personal assistance necessary for the successful completion of this work.

Publications and conference presentations

P1. B. Senyange, B. Many Manda, and Ch. Skokos, Characteristics of chaos evolution in one-dimensional disordered nonlinear lattices, Physical Review E, 2018, 98, 052229. DOI: doi.org/10.1103/PhysRevE.98.052229.

P2. C. Danieli, B. Many Manda, T. Mithun, and Ch. Skokos, Computational efficiency of numerical integration methods for the tangent dynamics of many-body Hamiltonian systems in one and two spatial dimensions, Mathematics in Engineering, 2019, 1, 447. DOI: doi.org/10.3934/mine.2019.3.447.

P3. B. Many Manda, B. Senyange, and Ch. Skokos, Chaotic wave-packet spreading in two-dimensional disordered nonlinear lattices, Physical Review E, 2020, 101, 032206. DOI: doi.org/10.1103/PhysRevE.101.032206.

Other publications:

P4. M. Hillebrand, B. Many Manda, G. Kalosakas, E. Gerlach, and Ch. Skokos, Chaotic dynamics of graphene and graphene nanoribbons, Chaos: An Interdisciplinary Journal of Nonlinear Science, 2020, 30, 063150. DOI: doi.org/10.1063/5.0007761.

P5. B. Senyange, J.-J. Plessis, B. Many Manda, and Ch. Skokos, Properties of Normal Modes in a Modified Disordered Klein-Gordon Lattice: From Disorder to Order, Nonlinear Phenomena in Complex Systems: An Interdisciplinary Journal, 2020, 23, 165. DOI: doi.org/10.33581/1561-4085-2020-23-2-165-171.

P6. A. Bountis, K. Kaloudis, Th. Oikonomou, B. Many Manda, and Ch. Skokos, Stability properties of 1-dimensional Hamiltonian lattices with non-analytic potentials, International Journal of Bifurcation and Chaos, 2020, 30, 2030047. DOI: doi.org/10.1142/S0218127420300475.

P7. T. Mithun, A. Maluckov, B. Many Manda, Ch. Skokos, A. Bishop, A. Saxena, A. Khare, and P. G. Kevrekidis, Thermalization in the one-dimensional Salerno model lattice, Submitted to Physical Review E, 2020. Preprint version: physics.class-ph/2012.03652.

Conference presentations:

  1. Poster presentation: Dynamical behavior of a 2D planar model of graphene. The International Conference on Nonlinear Localization in Lattices, Spetses, Greece, June 2018.

  1. Poster presentation: Chaotic wave packet spreading in nonlinear disordered media. The UCT & SU’s Science Postgraduate Symposium, Stellenbosch, South Africa, September 2019.

  1. Oral presentation: Chaotic wave packet spreading in the one- and two-dimensional disordered nonlinear Schrödinger equations. The 62nd Annual Congress of the South African Mathematical Society, Cape Town, South Africa, December 2019.

Chapter 1

General introduction

Dynamics of disordered systems

Disordered dynamical systems intend to model the heterogeneity of nature. They are obtained by breaking the spatial translational symmetry (isotropy) of Hamiltonian models by attributing to one or many of their parameters random values to each single degree of freedom. In principle, the random numbers are drawn from any chosen probability distribution and could directly affect the eigen-energies (see e.g. [Anderson, 1958; Gredeskul and Kivshar, 1992; Kramer and MacKinnon, 1993; Shepelyansky, 1993; Kopidakis et al., 2008; Pikovsky and Shepelyansky, 2008]), the on-site interactions (see e.g. [Kopidakis et al., 2008; Skokos et al., 2009; Skokos and Flach, 2010; Laptyeva et al., 2010; Bodyfelt et al., 2011a]), the inter-particle interactions (see e.g. [Prosen and Robnik, 1992; Dauxois et al., 1993; Barré and Dauxois, 2001; Roy and Pikovsky, 2012; Martínez et al., 2016; Hillebrand et al., 2019; Ngapasare et al., 2019; Pikovsky, 2020]) and the particle’s masses (see e.g. [Achilleos et al., 2016, 2018; Ngapasare et al., 2019; Hillebrand et al., 2020]) among others. The study of disordered systems has a long standing history, starting in the mid 20{sup}th century with the explosion of modern technologies such as semiconductors. Studying a linear electron tight-binding approximation of the latter material, P. W. Anderson [Anderson, 1958; Kramer and MacKinnon, 1993] discovered that initially localized electron matter waves remain in that state for an infinitely long time. This phenomenon is now referred as Anderson localization (AL) and has been experimentally shown to occur in a vast number of physical systems (see e.g. [14, 1] (Chabanov et al. 2000; Runge and Zimmermann 2003; Genack and Chabanov 2005; Billy et al. 2008; Hu et al. 2008; Kondov et al. 2011; Sheinfux et al. 2017)).

The question of what happens to AL when nonlinearity is introduced has attracted extensive work in theory and simulations [Shepelyansky, 1993; Shapiro, 2007; Kopidakis et al., 2008; Pikovsky and Shepelyansky, 2008; Flach et al., 2009; Skokos et al., 2009; Garcia-Mata and Shepelyansky, 2009; Veksler et al., 2009; Mulansky and Pikovsky, 2010; Laptyeva et al., 2010; Skokos and Flach, 2010; Krimer and Flach, 2010; Flach, 2010; Johansson et al., 2010; Basko, 2011; Bodyfelt et al., 2011a,b; Aubry, 2011; Ivanchenko et al., 2011; Molina et al., 2012; Vermersch and Garreau, 2012; Michaely and Fishman, 2012; Basko, 2012; Mulansky and Pikovsky, 2012; Laptyeva et al., 2012; Milovanov and Iomin, 2012; Lucioni et al., 2013; Vermersch and Garreau, 2013; Mulansky and Pikovsky, 2013; Skokos et al., 2013; Laptyeva et al., 2013; Tieleman et al., 2014; Ivanchenko et al., 2014; Antonopoulos et al., 2014; Ermann and Shepelyansky, 2014; Laptyeva et al., 2014; Basko, 2014; Flach, 2015; Achilleos et al., 2016; Flach, 2016; Martínez et al., 2016; Achilleos et al., 2018; Antonopoulos et al., 2017; Iomin, 2017; Sato and Klages, 2019; Vakulchyk et al., 2019; Rivas and Molina, 2020], as well as in experiments [Schwartz et al., 2007; Roati et al., 2008; Lahini et al., 2008; Lucioni et al., 2011]. Two typical disordered Hamiltonian models are at the center of these studies, namely the disordered Klein-Gordon (DKG) chain of nonlinear oscillators and the disordered discrete nonlinear Schrödinger equation (DDNLS). These systems were extensively studied in several of the works mentioned above. There it was shown that for large enough nonlinearity and disorder strengths, AL is destroyed. This leads to different dynamical subdiffusive spreading regimes which are characterized by a growing second moment of the wave packet, according to the power law tamt^{a_m} with 0<am<10 < a_m < 1. In particular, the weak chaos with am=1/(1+2D)a_m = 1/(1 + 2D) (DD being the lattice’s dimensionality) [Flach et al., 2009; Skokos et al., 2009; Flach, 2010] and strong chaos dynamical regimes where am=1/(1+D)a_m = 1/(1 + D) can occur [Laptyeva et al., 2010; Flach, 2010; Bodyfelt et al., 2011a].

The remarkable success of the theoretical framework explaining the diffusion process comes from the fact that nonlinearity couples normal modes, which interact with each other and allow the transfer of energy [Gredeskul and Kivshar, 1992; Shepelyansky, 1993; Flach et al., 2009; Flach, 2010]. This transfer happens in a chaotic manner whose evolution allows the randomization and thermalization of the wave packet [Flach, 2010; Basko, 2011; Skokos et al., 2013; Senyange et al., 2018; Many Manda et al., 2020].

Nevertheless, whether the system can sustain the underlying spreading for all times remains a controversial issue. Indeed a possible slowing down of the wave packet spreading is suggested by [Mulansky et al., 2011; Pikovsky and Fishman, 2011] as well as an eventual halt of it by [Johansson et al., 2010; Aubry, 2011], but the vast majority of the numerical simulations do not support such behavior [Shepelyansky, 1993; Pikovsky and Shepelyansky, 2008; Flach et al., 2009; Skokos et al., 2009; Laptyeva et al., 2010; Skokos and Flach, 2010; Bodyfelt et al., 2011a; Laptyeva et al., 2012; Skokos et al., 2013; Kati et al., 2020].

A notable distinction between all these possible outcomes is related to the tangent dynamics of the system which supports its chaotic behavior. Indeed, indicators of chaos often appears to be very sensitive to changes in the system and therefore could act as effective dynamical order parameters in several complex models [Pikovsky and Politi, 2016] and in particular in multidimensional Hamiltonian systems (see for e.g. [Barré and Dauxois, 2001; Hillebrand et al., 2019]). Therefore, understanding the characteristics of chaos in this context appears to be a fundamental, open question. Several works have been performed trying to fill this void. The connection between chaos and nonlinear wave spreading is investigated in [Mulansky et al., 2011] for small sized systems after their regularization, while an interesting chaotic behavior has been observed in the spreading of wave packets in [5]. In addition, [Michaely and Fishman, 2012] analyzed the temporal and spectral properties of the effective nonlinear strengths within the wave packet and identified it as noisy. Furthermore, [12] computed the spectral entropy, which is related to chaotic and ergodic properties of systems and measured the temporal evolution of chaos, but only for relatively small times. Much closer to what we have studied in this thesis, the asymptotic temporal evolution of chaos was studied for the one-dimensional (1D) DKG system in the weak chaos regime by [Skokos et al., 2013]. There, it was depicted a constantly decaying chaos strength during the wave packet’s evolution, and was visualized the fluctuations of localized chaotic hotspots mentioned in previous studies (see e.g. [Basko, 2011, 2012; Michaely and Fishman, 2012]) as necessary for the wave packet thermalization.

Subject of the thesis

One aim of this thesis is to show that the chaotic processes which were observed in [Skokos et al., 2013] for the weak chaos regime in the 1D DKG system, which possesses a single integral of motion, also appear in the 1D DDNLS model which admits an extra integral of motion compared to the 1D DKG system, thus generalizing in some sense the results obtained in that paper. In addition, we will take that study a step further, investigating also the dynamics in the strong chaos case; as well as providing a ‘complete’ quantification of the system’s chaoticity through the calculation of observables in the system’s tangent space, such as the maximum Lyapunov exponent (MLE) [Skokos, 2010; Pikovsky and Politi, 2016] or the deviation vector distribution (DVD) statistics [Dauxois et al., 1997; Barré and Dauxois, 2001; Skokos et al., 2013]. These two qualities are complementary: While the MLE estimates the mean chaos strength of the entire system, the DVD looks at local properties of chaos giving us information about which lattice sites contribute the most to the overall chaotic dynamics. A basic novelty of our work is that we do not only focus on the 1D version of the system, but we also expand our study to the two-dimensional (2D) version of the system whose subdiffusive spreading has barely been investigated [Garcia-Mata and Shepelyansky, 2009].

In order to reach our goals, we perform large scale numerical experiments in both the 1D and 2D DDNLS lattices, as well as semi-analytical calculations in some cases to interpret our findings. We note that the equations of motion of these DDNLS models are particularly difficult to integrate from a computational point of view. This is demonstrated by the fact that results for the DKG systems, whose numerical integration is a much easier task, are typically presented for final times which are one or two orders of magnitude larger than the DDNLS systems in both one (see e.g. [Senyange et al., 2018]) and two spatial dimensions (see e.g. [Many Manda et al., 2020]). Another important point we want to emphasize is the lack of efficient numerical integration techniques for the evolution of the tangent dynamics of the DDNLS systems in one and two spatial dimensions on top of the difficulty of numerically evolving the equations of motion of these systems. Therefore, we dedicate an important part of this work to creating novel numerical methods for the fast and accurate long time integration of the equations of motion and the variational equations for both the 1D and 2D models of the DDNLS system using the so-called tangent map method [Skokos and Gerlach, 2010; Gerlach and Skokos, 2011; Gerlach et al., 2012]. The utilization of the tangent map method takes advantage of symplectic integrators (SIs) which are specifically designed for the numerical evolution of trajectories of Hamiltonian systems. The implementation of the techniques we develop in this work allow us to reach computational performances comparable to the ones seen in previous studies, where only the integration of the equations of motion was done despite adding the significant computational load of the variational equations. Consequently this approach helps us to extract robust asymptotic behaviors of the wave packet chaotic dynamics (quantified through the evolution of the MLE for instance).

As we will see later in this thesis, systems exhibiting AL offer a perfect testbed for the analysis of chaotic processes in nonlinear dynamical systems as the destruction of AL is a purely nonlinear phenomenon [Shepelyansky, 1993; Garcia-Mata and Shepelyansky, 2009; Flach et al., 2009]. It is remarkable that the interplay between disorder and nonlinearity results in the localization of chaos, a behavior which will be discussed in some detail in our work. We will also explain how the combined action of the non vanishing chaotic strength along with the random fluctuations of localized chaos spots are characteristics of the chaotic dynamics of wave packet spreading in disordered media. Consequently, as long as they persist the propagation of wave packets never stops, even if the system is getting closer to the integrable limit. This result is general as it was also observed for the DKG model in one [Skokos et al., 2013; Senyange et al., 2018; Senyange, 2021] and two [Many Manda et al., 2020; Senyange, 2021] spatial dimensions.

The thesis is organized as follows. In Chap. 2, we discuss the notion of AL, emphasizing its manifestation in the 1D disordered discrete linear Schrödinger (DDLS) model, and review the theory of spreading of nonlinear wave packets. In Chap. 3, we introduce the numerical quantities we calculate in our study and describe in detail the numerical techniques we developed for the integration of the equations of motion and the variational equations of the DDNLS models. In Chap. 4, we present our numerical results for the chaotic behavior of the 1D DDNLS model, while Chap. 5 is dedicated to a similar analysis but for the 2D DDNLS model. Finally, in Chap. 6 we conclude our work, summarize our findings and present a few ideas for future investigations.

Localization and spreading of wave packets in disordered lattices

Introduction

In this chapter, we discuss the different regimes of wave packets’ dynamics we encounter in random disordered lattices. The word ‘disorder’ mentioned here could refer to impurities, vacancies, dislocations, distortions or defects which occur in idealistic perfect periodic lattices. A periodic lattice is invariant under space translation, which means that interactions are the same at every location within the lattice, and thus constitutes an idealistic representation of physical systems in nature.

The first part of this chapter is dedicated to a rather concise review of the extended work done on the analysis of models obtained through linear approximations of disordered systems. In this context we show how initially localized wave packets remain in that state forever. This exceptionally important result was initially obtained by Anderson in his seminal paper [Anderson, 1958] where he tried to understand the metal insulator transition. Nowadays it is well known that this so-called Anderson localization (AL) phenomenon appears in several other fields, making it an important problem to study in modern physics (see e.g. [Izrailev, 1990; Lahini et al., 2008, 2009; Krimer et al., 2009; Krimer and Flach, 2010; Savin et al., 2010]). We can depart from linear approximations to obtain more realistic models by adding nonlinear couplings between system’s components, through the inclusion of higher order terms from Taylor series expansions of the interaction potential functions. For example, high intensity photon beams propagating through optical waveguides create a nonlinear response of the medium, which in turn affects the wave propagation [Schwartz et al., 2007; Lahini et al., 2008]. Thus, the second part of this chapter focuses on the effects of nonlinearity on wave localization in nonlinear disordered systems. In that part, we review what we believe are the most important aspects of this behavior.

The chapter is organized as follows. In Sec. 2.1 we present a fundamental linear disorder model in order to define AL. In Sec. 2.2 we present the nonlinear models we are going to investigate, while in Sec. 2.3 we elaborate using (semi-) analytical arguments the different dynamical behaviors of the nonlinear wave packet in disordered lattices. Section 2.4 is similar to Sec. 2.3, but we base our analysis on statistical physics arguments. In Sec. 2.5 we explain the mechanism of effective wave packet propagation in the different dynamical regimes, while in Sec. 2.6, we discuss the implications of the existence of initial extended wave packets which turn out to be of practical importance for numerical simulations of two-dimensional systems. Finally, in Sec. 2.7 we summarize the content of the chapter.

Anderson localization

The classical theory of electron transport in conductive metals predicts that electron delocalization is due to inelastic multiple scatterings of the electron, which collides with atomic nuclei which are several times heavier, leading to the well known result of Ohm’s law with the Drude conductivity σE∝τ\sigma_E \propto\tau. Here τ\tau is the average time elapsed between two collisions and σE\sigma_E is the conductivity within the metal. Therefore, τv\tau v is the electron mean-free-path (with vv being the average electron’s velocity), which characterizes the mean distance between two successive scatterings of the electron. An interesting question which arises is what happens if disorder (in the form of a random potential around each nuclei generated from an arbitrary source) is introduced and tuned up higher and higher within the system? Will the conductivity be reduced due to the increasingly smaller electron mean-free-path? Or, will this process result to something more peculiar?

Anderson decided to tackle this problem in the mean-field approximation of the electron tight binding model whose Hamiltonian reads [Anderson, 1958; Kramer and MacKinnon, 1993]

HL=∑l=1Nϵl∣ψl∣2−Jl,l+1(ψl+1ψl⋆+ψlψl+1⋆).(1)\mathcal{H}_L=\sum_{l=1}^{N}\epsilon_l|\psi_l|^2-\mathcal{J}_{l,l+1}\left(\psi_{l+1}\psi_l^{\star}+\psi_l\psi_{l+1}^{\star}\right). \tag*{(1)}

Here ψl∈C\psi_l \in\mathbb{C} is the wave function, which describes the state of the electron at site ll and ψl⋆\psi_l^{\star} is its complex conjugate. The off-diagonal matrix elements Jl,m\mathcal{J}_{l,m}, 1≤l,m≤N1 \le l,m \le N, with NN denoting the size of the lattice, describes the strength of the hopping between sites ll and mm. In addition, ϵl\epsilon_l is a random parameter with units of the energy/frequency (here the system is rescaled such that ℏ=1\hbar=1). Its value at each site is drawn from a uniform probability distribution on the interval [−W/2,W/2][-W/2,W/2] with density PW=1/WP_W=1/W. Consequently, WW is the disorder strength as it defines the size of the interval from which the random values ϵl\epsilon_l are drawn. In Eq. (1) we consider that the electron can only diffuse to its nearest neighbors as also shown in Figure 2.1. In addition, we also assume that the probabilities for the electron to tunnel to

Figure 2.1. Schematic representation of the electron tight-binding model. The electron [grey dot] is attached to atom’s nucleus [black dots] which are considered to be fixed. Each atomic site possesses random on-site electron binding energies [red bars] selected from a uniform distribution on the interval [−W/2,W/2][-W/2,W/2] of measure WW [red vertical segment in the left]. The electron at site ll, can tunnel either to site l−1l-1 or site l+1l+1 with equal transmission probabilities J\mathcal{J} often referred as hopping [black arrow on the top of the electron]. Thus, the elements of J\mathcal{J} of the form Jl,l−1\mathcal{J}_{l,l-1} and Jl,l+1\mathcal{J}_{l,l+1} are equal and taken to unity (and zero otherwise).

the left or right site are equal and we set all the non-zero elements of the matrix J\mathcal{J} to Jl,m=1\mathcal{J}_{l,m}=1. In Eq. (1) the NN pairs (ψl,iψl⋆)(\psi_l,i\psi_l^{\star}) represent a set of canonical generalized coordinates of the system’s phase space. The equations of motion of the Hamilton function Eq. (1) are then obtained through

dψldt=∂HL∂(iψl⋆),(2)\frac{d\psi_l}{dt}=\frac{\partial\mathcal{H}_L}{\partial(i\psi_l^{\star})}, \tag*{(2)}

leading to the DDLS equation

iψ˙l=ϵlψl−(ψl−1+ψl+1),(3)i\dot{\psi}_l=\epsilon_l\psi_l-(\psi_{l-1}+\psi_{l+1}), \tag*{(3)}

where the dot refers to the derivative with respect to time tt.

We note that spatial localization means that the probability ∣ψl∣2|\psi_l|^2 of the presence of the electron at site ll exponentially decays, so that the electron does not tunnel out of the lattice [Kramer and MacKinnon, 1993; Krimer and Flach, 2010]. Over the years several methods have been developed in order to analytically and numerically observe such localization [Kramer and MacKinnon, 1993]. For instance, a crude approach consists of putting a lump of matter (preferably located at a single site) into the system and let it evolve for a very long time. If localization occurs, then at the end of the evolution we should expect a spreading of the matter to few subsequent lattice sites and a halting of this process. In this case, the number of sites on which the amount of matter is not exponentially small is a characteristic length of the system that we refer as localization length or localization volume. We shall emphasize the difference between these two appellations in the later paragraphs.

The normal modes (NMs) of a linear system represent an orthonormal basis of the system’s phase space. Therefore, any system’s wave packet can be written as a weighted sum over these modes and the wave packet localization implies that the modes are also localized. Following this observation, one can therefore focus in analyzing the characteristic of the NMs of the DDLS system (3).

Let us outline the steps of such procedure. The solution of linear wave equations consists of plane waves of the form

ψl=Ale−iEt,(4)\psi_l = A_l e^{-iEt}, \tag*{(4)}

where AlA_l is the amplitude of the wave at site ll and EE its frequency. Substituting Eq. (4) into Eq. (3) we obtain the equivalent eigenvalue problem

EAl=ϵlAl−Al−1−Al+1,(5)E A_l = \epsilon_l A_l - A_{l-1} - A_{l+1}, \tag*{(5)}

which reduces in a matrix form to

EIΨ=A~Ψ,(6)E I \boldsymbol{\Psi} = \widetilde{A}\boldsymbol{\Psi}, \tag*{(6)}

where Ψ=(A1,A2,…,AN)\boldsymbol{\Psi} = (A_1, A_2, \ldots, A_N),

A~=(ϵ1−10⋯00−1ϵ2−1⋯00⋮⋮⋮⋱⋮⋮000⋯ϵN−1−1000⋯−1ϵN)andI=(100⋯00010⋯00⋮⋮⋮⋱⋮⋮000⋯10000⋯01).(7)\widetilde{A} = \begin{pmatrix} \epsilon_1 & -1 & 0 & \cdots& 0 & 0 \\ -1 & \epsilon_2 & -1 & \cdots& 0 & 0 \\ \vdots& \vdots& \vdots& \ddots& \vdots& \vdots\\ 0 & 0 & 0 & \cdots& \epsilon_{N-1} & -1 \\ 0 & 0 & 0 & \cdots& -1 & \epsilon_N \end{pmatrix} \quad\text{and}\quad I = \begin{pmatrix} 1 & 0 & 0 & \cdots& 0 & 0 \\ 0 & 1 & 0 & \cdots& 0 & 0 \\ \vdots& \vdots& \vdots& \ddots& \vdots& \vdots\\ 0 & 0 & 0 & \cdots& 1 & 0 \\ 0 & 0 & 0 & \cdots& 0 & 1 \end{pmatrix}. \tag*{(7)}

Solving the eigenvalue problem (6) is equivalent to diagonalizing the matrix A~\widetilde{A}. We then obtain the normalized eigenstates Ψν=(Aν,1,Aν,2,…,Aν,N)\boldsymbol{\Psi}_\nu= (A_{\nu,1}, A_{\nu,2}, \ldots, A_{\nu,N}) with ∥Ψν∥2=∑lAν,l2=1\lVert\boldsymbol{\Psi}_\nu\rVert^2 = \sum_l A_{\nu,l}^2 = 1 associated with the eigenfrequencies EνE_\nu.1 An example of a computed NM Ψν\boldsymbol{\Psi}_\nu through this process of numerical diagonalization of A~\widetilde{A} (see Eqs. (6) and (7)) is shown in Figure 2.2. The parameters of system (1) are set to W=4W = 4 and N=1021N = 1021. The computed NMs are sorted by increasing order of their centers’ spatial position which is computed as lˉν=∑ll∣Aν,l∣2\bar{l}_\nu= \sum_l l|A_{\nu,l}|^2. In Figure 2.2, we plot the profile of the mode whose center is closest to the lattice’s middle (the mode’s number in the ordering in ν=357\nu= 357), associated with the eigenfrequency Eν≈−0.916E_\nu\approx-0.916. The spatial exponential localization of the mode is evident from the results of Figure 2.2(b).

Figure 2.2. A representative NM of Hamiltonian (1) with N=1021N = 1021 sites located around the center of the lattice (mode number ν=357\nu= 357). Amplitudes (a) Aν,lA_{\nu,l} and (b) ∣Aν,l∣2|A_{\nu,l}|^2. Panel (b) is in semi-logarithmic scale. The system disordered strength is W=4W = 4.

The NMs of a system exhibiting AL are characterized by an asymptotic exponential spatial decay from their center (here referred as lˉν\bar{l}_\nu) [Kramer and MacKinnon, 1993] of the form

References

  1. [1]M. Störzer, P. Gross, C. M. Aegerter, and G. Maret. Observation of the critical regime near Anderson localization of light. Physical Review Letters, 96(6):063904, 2006.
  2. [2]Z.-Y. Sun and X. Yu. Anomalous diffusion of discrete solitons driven by evolving disorder. Physical Review E, 101(6):062211, 2020.DOI
  3. [3]M. Suzuki. Fractal decomposition of exponential operators with applications to many-body theories and Monte Carlo simulations. Physics Letters A, 146(6):319–323, 1990.DOI
  4. [4]M. Suzuki. General theory of fractal path integrals with applications to many-body theories and statistical physics. Journal of Mathematical Physics, 32(2):400–407, 1991.DOI
  5. [5]O. Tieleman, C. Skokos, and A. Lazarides. Chaoticity without thermalisation in disordered lattices. EPL (Europhysics Letters), 105(2):20001, 2014.
  6. [6]D. Tong. Classical dynamics. University of Cambridge, Cambridge, 2004.
  7. [7]D. Tong. Statistical physics. University of Cambridge, Cambridge, 2012.
  8. [8]G. Tsironis and S. Aubry. Slow relaxation phenomena induced by breathers in nonlinear lattices. Physical Review Letters, 77(26):5225, 1996.pubmed.ncbi.nlm.nih.gov/10062747
  9. [9]I. Vakulchyk, M. V. Fistul, and S. Flach. Wave packet spreading with disordered nonlinear discrete-time quantum walks. Physical Review Letters, 122(4):040501, 2019.DOI
  10. [10]H. Veksler, Y. Krivolapov, and S. Fishman. Spreading for the generalized nonlinear Schrödinger equation with disorder. Physical Review E, 80(3):037201, 2009.pubmed.ncbi.nlm.nih.gov/19905249
  11. [12]B. Vermersch and J. C. Garreau. Spectral description of the dynamics of ultracold interacting Bosons in disordered lattices. New Journal of Physics, 15(4):045030, 2013.
  12. [13]R. B. White. The theory of toroidally confined plasmas. World Scientific Publishing Company, 2013.
  13. [14]D. S. Wiersma, P. Bartolini, A. Lagendijk, and R. Righini. Localization of light in a disordered medium. Nature, 390(6661):671–673, 1997.DOI
  14. [15]D. Wilczak, S. Serrano, and R. Barrio. Coexistence and dynamical connections between hyperchaos and chaos in the 4d Rossler system: a computer-assisted proof. SIAM Journal on Applied Dynamical Systems, 15(1):356–390, 2016.DOI
  15. [16]A. Wingen, K. Spatschek, and S. Abdullaev. Stochastic transport of magnetic field lines in the symmetric tokamak. Contributions to Plasma Physics, 45(7):500–513, 2005.DOI
  16. [17]H. Wolkowicz and G. P. Styan. Bounds for eigenvalues using traces. Linear algebra and its applications, 29:471–506, 1980.DOI
  17. [18]T. F. Xavier. Derivatives on unequally spaced grids, 2019. URL http://cococubed.asu.edu/code_pages/fdcoef.shtml.
  18. [19]H. Yoshida. Construction of higher order symplectic integrators. Physics Letters A, 150(5-7):262–268, 1990.
  19. [20]H. Yoshida. Recent progress in the theory and application of symplectic integrators. In Qualitative and Quantitative Behaviour of Planetary Systems, pages 27–43. Springer, 1993.
  20. [21]W.-C. Yueh. Eigenvalues of several tridiagonal matrices. Applied Mathematics e-notes, 5(66-74):210–230, 2005.

Paper details

vlnc:W_zgg6bhky
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 7

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents