LA-UR-93-4267 / MA/UC3M/10/93 Kink stability, propagation, and length scale competition in the periodically modulated sine-Gordon equation Angel S´anchez Theoretical Division and Center for Nonlinear Studies, Los Alamos National Laboratory, Los Alamos, New Mexico 87545, and Escuela Polit´ecnica Superior, Universidad Carlos III de Madrid, E-28913 Legan´es, Madrid, Spain 4 9 9 1 A. R. Bishop n Theoretical Division and Center for Nonlinear Studies, Los Alamos National Laboratory, a J Los Alamos, New Mexico 87545 4 2 Francisco Dom´ınguez-Adame 1 v Departamento de F´ısica de Materiales, Facultad de F´ısicas, Universidad Complutense, 5 E-28040 Madrid, Spain 0 0 (February 9, 2008) 1 0 4 9 Abstract / l o s - t We have examined the dynamical behavior of the kink solutions of the one- t a dimensionalsine-Gordonequation inthepresenceofaspatially periodicpara- p : metric perturbation. Our study clarifies and extends the currently available v knowledge on this and related nonlinear problems in four directions. First, i X we present the results of a numerical simulation program which are not com- r a patible with the existence of a radiative threshold, predicted by earlier calcu- lations. Second, we carry out a perturbative calculation which helps interpret those previous predictions, enabling us to understand in depth our numerical results. Third, we apply the collective coordinate formalism to this system and demonstrate numerically that it accurately reproduces the observed kink dynamics. Fourth, we report on a novel occurrence of length scale competi- tion in this system and show how it can be understood by means of linear stability analysis. Finally, we conclude by summarizing the general physical framework that arises from our study. Ms. number PACS numbers: 03.40.Kf, 85.25.Cp, 02.90.+p Typeset using REVTEX 1 I. INTRODUCTION. Technological progress has made possible the fabrication of highly ordered materials and structures for a very large number of applications. In parallel to those advances, it has also been realized that the special properties required for many purposes necessitate inhomo- geneous systems. Here inhomogeneity may mean spatial modulations, quasiperiodicity, or disorder of several kinds. In addition, there are other situations in which inhomogeneity is undesirable but unavoidable. In either case, the study of disordered systems acquires fundamental importance. This is even more so when the physical system in which disorder or inhomogenity is to be studied is described by a nonlinear model. Whereas the rˆole of disorder in linear problems is at least partially understood, much less is known about nonlin- ear disordered systems. In fact, even from a mathematical viewpoint, the understanding of these models, often related to stochastic partialdifferential equations (PDE), is very limited. Consequently, a great deal of research has been devoted to this topic [1–3]. A major part of the work done so far regarding nonlinear disordered systems has been concerned with some particular examples that are amenable to analytical treatment while capturing some essential physics. The sine-Gordon (sG) (actually, the whole family of non- linear Klein-Gordon equations, including, e.g., the φ4, double- and quadratic- sine-Gordon equations) and nonlinear Schr¨odinger (NLS) equations are often chosen as very suitable “canonical” examples. This is due to the fact that the basic mathematical structure under- lying them is well known and therefore provides a good starting point for theoretical work. This reason would not be sufficient if these models were not also related to a large number of phenomena that occur in quasi-one-dimensional physical systems, as is in fact the case. In the context of these two models, disorder is introduced through suitably chosen pertur- bation terms (see [2] for an extensive list of physically relevant perturbations). This is the usual procedure by which inhomogeneity of any kind is studied: The equation describing the problem is established, the terms relevant to the considered physical situation are identified, and a perturbation to those terms is introduced, representing the desired kind of disorder. Our viewpoint in this work is more generic: Although the system we deal with is indeed related to a number of applications, our aim is that we will be able to gain insight into un- derlying mechanisms of the phenomenology of nonlinear disordered systems. Therefore, we introduce a simple periodic perturbation which will allow us to study very interesting and general phenomena, such as length scale competition, and will provide information relevant to the more complicated processes occurring in random media (the periodic potential can be interpreted as a “color” of a general noisy one). The knowledge obtained will also be useful to tackle other problems where detailed studies including analytical treatment are not possible. In this paper we study the behavior of one-dimensional (1D) sG kinks when perturbed parametrically by a spatially periodic potential. Initially, this was motivated by our related research (from the above general point of view) on the sG [4,5] and NLS [6,7] models. As a preliminary step to the investigation of sG breather dynamics [4] on these kind of potentials, it is natural to first seek a good understanding of kink dynamics. Therefore we undertook that study, both analytically and numerically. Our point of departure was early theoretical work [8–10] on this problem, which we summarize for completeness in Sec. II. In particular, it had been predicted that a certain critical velocity exists at which the radiative power 2 emitted by the kink would diverge. Below that critical velocity, radiation would be zero, and above it, it would decrease with increasing speed (see Ref. [2] for a summary). Those results were obtained at a time where the main aim was to develop a perturbative approach to deal with soliton problems. That, and the fact that computers were not the easily available tool they are nowadays, meant that those results were never analyzed in depth or numerically checked. Therefore, as a first stage of our study, we devised a number of numerical experiments to check them, and we found no numerical evidence for the predicted divergence. In view of this result, we carried out an improved perturbative calculation, in the sense that it allowed us to interpret correctly the earlier results in Refs. [8–10] and to show that, although the earlier analyses were correct, the predicted divergence was actually unphysical. This theoretical analysis is reported in detail in Sec. III: A preliminary short report has been given elsewhere [11]. The work done to that point suggested to us the idea that, opposite to what was believed to date, sG kink dynamics on a periodic potential could be essentially that of a (pseudo-relativistic) particle. We thus applied a simple collective coordinate formalism to the problem, and it turned out to describe soliton behavior very accurately, even predicting unexpected new phenomena. The analytical approach and the numerical simulations are contained in Sec. IV. Finally, to complete our program studying lengthscalecompetitionanditseffectsinnonlineardisorderedsystems, weperformedfurther numerical experiments to clarify whether robust objects like kinks, which according to our collective coordinate theory behave mostly like particles, can still exhibit the destabilizing effect of length scale competition. We found that this was actually the case. Furthermore, the simplicity of kinks allowed us to carry out a (numerical) linear stability analysis which provided us with a clear explanation for the numerically observed features. We collect our results on this question in Sec. V. To conclude, we summarize the facts that we have learned, which considerably enhance the understanding of sine-Gordon kink propagation in disordered media and shed light in the so far unexplained phenomenon of length scale competition. Ourresults arealso ofrelevance to manynonlinear systems ofphysical interest, mainly in three directions: First, all non-numerically validated or non-physically interpreted predictions obtained through perturbative calculations should be treated with a degree of caution. Second, the collective coordinate formalism yields a very simple and accurate way to deal with perturbed nonlinear problems, especially those in which the perturbation enters parametrically rather than additively. And third, length scale competition is an ubiquitous phenomenon that may be responsible for many instabilities arising in different nonlinear disordered systems. II. BRIEF SUMMARY OF PREVIOUS RESULTS. We start by describing the picture of sG kink propagation on parametric periodic po- tentials that has been accepted to date. The problem, that has been studied by Mkrtchyan and Shmidt [8] and Malomed and Tribelsky [9,10] is given by a perturbed sG equation of the form u u +[1+ǫcos(kx)]sinu = 0, (1) tt xx − (modelingforinstancealongJosephsonjunctionwithmodulatedcriticalcurrent, tomention just one application); the question posed was whether kinks can propagate freely in such a 3 system, and if so, to describe this propagation. We will only record here a short summary of previous work. The reader is referred to the original papers [8–10] for details. Mkrtchyan and Shmidt [8] used a Green-function perturbation technique (GFPT). They derived a linearized equation for the first order correction to a kink moving with constant velocity, computed the Green function corresponding to its homogeneous version, and then used it to obtain the desired correction by integrating the source term with that Green function. They then noticed that radiation appeared only above a critical kink velocity v = (1 + k2)−1/2. At that particular value, the correction diverges, and the authors thr explain that their calculation become invalid in that region, as of course, it assumed the correction was small. On the other hand, the approach of Malomed and Tribelsky [9,10] was quite different. Its basis was the Inverse Scattering Perturbation Theory (ISPT). A meaningful summary of this kind of calculation would be quite lengthy, and hence we will omit it here, referring the reader to Ref. [2] which is mostly devoted to describing ISPT in detail. Let us just mention that the idea is that, if the amount of radiation emitted by the kink is small, as it should be if the perturbation is small, then the spectral density of the emitted energy can be computed following a Taylor expansion, and the total radiated energy is then derived by integration over all modes. Again, the result was that there was a critical velocity v = (1 + k2)−1/2 such that kinks traveling with velocities v < v thr thr did not emit any radiation at all, whereas in the opposite case the amount of emitted radiation decreased as v 1, diverging when v = v . Always within the framework of thr → ISPT, Malomed and Tribelsky [10] were also able to show that dissipation could play a regularizing rˆole, suppressing the divergence. As the results in Ref. [8] agreed with those in Refs. [9,10], the existence of this threshold for radiation with its associated divergence was accepted, and the question of kink propagation on periodic potentials was regarded as basically solved. As mentioned above, it has to be borne in mind that the main issue of those early researches was to establish the proper foundations for a perturbative theory for solitons. Hence, the question of the physical meaning and origin of the divergence was not addressed. Another unexplained point arises already from ISPT, which allows computation of the radiation nature. When this is done in our case, the radiation wavenumbers turn out to be related to the perturbation one by a complicated equation (see, e.g., [2]), which, in particular, implies that radiation is emitted with a non-intuitive wavenumber k−1 at the divergence. This prediction is difficult to understand physically. Let us recall at this point that a particlelike picture of kink propagation had been developed and had been largely successful so far [12] when compared to numerical experiments. If ISPT predictions for the radiation wavenumbers were true, the reason for them must come from the wave nature of kinks. Consequently, the particle picture should be regarded as a major simplification and valid only in limited cases. III. KINK PROPAGATION ON PERIODIC MEDIA. A. Numerical results. With the above scenario (and the question it poses) in mind, we carried out a num- ber of numerical simulations looking in the first place for the proposed threshold. All the 4 simulations we will be reporting on throughout the paper have been carried out taking pe- riodic boundary conditions. The integration was performed with two different procedures, an adaptation of the energy conserving Strauss-V´azquez finite-difference scheme [13] and a fifth order, adaptive stepsize, Runge-Kutta integration [14] of the discretization of the PDE. The results were independent of the procedure, which is a satisfactory checking. We performed a careful search, paying attention to the fact that the predicted value was a first order calculation, and that it may not be quantitatively accurate. On the other hand, the finite width of the simulated system may also be of relevance at this point, as its radiation spectrum structure is not identical to the continuum, infinite system (in particular, the low- est frequency in the model is restricted to be ω2 = 1 + (2π/L)2, L being the length of min the system). Hence, we monitored the amount of radiation emitted by the kink by making simulations with many different initial conditions, sweeping a range of initial velocities; if there was a threshold somewhere, there should be a change in the radiating power of the kink as it moved through it. The result was negative: No evidence for a threshold was found, even when the search was performed for a large range of initial velocities with a resolution of 10−2 for some choices of k. Examples of the outcome of the simulations are shown in Fig. 1 for three values of the potential wavelength, of the order (a) and smaller (b, c) than the kink width ( 6 in our dimensionless units) at v = v . It has to be stressed that the predicted thr ∼ divergence does not depend on the strength of the perturbing potential, ǫ, but we also tried to make the effect more visible by increasing this parameter. Indeed, in Fig. 1, ǫ = 0.4, a value that is not very small, and the kinks seem unaffected except for a small amount of radiation and an oscillatory motion superimposed on its trajectory, which is shown in Fig. 2. It is interesting to note that the kink traveling on the short wavelength potential (c) ap- pears not to be affected at all. This will be understood by means of the collective coordinate approach in Sec. IV. On increasing ǫ further, trapping behavior takes place, i.e., kinks are trapped by the potential and cannot propagate, but there is no strong emission of radiation (for an example, see Fig. 4(b), which will be discussed later). Actually, this trapping can be of two very different kinds, as we will discuss in sections IV and V. Another interesting remark is that we also observed that kinks always emit radiation, even when moving at a very low velocity, far below the predicted threshold. A similar result arises from the work of Peyrard and Kruskal on highly discrete sG systems [15], where kinks propagate on the peri- odic potential coming fromthe Peierls-Nabarro barrier, although this comparison should not be taken too literally as there are some differences between both problems, like the existence of a maximum allowed frequency in the discrete one, for instance. It thus becomes evident that the features of kink propagation on periodic potentials are qualitatively different from the above perturbative analytical results. Interestingly, numerical simulations on a similar perturbation of the φ4 problem [16] seem to confirm the absence of this divergence. We will elaborate more on this when presenting our conclusions in Sec. VI. B. Theory. In order to gain insight into the numerical observations, we developed a new perturbative approach for this problem, following a similar approach to that given by Fogel et al. [12]. To this end, we perform a Lorentz transformation and rewrite (1) in the rest frame of the soliton (i.e., the reference frame moving with the speed of the unperturbed soliton, v) 5 u u + 1+ǫcos[kγ(x+vt)] sinu = 0, (2) tt xx − { } with γ (1 v2)−1/2 the Lorentz factor. Here we consider ǫ 1, so the perturbative term ≡ − ≪ may be treated by assuming a solution of the form u(x,t) = u (x)+ǫu(1)(x,t), (3) v where u (x) 4tan−1(ex) is the unperturbed sG kink. For completeness, we now recall how v ≡ the most appropriate basis in which toexpand ǫu(1)(x,t) isobtained. Introducing the Ansatz (3) in Eq. (2) without the perturbation term, linearizing in the small quantity u(1)(x,t), and separatingtimeandspacebyintroducingu(1)(x,t) = f(x)e−iωt,weareleftwiththefollowing eigenvalue problem for f(x): d2 + 1 2sech2x f(x) = ω2f(x). (4) "− dx2 − # (cid:16) (cid:17) This is a well known eigenvalue problem [17]; there exists exactly one bound state, with ω = 0, and a continuum of scattering states with ω2 = 1+κ2; the corresponding normalized b κ eigenfunctions are f (x) = 2sechx (5a) b 1 f(κ,x) = eiκx(κ+itanhx). (5b) ω √2π κ These eigenfunctionshave avery clearphysical meaning. Theboundstatef (x)isassociated b to the Goldstone translation mode of the soliton, whereas the continuum eigenfunctions f(κ,x) are the radiation modes (see [12] for a detailed discussion). Besides, these functions form an orthogonal basis, since the corresponding operator is self-adjoint. We will make use of this fact to deal with our problem. In terms of this basis, the first order correction can be split into two parts, namely, u(1)(x,t) = u(trans)(x,t)+u(rad)(x,t), (6) where 1 u(trans)(x,t) = φ (t)f (x), (7a) b b 8 ∞ u(rad)(x,t) = dκφ(κ,t)f(κ,x). (7b) −∞ Z To find the amplitudes φ (t) and φ(κ,t), one again introduces the Ansatz (3) in Eq. (2), b linearizes and Fourier transforms in time; subsequent projection yields ∞ sinhx ¨ φ (t) = 4 dx cos[kγ(x+vt)] , (8a) b −∞ cosh3x Z ∞ e−iκx(κ itanhx) sinhx φ¨(κ,t)+(1+κ2)φ(κ,t) = 2 dx cos[kγ(x+vt)] − . (8b) −∞ 2π(1+κ2) cosh2x Z q It now remains to solve Eqs. (8) and invert the various transforms needed to arrive at them. In the following, we discuss translation and radiative parts in (6) separately. Let us 6 start with the simplest one, i.e., the translation mode contribution. Note that (8a) is, after performing the integration, nothing but the Newton’s law for a time-dependent force. Its solution may be readily found, and finally one obtains π u(trans)(x,t) = sin(kγvt)sechx. (9) 2v2sinh(kγπ/2) Recalling that we are working in the unperturbed soliton reference frame, this is a localized oscillatory motion superimposed on its otherwise constant trajectory. Now, let us remark thattheprefactorimpliesthatshortwavelength(k )perturbationswillhavenoeffecton → ∞ the motion of the center of the soliton, which is also in good agreement with our simulations in Fig. 2. This behavior can be understood in terms of a “smoothing” of the potential: The kink, having a width much larger than the perturbation wavelength, experiences only an effective averaged force, whose amplitude vanishes exponentially for large k (see Sec. IV; see also related comments in [4,6]). Equation (8b) for the κ-mode radiative contribution can also be solved. After computing the integral in the right hand side of Eq. (8b), one is left with the Newton’s law for a forced harmonic oscillator. This allows the determination of φ(κ,t), and substitution of it in Eq. (7b) to find the total radiative contribution: 1 ∂ u(rad)(x,t) tanhx (10) ≡ 4 " − ∂x# × ∞ 1+κ2 k2γ2 eikγvt e−ikγvt dκ − + eiκx. × −∞ (1+κ2)(1+κ2 k2γ2v2) "cosh[π(kγ κ)/2] cosh[π(kγ +κ)/2]# Z − − It is possible to deal with the integral in (10) in the complex plane: When x > 0, in the upper halfplane, andwhenx < 0inthelower half plane. Thepolestructure oftheintegrand will completely determine the total radiative contribution. In particular, we will see that radiation only appears for some special values of the system parameters. Wetakex > 0inwhatfollows(theoppositecaseistreatedinthesameway). Accordingly, the integral has to be analyzed in the upper half complex plane. The pole structure of the integrand is depicted in Fig. 3. All poles are simple, and their locations are z +i, 0 z iα +i√1 k2γ2v2, and z± κγ +i(2n+1), n being a non-negative intege≡r. For 1 ≡ ≡ − n ≡ ± the sake of clarity we treat each pole separately. i. The first pole, z = +i, is constant, and does not change when the system parameters 0 change. Since this pole is purely imaginary, it is immediately seen that the contribution of the residue at z is exponentially localized around the kink center. This term does not give 0 rise to any radiation, but rather to time-dependent corrections of the kink shape. ± ii. The family of poles z depends on the perturbation wavenumber k and on the kink n velocity through the Lorentz factor γ. However, they always have a positive imaginary part, ± thus leading again to exponentially localized contributions. Therefore, the z poles also do n not produce any true radiative correction. iii. The remaining pole is the key one. If α2 1 k2γ2v2 > 0, the same reasoning ≡ − applied to the other poles holds, and there is no radiation. It is worth mentioning that localized oscillations around the kink center, predicted from the contributions of z , z (α 0 1 ± real), and z , were already evident in our numerical simulations, as shown in Fig. 1. For n fixed k, as v increases, the pole moves down the imaginary axis, and at the critical value 7 v (1 + k2)−1/2 it lies at the origin of the complex plane. For kink velocities v > v thr thr ≡ the pole is purely real, and then it does give rise to a radiative contribution, whose form is given by (with β √k2γ2v2 1 a real number) ≡ − π i ei(kγvt+βx) e−i(kγvt−βx) (rad) u = 1 tanhx + . (11) β 4γ2v2 − β !"cosh[π(kγ β)/2] cosh[π(kγ +β)/2]# − This expression tells us that radiation occurs whenever β is real (v > v ), and this radia- thr tion is the superposition of two linear waves of different amplitudes, travelling in opposite directions but with the same phase velocity. C. Discussion. To this point, it appears that our perturbative calculation leads exactly to the same prediction as those in [8,10], namely that there is a critical velocity v (1+k2)−1/2 below thr ≡ which kinks do not radiate and above which they do. At that precise velocity, the amplitude of the emitted radiation diverges; notice that β vanishes as v approaches v from above and thr consequently the prefactor in Eq. (11) goes to infinity. However, this apparent equivalence is not so. The crucial difference arises when one looks more carefully at Eq. (11): As v thr is approached, not only the amplitude of the emitted wave diverges, but also its wavelength 2π/β. Then, we are faced with something similar to an “infrared” divergence, and usually those do not have a real physical meaning. We will show immediately that this is indeed the case here, but let us first comment on the reasons why our calculation provides us with this physically relevant result that was not transparent in the previous approaches. As to the GFPT computation [8], they compute the first order correction to the field much as we do here (actually the two approaches are basically the same in the beginning), but they do not use the natural translation mode-radiation basis, so they can not separate the different contributions and are therefore led to an expression they can not analyze in detail; as we already pointed out, they merely remark that their calculations are invalid in the vicinity of the divergence, as they assumed the correction should be small. On the other hand, ISPT [9,10] yields a different result than ours in spite of using a suitable basis because the integration over κ is made in an incoherent fashion, i.e., integrating over emitted energy instead of emitted amplitude (we notice in passing that many ISPT results are obtained by this same means). When the integration over radiation modes is made coherently as shown here, the result changes due to the superposition of different modes. These reasons lead us to believe that, although admittedly the early perturbative work was mathematically sound, the calculation we present here is the physically correct first order result. Now that we have a reliable perturbative calculation, we need to understand what is the nature of the divergence. To make progress, it is very important to turn to the form of our starting Eq. (1) with dimensions, namely u c2u +ω2[1+ǫcos(kx)]sinu = 0, (12) tt − 0 xx 0 where c and ω are a velocity and a frequency characteristic of the particular physical 0 0 context. Redoing the calculations with dimensions transforms the divergence condition kγv = 1 into kγ v = ω [γ = (1 v2/c2)−1/2]. This immediately clarifies what happens: thr 0 thr 0 0 − 0 8 The divergence occurs when the velocity of the kink is such that the time it takes to travel through a wavelength of the potential, T = λ/(γv ), λ = 2π/k, is exactly the period 0 thr of the lowest frequency phonon, T = 2π/ω . If the velocity is lower than v , the kink 0 0 thr will not be able to excite phonons, whereas when its velocity is higher it can and will subsequently radiate. At v , the excited radiation is that of the lowest phonon, and it has thr infinite wavelength and velocity, as predicted by our calculation. This natural picture of kinks exciting radiation according to the frequency of their propagation through a potential wavelength becomes therefore the likely candidate to explain the divergence. On the other hand, now it also becomes clear the divergence of the energy at v : It diverges because thr of the infinite contribution arising from the infinite wavelength mode when integrated over the whole x axis. This agrees with GFPT and ISPT results whose only difficulty was not to specifically identify the mode responsible for the divergence. In spite of this clarification, the most significant question is not answered yet: Why numerical simulations do not agree with this calculation, which seems to allow for a simple and physically reasonable interpretation? By looking again at Figs. 1 and 2, it is easy to realize that the flaw of the perturbative calculation is at its very root: We are computing first order corrections around a kink moving at a constant velocity v, and this condition never holds. Whatever the starting position of the kink is, it will behave like a particle in the sense that it will be accelerated or decelerated depending on whether it travels towards a minimum or a maximum of the potential. In fact, the translation mode correction itself is describing this: The kink velocity, in its reference frame, is not zero but rather it oscillates between positive and negative values. It is not a surprise, then, that first the resonance condition we have obtained is never matched, and second that the kink emits radiation at any velocity, because it is accelerating or decelerating. Of course, we should note that this is a perturbative calculation including only first order terms; the possibility still remains that the divergence is suppressed by higher order nonlinearities. IV. COLLECTIVE COORDINATE APPROACH. Theabovenumericalresultsandthesubsequent perturbativecalculationstronglysuggest that sGkinks behave aspoint-like particles inthepresence ofa periodicparametricpotential like the one we deal with here. Therefore, it is natural to try to describe those results by means of the collective coordinate formalism. This approach was first proposed in [12], and it has been applied recently to sG breathers on periodic potentials [5] as well as to NLS [7] equations with the same perturbation. In both cases the analytical predictions turned out to be in very good quantitative agreement with numerical simulations: For instance, in Ref. [5] the threshold for breather breakup into a kink-antikink pair was predicted with an accuracy better than 0.1%. On the other hand, the calculation in Ref. [7] predicted the appearance of the so called “soliton chaos,” verified by simulations of the full PDE. In our present problem, the advantage we have is that, due to the simpler nature of the kink, we will be able to compute the effective potential not only for kinks at rest but also for moving kinks. The basic idea of the collective coordinate formalism is very simple: To reduce a compli- cated problem with an infinite number of degrees of freedom, posed in terms of a PDE, to a much less complex problem with a few degrees of freedom (and correspondingly described 9 in terms of ordinary differential equations, ODE’s). There are a number of ways to do this, and different quantities can be chosen to play the rˆole of collective coordinates describing the motion of the nonlinear excitation as a whole. For our problem, it is enough to simply consider the kink center as our collective coordinate for the kink. Its motion will be then governed by an effective potential that can be computed by integrating the perturbative contribution to the hamiltonian over the kink profile, i.e., ∞ V (x ,t) = ǫ dx[1 cosu (x x ,t)]coskx, (13) eff 0 v 0 −∞ − − Z where u (x x ,t) denotes now a kink moving with constant velocity v and centered at x . v 0 0 − This integral can be easily evaluated and yields kπ V (x ,t) = 2ǫ cos[k(x +vt)]. (14) eff 0 γ2sinh(kπ/2γ) 0 From equation (14) we see that the potential experienced by the particle equivalent to the kink is basically the same perturbation potential that appears in the PDE (1), although the prefactor in front of it is quite complicated. The simplest dependence of this prefactor is on the wavenumber. It can be immediately seen that when k 0 (long wavelength limit) → the effective potential prefactor reduces to 4/γ and subsequently V becomes closer to the eff perturbative one; in the opposite limit, k , the sinh term makes the effective potential → ∞ vanish exponentially. This is in agreement with what we have learned so far: Looking at Fig. 1,itcanbeseenthattheshortwavelength potentialhasnoeffectonthekink(c), whereasthe motion on long wavelength perturbations resemble that of a particle on the bare potential. To phrase in the terminology introduced in Ref. [4], the behavior of the kink in these cases is that of a “bare” (long wavelength) or a “renormalized” (short wavelength) particle. It is also important to notice that this result agrees with the perturbative calculation we described in Sec. IIIb [see Eq. (9)] as it was to be expected. There we showed that the correction to the center of mass motion was basically an oscillatory term, implying that the velocity of the center of mass oscillates around some mean value. This is precisely the same kind of trajectory followed by a point-like particle in the potential in Eq. (14) (at least if the velocity is not too close to 1). Nevertheless, it is worth pursuing this agreement a bit further, by studying the threshold for kinks to propagate in this kind of potential. The easiest way to compute the threshold is by equating the kinetic energy of the kink to the maximum of the effective potential, provided we restrict ourselves to the non-relativistic limit (v2 not too close to 1) to keep the kink mass constant. This will give us the maximum potential height over which a kink that starts from a point at which the perturbation is zero with a certain velocity is able to overcome the nearest top point. Using the fact that the mass of a not too fast kink is 8 in our units, we find that the threshold is given by 2v2γ2 kπ ǫ = sinh . (15) thr kπ 2γ In the same way, we could have computed the threshold velocity for a given strength of the potential, butweprefertocheck ourpredictionsthiswaybecausethepresence ofγ makesthe other possibility more complicated. We compared this prediction to numerical simulations. 10