ebook img

Compactons versus Solitons PDF

0.1 MB·
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Compactons versus Solitons

Compactons versus Solitons Paulo E.G. Assis and Andreas Fring (Dated: January 9, 2009) WeinvestigatewhethertherecentlyproposedPT-symmetricextensionsofgeneralizedKorteweg- de Vries equations admit genuine soliton solutions besides compacton solitary waves. For models which admit stable compactons having a width which is independent of their amplitude and those which possess unstable compacton solutions the Painlev´e test fails, such that no soliton solutions canbefound. ThePainlev´etestispassedformodelsallowingforcompactonsolutionswhosewidth isdeterminedbytheiramplitude. Consequentlythesemodelsadmitsoliton solutionsinadditionto compactons and are integrable. I. INTRODUCTION for the Hamiltonian density in (1) is l,m,p 9 H 0 u +ul−2u +gimup−2um−3 p(p 1)u4 (3) 0 In a recent investigation Bender, Cooper, Khare, Mi- t x x − x 2 hailaandSaxena[1]havefoundcompactonsolutions,i.e. +2pmuu2u +m(m 2)u2u(cid:2)2 +mu2u u =0. n solitary wave solutions with compact support, for - x xx − xx x xxx(cid:3) PT a symmetric extensions of generalized Korteweg-de Vries The main aim of this manuscript is to investigate J (KdV) equations. The proposed models generalize vari- whether this equation admits soliton solutions and is 9 ous systems previously studied and are described by the therefore integrable for some specific choices of the pa- Hamiltonian density rameters l,m,p. We will also address the question of ] whether it is possible to find solitons and compactons h t ul g in the same model or whether only one type of solu- p- Hl,m,p =−l(l 1) − m 1up(iux)m. (1) tions may exist. To answer these questions one could e − − of course construct explicitly the soliton solutions, con- h served charges, Lax pairs, Dunkl operators, etc., which The density reduces to a modification of a Hamil- [ tonian descrHiplt,i2o,pn [2, 3] of generalized KdV-equations is usually a formidable task. Instead we will carry out the Painlev´e test following a proposal originally made 1 [4], which are known to admit compacton solutions. For v by Weiss, Tabor and Carnevale [14]. The test provides l = 3, p = 0 and m = ε+1 one obtains a re-scaled ver- 7 anindicationabouttheexistenceofsolitonsolutionsand sionofthe -symmetricextensionoftheKdV-equation 6 PT discriminatesbetweenmodels,whichareintegrablethose 2 (ε=1)introducedin[5]. ThefirstPT-symmetricexten- which are not. 1 sionsofthe KdV-equationproposedin[6]cannotbe ob- . tained from (1) as they correspond to non-Hamiltonian 1 0 systems. II. THE PAINLEVE´ TEST 9 Thevirtueof -symmetry,i.e. invarianceunderasi- 0 multaneous pariPtyTtransformation : x x and time : P → − The basic assumptionfor the existence ofa solitonso- v reversalT :t→−t,i→−i,foraclassicalHamiltonianis lution is that it acquires the general form of a so-called i thatitguaranteestherealityoftheenergyduetoitsanti- X Painlev´e expansion [14] linear nature [5]. When quantizing one also needs to r ensure -symmetryofthecorrespoHndingwavefunctions ∞ a PT inordertoobtainrealspectra[7,8,9,10]. Themostnat- u(x,t)= λ (x,t)φ(x,t)k+α. (4) k uralwaytoimplement -symmetryin(1)istokeepthe Xk=0 PT interpretationfromthestandardKdV-equationandview thefielduasavelocity,suchthatittransformsasu u. One further demands that in the limit φ(x,t) 0, the Then is -symmetric for real coupling cons→tant functionu(x,t)ismeromorphic,suchthatthele→adingor- l,m,p g andHall possibPleTrealvalues of l,m,p. Alternatively, we dersingularityαisanegativeintegerandthe λk(x,t)are could also allow a purely complex coupling constant, i.e. analyticfunctions. ThegeneralprocedureofthePainlev´e g iR, by transforming the field as u u, such that test consists in substituting the expansion (4) into the ∈ is -symmetric when l is even→an−d p+m odd. equation of motion, (3) for the case at hand, and deter- l,m,p HForgeneraPlTreviewson -symmetryandnon-Hermitian miningthefunctionsλk(x,t)recursively. Apartialdiffer- Hamiltonian systems sePeT[11, 12, 13]. ential equation is said to pass the Painlev´e test when all λ (x,t) can be computed, including enough free param- The equation of motion resulting from the variational k eters to match the order of the differential equation. In principle [15] we recently applied this method to -symmetric PT ∞ extensions of Burgers and the standard KdV-equation, δ dx dn ∂ u = H = ( 1)n H (2) wheremoredetailsonthegeneralitiesandliteraturemay t (cid:18) Rδu (cid:19)x nX=0 − (cid:18)dxn∂unx(cid:19)x be found. 2 A. Leading order singularities oftherecurrencerelationbecomesanidentity,theλ be- r comesafreeparameter,otherwisethePainlev´etestfails. The Painlev´e test stays and falls with the possibility The possible values for r can be found by substituting that the initial condition λ0 can be determined, which u(x,t) λ0(x,t)φ(x,t)α+λrφ(x,t)r+α (7) is essential to commence the iterative procedure to solve → into (3) and computing all possible values of r for which the recurrence relation. We compute λ by substitut- 0 λ becomes a free parameter. Considering the case iii) ing the first term in the expansion (4), i.e. u(x,t) r λ (x,t)φ(x,t)α, into (3) and evaluating the values f→or for integer values l,m,p the coefficients of the leading al0l possible leading order singularities α. The individual order φr+α(l−1)−1 is proportional to terms in (3) have the following leading order behaviour: λ gα(2−l)/m(r+1)(r+αl)[r+α(l 1)]φα(2−l)+1. (8) u φα−1, ul−2u φα(l−1)−1 and all remaining terms r − x t x are∼proportionalto∼φα(m+p−1)−m−1. Therefore the lead- This means that besides the so-called fundamental res- onance at r = 1, we also find two more resonances at ing order terms may only be cancelled if any of the fol- r = αl,α(1 l). Since thedifferentialequation(3)isof lowing three conditions hold: − − order three all these models fully pass the Painlev´e test i) α 1=α(l 1) 1 α(m+p 1) m 1, provided λ−αl and λα(1−l) can indeed be chosen freely. − − − ≤ − − − Thestandardproceduretoverifythiswouldbenowto which results from assuming that ut and ul−2ux derive the recursive equation resulting from combining constitute the leading order terms. In this case we (4) and (3). Clearly for generic values of l,m,p this will obtainl=2andtheinequalityα(2 m p) m. be extremely lengthy, but even for specific choices it is − − ≤− Thus α remains undetermined. fairly complicated. It suffices, however, to compute the λ up to k > αl. We will present these values for vari- ii) α 1=α(m+p 1) m 1 α(l 1) 1, k − − − − − ≤ − − ousexamplesforseveralchoicesofthe parametersl,m,p which corresponds to the assumption that ul−2ux correspondingtoscenariosleadingtosolutionswithqual- isthe leastsingulartermandmatchingthe leading itatively different kinds of behaviour. ordersofalltheremainingones. Thenweconclude that l 2 and α is fixed to α=m/(m+p 2). ≤ − III. GENERALIZED KDV-EQUATION iii) α(l 1) 1=α(m+p 1) m 1 α 1. − − − − − ≤ − which is the consequence of u being least singular Cooper, Khare and Saxena [16] found that in the gen- t termandthematchingoftheremainingones. This eralizedKdVequation,i.e. m=2,a necessarycondition means the leading order singularity of u(x,t) is of for compactons to be stable is to consider models with the order 2<l<p+6. Thismeansnoneoftheconditionsi)orii) fortheleadingordersingularitytocancelcanbesatisfied. m α= Z− and l 2. (5) The special choice l = p+2,0 < p 2 guarantees that p+m l ∈ ≥ ≤ − the compacton solutions have in addition a width which Cancelling the leading order terms then yields isindependentoftheiramplitude[2]. Forthatparticular case alsothe conditioniii)admits no solution, suchthat λ(n) =e2πinα/m[gl(l 1)]−α/m(iαφ )−α, (6) the Painlev´e test fails. 0 − x However, for models which admit stable compacton where 1 n p+m l indicates the different solutionshavinga widthdepending onthe amplitude we roots of t≤he de≤termining−equation. can find solutions to the condition iii) and proceed with the Painlev´e test. For instance, m = 2,p = 1,l = 5 is In principle we could alsoenvisagea scenarioinwhich such a choice. In this case we find from (5) that α= 1 − u andul−2u aretheleastdominanttermsandthelead- and the leading order singularity of the corresponding intg order sinxgularity is cancelled by all the remaining differential equation is φ−5. Computing now order by terms. However, all these terms only differ by an overall order the functions λk we find the two solutions numericalfactor,suchthatλ0 turnsouttobezerointhis λ± = 2i 5gφ , λ± = i 5gφxx, case and we can therefore discard this case. 0 ± x 1 ∓ φ p p x λ± = i√5g 3φ2xx−2φxφxxx , (9) B. Resonances 2 ∓ 6 (cid:0) φ3 (cid:1) x 3φ φ2 4i 5g3 6φ3 6φ φ φ +φ2φ A key feature of the Painlev´e test is the occurrence λ±3 = t x∓ p (cid:0) 4x8xg−φ5 x 3x xx x 4x(cid:1). of so-called resonances, which arise whenever the coeffi- x ± ± cient in front of a specific λ in the recurrence relations Cruciallyweobservenextthatλ andλ canbe chosen r ± 4 5 becomes zero. This implies that λ can not be deter- arbitrarily. The remaining λ for k > 5 can all be com- r k mined recursively. When in this case the remaining part puted,buttheexpressionsareallextremelycumbersome 3 andwewillthereforenotreportthemhere. Making,how- the λ(n) with n=1,2,3, of which the first terms are k ever,thefurtherassumptiononφtobeatravellingwave, i.e. φ(x,t) = x ωt, simplifies the expressions consider- λ(n) = ie2πin/3(42g)1/3φ , −± ± 0 − x sacbelny.arCiohroeodsuincge tλo4 = λ5 = 0 the two solutions for that λ1(n) = ie2πin/232(/231φg)1/3φxx, (11) x ie2πin/3(7g)1/3 3φ2 2φ φ λ± = λ± =0 for κ=0,1,2,... λ(n) = xx− x xxx , 3κ+1 3κ+2 2 2(6)(cid:0)2/3φ3 (cid:1) ± ± ω x λ0 = ±2ip5g, λ3 =−16g, λ(n) = ie2πin/3(7g)1/3 6φ3xx−6φxφxxxφxx+φ2xφxxxx . λ± = 3iω2 , λ± = ω3 , (10) 3 (cid:0) 4(6)2/3φ5x (cid:1) 6 ∓3584√5g5/2 9 573440g4 From (8) we know that we should encounter resonances ± 33iω4 at the level 6 and 7, which is indeed the case as we find λ = , 12 ±1669857280√5g11/2 that λ(n)and λ(n) can be chosen freely. The remaining 6 7 ± 3ω5 λk(n) for k > 7 can all be computed iteratively and the λ = , ... 15 −66794291200g7 Painlev´e test is passed for this example. For a travelling wave ansatz φ(x,t)=x ωt with the − choice λ(n) = λ(n) =0 the expressions simplify to We conclude that the Painlev´e test is passed for this 6 7 choices of parameters, which means that besides stable λ(n) = λ(n) =λ(n) =λ(n) =0, for κ=0,1,... compacton solutions, whose width depends on their am- 5κ+1 5κ+2 5κ+3 5κ+4 plitude,wealsofindgenuinesolitonsinthesemodelsand, λ(n) = ie2πin/3(42g)1/3, λ(n) = e4πin/3ω , (12) provided the series (4) converges, they are therefore in- 0 − 5 36(42)1/3g4/3 tegrable. 17iω2 λ(n) = , In the unstable compacton regime, i.e. l 2 or 10 598752g3 ≤ l p+6, the condition iii) can not be satisfied. Con- se≥quently we do not expect to find genuine soliton solu- λ(n) = 53e2i3nπω3 ,... tions. We have also verified this type of behaviour for 15 −21555072(42)2/3g14/3 other representative examples which we do not present Thus we observe no qualitative difference in the - here. PT symmetric extensions in comparison to the case m = 2 andfindthatalsointhisonemayhavestablecompacton solutions, whose width depends on their amplitude and genuine solitons at the same time. IV. PT-SYMMETRIC GENERALIZED In the unstable compacton regime, that is l 2 or ≤ KDV-EQUATION l p+3m, the condition iii) can not be satisfied and ≥ the Painlev´e test fails. Once again we do not represent hereotherrepresentativeexamplesforwhichweobtained For the -symmetric extensions of the general- PT the same type of behaviour. ized KdV-equation (3) the necessary condition for com- pactons to be stable was extended by Bender et al [1] to 2<l <p+3m. Thusalsoforgenericvaluesofmnoneof V. DEFORMATIONS OF BURGERS EQUATION theconditionsi)orii)fortheleadingordersingularityto cancelcanbesatisfied. Furthermore,therequirementfor Considering m = 1,p = 1,l = 3 in the equation of stablecompactonsolutionstopossessalsoawidthwhich motion(3)isaverysimpleexampleleadingtoaPainlev´e is independent of their amplitude was generalized in [1] expansion for u(x,t), which can even be truncated after tol=p+m. Asforthespecialcasem=2thisvalueco- thesecondterm. Asthistypeofbehaviourisreminiscent incides with the leading order singularity resulting from ofB¨acklundtransformationgeneratingsolutionsfoundin the condition iii) tending to infinity and therefore the other models [14], we present this case briefly. For this Painlev´e test fails. choice (3) simply reduces to As in the previous case, for models which have stable compacton solutions whose width is a function of their u +uu 2igu + igu u2 u u =0, (13) amplitude the Painlev´etest has a chance to pass, as one t x− xx u2 xx− x xxx x (cid:0) (cid:1) canfind a value forthe leading ordersingularityandpo- whichcanbeviewedasadeformationofBurgersequation tentiallyhasthecorrectamountofresonances. Weverify [15] corresponding to the first three terms. Proceeding this for the example m = 3,p = 1,l = 7, for which we obtain α = 1 and φ−7 as the leading order singularity as in the previous sections, we find the solution − in(3). Since α/m=1/3inthiscase,wefindnowthree 6igφ 6igφ 3φ − u(x,t)= − x + xx− t (14) non-equivalentsolutionsrelatedto the differentrootsfor φ 2φ x 4 provided that φ satisfies the equation integrable. We found thatthe generalizedKdVequation resultingfrom andtheir -symmetricextensions l,2,p φ2φ +φ2φ =2φ φ φ . (15) have theHsame qualitatiPveTbehaviour in the three x tt t xx tx t x l,m,p H differentregimes. Forconveniencewesummarizethedif- Atravellingwaveφ(x,t)=x ωtisforinstanceasolution ferent qualitative behaviours in the following table: − of (15), such that we obtain the simple expression compactons solitons 6igφ 3 l,m,p u(x,t)= x + ω (16) H l =p+m stable, dependent A,β no ωt x 2 − 2<l<p+3m stable, independent A,β yes for the solution of (13). Incidentally, the travelling wave l 2 or l p+3m unstable no solution for Burger’s equation [14] coincides with (16). ≤ ≥ Table 1: The models and their solutions. l,m,p H Clearly our investigations do not constitute a full VI. CONCLUSIONS fletched mathematical proof as we based our findings on various representative examples for the different classes In previous investigations [2, 3, 16] various criteria and it wouldbe very interesting to settle this issue more have been found, which separate the models into Hl,m,p rigorously with a generic argumentation not relying on three distinct classes exhibiting qualitatively different case-by-casestudies. At the same time such a treatment types of compacton solutions, unstable compactons and would probably provide a deeper understanding about stablecompactons,whichhaveeitherdependentorfreely the separation of the different models. Nonetheless, our selectable width A and amplitude β. We have carried findings provide enough evidence to make it worthwhile out the Painlev´e test for various examples for each of to investigate the models which pass the test with other these classes and found that all models which allow sta- techniques developed in the field of integrable models, ble compactons for which the width can not be chosen whereas models which do not pass the test may be ex- independently from their amplitude pass the Painlev´e cluded from such investigations. test. AssumingthatthePainlev´eexpansion(4)converges these models possess the Painlev´e property [17] and al- Acknowledgments: P.E.G.A. is supported by a City low therefore for genuine soliton solutions and are thus University London research studentship. [1] C. Bender, F. Cooper, A. Khare, B. Mihaila, and downexplainedbyanti-linearity,J.Phys.B5,S416–S419 A. Saxena, Compactons in PT-symmetric general- (2003). ized Korteweg-de Vries Equations, arXiv.org:0810.3460 [10] C. M. Bender, D. C. Brody, and H. F. Jones, Must a (2008). HamiltonianbeHermitian?, Am.J.Phys.71,1095–1102 [2] F.Cooper,H.Shepard,andP.Sodano, Solitarywavesin (2003). aclassofgeneralizedKorteweg-deVriesequations, Phys. [11] C. Figueira de Morisson Faria and A. Fring, Non- Rev.E48, 4027–4032 (1993). HermitianHamiltonianswithrealeigenvaluescoupledto [3] A.KhareandF.Cooper,One-parameterfamilyofsoliton electricfields: fromthetime-independenttothetimede- solutions with compact support in a class of generalized pendentquantummechanicalformulation, LaserPhysics Korteweg˘deVriesequations, Phys.Rev.E48(6),4843– 17, 424–437 (2007). 4844 (1993). [12] C. M. Bender, Making sense of non-Hermitian Hamilto- [4] P. Rosenau and J. M. Hyman, Compactons: Solitons nians, Rept.Prog. Phys. 70, 947–1018 (2007). with finite wavelength, Phys. Rev. Lett. 70(5), 564–567 [13] A. Mostafazadeh, Pseudo-Hermitian Quantum Mechan- (1993). ics, arXiv:0810.5643 (2008). [5] A.Fring, PT-Symmetricdeformations of theKorteweg- [14] J. Weiss, M. Tabor, and G. Carnevale, The Painlev´e deVries equation, J. Phys. A40, 4215–4224 (2007). propertyforpartialdifferentialequations,J.Math.Phys. [6] C.M.Bender,D.C.Brody,J.Chen,andE.Furlan, PT- 24, 522–526 (1983). symmetric extension of the Korteweg-de Vries equation, [15] P. Assis and A. Fring, Integrable models from PT- J. Phys. A40, F153–F160 (2007). symmetric deformations, arXiv.org:0810.3628 (2008). [7] E. Wigner, Normal form of antiunitary operators, J. [16] F. Cooper, A. Khare, and A. Saxena, Exact Ellip- Math. Phys.1, 409–413 (1960). tic Compactons in Generalized Korteweg-DeVries Equa- [8] C. M. Bender and S. Boettcher, Real Spectra in Non- tions, Complexity 11, 30–34 (2006). Hermitian Hamiltonians Having PT Symmetry, Phys. [17] B.GrammaticosandA.Ramani, Integrability-andHow Rev.Lett. 80, 5243–5246 (1998). to detect it, Lect. Notes Phys. 638, 31–94 (2004). [9] S. Weigert, PT-symmetry and its spontaneous break-

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.