Symmetry, Integrability and Geometry: Methods and Applications SIGMA 13 (2017), 018, 20 pages Ermakov–Painlev´e II Symmetry Reduction of a Korteweg Capillarity System Colin ROGERS † and Peter A. CLARKSON ‡ † Australian Research Council Centre of Excellence for Mathematics & Statistics of Complex Systems, School of Mathematics, The University of New South Wales, Sydney, NSW2052, Australia E-mail: [email protected] ‡ School of Mathematics, Statistics & Actuarial Science, University of Kent, Canterbury, CT2 7FS, UK 7 1 E-mail: [email protected] 0 2 Received January 13, 2017, in final form March 15, 2017; Published online March 22, 2017 r https://doi.org/10.3842/SIGMA.2017.018 a M Abstract. A class of nonlinear Schr¨odinger equations involving a triad of power law terms 2 togetherwithadeBroglie–Bohmpotentialisshowntoadmitsymmetryreductiontoahybrid 2 Ermakov–Painlev´e II equation which is linked, in turn, to the integrable Painlev´e XXXIV equation. A nonlinear Schr¨odinger encapsulation of a Korteweg-type capillary system is ] I thereby used in the isolation of such a Ermakov–Painlev´e II reduction valid for a multi- S parameter class of free energy functions. Iterated application of a B¨acklund transfor- . n mation then allows the construction of novel classes of exact solutions of the nonlinear i capillarity system in terms of Yablonskii–Vorob’ev polynomials or classical Airy functions. l n A Painlev´e XXXIV equation is derived for the density in the capillarity system and seen to [ correspond to the symmetry reduction of its Bernoulli integral of motion. 3 Key words: Ermakov–Painlev´e II equation; Painlev´e capillarity; Korteweg-type capillary v system; B¨acklund transformation 8 3 2010 Mathematics Subject Classification: 37J15; 37K10; 76B45; 76D45 2 3 0 . 1 1 Introduction 0 7 Giannini and Joseph [39], in a nonlinear optics context, introduced a class of symmetry reduc- 1 : tions for a cubic nonlinear Schr¨odinger (NLS) equation v i X iΨ +Ψ +ν|Ψ|2Ψ = 0, (1.1) t xx r a with subscripts denoted partial derivatives, which resulted in the Painlev´e II equation but with zero parameter α d2q = 2q3+zq, (1.2) dz2 see also [6, 49, 61]; the reduction of the NLS equation (1.1) to equation (1.2) was derived in [37, 86]. Numerical integration led to the isolation of interesting non-stationary solutions both bounded and stable in shape for a restricted range of ratio of nonlinearily to dispersion. However, the absence of the Painlev´e parameter α in that reduction does not allow the iterative construction of sequences of exact solutions via the B¨acklund transformation for the canonical Painlev´e II (P ) II d2q = 2q3+zq+α, (1.3) dz2 2 C. Rogers and P.A. Clarkson withαaparameter,asgivenbyGambier[38]andLukashevich[59]. Here,symmetryreductionto P (1.3) or to a hybrid Ermakov–Painlev´e II equation linked to the integrable Painlev´e XXXIV II (P ) equation XXXIV d2p 1 (cid:18)dp(cid:19)2 (α+ 1)2 = +2p2−zp− 2 , (1.4) dz2 2p dz 2p with α a parameter, is obtained for a wide class of NLS equations which incorporates a triad of power law terms together with a de Broglie–Bohm potential term. It is remarked that NLS equations involving such triple power law nonlinearities arise in nonlinear optics (see [31, 91] and literature cited therein). Moreover, NLS equations containing a de Broglie–Bohm term also arise in the analysis of the propagation of optical beams [75, 90] as well as in cold plasma physics [56]. Under appropriate conditions, such ‘resonant’ NLS equations admit novel fusion or fission solitonic behaviour [55, 56, 67, 68]. The Ermakov–Painlev´e II symmetry reduction is applied here to an NLS encapsulation of anonlinearcapillaritysystemwithorigininclassicalworkofKorteweg[54]. Iteratedapplication of a B¨acklund transformation admitted by P (1.3) permits the construction via the linked II P equation of novel multi-parameter wave packet solutions to the capillarity system in XXXIV terms of either Yablonskii–Vorob’ev polynomials or classical Airy functions. These are shown to be valid for a multi-parameter class of model specific free energy relations. An invariance of the (1+1)-dimensional Korteweg capillarity system under a one-parameter class of reciprocal transformations as recently set down in [80] allows the extension of the reduction procedure to a yet wider class of capillarity systems. 2 A Ermakov–Painlev´e II symmetry reduction Here, a class of (1+1)-dimensional nonlinear Schr¨odinger equations of the type (cid:20) (cid:21) |Ψ| |Ψ| iΨ +Ψ − (1−C) xx −ic x +λ|Ψ|2+µ|Ψ|2m+ν|Ψ|2n Ψ = 0, (2.1) t xx |Ψ| |Ψ|2 which incorporates a de Broglie–Bohm potential |Ψ| /|Ψ| and a triad of power law terms is xx investigated under a symmetry reduction. Thus, constraints on the parameters in (2.1) are sought for which the class of NLS equations admits symmetry reduction either to the P (1.3), II with non-zero parameter α, or to a hybrid Ermakov–Painlev´e II equation under a wave packet ansatz Ψ = [φ(ξ)+iψ(ξ)]exp(iη), (2.2a) with ξ = αt+βt2+γx, η = γt3+δt2+εγtx+ζt+λx. (2.2b) In the nonlinear optics context of [39] such a similarity transformation was used to reduce a standard cubic NLS equation to Painlev´e II but with zero parameter α and resort was made to a numerical treatment. Asymptotic properties of P (1.2) with α = 0 have been discussed II by various authors, see, e.g., [2, 8, 23, 28, 34, 41, 62]. In the present case, on introduction of the wave packet ansatz (2.2) into (2.1), it is seen that γ2d2φ − dψ(cid:2)2(cid:0)β+εγ2(cid:1)t+α+2λγ(cid:3)+ cγψ (cid:18)φdφ +ψdψ(cid:19)−∆φ = 0, (2.3a) dξ2 dξ |Ψ|3 dξ dξ γ2d2ψ + dφ(cid:2)2(cid:0)β+εγ2(cid:1)t+α+2λγ(cid:3)− cγφ (cid:18)φdφ +ψdψ(cid:19)−∆ψ = 0, (2.3b) dξ2 dξ |Ψ|3 dξ dξ Ermakov–Painlev´e II Symmetry Reduction of a Korteweg Capillarity System 3 where ∆ = 3γt2+2δt+εγx+ζ +(εγt+λ)2+λ|Ψ|2+µ|Ψ|2m+ν|Ψ|2n (cid:40)(cid:34) (cid:35) (cid:41) + sγ2 φd2φ +ψd2ψ +(cid:18)dφ(cid:19)2+(cid:18)dψ(cid:19)2 (cid:0)φ2+ψ2(cid:1)−(cid:18)φdφ +ψdψ(cid:19)2 , (2.4) |Ψ|4 dξ2 dξ2 dξ dξ dξ dξ with s = 1−C. The relations (2.3) together show that γ2(cid:18)d2φψ− d2ψφ(cid:19)−(cid:18)φdφ +ψdψ(cid:19)(cid:2)2(cid:0)β+εγ2(cid:1)t+α+2λγ(cid:3) dξ2 dξ2 dξ dξ (cid:18) (cid:19) cγ dφ dψ + φ +ψ = 0, (2.5) |Ψ| dξ dξ whence it is required that β+εγ2 = 0, in which case equation (2.5) admits the integral (cid:18) (cid:19) dφ dψ γ2 ψ− φ − 1(α+2λγ)|Ψ|2+cγ|Ψ| = I, (2.6) dξ dξ 2 where I is an arbitrary constant of motion. On use of the relation (cid:34) (cid:35) d2φ d2ψ (cid:18)dφ(cid:19)2 (cid:18)dψ(cid:19)2 (cid:18) dφ dψ(cid:19)2 d2|Ψ| φ +ψ + + (φ2+ψ2)− φ +ψ = |Ψ|3 , (2.7) dξ2 dξ2 dξ dξ dξ dξ dξ2 it is seen that (2.4) yields, if β (cid:54)= 0, ∆ = γβ−1(cid:0)ε2γ +3(cid:1)(ξ−αt−γx)+2(δ+εγλ)t+εγx+ζ +λ2 sγ2d2|Ψ| +λ|Ψ|2+µ|Ψ|2m+ν|Ψ|2n+ |Ψ| dξ2 sγ2d2|Ψ| = εξ+ζ +λ2+λ|Ψ|2+µ|Ψ|2m+ν|Ψ|2n+ , (2.8) |Ψ| dξ2 on setting βε = γ(cid:0)3+ε2γ(cid:1), αε = 2(δ+εγλ). (2.9) Moreover, equations (2.3) again combine to show that (cid:18) d2φ d2ψ(cid:19) (cid:18)dφ dψ (cid:19) γ2 φ +ψ + ψ− φ (α+2λγ)−∆|Ψ|2 = 0, dξ2 dξ2 dξ dξ whence, on use of the identity (cid:34) (cid:35) (cid:18)dφ(cid:19)2 (cid:18)dψ(cid:19)2 (cid:18)dφ dψ (cid:19)2 (cid:18) dφ dψ(cid:19)2 (φ2+ψ2) + − ψ− φ ≡ φ +ψ , dξ dξ dξ dξ dξ dξ together with (2.7), it is seen that (cid:34) (cid:35) d2|Ψ| (cid:18)dφ dψ (cid:19)2 (cid:18)dφ dψ (cid:19) γ2 |Ψ|3 − ψ− φ +(α+2λγ) ψ− φ |Ψ|2−∆|Ψ|4 = 0. dξ2 dξ dξ dξ dξ 4 C. Rogers and P.A. Clarkson The latter, by virtue of the integral of motion (2.6) and the expression (2.8) for ∆ now produces a nonlinear equation in the amplitude |Ψ|, namely d2|Ψ| c c +[c +c ξ]|Ψ|+c |Ψ|3+c |Ψ|2m+1+c |Ψ|2n+1+ 6 + 7 dξ2 1 2 3 4 5 |Ψ| |Ψ|2 I2 = , (2.10) (1−s)γ4|Ψ|3 where the constants c ,c ,...,c are given by 1 2 7 (α−δ/ε)2−γ2(ζ +λ2) ε λ c = , c = , c = , (2.11a) 1 (1−s)γ4 2 (s−1)γ2 3 (s−1)γ2 µ ν c2 2cI c = , c = , c = , c = , (2.11b) 4 (s−1)γ2 5 (s−1)γ2 6 (s−1)γ2 7 (1−s)γ3 and it is required that s (cid:54)= 1. Below a triad of cases is set down in which the amplitude equation (2.10) reduces either directlytoP ortoahybridErmakov–Painlev´eIIequationsubsequentlyshowntobeintegrable. II Case (i) I = 0; m = −1; n = −1. 2 In this case with c = 0, c = −1, c = −2, c = −α, c +c = 0, 1 2 3 4 5 6 the amplitude equation reduces directly to the Painlev´e II equation d2|Ψ| = 2|Ψ|3+ξ|Ψ|+α, (2.12) dξ2 corresponding to the symmetry reduction via the ansatz (2.2) of the class of NLS equations (cid:20) |Ψ| |Ψ| Cγ2α c2 (cid:21) iΨ +Ψ − (1−C) xx −ic x +Cγ2|Ψ|2+ − Ψ = 0. t xx |Ψ| |Ψ|2 |Ψ| |Ψ|2 Case (ii) I = 0; c = 0; m = −1; n = 0. 2 Here, with c = −c , c = −1, c = −2, c = −α, 1 5 2 3 4 the P equation (2.12) again results, while the associated class of NLS equations (2.1) becomes II (cid:20) |Ψ| Cγ2α (cid:21) iΨ +Ψ − (1−C) xx +Cγ2|Ψ|2+ +ν Ψ = 0. (2.13) t xx |Ψ| |Ψ| It is remarked that in the absence of the de Broglie–Bohm term, a time-independent NLS equation of this type (2.13) incorporating a nonlinearity ∼ |Ψ|−1 has been derived ‘ab initio’ in [84] via a geometric model which describes stationary states of supercoiled DNA. Ermakov–Painlev´e II Symmetry Reduction of a Korteweg Capillarity System 5 Case (iii) I =(cid:54) 0; c = 0; m = −1; n = 0. 2 In this case, equation (2.10) reduces to a hybrid ‘Ermakov–Painlev´e II’ equation of the type d2|Ψ| σ +ε|Ψ|3+(δξ+ζ)|Ψ| = , (2.14) dξ2 |Ψ|3 and which will be subsequently seen to be linked to P (1.4). It is recalled that the classical XXXIV Ermakov equation with roots in [30], namely d2E σ +ω(ξ)E = , dξ2 E3 admitsanonlinearsuperpositionprinciplereadilyderivedviaaLiegroupapproachasin[76,81]. In the subsequent application to the Korteweg capillarity system it will be the Ermakov– Painlev´e II symmetry reduction that will be exploited. With a positive solution |Ψ| of (2.14) to hand, the corresponding class of exact solutions for Ψ in the wave packet representation (2.2) is obtained via the integral of motion (2.6). Thus, the latter yields (cid:20) (cid:18) (cid:19)(cid:21) d φ δ cγ I γ2 tan−1 −α+ + = , dξ ψ ε |Ψ| |Ψ|2 whence, on integration (cid:18) (cid:19) (cid:18) (cid:19) (cid:90) (cid:90) φ δ 1 1 γ2tan−1 = α− ξ−cγ dξ+I dξ, (2.15) ψ ε |Ψ| |Ψ|2 where use has been made of the relation (2.9). Accordingly, with V = φ/ψ, it is seen that φ, ψ in the original wave packet representation are given by the relations |Ψ|V |Ψ| φ = ±√ , ψ = ±√ . 1+V2 1+V2 In the sequel, the link between the Ermakov–Painlev´e II equation (2.14) and P (1.4) XXXIV is used to construct novel classes of wave packet solutions of a nonlinear Korteweg capillarity system in terms of Yablonskii–Vorob’ev polynomials or classical Airy functions via the iterated application of the B¨acklund transformation for P due to Gambier [38] and Lukashevich [59]. II 3 The capillarity system In [4], Antanovskii derived the isothermal capillarity system with continuity equation ρ +div(ρv) = 0, (3.1a) t augmented by the momentum equation (cid:20) (cid:21) δ(ρE) vt+v•∇v+∇ −Π = 0, (3.1b) δρ where ρ is the density, v velocity and E(ρ,|∇ρ|2/ρ) is the specific free energy. Herein, the standard variational derivative notation (cid:20) (cid:21) δΘ ∂Θ ∂Θ = −∇• ∇ρ , δρ ∂ρ ∂A 6 C. Rogers and P.A. Clarkson is adopted with A = 1|∇ρ|2. In the above Π is an external potential, commonly taken to be 2 that due to gravity, in which case Π = −ρg. The classical Korteweg capillarity system as set down in [54] is retrieved in the specialisation κ(ρ)A E(A,ρ) = , κ(ρ) > 0, ρ in which case, the momentum equation becomes (cid:20) (cid:21) dκ(ρ) vt+v•∇v−∇ κ(ρ)∇2ρ+ 12|∇ρ|2 dρ +Π = 0. The classical Boussinesq capillarity system, in turn, is retrieved as the specialisation with κ constant in this Korteweg system. A system analogous to the Boussinesq model arises ‘mutatis mutandis’ in plasma physics [9]. In the case of irrotationality with v = ∇Φ, the momentum equation (3.1b) admits the Bernoulli integral δ Φ + 1|∇Φ|2+ (ρE)−Π = B(t), t 2 δρ and on introduction of the Madelung transformation [60] Ψ = ρ1/2exp(cid:0)1iΦ(cid:1), (3.2) 2 the capillarity system (3.1) may be encapsulated in the generalised NLS-type equation (cid:20) ∇2|Ψ| δ(ρE) (cid:21) iΨ +∇2Ψ+ − − 1 + 1Π Ψ = 0, (3.3) t |Ψ| 2 δρ 2 incorporating a de Broglie–Bohm potential term. It was observed by Antanovskii et al. in [5] that if Π = 0 and E(cid:0)1|∇ρ|2,ρ(cid:1) = C|∇ρ|2 +νρ+ τ, (3.4) 2 2ρ2 ρ then (3.3) reduces, if C = 1, to the cubic nonlinear Schr¨odinger equation iΨ +∇2Ψ−ν|Ψ|2Ψ = 0. t If, on the other hand, C =(cid:54) 1, it is seen that reduction is obtained to a ‘resonant’ NLS-type equation [78] (cid:20) ∇2|Ψ| (cid:21) iΨ +∇2Ψ+ (C −1) −ν|Ψ|2 Ψ = 0. (3.5) t |Ψ| Moreover, if C > 0 as in the present capillarity context, (3.5) may be transformed to a standard cubic NLS equation with the de Broglie–Bohm term removed (see, e.g., [73]). Thus, in 1+1 dimensions with three-parameter model energy E(1|∇ρ|2,ρ) of the type (3.4) reduction is made 2 to a canonical integrable NLS equation. The capillarity system encapsulated in the (1 + 1)- dimensional version of (3.5) then becomes amenable to established methods of soliton theory such as inverse scattering procedures and inherits admittance of invariance under a B¨acklund transformation together with concomitant nonlinear superposition principle (see, e.g., [1, 77, 79] and literature cited therein). It is noted that a gravitational potential term Π = −ρg is readily Ermakov–Painlev´e II Symmetry Reduction of a Korteweg Capillarity System 7 accommodated in the above reduction. Detailed qualitative properties of capillarity systems with model laws of the type (3.4) with K´arm´an–Tsien-type law κ(ρ) = C/ρ, C > 0, (3.6) have been recently set down in [16] while travelling wave propagation in (1 + 1)-dimensional capillarity theory has been investigated in [10]. Here, a more general class of model energy E(1|∇ρ|2,ρ) laws is considered, namely that with 2 κ(ρ)|∇ρ|2 R(ρ) E = + , 2ρ ρ so that, with Π = 0, (3.3) produces the class of NLS equations (cid:20)(cid:18) dκ (cid:19) ∇2|Ψ| (cid:18) dκ (cid:19) dR(cid:21) iΨ +∇2Ψ− 1+ |Ψ|4 − 1 κ(ρ)+ |Ψ|2 (∇|Ψ|)2+ 1 Ψ = 0. (3.7) t dρ |Ψ| 2 dρ 2 dρ Thus, capillarity systems encapsulated in (3.7) are isolated which may be aligned with NLS equations of the type (2.1) in the case c = 0, m = −2, n = 0. This occurs for the multi- parameter class of model energy laws with E(cid:0)1|∇ρ|2,ρ(cid:1) = C|∇ρ|2 +λρ− 2µ +2ν + τ, (3.8) 2 2ρ2 ρ2 ρ where λ, µ, ν and τ together with C > 0 are real constants. Importantly, this includes in the case µ = 0 the class which has been recently subject to a detailed qualitative analysis in [16]. The κ(ρ) capillarity relation is seen to be of the K´arm´an–Tsien type (3.6). Here, λ = −Cγ2c , µ = −Cγ2c , ν = −Cγ2c , (3.9) 3 4 5 in accordance with the relations (2.11). The associated class of NLS equations (cid:20) (cid:21) |Ψ| µ iΨ +Ψ − (C −1) xx +λ|Ψ|2+ +ν Ψ = 0, t xx |Ψ| |Ψ|4 hence, admits symmetry reduction via the wave packet ansatz (2.2) to the hybrid Ermakov– Painlev´e II equation (cf. (2.10)) d2|Ψ| σ +[c +c +c ξ]|Ψ|+c |Ψ|3 = , (3.10) dξ2 1 5 2 3 |Ψ|3 where (cid:34) (cid:35) 1 (cid:18)I(cid:19)2 σ = +µ . (3.11) Cγ2 γ Interestingly, this symmetry reduction to an integrable Ermakov–Painlev´e II equation will be admitted by Korteweg-type capillarity systems with the particular model energy laws of the type discussed in [16]. Under the translation ζ = ξ+(c +c )/c , with c (cid:54)= 0, (3.10) becomes 1 5 2 2 d2|Ψ| σ +c ζ|Ψ|+c |Ψ|3 = , (3.12) dζ2 2 3 |Ψ|3 8 C. Rogers and P.A. Clarkson where the Madelung relation (3.2) shows that, in the present capillarity context |Ψ| = ρ1/2. Thus, ρ1/2 is governed by a hybrid Ermakov–Painlev´e II equation while in terms of the density ρ it is seen that (3.12) produces d2ρ 1 (cid:18)dρ(cid:19)2 2σ = −2c ρ2−2c ζρ+ , (3.13) dζ2 2ρ dζ 3 2 ρ which is equivalent to P (1.4) (through a rescaling of the variables). This link between the XXXIV Ermakov–Painlev´e II equation (3.12) and P (1.4) has been noted previously in the context XXXIV of a Painlev´e reduction of a classical Nernst–Planck electrodiffusion system in [3]. We remark that the special case of equation (3.13) with c = 0 was considered by Gambier [38, pp. 27–28], 3 who linearised the equation. Multiplying (3.13) with c = 0 by ρ and differentiating gives 3 d3ρ dρ = 4ζ +2ρ, dζ3 dζ which has solution ρ(ζ) = C Ai2(z)+C Ai(z)Bi(z)+C Bi2(z), z = −c1/3ζ, (3.14) 1 2 3 2 with C , C and C constants. The solution ρ(ζ) given by (3.14) satisfies (3.13) only if c = 0, 1 2 3 3 σ = 0 and 4C C = C2. 1 2 3 In the sequel, it is convenient to proceed with c = −1, c = −1, σ = −1 (cid:0)α+ 1(cid:1)2, 2 2 3 4 2 whence (cid:34) (cid:35) ε = 1Cγ2 > 0, λ = Cγ2 > 0, (cid:0)α+ 1(cid:1)2 = −4 (cid:18)I(cid:19)2+µ , (3.15) 2 2 λ γ where the latter requires that µ < −CI2/λ < 0. The Ermakov–Painlev´e II equation (3.12) is then linked to P (1.4) via the relation ρ = |Ψ|2 > 0. XXXIV The well-known connection, in turn, between P (1.3) and P (1.4) is readily derived II XXXIV via the Hamiltonian system dq ∂H dp ∂H II II = , = − , dz ∂p dz ∂q where the Hamiltonian H (p,q,z;α) is given by II H (p,q,z;α) = 1p2−(cid:0)q2+ 1z(cid:1)p−(cid:0)α+ 1(cid:1)q, II 2 2 2 leading to the coupled pair of nonlinear equations dq dp = p−q2− 1z, = 2qp+α+ 1, (3.16) dz 2 dz 2 (see [44, 66]). Elimination of p and q successively in (3.16) duly leads to the P (1.3) and II P (1.4). Thus, in the present capillarity context, the density distribution ρ(ζ) is given by XXXIV dw ρ(ζ) = +w2+ 1ζ, dζ 2 Ermakov–Painlev´e II Symmetry Reduction of a Korteweg Capillarity System 9 where w(ζ) is governed by the P equation II d2w = 2w3+ζw+α. dζ2 Here, the concern is necessarily restricted tosolutions of P (1.4) in regions in whichρis XXXIV positive. Interestingly, the importance of positive solutions of P (1.4) also arises naturally XXXIV in the setting of two-ion electro-diffusion. Thus, in the electrolytic context of [7, 13], the scaled electric field Y was shown to be governed by the P equation II d2Y = 2Y3+zY +α, dz2 and associated ion concentrations by dY p = ± +Y2+ 1z, ± dz 2 with parameter 1−A /A − + α = , 2(1+A /A ) − + and A = −Φ /D , Φ being the fluxes of the ion concentrations and D diffusivity constants ± ± ± ± ± arisingintheEinsteinrelation. Thus,itisseenthattheionconcentrations,whicharenecessarily positive, are governed by P (1.4). This positivity constraint was examined in detail in [7] XXXIV for exact solutions in terms of either Yablonskii-Vorob’ev polynomials or classical Airy functions asinducedbytheiteratedactionoftheB¨acklundtransformationof[59]forP (1.3). Theresults II apply ‘mutatis mutandis’ in the present capillarity context. 4 Iterated action of a B¨acklund transformation Here, the consequences of the following well-known B¨acklund transformation for P (1.3) are II applied in the present capillarity context. Theorem 4.1. If q (z) = q(z;α) is a solution of P (1.3) with parameter α, then α II 2α+1 q (z) = −q (z)− , (4.1a) α+1 α 2q(cid:48) (z)+2q2(z)+z α α 2α−1 q (z) = −q (z)− , (4.1b) α−1 α 2q(cid:48) (z)−2q2(z)+z α α are solution of P with respective parameters α+1 and α−1. II Proof. See Gambier [38] and Lukashevich [59]. (cid:4) The iteration of the B¨acklund transformations (4.1) allows the generation of all known exact solutions of P (1.3). II We note that eliminating q(cid:48) (z) in (4.1) yields the nonlinear difference equation α α+ 1 α− 1 2 + 2 +2q2 +z = 0, q +q q +q α α+1 α α α−1 which is known as an alternative form of discrete Painlev´e I [33]. 10 C. Rogers and P.A. Clarkson Theorem 4.2. If q = q(z;α) and p = p(z;α) are solutions of P (1.3) and P (1.4) α α II XXXIV with parameter α respectively, then 2α+1 q = −q − , α+1 α 2p α 2α−1 q = −q + , α−1 α 2p −4q2 +2z α α (cid:18) 2α+1(cid:19)2 p = −p + q + +z, α+1 α α 2p α p = −p +2q2 +z. α−1 α α Proof. See Okamoto [66]; also [35]. (cid:4) 4.1 Rational solutions The iterative action of the above B¨acklund transformation on the seed solution q = 0 of P with II α = 0 produces the subsequent sequence of rational solutions d Q (z) q (z) = ln n−1 , n ∈ N, (4.2) n dz Q (z) n correspondingtothePainlev´eparametersα = n, forn ∈ N, wheretheQ (z)aretheYablonskii– n Vorob’ev polynomials determined by the quadratic recurrence relations (cid:40) (cid:41) (cid:18)dQ (cid:19)2 d2Q Q Q = zQ2 +4 n −Q n , (4.3) n+1 n−1 n dz n dz2 with Q (z) = Q (z) = 1 [89, 92]; see also [18, 19, 22, 48, 50, 87]. The Q (z) are monic −1 0 n polynomialsofdegree 1n(n+1)witheachtermpossessingthesamedegreemodulo3. Moreover, 2 on use of the invariance under q(z,α) → −q(z;−α), it is seen that P (1.3) also admits the associated class of rational solutions II d Q (z) q (z) = ln n , n ∈ N, (4.4) −n dz Q (z) n−1 corresponding to the Painlev´e parameters α = −n, for n ∈ N. The rational solutions of P XXXIV (1.4) are given by d2 Q (z)Q (z) p (z) = 1z−2 lnQ (z) ≡ n+1 n−1 , n ∈ N, n 2 dz2 n 2Q2(z) n corresponding to the parameters α = n, with n ∈ N. It is clear from the recurrence relation (4.3) that the Q (z) are rational functions, though n it is not obvious that they are polynomials since one is dividing by Q (z) at every iteration. n−1 In fact it is somewhat remarkable that the Q (z) are polynomials. Taneda [87], used an alge- n braic method to prove that the functions Q (z) defined by (4.3) are indeed polynomials, see n also [36]. The Yablonskii–Vorob’ev polynomials Q (z) can also be expressed as determinants, n see [22, 46, 47]. Clarkson and Mansfield [22] investigated the locations of the roots of the Yablonskii–Vorob’ev polynomials in the complex plane and showed that these roots have a very regular, approximately triangular structure; the term “approximate” is used since the patterns