Master of Science Thesis
Soliton solutions to Calogero-Moser systems
Axel Reis Philip
Supervisor: Professor Edwin Langmann
Mathematical Physics, Department of Theoretical Physics School of Engineering Sciences Royal Institute of Technology, SE-106 91 Stockholm, Sweden
Stockholm, Sweden 2019 Typeset in LATEX
Akademisk avhandling för avläggande av teknologie masterexamen inom ämnesområdet teoretisk fysik. Scientific thesis for the degree of Master of Engineering in the subject area of Theoretical physics.
TRITA-SCI-GRU 2019:383
c Axel R. Philip, November 2019 Printed in Sweden by Universitetsservice US AB, Stockholm August 2018 Abstract
We develop a general method of generating M-soliton solutions to the AN−1 classical N-body Calogero-Moser (CM) systems of rational, hyperbolic and trigonometric type, for arbitrary M < N. This method is based exclusively on properties shared between the Lax representations of these systems, and in particular on their relation to a certain functional equation. The formulae are obtained from the real and imaginary parts of Bäcklund transformations between systems of the same type. We also extend the method for treatment of the hydrodynamic limits (N → ∞) of these systems (not including the hyperbolic type), by reformulating our N-body formulae partly in terms of density and fluid-velocity fields. This is achieved by exploiting a connection between the CM Bäcklund transformations, and soliton solutions to the Benjamin Ono equation, which can be shown to exist virtue of the above-mentioned properties of the Lax representations. Lastly, though it is not included in our general method, we discuss a formula for the gener- ation of M-soliton solutions to the more general elliptic CM systems of classical AN−1 type.
These developments generalize particular results, recently obtained by Abanov et al. [1] [2], pertaining separately to the rational and trigonometric Calogero-Moser systems. We likewise expect that our results indicate a promising route toward further generalization, for instance to include the elliptic CM systems, and for adaption to other kinds of integrable systems with Lax representations, such as the classical relativistic CM systems.
We illustrate and examine the typical behaviour of our CM soliton solutions, in several in- structive and representative solution plots. We also provide brief reviews to the wide array of subjects pertinent to the discussion, including the CM systems themselves, their Kazhdan- Kostant-Sternberg construction by Hamiltonian reduction, Lax representations, the Olshanetsky Perelomov projection method, Bäcklund transformations, the pole ansatze to the KdV, Burger’s and Benjamin Ono equations, and solitons within many-body and hydrodynamic systems.
i Abstrakt
Vi utvecklar en generell metod för alstrande av M-soliton lösningar till klassiska AN−1 Calogero- Moser (CM) system i N kroppar, av rationell, hyperbolisk och trigonometrisk typ, för godty- ckligt M < N. Metoden är baserad uteslutande på gemensamma egenskaper hos Lax rep- resentationer av systemen, och då särskilt deras anknytning till en viss funktionalekvation. Ifrågavarande solitonlösningar erhålls som lösningar till real- respektive imaginärdelen av Bäck- lundtransformer mellan system av samma typ. Vi utökar även metoden till att kunna behandla M-solitonlösningar i de hydrodynamiska gränserna (N → ∞) av systemen (exklusive det hyperboliska systemet), genom att delvis omfor- mulera våra tidigare N-kroppsresultat i termer av densitets- och flödeshastighetsfält. Det senare uppnås genom ett samband mellan Bäcklundtransformerna och Benjamin-Ono-ekvationen, vilken kan visas existera till följd av ovan nämnda egenskaper hos Lax-representationer av systemen. Slutligen, även om dessa för närvarande inte ingår vår generalla metod, diskuterar vi M- solitonlösningar till de mer generella elliptiska CM-systemen av AN−1 typ.
Sammantaget utgör detta en generalisering av speciella resultat från Abanov m.fl. [1] [2], som hänför sig till det rationella- respektive det trigonometriska CM systemet. Vi håller det för troligt att vår metod och våra resultat anger möjlighet till ytterligare generalisering, till att omfatta exempelvis de elliptiska CM-systemen, och för anpassning till andra typer av integrabla system, såsom de klassiska relativistiska CM systemen.
Vi illustrerar det typiska beteendet hos solitonlösningarna i ett instruktivt antal världslinje- diagram, från vilka vi drar slutsatser av både kvalitativ och kvantitativ art. Vi ger även ett antal allmänna introduktioner till de många ämnen som berörs i det följande, däribland till Calogero-Moser systemen själva, deras Kazhdan-Kostant-Sternberg uppbyggnad genom sym- metrireduktion, Lax representationer, Olshanetsky & Perelomovs projektionsmetod, Bäcklund- transformer, pol-ansatser till KdV-, Burgers- och Benjamin-Ono ekvationerna, samt till solitoner inom flerkropps- och hydrodynamiska system.
ii Preface
Acknowledgements
I am grateful to Professor Edwin Langmann for his interested supervision and helpful advice, as well as for introducing me to the subjects on which this thesis is based.
Outline
This thesis is divided into two major parts. The first part, comprising chapter 2, pro- vides the theoretical background necessary to the discussion in the second part. In this part we review and list important results pertaining to the Calogero-Moser systems, their Kazhdan-Kostant-Sternberg construction by Hamiltonian reduction, Lax representa- tions, the Olshanetsky-Perelomov projection method, Bäcklund transformations, the pole ansatze to the KdV, Burger’s and Benjamin Ono equations, and solitons within many- body and hydrodynamic systems. The second part comprises chapter 3, and concerns soliton solutions to the Calogero-Moser systems. Herein we derive general equations for the generation of soliton solutions to a subset of many-body classical Calogero-Moser systems of AN−1 type, and to their hydrodynamic generalizations. In relation to this, we also illustrate the typical behaviour of these soliton solutions in a number of worldline diagrams. A few relevant derivations and proofs are not included in the main text, but are instead relegated to the appendices. In particular, the proofs on which our general derivation of soliton solutions is based, can be found in appendices B.1 and C.3. The structure of the thesis is explained further at the end of the introduction in chapter 1.
iii Glossary of technical terms and symbols
• I, II, III, IV: Labels the Calogero-Moser systems of, respectively, rational, hyper- bolic, trigonometric, and elliptic class, see §2.2
• I+: Labels the rational Calogero-Moser system subject to a harmonic external po- tential, see §2.2
• Auxiliary system: In chapter 3 refers to the Calogero-Moser system that governs the soliton solutions to a principal Calogero-Moser system.
• Calogero-Moser systems: Refers, for the most part in this thesis, to the classical Calogero-Moser systems of AN−1 type, see §2.2. • Hamiltonian reduction: See §2.3.
• KKS construction: Method of constructing certain Caloger-Moser systems by way of Hamiltonian reduction, see §2.3.
• Lax pair: A pair of operators or matrices L and M that satisfy L˙ = [L, M] and often comprise a Lax representation, see §2.4
• Lax representation: An alternative formulation of a dynamical system, in terms of linear operators of matrices, see §2.4.
• OP projection method: Method of explicitly integrating certain Calogero-Moser systems, see §2.5.
• Principal system: In chapter 3 refers to the Calogero-Moser system in which solitons appear.
• Solitons: Nonlinear wave-phenomenon, see §1 and §2.7.
• Special functional equation: α(x)α0(y)−α(y)α0(x) = α(x+y) α0(y)−α0(x) , where α(x) is a special Lax function.
• Special Lax function: An odd function α(x) is a special Lax function of the CM system with interparticle potential V (x) = α(x)α(−x) if α() = c−1/ as → 0, 0 2 α (x) = −α (x)/c−1 + const., and if it satisfies the special functional equation.
iv Contents
Abstract ...... i Abstrakt (på svenska) ...... ii Preface ...... iii Glossary of technical terms and symbols ...... iv
1 Introduction 1
2 Calogero-Moser (CM) Systems 5 2.1 Chapter summary ...... 5 2.2 Introduction to systems of Calogero-Moser type ...... 8 2.3 The Kazhdan-Kostant-Sternberg (KKS) construction of CM systems . . . 11 2.4 Lax representations ...... 14 2.4.1 Lax representations of many-body Hamiltonian systems ...... 14 2.4.2 Lax representations of isolated CM systems ...... 15 2.4.3 Examples of Calogero-Moser Lax representations ...... 17 2.5 Olshanetsky Perelomov (OP) Projection method ...... 19 2.5.1 I: Rational CM system ...... 19 2.5.2 I+: Rational CM system with external harmonic potential . . . . 20 2.5.3 II: Hyperbolic CM system ...... 21 2.5.4 III: Trigonometric CM system ...... 21 2.6 Bäcklund transformations ...... 22 2.6.1 General discussion ...... 22 2.6.2 Bäcklund transformations of Calogero-Moser systems ...... 24 2.6.3 Examples of BTs for CM systems ...... 25 2.7 Hydrodynamics and Solitons ...... 27 2.7.1 Calogero-Moser solitons in hydrodynamic systems ...... 28
3 Soliton solutions to Calogero-Moser systems 30 3.1 Solitons from Bäcklund transformations of CM systems ...... 32 3.1.1 General procedure ...... 32 3.1.2 I+: Rational model on the line ...... 33 3.1.3 I+: Examples of soliton solutions ...... 35 3.1.4 II: Hyperbolic model ...... 41 3.1.5 III: Trigonometric model on the unit circle ...... 42 3.1.6 III: Examples of soliton solutions ...... 43 3.2 Solitons from semi-hydrodynamic Bäcklund transformations ...... 48 3.2.1 Bidirectional Benjamin Ono equation ...... 48 3.2.2 General procedure ...... 49 3.2.3 I: Rational model ...... 50
v CONTENTS CONTENTS
3.2.4 III: Trigonometric model ...... 51 3.3 Concluding remarks on soliton solutions to the elliptic CM systems . . . 52 3.3.1 Summary of the unified method ...... 52 3.3.2 Soliton solutions to elliptic CM model (IV) ...... 53 3.3.3 Hydrodynamic soliton solutions to the elliptic model? ...... 54 3.3.4 Elliptic projection method and KKS construction? ...... 54
A Appendix on Lax representations. 55 A.1 A formal construction of the Lax pair of an integrable system ...... 55 A.2 Derivation of the CM Lax representations from Calogero’s ansatz . . . . 56 A.3 Local analysis of the functional equation and derivation of the general CM potential ...... 57
B Appendix on Bäcklund transformations 59 B.1 Proving the form of the CM Bäcklund Transformations (CMBTs) . . . . 59 B.2 Derivation of the CMBTs from the Schrödinger equation ...... 62
C Appendix on hydrodynamics and solitons 64 C.1 Some facts about the Hilbert transform ...... 64 C.2 Proof of theorem 2.7.1 ...... 64 C.3 Proof of theorem 2.7.2 ...... 66
Bibliography 68
vi Chapter 1
Introduction
To the classical Calogero-Moser (CM) systems belong a selection of dynamical systems, comprised of N interacting bodies in 1 spatial dimension, with the added virtue of being integrable. This includes some particularly simple examples of interacting particles in one dimension, such as N particles on the line interacting repulsively via an inverse cubed −3 force F ∝ [qi − qj] , as described by the Hamiltonian
N N 1 X X0 1 2 H = p2 + W (q ) + g2 , (1.1) rCM 2 i i q − q i=1 i,j=1 i j where g > 0 and the 0 on the sum implies that j 6= i. A further example is the Calogero- Sutherland system of N particles on the circle, interacting through an inverse trigono- 3 metric force F ∝ cos(qi − qj)/ sin (qi − qj), as obtained from
N N 1 X X0 1 H = p2 + W (q ) + g2 . (1.2) tCM 2 i i sin2 (q − q ) i=1 i,j=1 i j The examples, in the absence of any external potential (whereby W (x) = 0), constitute Liouville integrable systems; and (1.1) is known to be Liouville integrable with a harmonic 1 2 2 external potential included, W (x) = 2 ω x . They are however not integrable for arbitrary W (x). The aim of this thesis is the derivation of soliton solutions to the Calogero-Moser systems. The term soliton here refers to a phenomenon, present in nonlinear media, of localized waves bearing permanent form and propagating at constant speed. Moreover, the form (and hence identity) of the solitons should remain the same even after soliton- soliton collision, though the colliding soliton waves would then generally acquire a phase shift. The solitons might thus to some extent be characterized as waves with particle-like behaviour. The first analytically derived soliton waveform would in present day be recognized as the hyperbolic 1-soliton solution 1 1√ u(x, t) = v sech2 v(x − vt − x ) , (1.3) 2 2 0 1 to the Korteweg de Vries equation ut −12cuux +cuxxx = 0. This solution is meromorphic, with an infinite line of poles, pn(t), that is parallel to the imaginary axis and centered
1complex analytic except for isolated points which are poles
1 A. R. Philip Soliton solutions to Calogero-Moser systems
evenly about the real line [25] iπ 1 p (t) = x − vt − √ n + , n ∈ Z. (1.4) n 0 v 2 Moreover, its evolution in time is determined by the motion of these poles (or rather the motion of the pole-line along the real axis). This pole-evolution in turn is determined by a 1-particle Hamiltonian system (that is periodic along the imaginary axis), as was established by H. Airault, H.P. McKean & J. Moser in [9] and G.V. Choodnovsky & D.V. Choodnovsky in [3]. We provide a more complete review of hydrodynamic solitons in §2.7.
Soliton solutions to (AN−1) Calogero-Moser systems have been derived previously for two special cases, the rational CM system with external harmonic potential on the line and the trigonometric CM system on the circle, respectively, by Abanov, Gromov & Kulkarni in [1] and Abanov, Bettelheim & Wiegmann in [2]. The former is given by (1.1) with 1 2 2 W (x) = 2 ω x and qi ∈ R, and the latter by (1.2) with W (x) = 0 and qi ∈ [0, 2π)/2πZ (angles from 0 to 2π). Waves within these discrete systems of particles in one dimension must, of course, be longitudinal and generated by some pattern of coordinated concentration and dilution among the particles in the system. Hence they are waves of particle-density. A soliton wave is in addition permanent, wherefore its associated particle motions must be highly interdependent at all times and belong to only a restricted part of the phase-space avail- able to the system. Indeed, as will be shown in §3.1, the soliton solutions effectively reduce the number of degrees of freedom of the system, in the sense that, for every initial soliton-wave configuration, the continued evolution of the soliton solution can be encoded in an auxiliary many-body system of fewer degrees of freedom. This should be compared to the hydrodynamic soliton solutions mentioned above, and discussed at depth in §2.7, whose evolution is encoded in a finite many-body system of poles. In [1] the ’reduc- tion’ is achieved, for the rational CM system with harmonic external potential in N real coordinates xi ∈ R, by the ’soliton equation’
N M X 1 g X 1 ωxi = g − Re , (1.5) xi − xj 2 xi − zn j6=i n
g PM −1 together with momenta x˙ i = − 2 n Im [xi − zn] . The M < N coordinates zn ∈ C themselves belong to an auxiliary CM system of the same type, but defined on the complex plane. Solutions to this soliton equation are also M-soliton solutions to the rational CM system, as will be shown in §3.1. The soliton equation (1.5) does not possess any known analytic solutions, but in [1] it is solved approximately for M = 1. In this case, the single ’z1’ particle traces an elliptical trajectory on the complex plane (more generally each zn traces a different ellipse, so the zn system is not inherently one-dimensional), and an initial soliton configuration is determined entirely by an initial value z1(0) and the N equations of (1.5). We are at present able to reproduce the worldline diagram given in [1] with some additional detail, see fig. 1.1. In the figure, the worldlines to the principal system of particles {xi}, are represented by the set of black lines at iy = 0 (the real line). The blue curve around the principal system represents the elliptical trajectory z1(t), on the complex plane, of the auxiliary 1-particle system. As will be explained in §3.1.3, this auxiliary system can be considered as a positively charged particle on the complex plane, orbiting a system
Chapter 1 2 Section 1.0 A. R. Philip Soliton solutions to Calogero-Moser systems
of negative charges constrained to the real line. The trajectory of z1 then induces the soliton wave among the negative charges through electrostatic attraction.
1
0.8
π 0.6 t/2 ω 0.4
0.2
0 1.5 1 1.5 0.5 1 0 0.5 -0.5 0 iy -0.5 -1 -1 x/R -1.5 -1.5 Figure 1.1: World lines for 1-soliton solution to the rational Calogero-Moser system with harmonic external potential, comprising 40 particles on the real line (in black), with g = ω = 1. The soliton wave can be interpreted as being electrostatically generated by the elliptic 2D trajectory of an oppositely charged particle (in blue) starting at the complex coordinate z(0) = 1 + i (indicated by red dot), see §3.1.
Additionally, in §3.1, we offer a number of graphs pertaining to soliton solutions with M > 1 auxiliary particles and soliton waves (wherein soliton interaction comes into play), and also to soliton solutions of the trigonometric CM system constrained to the unit circle, as illustrated by fig. 1.2.
3
3 2.5
2 2 t t 1.5 1
1 0 4 0.5 2 4 0 2 0 Im -2 0 -2 -4 -3 -2 -1 0 1 2 3 4 Re -4 -4 phase (a) fig 1 (b) fig 2
Figure 1.2: Worldlines for 3-soliton solution to the trigonometric Caloger-Moser system of 8 particles on the complex unit circle (in red), with g = ω = 1. (a) The soliton excitation can be seen as being electrostatically generated by the orbit trajectories of three particles of opposite charge (in blue), starting at complex coordinates exp{izn(0)}, where z(0) = [ 0.6(1 + i), 0.6(1 − i), 1.2(4 − i)]. In (b) is plotted the evolution of the real angles of the particles on the unit circle. See §3.1 for details.
In [1] and [2] are also derived soliton equations that apply to the hydrodynamic limits of the rational and trigonometric CM systems. The method used is essentially based on
Chapter 1 3 Section 1.0 A. R. Philip Soliton solutions to Calogero-Moser systems
a discovery by D.V. Choodnovsky [8] (1978), namely that the soliton solutions
N M X 1 X 1 u(x, t) = − , (1.6) x − x x − z i i n n
to the Benjamin Ono equation2
ut + 2cuux + cH[uxx] = 0, (1.7) are governed by a pair of separate and independent rational CM systems in respectively N particles {xi}, and M particles {zn}. An analogous statement for trigonometric type solutions was found by H. H. Chen, Y. C. Lee & N. R. Pereira in [10] (1979). The Benjamin Ono equation is independent of N and M, which allows the soliton equations to be re-derived from it after the limit N → ∞ has been taken, such that the principal CM system is fashioned into a liquid, described by fields of density and flow-velocity.
∗ ∗ ∗
Present thesis is structured as follows:
• Chapter 2 introduces (§2.2) the Calogero-Moser systems and a few of their rel- evant properties, including (§2.3) their Kazhdan-Kostant-Sternberg construction by Hamiltonian reduction, (§2.4) their Lax representations, (§2.5) the projection method due to Olshanetsky & Perelomov, and their Bäcklund transformations (§2.6.2). The chapter concludes with solitons in integrable nonlinear hydrodynamic systems (§2.7), where we show that the CM systems are intimately related to the soliton solutions of the Burger’s Hopf and Benjamin Ono equations.
• Chapter 3 derives soliton solutions to the Calogero-Moser systems, from the back- ground knowledge of ch. 2. Soliton solutions to many-body CM systems of rational, hyperbolic and trigonometric type (§3.1) are derived, and their behaviour examined in a number of representative worldline diagrams. In addition, soliton solutions to the hydrodynamic extensions of the systems are derived (§3.2). The chapter con- cludes with a discussion on soliton solutions to the elliptic type CM system (§3.3.2).
• The appendices contain an assortment of calculations and proofs. In particular, we show that the auxiliary function β(x) to the general CM Lax representation may be written β(x) = −α(x)α(−x)/c1 + const. in terms of the Lax function α(x) and its asymptotic coefficient c−1. We also show that the Bäcklund transformations of the rational, hyperbolic and trigonometric CM systems exist virtue of the properties of the Lax functions belonging to their Lax representations (appendix B.1). For the same set of systems and again using only the properties of their Lax functions, we prove that the Bäcklund transformations govern the pole dynamics of certain solutions to the Benjamin Ono equation (appendix C.3), and that the dynamics take the form of the Bäcklund transformations mentioned above.
2 p.v. R ∞ f(z) H[f] denotes the Hilbert transform, H[ f ](x) = π −∞ x−z dz, where p.v. stands for principal value.
Chapter 1 4 Section 1.0 Chapter 2
Calogero-Moser (CM) Systems
2.1 Chapter summary
Below we summarize those principal conclusions and results, of present chapter, that are necessary to the discussion in chapter 3.
Systems of Calogero-Moser type (§2.2) The classical Calogero-Moser (CM) systems are described by a Hamiltonian of the form
H(q1, ...qN , p1, ...pN ), (2.1)
with qj and pj the positions and momenta of particles moving in one spatial dimension. In this work we restrict ourselves to the AN−1 type CM systems, with some external potential W (x), described by
N N 1 X X0 H = p2 + W (q ) + g2 V (q − q ), (2.2) CM 2 j j j k j jk where the 0 above the sum implies j 6= k, and the interparticle potential 1/x2, (rational), (I) a2/ sinh2 ax, (hyperbolic), (II) V (x) = 2 2 a / sin ax, (trigonometric), (III) ℘(x), (elliptic), (IV)
where ℘( x | ω1, ω2 ) denotes the elliptical Weierstrass function. Prominent special cases 2 include particles on the line interacting via a rational interparticle potential VI(x) = 1/x and subject to a harmonic external potential W (x) = 2ω2x2; and particles on the circle 2 interacting via a trigonometric interparticle potential VIII(x) = 1/ sin (x). We label the former system by I+
N N 0 1 X 2 2 2 2 X 1 2 + HI+ = pj + 2ω qj + g . (I ) 2 qj − qk j jk The systems of class I - IV with W (x) = 0, and the class I+ system, are Liouville integrable [14]. • The CM systems with W (x) = 0, we in the following will be referred to as the isolated CM systems as they are free from external influences.
5 A. R. Philip Soliton solutions to Calogero-Moser systems
Lax representations (§2.4.2) The isolated Calogero-Moser systems permit Lax representations in Lax pairs on the form L = δ p + (1 − δ )α(q − q ), nm nm n nm n m N (2.3) X M = δ β(q − q ) + (1 − δ )α0(q − q ), nm nm n l nm n m l6=n
where the N × N Lax matrices L and M obey the Lax evolution equation L˙ = [L, M]. The functions α(x), β(x) in turn satisfy V (x) = α(x)α(−x), β(x) = −V (x)/c−1 + β0 and the functional equation α(x)α0(y) − α(y)α0(x) = α(x + y) β(x) − β(y) , (2.4)
where the constant β0 is arbitrary but c−1 is the asymptotic coefficient of α(x) as x → 0. Note that, for any V (x), an odd Lax function may be chosen such that V (x) = −α2(x) 2 and β(x) = α (x)/c−1 + β0.
Special Lax representations of CM systems (§2.4.2) For systems of type I, II and III an odd Lax function may be chosen such that β(x) = 0 −α (x) + β0, and the functional equation then reduces to α(x)α0(y) − α(y)α0(x) = α(x + y) α0(y) − α0(x) , (2.5) which we refer to as the special functional equation. The Lax function we likewise term the special Lax function. Provided a special Lax function a CM Lax pair is given by L = δ p + (1 − δ )α(q − q ), nm nm n nm n m N (2.6) X M = −δ α0(q − q ) + (1 − δ )α0(q − q ), nm nm n l nm n m l6=n which constitutes what we term the special Lax representation.
The Olshanetsky Perelomov projection method of solution (§2.5) The projection method allows the isolated systems of type I, II and III, and the I+ system, to be solved exactly for qn(t), from an initial phase profile qn(0), pn(0). The qn(t) are obtained as the eigenvalues of a matrix that is directly integrable. For example, the solutions of the I system correspond directly to the eigenvalues of the N × N matrix with components t Xnm(t) = δnm qn(0) + pn(0)t + 1 − δnm ig . (2.7) qn(0) − qm(0) The same holds for the I+ system with respect to sin ωt ig sin ωt Xnm(t) = δnm qn(0) cos ωt + pn(0) + 1 − δnm . (2.8) ω ω qn(0) − qm(0) Analogous formulas for the type II and III systems can be found in §2.5.
Chapter 2 6 Section 2.1 A. R. Philip Soliton solutions to Calogero-Moser systems
Bäcklund transformations (§2.6.2)
Provided the canonical coordinates pj and qj are taken complex the subset of systems associated with the special functional equation permit Bäcklund transformations on the form N M X X x˙ j = − i α(xj − xk) + i α(xj − zn) + iωxj, k6=j n (2.9) N M X X z˙ = i α(z − z ) − i α(z − x ) + iωz , n n m n l n m6=n l where N and M are independent. Note that ω = 0 for all but the class I (rational) CM system, where α(x) ∝ 1/x.
Pole ansatze of the Benjamin Ono equation (§2.7)
Provided a special Lax function α(x), with asymptotic coefficient c−1, soliton solutions to the Benjamin Ono equation, ut+2cuux+c−1c H[uxx] = 0, are provided by the meromorphic functions
N M X X u(x, t) = i α(x − xj) − i α(x − zn) − iωx, (2.10) j n where, ω = 0 for all but the class I (rational) CM system. From the residue of the solutions inserted into the Benjamin Ono equation, as x → xi and x → zn respectively, it is found that the motion of these poles is governed by the Bäcklund transformations of (2.9).
Chapter 2 7 Section 2.1 A. R. Philip Soliton solutions to Calogero-Moser systems
2.2 Introduction to systems of Calogero-Moser type
A many-body integrable Hamiltonian system is said to be an AN−1 type Calogero-Moser (CM) system if it is expressible in the form
N N 1 X X0 H = p2 + W (q ) + g2 V (q − q ), (2.11) CM 2 j j j k j jk with interparticle potential V (x) in the classified set of meromorphic functions [12] 1/x2, (rational), (I) a2/ sinh2 ax, (hyperbolic), (II) V (x) = 2 2 a / sin ax, (trigonometric), (III) ℘(x), (elliptic), (IV) where ℘(x) is shorthand for the elliptic and doubly periodic Weierstrass function [18]
1 X 1 1 ℘( x | ω1, ω2 ) = 2 + 2 − 2 , (2.12) x (x + 2mω1 + 2nω2) (2mω1 + 2nω2) {m,n}6={0,0}
with primitive periods 2ω1 and 2ω2. More generally the interparticle potentials of Calogero- Moser systems are constructed from root systems of Lie algebras [17]. So for other types of CM systems the interparticle potential is not simply V (qi − qk), but rather some more complicated function based on the functions V (x). For example, the DN type system keeps the interparticle potential [17]
2 U(qi, qj) = g V (qi − qj) + V (qi + qj) . (2.13) A more complete classification of these root system Calogero-Moser systems is provided by Perelomov in [17]. There are also relativistic, quantum, and relativistic quantum Calogero-Moser sys- tems. For example, the nonrelativistic quantum AN−1 type system (without external potential) is written [13]
N N 1 X 2 X0 Hˆ = − i ∂ + g(g − ) V (q − q ). (2.14) 2 ~ xj ~ j k j jk For more on relativistic and quantum CM systems, see e.g. Ruijsenaars in [13]. Needless to say, the theory of Calogero-Moser systems is rich and extensive, while our time is finite. Thus, in the rest of this work we restrict ourselves to the classical Calogero Moser systems of AN−1 type with external potential, as defined by (2.11). These systems will in the following be referred to simply as the Calogero-Moser systems.
Note that pole at x = 0 of the interparticle potential V (x) implies a divergence for vanishing interparticle separation: qi − qj → 0 ⇒ V (qi − qj) → ∞. Hence a CM system with repulsive interaction, modelled by g2 > 0, affords a stable system with particle trajectories that cannot intersect (so the ordering of particles in one dimension is conserved). For the rest of this work we assume g2 > 0, such that our CM systems are composed of mutually repulsive particles.
Chapter 2 8 Section 2.2 A. R. Philip Soliton solutions to Calogero-Moser systems
2 Remark. The two most prominent CM models are given by V (qi − qj) = 1/(qi − qj) , W (x) = 2 2 2 2ω qi with qi ∈ R, and V (qi − qj) = 1/ sin (qi − qj) with qi ∈ [0, 2π)/2πZ (so the qi represent angles on the the unit circle). More generally, qi ∈ M where M is some smooth connected 1-manifold, referred to as the configuration manifold of the Hamiltonian system. The set of possible M then comprises all spaces topologically isomorphic to either the line or the cir- cle. One benefit with the two models above is that they are constrained, whereas the model 2 V (qi − qj) = 1/(qi − qj) , W (qi) = 0 with qi ∈ R, for example, yields a system of particles that drift apart with time.
In the absence of an external potential, i.e. W (x) = 0, the systems included in the Hamiltonian (2.11) are all Liouville integrable, in that they possess (a minimum of) N distinct conserved quantities, H1, ..., HN , in involution:
{Hi,Hj} = 0, where { , } here denote Poisson brackets. Out of convenience in the following we refer to these systems as the isolated Calogero-Moser systems, as their particles interact only with each other, and are free from external influences. We may then state that the isolated CM systems are Liouville integrable. This can be proven, for example, by using their Lax representations. The integrability of the elliptic class IV system (which is also the most general of the isolated systems included in (2.11)) has been proven by Perelomov in [14], see also [17]. Alternatively, for the isolated systems of type I, II and III it follows trivially from the Kazhdan-Kostant-Sternberg construction [16] (as Tr{Y n} : n = 1, ..., N yields N involute integrals of motion, see §2.3). With an arbitrary external potential in (2.11) the resulting system would generally not be integrable. However, it is known that the class I interparticle potential combined with a harmonic external potential, W (x) = 2ωx2, yields an integrable system. For convenience we classify this system, i.e.
N N 0 1 X 2 2 2 2 X 1 2 + H + = p + 2ω q + g , (I ) I 2 i i q − q i=1 i,j=1 i j
+ as the I Calogero-Moser system. Beyond the AN−1 type Hamiltonian (2.11) a few more integrable CM systems with external potential are known. For the quantum case a few can be found in table 1 of [15].
Lastly, the Weierstrass function is only determined as an ordinary function of x after its two primitive periods 2ω1, 2ω2 have been specified, whence different combinations of (ω1, ω2) in ℘(x), correspond to distinct potentials V (x). So the class IV system comprises a number of inequivalent CM systems. In fact, as evidenced by the following limits of the Weierstrass function [12]
2 ω1 → ∞, ω2 → i∞ : ℘(x) → 1/x , (2.15) 2 2 2 ω1 → ∞, ω2 = iπ/(2a): ℘(x) → a /3 + a / sinh ax, (2.16) 2 2 2 ω1 = π/(2a), ω2 → i∞ : ℘(x) → −a /3 + a / sin ax, (2.17)
Chapter 2 9 Section 2.2 A. R. Philip Soliton solutions to Calogero-Moser systems
wherein the V (x) of classes I, II and III are recovered, it comprises all potentials permitted by (2.11). Thus (2.11) can be written more compactly as
N 1 X X0 H = p2 + W (q ) + g2 ℘(q − q ), (2.18) CM 2 j j j k j jk which up to scaling corresponds to the most general AN−1 type Calogero Moser system known [11] [12].
Chapter 2 10 Section 2.2 A. R. Philip Soliton solutions to Calogero-Moser systems
2.3 The Kazhdan-Kostant-Sternberg (KKS) construc- tion of CM systems
Our presentation in the following is based on P. Etingof in [16] and Perelomov in [17].
On Hamiltonian reduction. The presence of continuous symmetries within a Hamiltonian system (T ∗M, H) reduces its effective degrees of freedom in that any initial configuration ( q(0), p(0) ) ∈ T ∗M de- ∗ fines a submanifold T Mred of lower dimension, within which the continued evolution of the system is constrained. So for a pair of initial q(0), q˜(0) related by a symmetry transformation T , such that q˜(0) = T q(0), it holds in continuation that q˜(t) = T q(t). Upon repeated transformations along a symmetry ’direction’, we pass through redundant descriptions of the same physical configuration of the system. This redundancy can be removed if the Hamiltonian system is explicitly reformulated as a system of fewer degrees of freedom, by the procedure of Hamiltonian reduction.
In essence, Hamiltonian reduction amounts to expressing the Hamiltonian system in terms of one of its degenerate configurations, after which the reduced symmetry becomes ’hidden’1 and the Hamiltonian function itself typically becomes more complicated. First however, the reduced manifolds must be determined. To this end consider a symmetry group G with Lie algebra g, acting on the symplectic phase manifold T ∗M of the Hamiltonian system. To the coadjoint algebra g∗ there is associated a map µ(x): T ∗M → g∗, called the moment map, satisfying certain properties (see [17] for details). The moment map is invariant under the action of G. Hence, if G is a symmetry of the Hamiltonian function, µ(x) generates the conserved quantities c ∈ g∗ associated with the symmetry. If the moment map can be found, its reverse for fixed values c, µ−1(c), yields the submanifolds related to conserved quantity c, and its quotient with respect to the −1 c-invariant subgroup Gc, µ (c)/Gc, yields the reduced phase manifold.
KKS construction of the CM systems. That the reduced Hamiltonian typically is more complicated than the original one, sug- gests in reverse a use of Hamiltonian reduction to construct complicated Hamiltonian systems from simpler ones, with more degrees of freedom, and where the ’hidden’ sym- metries of the complicated system are apparent. The isolated CM systems of class I, II and III afford just such a construction, from Hamiltonian functions on the larger configu- 0 ration space XN = {X} of N × N traceless matrices X, known as the Kashdan, Kostant & Sternberg (KKS) construction. ∗ 0 In the KKS construction, the phase manifold is the tangent bundle T XN , which might itself be mapped by canonical pairs of traceless N × N matrices (X,Y ) (since covectors are identified with vectors by the inner product Tr{XY }), and the Hamiltonian functions are invariant with respect to similarity transformations (X,Y ) → (UXU −1,UYU −1), (2.19) where U is an invertible N × N matrix. These similarity transformations constitute the action of the symmetry group P GLN (C) with Lie algebra g = slN (C) [16]. The action of
1That is, not apparent prima facie, though still technically present.
Chapter 2 11 Section 2.3 A. R. Philip Soliton solutions to Calogero-Moser systems
∗ 0 the group induces a vector field on T XN , which equivalently could be generated by the Hamiltonian function [17] F = Tr{ξ[X,Y ]}, ξ ∈ g. (2.20) The above must constitute the inner product between a vector ξ and a covector [X,Y ]. Thus a moment map µ : M → g∗, associated with the symmetry, is given by the com- mutator µ = [X,Y ]. (2.21) Then the set of reduced phase manifolds along the symmetry direction c, comprises (X,Y ) such that µ = [X,Y ] = c, (2.22) with the general solution [17] X = UQU −1,Y = ULU −1, (2.23) where Qnm = δnmqn(t), (2.24) ig Lnm = δnmpn(t) + 1 − δnm . qn − qm Remark. The L might be recognized as the Lax matrix of the rational Calogero Moser system, see §2.4.2.
I: Rational CM system The rational CM system derives from the Hamiltonian function 1 H = TrY 2 , (X,Y ) ∈ T ∗X0 , (2.25) 2 N 0 describing free motion on XN . Due to the cyclic property of the trace operation, the Hamiltonian must be invariant with respect to the group of similarity transformations. Consequently, the system may be reduced by projecting for instance X → Q, Y → L, and that this yields the rational CM system is demonstrated by the fact that
1 1 X X0 −2 H = TrL2 = p2 + g2 q − q . (2.26) 2 2 i i j i ij Remark. Tr{Y n} : n = 1, .., N yields N integrals of motion in involution. Hence the N-dimensional reduced system must be integrable.
I+: rCM in external harmonic potential Including an external harmonic potential, the isolated rCM Hamiltonian is recast into 0 1 X 2 2 2 X g 2 H + = p + ω q + . (2.27) I 2 i i q − q i ij i j 0 The system derives from harmonic motion on XN 1 H = TrY 2 + ω2X2 , (X,Y ) ∈ T ∗X0 , (2.28) 2 N through the same projection X → Q, Y → L, with X and L as in the previous.
Chapter 2 12 Section 2.3 A. R. Philip Soliton solutions to Calogero-Moser systems
II: Hyperbolic CM system
0 The hyperbolic CM system derives from motion along hyperbolic geodesics on XN , ex- pressed by the Hamiltonian function [16] 1 H = Tr XY 2, (X,Y ) ∈ T ∗X0 . (2.29) 2 N The projection in this case yields the reduced system
N 1 X X0 qiqj H = (q p )2 + g2 . (2.30) tCM 2 i i (q − q )2 i ij i j
However, by a canonical transformation to hyperbolic coordinates 2aqˆi = ln qi, pˆ = qipi it follows that
N 1 X X0 1 H = pˆ2 + g2a2 , (2.31) tCM 2 i sinh2{a(ˆq − qˆ )} j ij i j
and that