Добавил:
Upload Опубликованный материал нарушает ваши авторские права? Сообщите нам.
Вуз: Предмет: Файл:
Diss / 6 - 753p.pdf
Скачиваний:
64
Добавлен:
27.03.2016
Размер:
15.34 Mб
Скачать

ABCD and ODEs

689

K/Baa ; the model-dependent integer K corresponds to the degree of fusion (see for example [16]).

3 The Pseudo-Differential Equations

To describe the pseudo-differential equations corresponding to the An1, Bn, Cn and Dn simple Lie algebras we first introduce some notation. We need an nth-order differential operator [9]

Dn(g) = D(gn1 (n1)) D(gn2 (n2)) . . . D(g1 1) D(g0), (19)

D(g) =

d

g

,

(20)

 

 

dx

x

depending on n parameters

g = {gn1, . . . , g1, g0}, g= {n 1g0, n 1g1, . . . , n 1gn1}. (21)

Also, we introduce an inverse differential operator (d/dx)1, generally defined through its formal action

 

d

1

 

xs

1

 

 

 

 

xs =

 

+

,

(22)

dx

s

1

 

 

 

+

 

 

and we replace the simple ‘potential’ P (E, x) = (x2M E) of (5) with

 

PK (E, x) = (xh M/K E)K .

(23)

Using the notation of Appendix B in [12] the proposed pseudo-differential equations are reported below.

An1 models

The An1 ordinary differential equations are

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Dn(gn1(x, E) = PK (x, E)χn1(x, E),

 

 

 

 

(24)

with the constraint

 

n1 g

i =

n(n1)

and the ordering g

i

< g

j

< n

1,

 

i < j .

 

 

 

i=0

2

 

 

 

 

 

γ (g)

 

γ

 

 

 

 

We introduce the

alternative set of parameters γ

 

 

 

 

a

(g)

 

 

 

 

 

 

 

 

 

 

 

 

 

=

 

 

= {

 

 

}

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2K

a1

a(h

1)

 

 

 

 

 

 

 

 

 

 

 

 

 

γa =

 

gi

 

 

 

.

 

 

 

 

 

 

 

 

(25)

 

 

 

h M

 

 

2

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

i=0

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

690 Patrick Dorey, Clare Dunning, Davide Masoero, Junji Suzuki and Roberto Tateo

The solution χn1(x, E) is specified by its x 0 behaviour

 

χn

1 xn1g0 + subdominant terms, (x 0+).

(26)

 

 

In general, this function grows exponentially as x tends to infinity on the positive real axis. In Appendix B of [12], it was shown that the coefficient in front of the leading term, but for an irrelevant overall constant, is precisely the function Q(1)(E, γ ) appearing in the Bethe Ansatz, that is

 

M xM+1

χn1

Q(1)(E, γ (g)) x(1n) 2 e M+1 + subdominant terms, (x → ∞).

Therefore, the set of Bethe ansatz roots

 

 

{Ei(1)} ↔ Q(1)(Ei(1), γ ) = 0

coincide with the discrete set of E values in (24) such that

 

M xM+1

 

χn1

o x(1n) 2 e M+1 , (x → ∞).

This condition is equivalent to the requirement of absolute integrability of

x(n1)

M

xM+1

 

2 e

M+1

χn1

(x, E)

(27)

(28)

(29)

(30)

on the interval [0, ). It is important to stress that the boundary problem defined above for the function χn1 (26) is in general different from the one discussed in Sects. 3 and 4 in [12] involving ψ (x, E). The latter function is instead a solution to the adjoint equation of (24) and characterised by recessive behaviour at infinity. Surprisingly, the two problems are spectrally equivalent and lead to identical sets of Bethe ansatz roots.

Dn models

The Dn pseudo-differential equations are

Dn(g)

dx

1

Dn(g2n1(x, E)

 

 

d

 

= PK (x, E)

d

PK (x, E) χ2n1(x, E).

(31)

dx

Fixing the ordering gi < gj < h /2, the g γ relationship is

ABCD and ODEs

691

γa

 

 

 

 

2K

 

 

a1 gi

a

h ,

 

(a

=

1, . . . , n

2)

(32)

 

 

 

 

 

 

 

2

 

= h M

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

i=0

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

γn

 

1

 

 

 

 

K

 

n1 gi

 

 

n

h ,

 

 

 

 

 

 

= h M

 

 

 

 

 

 

 

 

 

 

 

 

 

2

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

i=0

 

 

 

 

 

 

 

 

 

 

 

 

 

 

(33)

γn

 

 

 

 

K

 

 

 

n2 gi

 

gn 1

 

 

n 2

h .

 

 

 

= h M

 

 

 

 

 

 

 

 

 

2

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

i=0

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

The solution is specified by requiring

 

 

 

 

 

 

 

 

 

 

 

 

 

χ2n1 xh g0 + subdominant terms,

 

(x 0+),

 

 

(34)

 

 

 

 

 

 

 

 

 

 

 

 

M xM+1

 

 

 

 

 

 

 

(x → ∞).

χ2n1 Q(1)(E, γ (g)) xh 2 e M+1 + subdominant terms,

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

(35)

Figure 1 illustrates Ψ (x, E)

=

xh

M

 

xM+1

 

(x, E) for the first three eigen-

 

 

 

2 eM+1 χ2n 1

values of the D4

pseudo-differential equation defined by K=1, M = 1/3 and

g = (2.95, 2.3, 1.1, 0.2).

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Fig. 1 Lowest three functions Ψ (x, E) for a D4 pseudo-differential equation

Bn models

The Bn ODEs are

 

 

 

 

 

 

 

 

 

dx

PK (x, E)χ2n

 

 

Dn(g)Dn(g2n

 

1(x, E)

 

 

 

PK (x, E)

 

1(x, E). (36)

 

 

=

 

 

 

 

 

d

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

With the ordering gi

< gj < h /2, the g γ relation is

 

 

 

 

 

 

 

2K

 

a1

 

 

a

 

 

 

 

 

γa

=

 

 

 

 

 

 

gi

 

 

h .

 

(37)

 

 

h

 

M

2

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

i=0

The asymptotic behaviours about x = 0 and x = ∞ are respectively

692

 

Patrick Dorey, Clare Dunning, Davide Masoero, Junji Suzuki and Roberto Tateo

 

 

χ2n1 xh g0 + subdominant terms, (x 0+) ,

 

(38)

and

 

 

 

 

 

 

 

 

2n1

 

 

M

xM+1

+

 

→ ∞

 

χ

 

Q(1)(E, γ (g)) x

h 2

e M+1

 

subdominant terms, (x

).

(39)

Cn models

The pseudo-differential equations associated to the Cn systems are

Dn(g)

dx

Dn(g) χ2n+1

(x, E) = PK (x, E) dx

1

PK (x, E) χ2n+1(x, E)

 

 

d

 

 

d

 

(40) with the ordering gi < gj < n. The relation between the g’s and the twist parameters in the BAE is

γa

 

2K

a1 gi

 

an ,

γn

K

n1 gi

 

n2

 

(41)

= h M

= h M

 

 

 

 

 

 

 

 

 

 

 

i=0

 

 

 

 

i=0

 

 

 

 

and

 

 

 

 

 

 

 

 

 

 

 

 

χ2n 1 x2ng0 + subdominant terms,

(x 0+),

 

 

 

(42)

+

 

 

 

 

xM+1

 

 

 

 

 

 

 

 

 

 

 

+ subdominant terms,

(x → ∞).

(43)

χ2n+1

Q(1)(E, γ )xnM e M+1

Using a generalisation of Cheng’s algorithm, the zeros of Q(1)(E, γ ) can be found numerically and shown to match the appropriate Bethe ansatz roots [12].

In general, the ‘spectrum’ of a pseudo-differential equation may be either real or complex. In the An1, Bn, Dn models with K = 1,1 the special choice gi = i leads to pseudo-differential equations with real spectra, a property which is expected to hold for a range of the parameters g (see, for example, [9]). The K > 1 generalisation of the potential (23), proposed initially by Lukyanov for the A1 models [17] but expected to work for all models, introduces a new feature. The eigenvalues corresponding to a K = 2, 3 and K = 4 case of the SU(2) ODE are illustrated in Fig. 2.

The interesting feature appears if we instead plot the logarithm of the eigenvalues as in Fig. 3. We see that the logarithm of the eigenvalues form ‘strings’, a wellknown feature of integrable models. The string solutions approximately lie along lines in the complex plane, the deviations away from which can be calculated [12] using either WKB techniques, or by studying the asymptotics of the Bethe ansatz equations directly.

1 The Cn spectrum is complex for any integer K 1.

Соседние файлы в папке Diss