Home Search Collections Journals About Contact us My IOPscience

Quantum dynamical aspects of two-dimensional spaces of constant negative curvature

This article has been downloaded from IOPscience. Please scroll down to see the full text article.

1989 J. Phys. A: Math. Gen. 22 3577

(http://iopscience.iop.org/0305-4470/22/17/022)

View the table of contents for this issue, or go to the journal homepage for more

Download details:
IP Address: 143.233.244.88
The article was downloaded on 06/07/2010 at 11:21

Please note that terms and conditions apply.

J. Phys. A: Math. Gen. 22 (1989) 3577-3596. Printed in the UK

Quantum dynamical aspects of two-dimensional spaces of

constant negative curvature
E N Argyrest, C G Papadopoulos$, E PapantonopoulosD
and K TamvakisII
T Institute of Nuclear Physics, NRC Demokritos, Ag Paraskevi, Attiki, Greece
t Department of Physics, University of Athens, G R 157 71 Athens, Greece
5 Department of Physics, National Technical University of Athens, Zografou Campus, G R
157 73 Athens, Greece
11 Physics Department, University of Ioannina, Ioannina, Greece

Received 6 February 1989

Abstract. We study the quantum mechanics of two-dimensional spaces of constant negative

curvature. The connection between different formulations is analysed. We derive the
propagator and find that in the semiclassical approximation, the time evolution of observ-
ables exhibits similar features as in the classical case.

1. Introduction

In modern theories of particle physics, such as superstrings or supergravity, chiral

matter fields invariably parametrise a curved Kahler manifold. Apart from perturbative
treatments [ 11, little is known about the quantum behaviour of such field theories with
non-minimal kinetic terms. In the one-dimensional case such systems correspond to
the motion of particles in curved spaces. Quantum mechanics on curved spaces is still
an open field with many unanswered questions.
A typical example, and perhaps the simplest, is the case of an SU( 1, l ) / U ( 1) space
parametrised with a simple complex field. It is related to the dilaton and is contained
in all N = 1 supergravity theories that appear in the local limit of compactified
ten-dimensional or four-dimensional superstrings. The one-dimensional case of this
field theory corresponds to a point particle moving in a two-dimensional space of
constant negative curvature. In this paper we study the spectrum and compute the
propagator that determines the evolution of amplitudes of such a system. In previous
papers [2], where we studied the classical system and some of its cosmological
implications, we noted the characteristic divergence of geodesics, a feature that is well
known to be related to chaoticity. In the quantum case similar features emerge when
we study the time evolution of expectation values on the full non-compact manifold.

2. General considerations

Consider an N-dimensional manifold viewed as a set of points in patchwise one-to-one

correspondence with open subsets of the Euclidean space R N . Let us denote the
coordinates as #I',. . . , c,bN. We assume the existence of a smooth symmetric positive

0305-4470/89/173577+20%02.50 @ 1989 IOP Publishing Ltd 3577

3578 E N Argyres et a1

definite metric gij(q5) in terms of which we define a line element, a connection and a
curvature tensor as
ds2= g v ( $ ) d 4 i d &
ri =1 Im
Jk 2g (ajgkm +akgjm -amgjk)
Rikl= akr;,- alrfk+ rLmrT- - r i m r ; .
+
In the case N = 2 n , a set of n complex coordinates can be introduced as Z A= qbA i&A+n
and zA= 2" = 4 , -i$A+n ( A = 1 , . . . , n ) . In a so-called Hermitian manifold there
exists a preferred class of coordinate systems for which gAB=gAB=O and d s 2 =
2gas d Z A d Z B= 2gf: d Z A d z B . In the special case in which

the manifold is called a Kahler manifold and the real function G ( Z , z ) is called a
Kahler potential. Chiral superfields coupled to N = 1 supergravity automatically span
a Kahler manifold. For a Kahler manifold the only non-vanishing components of the
connection are the unmixed ones r& and &. In addition we can show that
rf:B= a, In( det g,a)
which leads to
RAE= - d , a ~ ln(det g,a)
for the Ricci tensor RAE = R&. As an example of a Kahler manifold, consider the
homogeneous space SU( n, l)/SU(n ) x U( 1) with a Lagrangian

defined in terms of the Kahler potential G = -In( 1 - k2ZAZA)/ k2. It is easy to check
that this is a space of constant negative curvature (see appendix 1 )
R f : = - ( n- l)k2gf:. (2.2)
Consider now the simplest possible case of a point particle moving freely in a
general Riemannian manifold with a line element ds2= g, dq' dq'. From the
Lagrangian

we obtain the equations of motion

$+rj,gJqk=o. (2.4)
The distance travelled by the particle along a geodesic in a time interval [0, TI is
obtained by substituting the solution of the equations of motion in
rT
D [ q ' (T ) ,q i ( 0 ) ] = J dt(gvqiq')1'2. (2.5)
0

Since the energy is a constant of the motion D = T m , and therefore the value of
the classical action is connected to the geodesic distance through the formula

S.[q'(T),q'(0)]=~oTdr~g,qiqJ=ET=
D2/2T. (2.6)
 Two-dimensional spaces of constant negative curvature 3579

A Hamiltonian is defined in terms of the canonical momenta ri= gijq' as H =4g'rirj.

To quantise such a system, we can either impose canonical commutation relations
[ q i ,rj]= ih8; or consider the Feynman path integral. In any case the definition of the
Hilbert space must be made in terms of the inner product

(2.7)

where g=det(gi,) appears in the invariant volume element d"q&. In the { q i }

representation the momenta are represented by the operators

Hermiticity is proven as follows:

a
( ~ , ! / r ~-ih d k q g ' i 2 + * g - ' / 4 z ( iP4x)
/,y)=

= (XI Til +)*

provided +*g1I2X vanishes at infinity.

In the transition from the classical to the quantum Hamiltonian care should be
taken of the operator ordering. With the usual midpoint definition of the Feynman
path integral the various terms in the classical action correspond to Weyl-ordered
operators on the Hilbert space. The relation between the differently ordered quantum
Hamiltonians is not always simple. For example, the Laplacian operator, in terms of
our representation of momenta, is

For the simple Lagrangian L = f(q:+ q i ) / (1 - q: - q:)2, the extra terms reduce to a
constant and the Hamiltonian is just the Weyl-ordered one
h2
H =S(l - q : - - q : ) p * ( l - q : - 4:) = -1( 1 - q:- q y ($+$) (2.9)

with

(2.10)

In a similar fashion one can write down the Hamiltonian formalism and proceed
with the quantisation in the case of a Kahler manifold as well. Dynamics in any case
is governed by the Hamiltonian in terms of the propagator
G(q", q i ; T) = (q'jl exp(-iTH/h)lq'). (2.11)
3580 E N Argyres et a1

3. Classical motion in two-dimensional spaces of constant negative curvature

There are various models which are known to be equivalent descriptions of a two-
dimensional space of constant negative curvature [3]. One can go from one to the
other in terms of suitable coordinate transformations. An infinite two-dimensional
surface of constant negative curvature can be thought as one sheet of a two-sheeted
hyperboloid embedded in Minkowski space. A natural parametrisation of such a
surface (a pseudosphere) can be given in terms of pseudospherical coordinates 0 < 6 <
CO and 0 6 4 <21r. The line element on the pseudosphere is ds2 = (dO)’+sinh2 O(d4)2
while the invariant volume element is d cosh O d4. The Lagrangian of a free particle
moving on the pseudosphere is
L = f(b2 + sinh’ e&’),. (3.1)
Another description of the same space can be obtained by introducing K = r exp(i4) =
tanh(6/2) exp(i4). In this parametrisation (the Poincari disc) the Kahler property of
the manifold is explicit and the line element is
ds2= 4 d K d K ( 1 - K K ) - 2 = 4 [ ( d ~ ) ~
( d+~ ) ~ ] ( Yl ~- ) - ~ .
The geodesic distance of two points on the PoincarC disc is given by
D[K, K’]=cosh-’[l+21K - K ‘ 1 2 ( 1 - K K ) - ’ ( 1 - K ‘ K ’ ) - ’ ] . (3.2)
The corresponding formula in pseudospherical coordinates is
D [ 6 , 4 , e’, 4’1 = cosh-’[cosh 6 cosh 0’-sinh O sinh 8’ cos(4 - 4’)l. (3.3)
The continuous symmetries of the PoincarC disc can be realised by
K‘= (aK +p)/(p”K + a*) la12-Ip12=1
which correspond to the pseudo-unitary group SU( 1 , l ) . Another model describing
the same space is the PoincarC complex half-plane, which is obtained through the
change of variables
5 = (1 - K ) / (1 + K ) K = (1 - 5)/(1+ 5).
Thelineelementis d s 2 = 4 d l dc((5+f))-2=[(dx)2+(dy)2]x-2( - c o < ~ < c Oo<,X < C O ) .
The metric is again explicitly conformal.
The geodesic distance is
D ( f ;5’)= cosh-’[ 1 +2lf - f’I2(l+ si)-’].
c)-’(l‘+ (3.4)
The symmetries are realised by the transformation
5’= ( af - iP )/ (i rC + fi ) afi-p7=1
which again corresponds to SU(1,l) or PSL(2, R ) .
The classical equations of motion in pseudospherical coordinates are
e - ; sinh(2O)d = 0 (3.5)
& sinh2 6 = L (3.6)
where L is a constant of the motion. In addition, the energy is E =$(e2+sinh2 6 6 ’ ) .
The solutions of the equations of motion can be expressed as O =
O(O(O), +(O), d(O), d(0); t ) , 4 = 4 ( 6 ( 0 ) , 4(0),6(0), d(0);t ) or in terms of the con-
stants E and L as 6 = e(O(O), 4 ( 0 ) ,E, L;t ) 4 = +(e(o), +(O), E, L;2 ) .
 Two-dimensional spaces of constant negative curvature 3581

They are explicitly

cosh e( t ) = a cosh( tm+b ) (3.7)
exp(2b + 2 t m ) + (L’-2E)/(L2+2E)
+ ( t ) = +(O)+tan-’
2 L m / ( L2+ 2 L F )
exp(2b) + ( L2- 2E)/ ( L2+ 2 E )
-tan-’

where
b=cosh-’( cosha@(O) ).
The form of the solutions in terms of PoincarC plane coordinates was given in [2].
Although e( t ) and C#J(
t ) appear to describe a complicated motion, we are dealing with
a free particle, and since the geodesic distance is
a e ( t ) , d(t), + ( O ) l = tm
the corresponding equation for the geodesic distance will have the trivial form
d2
- m e ( t ) , 4(tL e(o), 44011 =o.
dt2
The phase space of the classical motion can be parametrised in various ways using
suitable variables, which exhibit different geometrical aspects of the motion. In general,
a classical trajectory could be fully parametrised by three quantities, for example e(O),
8(0), $(O), ( 4 ( 0 )plays a trivial role). Equivalently, we could use the values of the
conserved quantities of each model. In the PoincarC disc, three conserved quantities
that represent the symmetry algebra are, in Cartesian PoincarC-disc coordinates,
B * -1
- 2 ( 1 - x2 - y ’) My + Mx B2 =-i(L - x2y2)Mx+ My
M = XPy -YPx p x = X / ( l -x2-y2)2 etc.
These three constants satisfy the S U ( 1 , l ) algebra
(4,
B21= - M {Ml B,) = B2
Y { M , 8 2 ) = -B1.
The Hamiltonian, being a Casimir of SU(1, l ) , is
H = f ( B : + B: - M2).
To have a description of the trajectories, we can express B1,B2 and M in terms of
two angles ++, +-
and the energy E as
B~= 2v‘Z cos i(++ + +-)/sin f(4+- +-)
B~ = 2 d E sin f(4++ +-)/sin f(++ - 4 - )
M =2LF cot f(++ - +-).
The angle +- gives the angular location of a particle which, starting from the boundary
of the PoincarC disc and moving on a geodesic with constant energy E, reaches again
the boundary with an angular location ++. If we go to half-plane coordinates and
put the point ++ at infinity, we find exponentially fast ( e x p ( - t m ) ) converging
trajectories, while a slightest variation a++ results in exponentially (exp( tm))
separating trajectories as t + 00. As is usually said, the model exhibits hyperbolic flow,
meaning roughly that the phase-space trajectories diverge.
3582 E N Argyres et a1

4. Quantum mechanics in two-dimensional spaces of constant negative curvature

The classical Hamiltonian describing a particle on the PoincarC disc is

1
H =- (1 - KR)’iin (4.1)
2m
in terms of the canonical momenta n = K/(1 - KK)2, 71. = E/(1 - KK)2. The quantum
Hamiltonian is given by the operator
1
f f = - ( i - ~ E ) ~ ~ (KK)=--
i--
h2 a2
(1 - KR)’- (4.2)
2m 2mR2 aKaK
which is equal to the Weyl-ordered one up to a constant. The momenta are represented
by
ih
n = -- (1 - KK)-
a (1 - KR)-’
R aK
(4.3)
ih
71.= -- (1 - KR):
a (1 - K K- ) -1 .
R dK
On the other hand in the (0, 4} representation we have

(4.4)

Here R is the radius of curvature of our space, so that all variables are dimensionless.
The Hilbert space is defined with an inner product

($l,y)=
I d K dK(1-KR)2$*(K,K),y(K, K ) = I l m d c o s h 0 [02‘d4$*(0, 4)~(0,4).
(4.5)
The energy spectrum can be obtained from the solution of the eigenvalue problem
a 2mR’
$E(O, 4 ) = - FE ~ $ ~ ( O4, ) . (4.6)
This equation is separable and has delta-function normalisable solutions [4]
$:(O, 4 ) = NT exp(im4)P{-,,,(cosh 0 ) (4.7)
where m = 0, * l , , . , , *CO,0 s A c CC and = (A2+b)h2/2mR2.The functions PK-1/2
are the conical functions [ 5 ] . The energy eigenvalues form a continuum starting from
E ~ =h2/8mR2 and extending to infinity. The eigenfunctions 4:, describing the free
particle on the entire pseudosphere, form a complete orthonormal set

joEdA $:(e, 4)$:*(0‘,+’)=S(+-+’)S(cosh 0-cosh 0’) (4.8)

m=-x

d cosh f3 {o’ dr+ $?*( 0, 4)$::( 8, 4 ) = 6,mtS(A -A’) (4.9)

provided
2n
N : = ( A tanh(.rrA) ) T(iA +$)
T(iA + m +;)’ (4.10)
 Two-dimensional spaces of constant negative curvature 3583

The time evolution will be described by the propagator

G(cosh 8,4, cosh e', 4'; t )
= 2
m=-m
jomdA
I
r(iA+;)
n h ( d ) T(iA + m +$)
A t a 21T
exp[im(+ - +')]Ps-l/2(cosh 8 )

x P ~ - , 1 2 ( c o s he') exp[ -i(A2 +$)ht/2mR2]. (4.11)

The summation over the angular momenta m can be performed (appendix 2) and we
are led to

G(D; t)
21T
Io" D) exp[-i(A2+a)Rt/2mR2]
dA A tanh(l~A)P,,-~/~(cosh (4.12)

where
D = cosh-'[cosh e cosh 8'-sinh 8 sinh 8' cos( $I - +')I (4.13)
is the geodesic distance between the initial and final point on the pseudosphere.
The propagator solves the Schrodinger equation

(H - i - aaJ G ( D ; t )
h2 1 a
2mR2 sinh D aD aD at
1 "
=- dA A tanh(l~h)P,,-,/~(cosh
D)
2.rr 0

{ [
x -ihG(t) - ihO(t)( -%)(A2+:)

R2
-8(t)-
2mR2
(A +$>I exp[-i( A ++)ht/2mR2

= -iAG(t)G(cosh @-cosh @')a(+-+'). (4.14)

Here we used the property

sinh a
ae )
f(D(8, 4, e', 4') =-sinh1 D -
dD (sinh DS)
and [ 5 ] for the representation of the delta function in terms of conical functions.
The Fourier transform of G ( D ; t ) takes the form

6(D ; w ) = I-,
m
d t exp(iwt)G(D; t )

=-
1 "
2.rr 0
jdA A tanh(nA)P,,-,,,(cosh D) d t exp[iw -i(A2+a)ht/2mR2]

=
21T
lom dA A tanh( T A )
w
CA-l,2(cosh Dl
- ( h / 2 mR2)( A +$) + iE *
(4.15)
3584 E N Argyres et a1

where Q is the conical function of the second kind, which although a solution of the
time-independent Schrodinger equation, is not an eigenfunction since it is not normalis-
able on the entire pseudosphere, (4.15) becomes

-$!"
-
857 h dx(
Qifi-I/*(COSh D ) - Q-iv~-1/2(cosh D )
x - f ( w ) -iE x-f(w)-iE
where
f(w) =2mR2(w-h/8mR2)/h.
Using the asymptotic forms
Q i f i - 1 / 2 ( ~D - x-'l4 e x p ( - i D h )
~ ~) h
X-03

~ ~) h= [ Q i ~ - 1 / 2 (D)]*
Q - i f i . - 1 , , 2 ( ~D -
~ ~ ~ h x - " ~exp(iD&)
x-m

and going to the complex x plane, we find that the first integral is zero (by closing the
contour in the lower half-plane) whereas the second integral (by closing the contour
in the upper half-plane) gives

(4.16)

Notice that the Fourier transform 6 has a branch point at wb = h/8mR2, implying
that the energy spectrum is continuous with a lower bound E,,, = hwb = h2/8mR2,in
agreement with the eigenvalues of the Schrodinger equation.
In terms of the propagator G ( D ;t ) the time evolution of states can be written as

(L( e, 4, t ) = lIm d cosh 8' d 4 ' G ( D (8, 4 ; e', 4'; t))rL( e', 4'; 0). (4.17)

Equivalently, time evolution could be studied in the Heisenberg picture starting

from the Heisenberg equations of motion, which, however, are rather difficult to solve.
The classical action between two configurations q ' ( T ) and q ' ( 0 ) is given for a free
particle in terms of the geodesic distance
S c [ q ' ( T I , q1(0)1= D2[q'(T I , qi(O)1/2T.
In the semiclassical approximation, the propagator is given by
G(q'(7-Lq ' ( 0 ) ; T )

In our case, we compute

fj2SC 62Sc
6 cosh e( T)S cosh e(0) 6 cosh e( T)S4(O) - 1
--- D
det
S2Sc S2S, T 2 sinh D
 Two-dimensional spaces of constant negative curvature 3585

which gives

The power of the time factor in front reveals the dimensionality of the system.
In order to estimate higher orders in h, we can substitute G( D ; T) in the Schrodinger
equation (4.14) and compare powers in h :
h2 1 d
2mR2 sinh D aD

x--
mR2 1 (D )”2
-
- exp(imR2D2)[1+ i h TF( D)/ mR2]= 0
27rih T sinh D 2hT
where F ( D) is an unknown function to be determined. It is important to keep in mind
that the small-h approximation is a large-distance approximation and should not be
valid at small D. The order-h correction can be obtained as

plus, of course, other terms of order h2.

The semiclassical propagator falls off at large distance exponentially like
1
-aexp(-D/2)
T
( l-- 4TR2 T) exp(imR2D2/2hT)

in contrast to a particle on a plane, which propagates with (1/ T) exp(imR2D2/2hT).

The semiclassical approximation to the propagator can also be obtained from the
integral expression of the conical function [ 5 ] :

P i h - 1 / 2 ( ~D~ )~=-
h
a coth(7rA) du sin(hu)(cosh U -cosh D)-’”
7T

Substituting this expression in the integral form of the propagator, (4.12), we get

du(cosh u-cosh D)-”2 dh A sin(hu) exp(-i&,t/A)

&( -
=--e ( t ) iht )-3’2/1
du u(cosh -cosh D)-1’2 U
a r ’ 4 2mR2
xexp( --)mR2u2 exp( -2).
(4.19)
2iht
With the change of variables U = w2+ D, the integral becomes
m
cosh(w2)- 1
I =exp(-mR2D2/2iht) do(”’+ D)( cosh D S sinh(w’) sinh D)-”’
o2 6J2

= exp(imR2D2/2ht) d o f ( w ) exp(-mR2w2D/iht).
3586 E N Argyres et a1

Expanding f ( w ) =f(O) + f ’ ( O ) w +. . . and noticing that f(0)= D / V ” we obtain

(1
CO
D
I =e xp(im~~~~/2ht) dw exp(-mR2Dw2/iht)+itO(h)
-
4
Thus

x exp(-iht/8mR2) exp(imR2D2/2ht)(l+itO(h)) (4.20)

which agrees with the result of the semiclassical formula apart from the zero-point
energy term which, in any case, to this order is unity.
A test of how good an approximation this is to the exact propagator is provided
by a comparison of (4.20) with (4.19), the latter being evaluated numerically to any
degree of accuracy desired. Introducing the dimensionless Euclidean time
7=iht/2mR2=ii
the exact propagator, (4.19), can be written as
G ( D ;T ) = (2.rr)-3/2

J
x e x p ( - ~ / 4 ) ~ - ~ ’ ~ du
D
$ u exp(-u2/4~)(coshu -cosh D ) - * l 2

exp( - u 2 / 4 ~ )
I:
= ( 2 ~ ) - ~~x ’P (~- T / ~ ) T - ~ ’ ~ du(cosh U -cosh D)’l2

X
sinh u
U coth U -1 +-
by integration by parts. The integrand of the above expression is a smooth function
of U which vanishes at the lower and upper bounds and can thus be calculated
numerically to any degree of accuracy.
The result of this calculation Gexact(D; T ) ( ~ T ) ~e ’x~ p ( ~ / 4 ) is ~ ’ ~ in figure 1
~ shown
(full curve) together with the semiclassical approximation (broken curve) as a function
of D for different values of T. We observe that for T S 1, the classical approximation
is very good for all D. It is also very good for any D and T such that D/&.- ’I. It
fails progressively more and more as T increases ( T >> 1) and D/&< 1. For example,
for T = 10’ it fails for D < 10 by a factor of 2-4.
A more direct way of obtaining the propagator is by making use of the Mehler-Fock
transform. Consider the Schrodinger equation for G(D, t ) (equation (4.14)). Take
the Mehler-Fock transform [3] in D and the Fourier transform in t, i.e.

Id
X

G ( D ;t ) = dA A t a n h ( ~ A ) P - ~ ~ ~ + ~D~)( c o sd ho exp(iwt)G(A, w )

and
8(t)b(cosh 6-cosh 0‘)8(4-+’)
1 1 ”
dA A tanh(~A)P-,/2+i,(coshD ) 1 7 --3o
dw exp(iot).
 Two-dimensional spaces of constant negative curvature 3587

Figure 1. The exact (equation (4.19), full curve) and the semiclassical (equation (4.20),
broken curve) propagators as functions of D for different values of r ; for T = the two
cannot be distinguished in this scale.

Then
(H-ihd/dt)G(D; t )

= los.dA A tanh(.rrA) 1s.

--cc
d w exp(iwt)@A, w)[H-ih(iw)]P-llz+,,

- --
ih jOm dA A tanh(.rrA)
a:

dw exp(iwf)P-l/Z+iA.

Using the fact that [2]

Hp-l/Z+iA = EAP-1/2+iA
where EA = ( A 2 + $ ) h 2 / 2 m R ’ we
, obtain

G(A, w ) = --( 2ih~ ) fiw 1

’ + E~ - i E ’
3588 E N Argyres et a1

The ie prescription follows from the fact that

=-exp(-ieAt/h)B(t)

so that G obeys the causality condition. We thus obtain

dh A tanh(~h)pl,,-,~,(cosh
D ) exp(-is,t/h)
2T
which is (4.12), since e A / h= (A2+a)h/2mR2. If we now use the addition formula for
the P (appendix 2) in (4.12) and compare the resulting expression with (4.11) for G,
we conclude immediately that the normalised energy eigenfunctions are those given
by (4.7).
We carry out next the quantisation in the Poincar6 half-plane in order to compare
the results with those obtained on the pseudosphere. In half-plane variables the
classical Lagrangian is
L, = imR2(x2 y 2 ) / x 2+ (4.21)
leading to the classical Hamiltonian
H,= (1/2mR2)x2(p:+p:). (4.22)
Using
1
- 0
X2
gij =
1
0 -
X2

we get the following representation of the momentum operators:

a
px = -if(:-) py = -ih--. (4.23)
dY
So that the time-independent Schrodinger equation becomes

(4.24)

Letting
44x9 Y ) = exP(i~Y)x”2g(x)
we obtain an equation for g(x):
x2g”+x g ’ - [k2x2- (2mR2/ h 2 ) E+a]g = 0.
Letting

P= [7I)&-
2mR2
(E

the equation becomes

x2g”+xg’ - [ k2x2+ ( i ~ ) ~=] 0.
g (4.25)
 Two-dimensional spaces of constant negative curvature 3589

This is the modified Bessel equation, which has solutions for any real p, i.e. E 2
h2/8mR2. The normalised solution is

(4.26)

Since x ' / ~ K ~ , ( x-+x+,3c

) 0, the value k = 0 is excluded.
We see then that in the PoincarC half-plane quantisation gives the same energy
spectrum, but both quantum numbers are now continuous, in contrast to the pseudo-
sphere case where m is integer. The same results have been obtained by the path
integral method [ 6 ] .

5. Expectation values and time evolution

Wavefunctions describing the system under study can be expressed as a superposition

of energy eigenfunctions

T(iA +f)
m=-m 271. T(iA + m +f)
x e ~ p ( i m 4 ) P ~ - , , ~ ( c o0s) .h
The coefficients C,(A) are expressible in terms of @(e,4 ) as

( d cosh Io2rd 4 exp(-im+)Bz?:/,

) Ilm
A tanh( T A )
C,(A) = 0 (cosh e)$(e, 4 ) .
2rr
In the case of a &independent wavefunction, the coefficients can have only m = 0.
Setting C,(A) = S,,C(h), a general q5 independent wavefunction can be written as a
Mehler-Fock transformation

(5.2)

It is also possible to consider wavefunctions depending on the geodesic distance from

a fixed centre eo, 4,. A specific choice of coefficients
r(iA + m +f)
Cm(A)= (?(A)(-1)" exp( - i m ~ o ) P ~ ? , ~ 2 ( c o0,)s h
T(iA +f)
with the help of the Mehler addition formula (appendix 2 ) gives

$(cosh D ) = lom dA C(A)(

A tanh( T A )
271. ) -'I2
D)
~A-,/2(COsh (5.3)
with
cosh D = cosh 8 cosh eo-sinh eo cos(4 - 4,).
A family of localised wavefunctions is given by Gaussian functions of the geodesic
distance from a fixed centre. For example, let us consider

(5.4)
3590 E N Argyres et al

The time-evolved wavefunction will be given by

$(D(O, 4 ; @ 0 , 4 0 ) ; t )
r“0 r 257
d cosh 8’ J
= J, 0
d 4 ’ G(D(8, 4; e’, 4 ’ ) ;t)$(D(8’, 4’; Bo, Cpo); 0) (5.5)

It is easy to prove that

= lom d cosh O3 d 4 3 G( DI3; 7 1 - 73)G(D32; 7 3 - 7 2 ) . (5.6)

Using the semiclassical expression, (4.18), for the propagator in the above equation,
we get

For 73 = 0 and 72= - a / 2 this gives

which means that (see equation ( 5 . 5 ) )

$(D; 7) = $(e, 4, 0 0 , 4 0 ; t )

-
a+27=
’( -
s i n h D )li2exp(8;/2a) e x p [ - D 2 / 2 a ( 1 + 2 ~ / a ) ]

ff
-
-

Since (I)(D ; ?)I$( D ; f ) ) = exp( 8;?’/a( 1+ f2)), normalisation is not quite preserved
for ~ > >inl this approximation, unless eo is very small. Expectation values of
 Two-dimensional spaces of constant negative curvature 3591

D-dependent observables are readily computed. For instance, we find

($(D; i)lDl!b(0; 0)
= eo exp(e;/a j exp[-&/(a'+ ?')I

(e;+-
($(D; ? ) / D 2 / $ ( D ; i ) ) = e x p ( 8 ; / a ) e x p [ - 8 ; a / ( a 2 + f2)]
a2:i2).

For t >> 1 we have

(02)i5- 1 ;*=-1 -fi2t2

a 4m2R4
AD)^=-1 I--
a

a
( 4"> ;( ">
?*=- -
h2t2
4 4m2R4'
We see that (0);grows linearly with time just as in the classical problem where
D( t ) = tm.
It is interesting to also study the time evolution of the position operators 0 and 4
or the Poincart-plane ones
5 R = ( l + & / 2 = ( c o s h 6-sinh 6 cos 4j-l
Jl=(S-[)/2i=-sin4(coth ~+COS~)-'.
Considering a Gaussian state, properly normalised for ?>> 1,

exp[ - D2/2(a + i ?)I

we get

x exp[-aD2/(a2+ i2)](cosh 8 -sinh 0 cos +)-'

(51);=-;
a 1
Iom
jo257
d cosh 0 d4-
D
sinh D
xexp[-crD2/(a2+ ?')I sin +(coth e+cos +)-I.

The simplest special case one can think of is the case that the wavefunction is centred
at eo= dO=O.In that case the 4 dependence disappears and we get

( 5 ) i --ae j+ oi 2m d O B ]0257d4 exp[-ae2/(a2+ i2)](cosh 0-sinh @ c o s

=-
2a
CY2+ T2
lom d e 0 exp[-ae2/(cu2+ ?*)I = 1 i>> 1

and
3592 E N Argyres et a1

It can actually be proven that (lR)i=1 and ( l l ) i = O for any +-independent

wavefunction.
In order to see what happens for &dependent wavefunctions let us consider the
limiting case of Gaussian wavefunctions centred at infinity (i.e. 4o= 0, Bo + CO). Then
cosh DE; exp( Oo)(cosh 6 - sinh 0 cos 4 )
and

( l d ; a=/ 77ml I x d cosh 6 ~ 0 2 T d 4D~ e x p [ - a D 2 / ( a i2)](2

2 + cosh D exp(-B,))-’

D
- exp[ -aD2/( a 2+ ?’)I

where we used appendix 3. Let A = [a/(o’+ f’)] and, since eo>> 2, we obtain
D
(lR)i=exp(Bo)A2 dD-
jox
cosh D
exp( -A2D2)

= 2 exp( Bo)A2 lox d D D exp(D ) exp( -A2D2)

m

= 2 exp( 0,) exp( 1/4A2)

I,,,+ 1/2A
dx(x-iA) exp(-x2)

= exp(eo) exp(1/4A2) { exp[ (Aeo+-~ A ) 2 ] - ~ e r f c ( A B , t ~ ) } .

-

Assuming we can extrapolate to large times and using the asymptotic form of erfc:

[ 212 I).:(
1
erfc(z) + -exp(-z ) 1--+0 -
Z+m ZJ;; 2

we see that time dependence persists, namely

(5R) i/(lR)O a ’/ ( + + O( ( T4)*
Thus we see that in the +dependent case (lR) depends on time, going to zero as t +-CO,
in contrast to the case of the +independent wavefunction, where (lR)i+ 1. This
situation is, of course, analogous to the classical behaviour [2]. Thus we can conclude
that motion in the semiclassical approximation does not deviate qualitatively from the
classical motion.

Appendix 1

As an example of a space of constant negative curvature consider the Kahler manifold

with G = -( 1/ K 2 )In( 1 - K2ZAZA)(A = 1, . . . , n ) . The metric is
a2 1 ZAP
g + - d Z A d Z B G = 1 - K ’ZAZA( S : f K 2 1 - K 2z A - )
Then

det g z = exp Tr In
1
1
[ - K’Z.
Z
=exp[-nln(l-K’ZZ)+Tr 2 -1 (ZAZB)’

v=l Y (1 - K’ZZ)”
=exp[-n ln(1- K 2 Z 2 ) - l n ( l - K’ZZ)].
 Two-dimensional spaces of constant negative curvature 3593

Therefore
det g : = (1 - K2,&-("+').
Finally the curvature will be

R AB-- -~a a
- In( 1 - K2ZZ)-("+1)
azAaZB
= - K 2 ( n + 1)g:.

Appendix 2

The functions B;""(z) form the canonical basis for the irreducible representations of
the group SL(2, C) and can be viewed as playing the same role for the group SU(1,l).
A convenient representation, which can serve as a definition for the functions B;""(z),
is

where r is the unit circle prescribed positively, m and n are integers and 1 can be
complex (typically of the form 1 = kip - f , p > 0). The generating function of the B;""
is
a2

B;""(cosh e ) exp(-im+)
m=-a2

= exp( -in+)
(
e
!)
cosh 2-+ exp( i4) sinh 2 ""(cosh :+e exp(- 4 ) sinh

In the case n = 0, we have

m

m=-m
BT;-,,,(cosh e ) exp(-im+) = (cosh 8 + sinh 8 cos 4)*iA-"2,
The following properties are useful
B;""(cosh 0 ) = B;m3-n(coshe )
B;""(coshe ) = (-l)"-"B?~-,(cosh 0)
[By"(cosh e)]* = B$"(cosh e).
The Legendre functions P;"(z) with 1 = i A -f are called conical functions and are
related to Bz!1/2 through
T(iA +f)
Bz:,,,(cosh e)= P~-,,,(cosh e).
T(iA + m +f)
The orthogonality relation
3594 E N Argyres et a1

enables us to prove orthonormality for the energy eigenfunctions I)Y( 0, 4 ) . Indeed,

j l m dcosh 0 IoZT d ,4"LC 4)CLh"(R#I

T(-iA + m +$) T(iA'+ m +$)

= 27rNfmN:.'S,,.
T(-iA +$) r(iA'+;)

provided
27r
N ; ' = ( A tanh T A ) li2 T(iA+b)
T(iA + m +i)'

On the other hand, the propagator contains the sum over angular momenta

r(-iA +f) r(iA+&I

m=l +
T(-iA m +;) T(iA - m +&)

x Ps-l/2(cosh 6)P,"_,lz(cosh e')

T(iA +j)
1
T(-iA +f)
m=l +
T(iA m +;) T(-ih - m ++)

1 m

n cosh( T A ) exp[im( 4 - 4')J exp[ -im( 4 - 4')]

sin-'(-i?rA + m 7 r S 5712)
= flA-l/Z(COSh e)e,-1,2(COSh
+
sin-'(irrh rrm + 7r/2)
e')
+ 3
cc
+2 (-1lm cos[m(4 - ~ ' ) ] P ~ - l i , ( c oB)P;"_,/,(cosh
~h e')
m=l

= qA-l/Z(COSh e)
where the last step is the Mehler addition formula with
cosh $=cosh 0 cosh e'-sinh 0 sinh e'cos(4 -4').
Finally, the following formula holds:

jowdA A tanh(nh)&-,/,(cosh 8)=2rrS(cosh e-cosh @')8(4-4')

 Two-dimensional spaces of constant negative curvature 3595

which is equivalent to the statement

G ( D ; 0) = G(c0sh O-cosh O ’ ) S ( $ -4’).
Most of these formulae can be found in [5].

Appendix 3

A useful property is that

J= d cosh 0 lo2- d$f( D ) = 27r [*cosh Bo

d cosh D f ( D ) .

The proof proceeds as follows:

a cosh e
J = I m d c o s h D f ( D ) [02TdQdCOShD.
cosh eo

The Jacobian is
COS(^ - 4 0 ) J G [ ~ o ~ 2-(4,)~s:-
4 1)+ T’-
(Si- 1) cos2(4- 40) - 6:
where
$=cosh D 6 = cosh 8 to= cosh Bo.
Differentiating, we get
at
-- [-50*Fcos(4-40)J~~cos~(4-40)(5:-1)+12-6i)1-”~
at- [(Si-l) co~2(4-40~-6:l
Performing the integration over 4 we obtain zero for the second term and for the first
we obtain

= -cosh eo jo2r d4
sinh’ Bo cos2 4 - cosh2 0

d4
sinh eo cos 4 -cosh eo- jOzrsinh Bo cos 4 +cosh eo
_-
-
2 sinh eo [jO2-4 cos
“
- coth eo- lo2’ +
cos 4 coth Bo

[Io2- -
- 1
--
2 sinh Bo coth Bo 1 tanh Bo cos 4 1+ tanh eocos 4
- 1 47r
2 cosh Bo 41 - tanh2 eo
= 27r.
3596 E N Argyres et a1

Therefore

=27ij dcoshDf(D).
cosh 00

References

[ l ] Alvarez-GaumC L and Ginsparg P 1985 Finiteness of Ricci flat supersymmetric non-linear a-models
Preprint Harvard University HUTP-85/A016
121 Enqvist K, Papantonopoulos E and Tamvakis K 1985 Phys. Lett. 16JB 299
Enqvist K, Nanopoulos D V, Papantonopoulos E and Tamvakis K 1986 Phys. Lett. 166B 41
Papantonopoulos E, Uematsu T and Yanagida T 1987 Phys. Lett. 183B 282
[3] Balazs N L and Voros A 1980 Phys. Rep. 143C 109
[4] Davis A C, Macfarlane A J and van Holten J W 1983 Nucl. Phys. B 216 493
[5] Robin L 1957 Fonctions Spheriques de Legendre et Fonctions Spheroidnls (Paris: Gauthier-Villars)
[6] Grosche C and Steiner F 1987 The path integral on the PoincarC upper half plane and for Liouville
quantum mechanics Preprint DESY 87-035