See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/223118217

Open Strings in Background Gauge Fiels

Article  in  Nuclear Physics B · December 1987

DOI: 10.1016/0550-3213(87)90164-7

CITATIONS READS

713 677

4 authors, including:

Scott A. Yost
The Citadel
139 PUBLICATIONS   4,606 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Skyrmions View project

String Theory View project

All content following this page was uploaded by Scott A. Yost on 03 March 2018.

The user has requested enhancement of the downloaded file.

Nuclear Physics B280 [FS 18] (1987) 599-624
North-Holland, Amsterdam

OPEN STRINGS IN BACKGROUND GAUGE FIELDS

A. ABOUELSAOODt, C.G. CALLAN 2, C.R. NAPPI 1 and S.A. YOST 1

Joseph Henry Laboratories, Princeton University, Princeton NJ 08544, USA

Received 3 September 1986

(Revised 6 November 1986)

Some properties of open bosonic strings in a background abelian gauge field are investigated.
We derive the equations of motion and effective action for the electromagnetic field, and discuss
the effect of the field on the string spectrum and partition function.

1. Introduction
The study of strings propagating in background fields has yielded interesting
results in the case of closed strings, where the conformal invariance has been used to
derive the variational equations of a spacetime effective action for the background
fields [1-6]. One purpose of this paper is to find the equations of motion of a
background gauge field in a theory of open bosonic strings. We will deal only with
abelian gauge fields, since the nonabelian case is somewhat more complicated.
Moreover, we will work in the approximation of slowly varying fields. In this
approximation, our result is exact to all orders in the inverse string tension or'. Our
method involves computing the beta function for the electromagnetic field at the
open-string tree level. Conformal invariance implies the vanishing of the beta
function, which gives the equations of motion for the field strength F~,. From these
equations, we can derive an effective action. Fradkin and Tseytlin [7] have obtained
an effective action for an abelian gauge field coupled to open bosonic strings by a
totally different procedure, using the Polyakov path integral. In the limit of a
constant field, the path integral reduces to a gaussian integral over fields on the
boundary, and can be exactly computed. Interestingly enough, the two procedures
yield the same action (just as they did in the closed string case) even though it is not
a priori obvious that they should*.

1Supported by NSF grant PHY80-19754.

2Supported by DOE grant DE-AC02-76ER-03072.
* The basic issue is that the beta function is associated with the one-particle-irreducible generating
functional while the Polyakov path integral naturally produces the connected generating functional.
The single-massless-particle poles in the connected functional lead to divergences in the path
integral which must be regulated. The most satisfying situation would be to find a unique regulator
which, in effect, truncated the connected functional to the one-particle-irreducible functional. Steps
in this direction have recently been taken by several authors [8-10].

0619-6823/87/$03.50©Elsevier Science Publishers B.V.

(North-Holland Physics Publishing Division)
600 A. Abouelsaood et al. / Open strings

An effective action for F~ has also been reconstructed up to terms quartic in /:

by calculating four-point scattering amplitudes, both in the case of the open bosonic
string [10] and in the case of the open superstring [11]. The result is the same in both
cases, and is compatible with ours. However, since our result is valid to all orders in
ct', one would have to sum the contributions of infinitely many string tree graphs to
obtain it using the S-matrix approach.
The method of extracting the effective action from the beta function cannot easily
be generalized to include the effect of string loops, since the beta function is
independent of the world-sheet topology. In an attempt to get some understanding
of the string loop corrections in the presence of a background field, we investigate
the vacuum-to-vacuum amplitude on the annulus, using two different methods. The
first, based on the stress-energy tensor, requires computing the exact propagator on
the annulus in the presence of a gauge field. The other, based on the operator
formalism, seems to yield more information about the dependence of the partition
function on the background field. This method requires computing the spectrum of
the open string in a background gauge field, an analysis which is interesting on its
own merits. The partition function on the annulus in the presence of gauge fields
has also been computed by Fradkin and Tseytlin [7], using path integral methods.
Our result agrees with the divergent part of their result.
This paper is organized as follows. In sect. 2, we derive the equations of motion
for the electromagnetic field, i.e. the string modification to the Maxwell equations.
This calculation is performed at the string tree level, using background field
methods. We show that this equation can be obtained by varying the Born-Infeld
lagrangian [7,12,13], which is therefore an effective lagrangian for the electromag-
netic field, at the string tree level. In sect. 3, we move to the one-loop level and
compute the Neumann propagator on the annulus when the boundary condition is
modified by the presence of gauge fields. This result is then used in sect. 4 to
compute the partition function on the annulus. In sect. 5, we find the spectrum of
the bosonic open string a constant background electromagnetic field. We study both
the case of strings with zero net charge and with nonzero net charge. The results are
used in sect. 6 to compute again the partition function on the annulus using the
operator formalism. Finally, in appendix A, we give an alternate derivation of the
beta function for F~,, and in appendix B, we discuss the effect of a toroidal
compactification on open strings in a magnetic field.

2. Modified Maxwell equations and effective action

In this section we will compute the equation of motion for an abelian gauge field
coupled to the open Bose string. We will work at the tree level in string theory and
represent the string world-sheet with the upper half-plane as in fig. 1. Throughout
this section we will adopt euclidean metrics both on the world sheet and in
spacetime.
 A. Abouelsaood et al. / Open strings 601

_T
0
Fig. I. Upper half-plane representation of the string tree-level world sheet.

The action for the string coupled on the boundary to the electromagnetic field is

S= 1 [~fMd2ZOaX~,OaX~+ifoMd.rA~O~Xg
,,~ ~ 2 ] (2.1)

where A n has been rescaled to contain a factor 2~ra'. In this calculation we will use
the background field approach [4,14]. If we expand the action (2.1) around an
arbitrary background ,~, i.e.

x,'(~-, o) = Y~(~-, o) + ~(~-, ,,), (2.2)

we get

1 v h

1
+~F,.~ v a.~ p + ~vv~W,~o,,,+ .

..) ] , (2.3)

where W~= O/aX ~ and F~ = ~7~A~- W~A~. Here we will be working in the ap-
proximation of slowly varying fields. Therefore we will neglect terms with more than
one derivative of F.
On-shell, the terms linear in ~ disappear since A" satisfies the equations of motion

~X~'= O,

O~X~ + iF~O,X'J aM = 0, (2.4)

602 A. Abouelsaood et al. / Open strings

where [] = 0 z + 0o2. Therefore the on-shell action is more simply

i 1 v ~ --t~

(2.5)
We want to compute the one-loop (in the field theory sense) counterterm to the
gauge coupling term in eq. (2.1), namely a counterterm of the form
i

From F, we will extract the beta function •a for the electromagnetic field A~,. By
imposing conformal invariance, i.e./~a = 0, we will derive the equations of motion
for F~.
The Neumann propagator in the upper half-plane satisfies the equations
1
2 ~ a , n O ( z , z') = - 8 ( z - z ' ) ,

OoG(z, z')lo= 0 = 0, (2.7)

where z = r + io. The solution is

G(z, z') = - a ' ( l n l z - z'l + lnlz - f ' l ) . (2.8)

To compute the counterterm (2.6) we could work with this propagator and sum up
all one-loop graphs with an external 0 , 2 and all possible insertions of the vertex
F~,~'O,~ ~'. This calculation is carried out in appendix A. A more straightforward
method is to compute the exact propagator in the presence of the gauge field F. In
the presence of the gauge fields the propagator must satisfy the following boundary
condition:
OoG( z, z" )~, + iF~XO,Gx,(z, z' )lo_o = O. (2.9)

We find that the solution to this equation is

1 1-F
at,,,(z,z')= - a ' 8,,~lnlz-z'l + 2k( l-+-F/l r , ~ ln( -
1/I+F\ ]
+ ~- [ -i--S-~) ~ l n ( f - z') . (2.10)
 A. Abouelsaoodet al. / Openstrings 603

tr=O
7 F #rX

Fig. 2. The only one-loop graph contributing to flA. The line . . . . indicates the exact F-dependent
propagator.

This is symmetric under the exchange z ~ z' and/~ <-->v. For F = 0 this propagator
reduces to the one of eq. (2.8).
The only counterterm to S t is given by the one-loop graph in fig. 2. The
evaluation of this graph gives

-i
AS,: 2~r-----7f d, ½~7.F,xa.~.G'X(,. (2.11)

where G"X(~, ~') is the propagator on the boundary (o = o ' = 0). In the limit ~---->r '
we get

G,x(~'--*r')-- -a' 1 + .2 I + F .+ . l n A. = - 2 ~ ' l n A ( 1 F 2 )~x

-1 ,
2 1-FJ~x

(2.12)
where A is a short-distance cutoff. The beta function can now be obtained by
differentiating with respect to the cutoff A,
0
= -F )x~- (2.13)

The equations of motion are therefore

V'F)(1- F )x~
- ' = 0. (2.14)

We should remember here that F is actually 2~ra'F and therefore eq. (2.14)
contains all orders in a'. We have obtained the exact beta function to lowest order
in derivatives of F. Graphs with two or more loops would introduce corrections of a
higher order in derivatives. At the leading order in et', eq. (2.14) reduces to
Maxwell's equations. Notice that, since F is a real antisymmetric tensor, F 2 has
only negative eigenvaluest, and hence eq. (2.14) makes sense.
A remark is in order. The propagator in eq. (2.10) can also be written as

I+F 2 F z-f']
a~,,(z,z')=-a' l n l z - z ' l + - - -l -n-l-z~- ~ ' l l - 1 - - F 2 ln~--S-~7]~ . (2.15)
604 A. Abouelsaood et al. / Open strings

This rewriting shows that we get an F-dependent logarithmic divergence from the
propagator (2.10) and hence a F-dependent contribution to the beta function only
when z ~ z' on the boundary of the world sheet. Suppose in fact that o = o' ~ 0.
Then the only logarithmic divergence comes from l n [ z - z ' [ in eq. (2.15), while the
F-dependent term l n l z - ~'[z ~ z' = ln(2o) is finite and gives no contribution to the
beta function. This means that the beta function for the dosed string fields (for
instance for the graviton and dilaton fields) are not affected by the presence of
gauge fields on the boundary.
We have computed the string correction to Maxwell equations and obtained eq.
(2.14). The obvious question now is to find the action whose variation gives these
equations. The answer is that although the beta function (2.13) is not the variation
of any action, it is possible to exhibit an action whose variation is X~,/3~, where
X ~ , ( F ) is an invertible tensor. Therefore, the variational equation is equivalent to
flA = 0. A general relation of this form between the effective action and beta
functions is suggested in [9,15], and a special case of this relation has been found to
occur for closed strings in [6]. Let us first notice that the following identity holds:

(1 - F2)~-~1/3~= W~ ]-7--F-5 ~,~- ~-F-~]~, x ~ p~ (2.16)

The second term in the r.h.s, of the above equation can be rewritten by making use
of the antisymmetry of F and of the Bianchi identity. Then eq. (2.16) becomes

(1 - ~ F )~,p+ I(I_~FF2)~,V~ t r l n ( 1 - r 2 ) .
F2)~-,1/3~=V~(11 (2.17)

This suggests that our equation of motion might be derived from an action that is a
function of trln(1 - F2). Along these lines one can indeed derive the lagrangian

~. = exP(¼trln(1- F 2 ) ) = exP(½trln(1 + F ) ) = (det(1 + F ) , (2.18)

whose Euler-Lagrange equations are

v/det(1 + F ) (1 - F 2 ) ; , l f l ] = O. (2.19)

Notice that this equation has the same solutions as flf = 0. The Born-Infeld action
~/det(1 + F ) is exactly the one obtained by Fradkin and Tseytlin using the Poly-
akov path integral [7]. A calculation of the gauge-field beta function has also been
carried out in [16] up to terms quadratic in a', with compatible results. In [16], the
nonabelian case is also considered.
 A. Abouelsaoodet al. / Open strings 605

3. Propagator on the annulus

We now would like to consider the effect of the gauge field at the next order in
the string loop expansion. To begin, we shall calculate the propagator on the
annulus in the presence of a background gauge field. As before, our result will be
exact to zeroth order in derivatives of F. Let us take the string world-sheet to be a
flat annulus with inner radius a and outer radius b, as shown in fig. 3. The
N e u m a n n function for the annulus satisfying the equations

1
2~ra;DG( z, z' ) = - 8 ( z - z' ) , (3.1)

O---r I
0 G ( z , z ' ) r=b = '~'
b ' ~C(z,z') ~-o = 0 (3.2)

is given by the infinite series [17]

o. l(b,

_la?": a?"z' 11"

N o t e that since the annulus has finite area, one cannot impose OG/Or = 0 both at
r = a and r = b. This is because eq. (3.1) and Gauss's theorem require
fr=a,b(OG/On)ds = - 2 e r a ' , which is incompatible with 9G/Or=--O. The simplest
allowable boundary condition is to set the normal derivative equal to a nonzero
constant on some boundary component, as in (3.2). This change in the N e u m a n n
b o u n d a r y conditions amounts to shifting the background field X ~ by its average
value on the boundary.

Or

Fig. 3. The word sheet for the one-loop orientable open string diagram may be represented by an
annulus. The modulus of the annulus is a/b.
606 A. Abouelsaoodet al. / Openstrings
We now would like to find the propagator on the annulus in the presence of the
gauge field F~. The boundary condition is then given by eq. (2.4). However, when
F 4=0, this boundary condition together with (3.1) and Gauss's theorem would not
permit the propagator to be single-valued. The simplest modification consistent with
a single-valued propagator is to require

0 i 0 0 at r=a
- a ' 8~.
O---~G~.,(z, z') - rrx,-~GX,( z, z') = ----~ at r = b . (3.4)

The propagator defined by (3.1) and (3.4) is then

G,.(,.,.): -,.~<< [~,.(:. z') + ~;p(~,,,)] (3.5)
where

G,~(z,z')=6~,ln(z-z')+8~,~ln
([ 1-(b ] z[[-(b}
al ":l) z']

'l-F' ~ 1 [[1 (a/2nb2][ 1 [al2nzS"]l

(3.6)

When deriving this, it is convenient to diagonalize F~. in 2 × 2 blocks,

0 i,
-fl 0
F~.,= 0 f2 , (3.7)
-f2 0

and to adopt a complex basis, so that each of these blocks is replaced by ifn for
n = 1... ½D, where D is the dimension of spacetime. Then the propagator becomes
a diagonal matrix with ½D complex entries. Note that the propagator (3.5) can also
be written as

1
a 'G~'p(z'z')=8~'lnlz-z'l+8~p~__tln (1 - b] z' kb] z
[ 1+F2] ( (a]2"b2 1 a]2"z~' )
+ t 1----2--~]~,~=1~1n 1 - b ] z2" -(-£] - ~

_ a/~"~' 1
F o~ \b} ze'J[ a~2,ez,] (3.8)
--(1---"~-~)~'n-~-lln [1 'b](al2nb2
][lzz'][ (~]/ !--~-]
 A. Abouelsaood et al. / Open strings 607

From eq. (3.8) it is obvious that this propagator reduces to the usual one (3.3) when
F~,,= O.
A remark on the beta function is in order. In sect. 2, we computed the beta
function for A~, on the upper half-plane, i.e. at open string tree level. Since the
conformal anomaly should be a local effect, the beta function should not depend on
the topology of the world sheet. We may now easily check this for the annulus.
Suppose we compute the graph of fig. 1 using the propagator (3.8). In the limit
z --->z', we find a logarithmic divergence in G(z, z') due to the first term lnlz - z'[.
Moreover, there are only two other terms in the series which can give rise to
logarithmic divergences. One of these is the term l n l l - aZ/z$'[ when z ~ z' on the
inner boundary, and the other is the term l n l l - z$'/b21 when z --+ z' on the outer
boundary. In either case, the propagator loop in fig. 1 gives the divergence

F2
1+ ] lnA,
G~,,(z + z')= -2~ra' 1 + -1--F-gl~,v

exactly as in the tree level case (2.12).

4. Vacuum amplitude on the annulus

In this section, we will compute the vacuum-to-vacuum amplitude for the annulus
by integrating the stress-energy tensor over the modulus a/b, as outlined in [18].
The stress-energy tensor may be computed using the propagator (3.8) on the
annulus in the presence of a gauge field found in the previous section.
The vacuum amplitude or partition function is given by the Polyakov path
integral

z= >2 fmg Xe-S g,x,, (4.1)

topologies

where

1 i 3
S[g, X] = ~fd2zv~gaOa=X"ObX~,+~fr:ab, dsA~,ff-fsX~

and gab is the world-sheet metric, with line element ds. Previously, we have always
set gab = gab" The gauge invariance of (4.1) allows us to choose this metric here as
well, provided we integrate over the modulus of the annulus. We will take the outer
radius b to be one, so the modulus is just the inner radius a. Therefore the partition
function can be written as an integral over d a of

Z(a) =f ~Xe -st~ob'xl

annulus of
modulus a
608 A. Abouelsaood et al. / Open strings

The partition function Z ( a ) for an annulus of modulus a can be computed using

the stress-energy tensor as follows. In terms of Z(a), the stress-energy tensor may
be written
2~ra' 8 In Z ( a )
ra = 8g

Therefore, we can write

0 1 8g ab
7alnZ(a) = 2~r~' fd2z¢~ - ~ a Tab. (4.2)

Therefore, if we know the stress-energy tensor, we can compute Z up to a

multiplicative constant (since the integration constant in (4.2) is undetermined).
The factor 8gaJSa in (4.2) requires some explanation. Above, we have fixed the
gauge in (4.1) by choosing a flat metric on the annulus and integrating over the
radius of the inner boundary. Alternately, we could fix the boundaries and integrate
over a one-parameter family of "Teichmiiller deformations" of the metric [19]. The
metric change 8gab induced by a change 8a of the inner radius is the Teichmiiller
deformation of gab which is conformally equivalent to changing the inner radius by
an amount 8a while keeping the metric fixed. The following calculation of 8gab/~a
will clarify this procedure.
Let us map the annulus of inner radius a to an annulus of radius a + 8a via a
mapping
z ~ z' = z + e(z, ~), (4.3)

which satisfies the boundary conditions

e(z, 2) = ( 0 at r = 1 (4.4)
eia~a at r = a,

so that the inner boundary is displaced a distance 8a perpendicular to itself. In

complex coordinates, the original flat metric is

ds 2 = [d2[ 2 = gz~dZ d~ + g~ d ~ d z ,
with
gz~ = g~.z = ! 2, g ~ = g ~ * = 2,

gz~=grr=gZZ=g~=O. (4.5)

The mapping (4.3) induces a metric on the new annulus, which to first order is

ds 2 = [dz'l 2 = I1 + 0zel21 dz + 0red~l 2 . (4.6)

 A. Abouelsaoodet al. / Openstrings 609

The traceless part of the change in the metric,

3g~ = O~e, 3g~z = O~---e, (4.7)

cannot be removed by a conformal transformation and is potentially a Teichmiiller

deformation.
In order for (4.7) to be a Teichmiiller deformation, it must be in the kernel of Pt*,
where P1 is the operator introduced in [19] to map a diffeomorphism into the
traceless part of the metric variation induced by it. This amounts to imposing the
condition
OzSg~ = O~3gzz= 0, (4.8)

which means that e(z, $) must be a harmonic function, i.e.

0~ 0~e = O. (4.9)

Eqs. (4.9) and (4.4) uniquely determine e to be

a(1-`z'2) (4.10)
a ( z , ~ ) = 3 a z i - - a----i "

Consequently,
1( a3a ]
]'
and we find that

O,n a, 4a f ,(1 1 }
The stress-energy tensor is determined by the two-point function

G=

Since the limit z ~ z' is singular, a regularization is required. Using the propagator
(3.5) found in the previous section, we get the regularized stress-energy tensor

OZO2' 2 ( 2 - - Z ' ) 2 = 4~rz2 =1 (1--a2n) 2' (4.12)

where D is, as usual, the spacetime dimension.

610 A. Abouelsaood et al. / Open strings

Notice that this result is completely independent of F~,. Substituting Tz~ and its
complex conjugate T~ into (4.11) and integrating over z gives
a 2n
OlnZ(a) 4aD ~ a2~ f ldr
Oa 1 - a2 n=l (1 -- a 2 n ) 2 r3 2Da n~= l (1---a--2~)2" (4.13)

Integration over d a gives, up to a constant [20],

In Z(a) = -D ~ ln(1 - a 2 ~ ) ,
n~l

and therefore,
oo
Z(a) = I-I (1 - a 2")-D (4.14)
n=l

So far, we have neglected the ghost contribution, which is obviously independent of

F~,~. Introducing it, one recovers the usual partition function for the annulus [21],

ida i~i ( 1 - a 2 n ) -24. (4.15)

Z = J a3 n ~ l

An interesting feature of this result is that it does not depend on F. However, note
that the above method would not be sensitive to an overall F-dependent factor. In
fact, using the operator formalism, we will find in sect. 6 that the partition function
actually has the form
Z F = det(1 + F)Z, (4.16)

where Z is the partition function (4.15).

5. Open string spectrum in a background gauge field

In this section we will discuss the spectrum of open strings in a constant
background abelian gauge field. In this limit, the oscillator modes, Virasoro algebra,
and spectrum may be found exactly. We assume that the electromagnetic field
couples to charges ql and q2 at the ends of the string. We will work in Minkowski
space and represent the world-sheet as the strip shown in fig. 4, with a spacelike and
~- timelike. The string action in Minkowski space is (setting a ' = ½)

rr -oo d~" [qlA~)/"(o= 0) + qEA,J/"(o = or)], (5.1)

 A. Abouelsaood et al. / Open strings 611

Fig. 4. Strip representation of the open string world sheet.

where X" = O,X" and XA = 0o X~,. Varying this action gives the usual wave equation
, ~ - X~' = 0 with the boundary conditions

X/, = qlFt,~f(~ at o = O,

X~ = - q2F,~3/"~ at a = ~r. (5.2)

We shall take F~ to have only spatial components. Since F is a real antisymmet-

ric tensor, we can make a rotation to put it into block diagonal form (3.7). The
blocks are independent, so we can concentrate on only one of them,

(0
In these two dimensions, it is convenient to introduce complex fields

X+= ~ ( X 1 + iX2), X_= ~()(1 -iX2) = X~.

(Note that these are spatial components, and have nothing to do with the light-cone
coordinates, which will never appear in our discussion.) The boundary conditions
(5.2) then take the form

X;(O) = -iaff+(O), X+(rr) = iflf(+(~r), (5.3)

where a = q l f and fl = q2f.

We first consider the case when q~ + q2 =~O, and then discuss the neutral string
case. When the net charge is nonzero, the normal mode expression for X+ is

X+ = x+ + i a,t~,('r, a) - btm,_m(r, o) . (5.4)

m=0
612 A. Abouelsaoodet at / Openstrings
Here the ~p, are the normalized mode functions that satisfy the wave equation and
the boundary conditions (5.3),

~bn -----In -- e I - 1/2cos[(n -- e)o + y]e -i("-~)', (5.5)

where n is an integer, and e = ( 1 / ~ r ) ( 7 + y ' ) with y = t a n - l a and y ' = t a n - l f l .
There are some important differences between (5.4) and the mode expansion for the
ordinary string [22]. Note that there is no term in the expansion (5.4) linear in ~.
This is because no function linear in ~- and o can satisfy the boundary conditions
when a + fl ~ 0. Also, since the integrals of the mode functions ~p, are nonzero, x+
is not identified with a center-of-mass coordinate.
The mode functions satisfy the orthogonality relation

f0,,--~-~,('r,
do alliO, + a S ( o ) + ~8(~r-a)]~p,,(r,o)=Sm, sgn(n-e 1. (5.6)
r

where q~0,q, -= q~0~q~- ~pa,~. Also, each ~, is orthogonal in this inner product to a
constant mode, i.e.

Tlio, + ~8(o) + B~(~-o)] ~,(,, o)=0.

f0~rdo r ,-, (5.7)

Moreover, these modes, together with the constant mode, are complete in the sense
that for any function q,(~-, o), the following relation holds:

~,o)= E sgn(.-0:.(~,o)

×Jo'~d:--'-~--
#"( ~" :)[iO".+aS(8)+flS(~r
- ~)] q~(~', ~)

i~ - .
a+/3

The canonical momentum P+(~, o) is defined in the usual way,

a.gP 1 . 1
e+(~', o ) = 0~?+
= -Ir- X _ ( ¢ , o ) - - -~rA (z,o)[q:8(o)+q28(cr-o)]. (5.8)

Because of the gauge interaction terms in (5.1), the canonical momentum contains a
boundary contribution. If we choose the gauge such t h a t

,4 = ~:(x::~- x~),
 A. Abouelsaood et al. / Open strings 613

we get
~P + = 2 _ + ~ i x _ [ ~ ( o ) + ¢~(~- o)1. (5.9)

Although the boundary contribution in (5.9) is gauge dependent, we shall see later
that physical quantities are gauge invariant.
We can use the orthogonality relations (5.6) and (5.7) to invert the Fourier
expansion (5.4) and write the Fourier coefficients x+, a, and b, in terms of X+,
)(+, X , and X_. We can then use eq. (5.9) and its complex conjugate to find )(_+
in terms of P:~ and X±. The result is

a n= £~ d o G( , r , o )[ P _ ( ¢ , o ) - - [(.- ~) + ~a(o) + ~fl~(~-

q./-
0)] x+(¢, o)],

b,,= doqJ_,.(,.o) [P + ( ~ ' , o ) - 7i[(m+e)-lza3(°)-~fl3(~'-°)]X-('r'°)] '

x+--~
a+/3
do[triP_(¢,o)+ [½a3(o)+½fl3(qr-o)lx+(%o)].
Using the canonical commutation relations of the theory,

[x.(.,.),x~(.,o,)]=0, [e.(.,o),e.(.,o')]=0,
[ x.(., ~), e.(~, °,)] = gGa(° - o,), (5.10)

for #, I, = + , we find that the Fourier modes satisfy

[a,, atm] = 8,,,, [a,,aml=[a~,a~]=O,

[b,, b~l = ~,m, [b,,b,,l=[b+,,b~]=O, (5.u)
and all of the a oscillators commute with all of the b oscillators. We also find the
zero-mode commutator
,'/r
Ix+, x_l = - - (5.12)

where x_ = xt+. The commutation relation (5.12) may seem surprising, since in
ordinary string theories, the constant modes x~ commute. In fact, eq. (5.12) implies
that (i/rr)(a + fl)x_behaves as a conjugate momentum operator for x+. It is easily
verified that eqs. (5.11) and (5.12) do not change if one makes a gauge transforma-
tion A ---,A + xTA. Therefore, the spectrum will be gauge invariant.
614 A. Abouelsaood et al. / Open strings

Next we study the Virasoro algebra of the theory. The Virasoro operators L , are
defined as usual [22] to be the Fourier components of the constraint equation, i.e.

D-1
1
~E :
(}~'g, ~ X ; ) 2
=--
1 E L. ei"(~ ±°) (5.13)
p,=O q'/' n = - o o

We can find the contribution of X 1 and X 2 to eq. (5.13) by using (5.4) and (5.5)
together with the fact that

2
1 • ,' 2 _ S ~ _ ) ( ~ j ~ "}- S t )
_ _ .

#=1

This gives for n >t 0,

L(n1'2)= E [(m-e)(n+m-e)]l/2a~an+m
m=l
o~
1/2 t
+ ~_~ [ ( m + e ) ( n + m + e ) ] brmbn+m
m=0

n-1
"~1 1 / 2 .
+ • [(m + e)(n - m - e)i Oman_ m (5.14)
m=O
and
;:) = L 1,2),

where the superscript (1, 2) refers to the contribution of these two spatial compo-
nents of X~, to which must be added the contributions of the other twenty-four
components.
It is a straightforward but tedious exercise to calculate the commutators of the
Virasoro operators. The result is

[L~I'Z),L~ '2)] = (n - m)L(14-z).,+ 6.+.,,ot12,

[2--(n3-n)+ne(1-e)]. (5.15)

This is the expected result, apart from the extra c-number piece n e ( 1 - e)rn+,,,o
which is linear in n, and hence can be eliminated by a shift L o --* L 0 + ½e(1 - e ) .
This corresponds to changing the normal ordering constant a 0 from its original
value of 1 to 1 - %(1 - ~).
We can now describe the spectrum of the theory. The effect of the magnetic field
is to shift the frequencies of the a oscillators by - e and those of the b oscillators by
+e. The zero mode operators have been drastically changed: the string total
momentum operators p ± do not appear in the mode expansion, while instead we
 A. Abouelsaood et al. / Open strings 615

get extra Fourier operators b*0 and b o which create and annihilate quanta of
frequency ~. In ordinary string theory, we would take our states to be eigenstates of
p _., which commute with L 0. We would not take them to be eigenstates of the
center-of-mass operators, since these do not commute with L o. However, when
F ~ 0, x + and x _ do commute with L0, since the zero modes are missing from the
mode expansion (5.14). Therefore, the states may be taken to be eigenstates of x+,
for example.
N o w L o does not depend on x _+, so there is an infinite degeneracy if the first two
dimensions are not compact. (The degeneracy is finite if these dimensions are
compact. This case is discussed in appendix B.) This situation is the same as that of
a charged particle moving in a constant magnetic field: the states form equally
spaced Landau levels of infinite degeneracy, the separation between consecutive
levels being proportional to qB. In our case, q = qx + q2 and B = f , and the
operators bo and b*o move the string from one Landau level to another. The
frequency separation e is proportional to qB when qB is sufficiently small.
The excited states of the string are created by acting on the ground-state
wavefunctions with the oscillator creation operators. At the first excited level we
have the states altlx+) which have (mass) 2 = - ½e(1 + e) and hence are tachyonic,
and the states b*llX+) which have (mass) 2 = ½e(3 - e) and hence are massive. This is
reminiscent of the behavior of Yang-Mills fields in the presence of a constant
chromomagnetic field condensate, where one gluonic polarization becomes tachyonic
and the other becomes massive [23].
We have so far considered the case where the string has nonzero total charge.
When the total charge vanishes, several changes occur. First, since e = 0, the string
has integer modes, as it does in the absence of F~, (although the phase shift ~, does
not vanish). More importantly, the zero modes change completely, as can be seen
from eq, (5.12), which is not well-defined when a + fl ---, 0. The m = 0 mode in (5.4)
also disappears. On the other hand, we now have the freedom to add to X+ a term
proportional to r - iao, which satisfies the boundary conditions when a +/3 = 0.
Now we can write the mode expansion as

x+ +p_[':-ia(o-½1r)] oo
X+(~',o) = Vt~+ a2 +i Z [a.~Pn('r,o)-b~b-.(T,o)],

(5.16)

where ~k. is given by (5.5) with e = 0.

The operator p _ which now appears in the mode expansion becomes the minus
component of the total momentum when F vanishes. The commutation relations
for the zero mode operators are

[x~,,x,] =0, [p~,p~] = 0 , [x~,p~] =i3~,~, (5.17)

616 A. Abouelsaoodet al. / Openstrings
where #, v = _+. The factor (1 + a2) -1 in front of the linear terms in (5.16) was
introduced to allow x~ and p, to satisfy the standard commutation relation. As in
the charged string case, the oscillator mode operators satisfy (5.11). There is,
however, no b o mode for the neutral string. We conclude that the Fourier modes ot
a neutral string satisfy the same commutation relations as they would in the absence
of an external magnetic field. The Virasoro operators, when written in terms of the
oscillators and the new zero modes, also have the same form as in the absence of the
field. In fact, using (5.16) we get

f( + + X (~= e ~ iV [ P + + ~ l v/-n( a"e-i"(~ ± °) + bt"e~n(~+°) ) ] (5.18)

Using (5.13) we get

00

Lo=P+P-+ E n(a~a~+b~b,,), (5.19)

with p+p_ = ~(Pl

1 2 +p22). Hence, the spectrum of a neutral string is not affected by
the magnetic field. However, as we shall see in the next section, the vacuum-to-
vacuum amplitude is changed by the field.

6. Vacuum amplitude in the operator formalism

In this section we will compute the vacuum-to-vacuum amplitude in the presence
of a constant gauge field, using the operator formalism [21]. Specifically, we will
compute the vacuum amplitude

Z = - ½Tr l n ( L o - ao) (6.1)

on an annulus, where a 0 is the normal ordering constant and the trace refers to a
sum over string states and an integration over the loop momentum. In the absence
of gauge fields, (6.1) gives the one-loop orientable contribution to the cosmological
constant. We shall restrict our attention to the neutral string case, and use the
results of the previous section to compute Z in the presence of a background
magnetic field.
We have learned that when the string is neutral, both the normal ordering
constant and the Virasoro operators are the same as in the absence of the gauge
field. The naive conclusion seems to be that Z should not depend on F~. However,
the momentum integration in (6.1) requires some care. The momentum integration
measure in (6.1) is determined by putting the system in a box of length L. From
(5.16) we see that x~ and p~ are rescaled by a factor of (1 + ~ 2 with respect to
X~,. If X, goes from 0 to L, then x~ goes from 0 to L~/1 + a 2 . Therefore the
 A. Abouelsaoodet al. / Openstrings 617

momentum eigenvalues are p, = (21rn/L)(1 + a2) -a/2. This means that, in the
thermodynamic limit, the sum over momentum becomes

Z 2
E ~ ~-~ lVri--ff~a2f d p .
n

Hence the integration over d p l and dp2 introduces a factor 1 + a z = 1 + f 2 (setting

qx-- - q 2 = 1), and similarly for each of the other pairs of transverse momentum.
The vacuum amplitude Z F in the presence of F will differ from the free amplitude
Z given by (4.15) by a factor of

D/2
I-I (1 +f,2) = det(1 + F ) .
i=l

This result does not depend on the specific way we rescale x, and p~ to satisfy
the commutation relations (5.17). All of the F-dependence could be put instead in
Pw Then instead of (5.16) we would have
oo
X+=x++ P- ia(o - ~r)] + i E [ a . + . - b.%_.] (6.2)
n=l

and again (5.17) is satisfied. Now the momentum integration in (6.1) is the usual
fdDp/(2rr) D. However, L 0 now depends on F. Instead of eqs. (5.18) and (5.19), we
get

)/'+ -1-X~_= P+ + ~ ~/n ( a,,e-i"°'+-") + b~ei"('~±") )

n~l

and
oo
P~+P- + Y'. n(a*.a, + bt.b.),
L° =v[l+ Or2 n ~ l

respectively. Therefore the same overall factor det(1 + F ) emerges when we perform
the momentum integration, via the jacobian of the transformation p ~
p~,(1 + a2) -1/2.
In conclusion, we find that in the case of the neutral string, the F-dependence of
the vacuum amplitude may be written

Z F = det(1 + F)Z. (6.3)

Note that this result agrees with the result of [7], up to finite terms. We believe that
the result of [7] corresponds instead to the case of equal, rather than opposite,
618 A. Abouelsaood et al. / Open strings

charges on the boundaries, and that a path integral computation of Z F in the case
of the neutral string would agree with (6.3). This has been checked by an indepen-
dent calculation which will be discussed in more detail in a subsequent paper [24].
It is tempting to interpret the cosmological constant Z F on the annulus as a string
loop correction to the effective action [7]. However, Z F is plagued with infinities,
since Z has the usual infinities associated with the tadpole divergence [21]. Actually,
at the one-loop level, we cannot really talk of open strings alone, since closed string
poles automatically appear in the scattering amplitudes associated with the annulus.
The inclusion of closed string modes may be needed to obtain a finite or renormaliz-
able theory, and a more complete approach should include closed string back-
grounds, such as the metric, from the start. We shall pursue this in a later paper [24].

7. Conclusions
In this paper, we have developed methods for doing exact calculations of open
string physics in a slowly-varying background field-strength. In this approximation,
we have done the generic string tree and one-loop calculations. The string tree
calculation was used to determine the beta function, and hence the equation of
motion, for the background gauge field, as well as an effective action from which
this equation of motion can be derived. By a straightforward extension of these
techniques, it would be possible to obtain corrections arising from the inclusion of
higher derivatives of the background gauge field strength.
The string-loop calculation, aside from being interesting in its own right, should
be thought of as a technical warmup for the problem of finding the gauge field
corrections to the equations of motion for the closed string background fields (in
particular, the metric). The point is that the partition function on the annulus has
background gauge field divergences coming from the integration over the modular
parameter. If we take the attitude of Fischler and Susskind [25] and renormalize
them by appropriate world-sheet sigma model counterterms, we will obtain gauge-
field dependent string loop contributions to the beta functions of the closed string
background fields. In particular, we should recover the right gauge field contribu-
tions to the metric beta function (or, equivalently, the gauge field contribution to
the matter energy-momentum tensor). The details of such calculations, which will
make heavy use of methods described here, will be reported in a subsequent
paper [24].

Appendix A

ALTERNATE DERIVATIONOF THE BETA FUNCTION

In this appendix, we derive the beta function for F~,~, starting from the free open
string propagator (2.8) and using the quadratic and cubic interaction vertices in
 A. Abouelsaoodet al. / Openstrings 619

(2.5) to form all possible divergent one-loop graphs proportional to 8~X ~. The
results are the same as those found in sect. 2. We set 2era' = 1 in this appendix.
The Fourier transform of the free propagator in the upper half-plane is [26]

G(z, z,)-- f dP2er

~eiP(~-T')2lp(e-lplIo-o'1+ e-lpl(o+o'))
Since the interactions are all at the boundary, we need this propagator only for
o = a ' = O, i . e .

f ~ eiP(r-'r')
(A.O
a(~, ~')= IPl
Notice that

G-l('r, ,r')- 8~.8,.,G('r, "r')= f-~}ple 'p°-''' (A.2)

is the inverse boundary propagator. In fact,

(A.3)

The one-loop divergence is obtained by summing the infinite series of graphs in

fig. 5. We are summing all tadpole graphs with any number of insertions of the
vertex ~tF~,~ 8 ~ ~. Each graph with n such insertions represents a sum of 2 n terms
with all possible orderings of the derivatives ( ~ O,~" or ( a , ~ ) ~ ~ at each vertex).
These 2 n terms are a l l equal, as is easily checked using the antisymmetry of F~, and
integration by parts. We therefore restrict attention to the ordering shown in fig. 5
and include a combinatorial factor 2 n. The contribution from the graph with n
insertions is

I,,=2 (5i)
n 1 n+l
fd~d~l .d%(vF,,,,O,X )F ,~
, . v --p. v1 . .
.F'.,

x c(~-, ~-,)a,,c(T,, ~-~)... o,fi (~-o, O.

+ ~ ~ a ~ + ¢ ~ a ~ T¢+
o-=0 " ' "

VFaT~ VFarX F ~Far X F F

Fig. 5. One-loop graphs for the gauge field beta function.
620 A. Abouelsaood et al. / Open strings

In the approximation of slowly-varying fields, the background fields F ( X ) all can

be evaluated at the point ~', as is usual in background field calculations [14]. Then
we can integrate over 0"a .... Tn, using the relation

f d0-k~,,_ G(z,_ 1, vk)3,G(~',, z,+1) = -- ~("fk+l - - ~/'k-1),

valid for n - 1 >t k >1 2, which may be derived by integrating by parts and applying
(A.2) and (A.3). A similar relation applies to 0"n, with z replacing the variable zk+ 1.
For even n, integrating over ~2. . . . . % just leaves a factor of ( - 1 ) ~ / 2 8 ( r - T1),
and therefore the integral reduces to

(A.4)

where

a(~-,~-)= ~
l fdp=
}Pl -21nA

is logarithmically divergent in the ultraviolet cutoff A of sect. 2. For odd n, we find

instead

zn = f d'r d'rl ( V'Fh~ oq,S/~)( F n ) ~vG('l', "rl)oqriG ( 7-1, "r),

which is proportional to an odd momentum integral f d p p / ( 2 ~ r l p ] 2 ) , and hence

vanishes identically. Thus we obtain the total one-loop counterterm

~S,= ~ In=½ifdza~V'F~x(1 - e 2 -1 a ( , , , ) . (A.S)

n=0

Therefore, the beta function is

f l ~ = v.F~,a(1 _ F 2 )ku
-1

as found in sect. 2.

Appendix B

STRINGS IN A MAGNETIC FIELD ON A TORUS

We shall consider what happens when x 1 and x 2 are compact, with length L. In
this appendix, we will rename x 1 and x : to be x and y. Since only the zero modes
are important in the analysis below, we will find a close analogy to the situation
 A. A bouelsaoodet aL / Openstrings 621

Y (or x a)
L
Region I
(2

Region XE
:-- X (or Xl)
o L
Fig. 6. The torus is represented as a square with opposite sides of length L identified. It is divided into
two regions I and II by the circles y = 0 and y = a.

when a charged particle moves in a magnetic field on a toms. To clarify the analogy,
it will be convenient to use the notations q = ql + q2, and B - f , since these
quantities will be analogous to the charge of the particle and the magnetic field,
respectively.
A constant magnetic field on a torus is actually a monopole field; no smooth
single-valued vector potential A exists in this case. We may separate the torus into
two regions I and II (see fig. 6), and place in the two regions smooth potentials that
are related b y a gauge transformation at their intersection, which consists of the
circles y = 0 and y = a. Specifically, we may choose

-By 0~<y<a,
AI= -B(y-L) a<.y<L,

A2=0. (B.1)

In real coordinates, the zero-mode operators x and y satisfy the commutation

relation

¢ri ~i
[x, y l = a + ~ qB' (B.2)

so that k - qBy/rr may be taken to be a conjugate m o m e n t u m operator for x. States

are created b y applying the a*~ and b~ to the " m o m e n t u m " eigenstates Ik ) . A gauge
transformation A ~ A - XJ~ has the effect of shifting the zero mode y to y + X,
while x is unchanged. Thus k ~ k + (1/~r)qB X, and it follows that the zero-mode
wave function ( x l k ) = e i~x is multiplied by a phase factor e tqBxx/~r. In particular, it
follows from (B.1) that the transition function from region I to region II is e iqBLx/*r.
622 A. Abouelsaood et al. / Open strings

The transition function e iqBLx/~r is single-valued only if the Dirac quantization

condition
1
--qBL 2 = 2~rn (B.3)

is satisfied for some integer n. Since the torus is not divided in the x-direction, the
zero-mode wave function (xlk) = e ~ must be single-valued. This implies that the
k eigenvalues are quantized, k,, = 27rm/L. Therefore, using the Dirac quantization
condition (B.3), we find that y can take only the values

m
Ym= - - L . (B.4)
n

Since y + L is the same as y, it follows that there are only n independent wave
functions (xlkm), and hence the degeneracy of any state is n. This is as should be
expected, since the degeneracy of a Landau level for a particle of charge q in a
uniform magnetic field B is also n, with n given by (B.3). We conclude this
appendix by demonstrating this fact.
Assume the gauge potential in the xy plane is A = -Byk. If the plane were
infinite, the eigenstates of the hamiltonian would have the form

~p,.,, (x, y ) = e-ikmx~pt(y -Ym), (B.5)

where ~t is the lth wave function of a linear harmonic oscillator. The energy
eigenvalues Et. k = qB(l + ½), are independent of k. However, on the torus of fig. 6,
the wave functions must be single-valued as functions of y, and should therefore be
related in the two regions by the appropriate gauge transformations on the y = 0
and y = a circles. We shall try to satisfy this condition by taking linear combina-
tions of the infinite-plane wave functions with the same value of l, but different
values of k, so they will still be eigenstates of the hamiltonian.
We begin by writing a wave function that is properly matched on the boundary of
the regions,

[ ~ Cme-i~"~xfo(y--ym), O<~y<b
~bt(x, y ) = { m=-oo ~ (B.6)
e iqBLx/~ ~_, cme-ikmxf~,(y-ym), a<~y<L.
Iq' l ~ - - 0 0

We now impose the single-valuedness conditions

~,(x, O) = ~,(x, L ) , ~,(0, y) = ~ , ( L , y ) . (B.7)

The first of these gives k,, = 2~rm/L for some integer m. Substituting (B.6) into the
 A. Abouelsaood et aL / Open strings 623

second condition gives

-2~2m ).
e iqBLx/~r cme-2~rimx/Lf2l L qBL = E Cme-2~rimx/Lt~l
m= --c~ m= --oo qBL

(B.8)

The Dirac quantization condition (B.3) can be written as qBL = 2~r2n/L, and hence
eq. (B.8) can be rewritten

~. cme 2"`<"-m)x/L~`
((m))
1- L = ~_. e-2"rimx/L~, _ n L
(m) . (B.9)
m= -- oo m= -oo

N o w we substitute m ~ m + n in the 1.h.s. of (B.9) to obtain the result

Cm+ ne - 2~rimx/L lff l ( - -~ t ) = ~ Cme- 2~rimx/L l~l ( -- -~ t )

~1~ - oo m ~ - oo

from which it follows that c,,+, = c m. There are precisely n linearly independent
solutions of this last condition, and hence n independent wave functions at the lth
Landau level. We may choose them to be
oo
fft,j(x, y ) = const E e-2~ri(j+mn)x/L ~,(Y - Yj+mn),

where j = 0,1 . . . . . n - 1.

References
[1] C. Lovelace, Phys. Lett. 135B (1984) 75
[2] P. Candelas, G.T. Horowitz, A. Strominger and E. Witten, Nucl. Phys. B258 (1985) 46
[3] E.S. Fradkin and A.A. Tseythn, Phys. Lett. 158B (1985) 316; Phys. Lett. 160B (1985) 69
[4] C.G. CaUan, D. Friedan, E.J. Martinec and M.J. Perry, Nucl. Phys. B262 (1985) 593
[5] A. Sen, Phys. Rev. Lett. 55 (1985) 1846
[6] C.G. Callan, I. Klebanov and M.J. Perry, Nucl. Phys. B278 (1986) 78
[7] E.S. Fradkin and A.A. Tseytlin, Phys. Lett. 163B (1985) 123
[8] C. Lovelace, Nucl. Phys. B273 (1986) 413
[9] A.A. Tseytlin, Phys. Lett. 176B (1986) 92
[10] B.E. Fridling and A. Jevicki, Phys. Lett. 174B (1986) 75
[11] D.J. Gross and E. Witten, Nucl. Phys. B277 (1986) 1
[12] M. Born and L. Infeld, Proc. Royal Soc. 144 (1934) 425
[13] J. Scherk and J.H. Schwarz, Nucl. Phys. B81 (1974) 118;
E.H. Nielsen and P. Olesen, Nucl. Phys. B57 (1973) 367;
H.C. Tze, Nuovo Cim. 22A (1974) 507
[14] L. Alvarez-Gaumd, D.Z. Freedman and S. Mukhi, Ann. of Phys. 134 (1981) 85
 624 A. A bouelsaood et al. / Open strings

[15] A.A. Tseytlin, Nucl. Phys. B276 (1986) 391;

D.J. Gross, in Unified string theories, eds. M. Green and D.J. Gross (World Scientific, 1986), p. 391
[16] H. Dorn and H.-J. Otto, DESY preprint 86-064 (June 1986)
[17] C.S. Hsue, B. Sakita and M.A. Virasoro, Phys. Rev. D2 (1970) 2857
[18] D. Friedan, Proc. 1982 Les Houches Summer School, Recent advances in field theory and statistical
mechanics, eds. J.-B. Zuber and R. Stora (North-Holland, 1984)
[19] O. Alvarez, Nucl. Phys. B216 (1983) 125
[20] E.J. Martinec, Princeton preprint
[21] J.H. Schwarz, Phys. Reports 89 (1982) 224
[22] J. Scherk, Rev. Mod. Phys. 47 (1975) 123
[23] H.B. Nielsen and P. Olesen, Nucl. Phys. B144 (1978) 376
[24] C. G. Callan, C. Lovelace, C.R. Nappi, and S.A. Yost, Nucl. Phys. B., to be published
[25] W. Fischler and L. Susskind, Phys. Lett. 171B (1986) 383; 173B (1986) 262
[26] R. Balian and C. Bloch, Ann. Phys. 64 (1971) 271

View publication stats