Abstract
The issue of holographic mapping between bulk and boundary in the planewave limit of
AdS/SYM correspondence is reexamined from the viewpoint of correlation functions. We first study
the limit of large angular momentum for the socalled GKPW relation in supergravity approximation, connecting directly the effective action in the bulk and the generating functional of correlation
functions on the boundary. The spacetime tunneling picture which has been proposed in our previous works naturally emerges. This gives not only a justification of our previous proposal, with some
important refinements, on the mapping between bulk effective interaction and the CFT coefficients
on the boundary in the planewave limit, but also implies various insights on the interpretation of
holography in the planewave limit. Based on this result, we construct a new holographic string
field theory. We confirm for several nontrivial examples that this gives the CFT coefficients derived
by perturbation theory on the gaugetheory side. Our results are useful for understanding how apparently different duality maps proposed from different standpoints are consistent with each other and
with our definite spacetime picture for the AdS holography in the planewave limit.
2005 Elsevier B.V. All rights reserved.
PACS: 11.25.w; 04.60.m
1. Introduction
An impressive amount of computations have been done following the BMN conjecture [1] as to the identification of stringy operators in AdS/CFT correspondence. In spite
of all those important works of two years, however, it seems that the question of holographic correspondence of correlation functions for the BMN operators still has not been
appropriately understood. In the case of the original AdS5 /SYM4 correspondence, the relation between the bulk fields {i (z, x)} and the gaugeinvariant operators {Oi (
x )} of 4D
YangMills theory has been concretely formulated as the famous GKPW relation [2]
Z[] exp
i (
x )Oi (
x) ,
d 4x
i
However, our original argument in [4] has not yet been completely satisfactory, since it
involves some ambiguities with respect to normalization and shortdistance cutoff when we
discuss 3point and higher correlation functions. One of the purposes of the present work
is to reexamine our basic ideas from an equivalent but a more systematic standpoint by
studying the planewave limit directly on the GKPW relation, and to strengthen our picture
by presenting further supports and extensions. In our first work [4], we have not started
from the GKPW relation. Instead, we have treated the bulk field equation in the WKB
approximation and then proposed a natural ansatz for relating correlation functions and
the Euclidean Smatrix in the same spirit as the GKPW relation. In the present work, we
study the limit of large Rcharge (J ) for the GKPW relation directly and confirm that our
original picture emerges automatically within the supergravity approximation. Then, on the
basis of the known relation [11] for chiral operators between supergravity and SYM gauge
theory, we establish a definite relation between 3point correlators and the bulk effective
interactions which are consistent with our original picture, including precise normalization
and cutoff.
It turns out that the resultant supergravity effective theory in the planewave limit cannot
be obtained from any versions of previously known string field theories. Our result can
be adopted as a strong constraint in constructing a string field theory describing higher
stringy modes in accordance with our picture. We propose a new holographic string field
theory, which reduces to the derived effective action when restricted in the supergravity
sector, and simultaneously gives the correct 3point correlation functions on the SYM side
derived by perturbation theory, via our holographic mapping. As a byproduct, we also give
some clarifications on the relation of our picture with other approaches which have been
discussed in some of recent works for mapping the known versions of string field theories
to gauge theory. We believe that our results not only resolve holography in the planewave
limit, but also lay a foundation for understanding relation among different proposals and
for investigating further extensions, on the basis of holography for correlation functions. In
particular, we give a definite prediction for impurity nonpreserving cases which have not
been treated in the literature before.
The present paper is organized as follows. In Section 2, we analyze the large J limit
directly on the usual diagrammatic rules (Witten diagrams) for computing 3point functions from the viewpoint of bulk theory. We demonstrate how our tunneling picture arises
in the WKB approximation and establish the validity of this picture by comparing with
exact computations. In Section 3, we derive directly the effective action in the planewave
limit from the bulk supergravity effective action given previously in Ref. [11]. We then
formulate the holographic relation which should be valid for (nonBPS) stringy operators.
In Section 4, we construct the holographic string field theory which is consistent with
our effective action for supergravity sector. In Section 5, we clarify the relation of our results with other approaches. In Section 6, we confirm our general discussion by explicit
examples. We conclude the paper in Section 7 by giving further remarks. Appendices A
and B are devoted to a detail of calculation and to a summary of the properties of stringinteraction vertices of string field theory, respectively.
R2 2
(2.2)
dz + d x2
2
z
using the Poincar coordinate. According to the GKPW dictionary, a 3point correlation
xi ) (conformal dimension = i = Ji + ki = ki ) correspondfunction of three operators Oi (
ing to the bulk fields i is given, up to an overall normalization factor and to the lowest
order with respect to the coupling constant g, as
O1 (
x1 )O2 (
x2 )O3 (
x3 )
4
d x dz
=
(2.3)
K1 (z, x; x1 )K2 (z, x; x2 )K3 (z, x; x3 ),
z5
ds 2 =
2
z3
+ z2 2 m2 K (z, x; y) = 0,
z5
z
z
x
(2.4)
for z > 0
(2.5)
and
lim z4 K (z, x; y) = (
x y).
z0
(2.6)
The PPwave limit amounts to taking the limit where Ji , R with Ji /R 2 being
kept fixed and ki O(1). The angular momentum which comes from an SO(2) part of
the SO(6) Rsymmetry must be assumed to be conserved, say, J1 = J2 + J3 . Obviously,
since i Ji , the integral in the expression (2.3) can be studied by saddlepoint
methods.2 For our purpose it is useful to do a warmup in the case of twopoint functions.
In what follows until stated otherwise explicitly, it is convenient to adopt the unit such that
the AdS5 (S 5 ) radius to be one, R = 1, since R is the only length scale characterizing this
system in the supergravity limit 0 with fixed R.
2 A preliminary discussion on the approach of the present paper has been given by one of the present authors
in a talk at the Strings 2003 conference [12].
+ ln 2
= 0,
ln 2
z
z + (
x x1 )2
z + (
x x2 )2
z
z
+
ln
= 0.
ln 2
x
z + (
x x1 )2
z2 + (
x x2 )2
(2.7)
(2.8)
(2.9)
(2.11)
x = x0 ( ) + x,
(2.12)
1
2
(dz0 + d z )2 + (d x0 + d x)
2
z
4 cosh2
cosh2 (d z )2 + (d x )2 ,
(d )2 +
2

x1 x2 
(2.14)
n x = 0,
(2.15)
with n being the unit vector along the direction of the vector x1 x2 , which connects two
points x1 , x2 on the boundary. Thus, the Jacobian is given as
4 cosh2 4
J ( ) =
(2.16)
cosh2 .

x1 x2 2
On the other hand, the effective secondorder action for the saddlepoint integral is
(2)
Seff =
4
cosh4 z 2 + cosh2 (x )2 + dependent factor.
2

x1 x2 
(2.17)
The classical part of the action gives only a factor which is independent of the collective
coordinate and has the correct dependence on the distance of boundary points, as can be
checked by using
K z0 ( ), x0 ( ); x1,2 =
()

x1 x2  e ,
2 ( 2)
(2.18)
where the sign on the exponentials depends on the points x1 or x2 , respectively. It is now
evident that the Jacobian factor is canceled by the integrations over the fluctuating coordinates, up to a dependent proportionality constant. Consequently, the twopoint function
in the large limit is given simply by
2
G2 (
x1 , x2 ) 2 
x1 x2 2()
+
d (2 cosh ) ,
(2.19)
where the dependent contributions come from the factor z in the defining expression
(2.7). Thus, in the limit 0+, we reproduce the correct behavior for twopoint correlators for conformal operators, up to the pole singularity
+
2
d (2 cosh ) .
(2.20)
22 1

x1 x2 2 .
2
(2.21)
Here we have used the general formula for this type of integral [9],
dz
0
d D x
za
(z2
+ (
x x1
)2 )b (z2
+ (
x x2 )2 )c
= 
x1 x2 1+a+D2b2c I (a, b, c, D),
I (a, b, c, D)
D/2
a
2
1
2
b + c D2 a2 12 12 + a2 + D2 b 12 +
(b)(c)(1 + a + D b c)
a
2
D
2
(2.22)
In reality, for the effective theory described by (2.1), it is more appropriate to consider
4
d x dz
g (z, x) K (z, x; x1 ) K (z, x; x2 ) + m2 K (z, x; x1 )K (z, x; x2 )
5
z
(2.23)
than (2.7). It is easy to repeat the above calculation for this case. Only difference of the
final result from the case of G2 is the multiplication of (m2 2 )/m2 4/.
2.2. 3point functions
Armed by this exercise, we now go back to the 3point function (2.3). The saddlepoint
equations are
3
i=1
3
i=1
1
2z
2
z z + (
xi x)2
= 0,
xi x
= 0.
z2 + (
xi x)2
(2.24)
(2.25)
It is easy to convince oneself that, for generic configurations of three boundary points, there
is no solution to these equations. However, if we take a limit where two of the boundary
points, say, x2 and x3 approach sufficiently to one point xc = (
x2 + x3 )/2, the same trajectory connecting x1 and xc as we have discussed in the previous subsection can be regarded
as an approximate solution. Therefore, let us try to reduce the integral to the one along the
following trajectory,
1
1

x1 xc 
x1 + xc ) (
x1 xc ) tanh ,
,
z0 ( ) =
x0 ( ) = (
(2.26)
2
2
2 cosh
with the fluctuations, x = x x0 and z = z z0 . For our purpose it is sufficient to evaluate
the integral to the leading order in the shortdistance limit,
x3 + xc 0.
x2 xc =
In order to avoid unnecessary complications, we assume that all three points x1 , x2 , x3 are
along a single line on the boundary.
The effective action for this computation
Seff =
3
i=1
is rewritten as
i ln
xi x)2
z2 + (
z
(2.27)
10
2
z + (
z2 + (
1 + 2 + 3
x1 x)2
xc x)2
ln
+ ln
2
z
z
2
2
2
z + (
z + (
x1 x)
xc x)2
1 2 3
ln
ln
+
2
z
z
2 (
xc x) + 2
2 (
xc x) + 2
+
.
ln
1
+
+ 2 ln 1 + 2
3
z + (
xc x)2
z2 + (
xc x)2
(2.28)
The first line of (2.28) can be treated in exactly the same way as for the 2point case, by
replacing by (1 + 2 + 3 )/2 ( J1 ). Next, since
1 2 3 O(1),
(2.29)
the second line can be approximated by its value on the classical trajectory
2
z + (
x1 x0 )2
x2 x0 )2
z2 + (
1 2 3
= (1 2 3 ).
ln 0
ln 0
2
z0
z0
(2.30)
The third line gives, to the zeroth order with respect to the fluctuations,
2
2
(2 + 3 )
2(2 3 )
+
(
+
)
e2
2
3

x1 xc 

x1 xc 2

x1 xc 2
+ O 3 .
(2.31)
Here we have kept the third term which is of second order in 0, since it shows that
for sufficiently large there is a natural cutoff for the range of the affine parameter for
0. The other terms which are independent of can be ignored in
arbitrary small = 
the limit 0.
As for the contribution of fluctuating coordinates in the third line, it is easy to convince
ourselves that keeping only the first order term with respect both to the fluctuations and to
the shortdistance cutoff is sufficient for our purpose. The relevant terms are arranged as
xc x0 )) 1
x
(
xc x )(z z0 x (
4(2 3 )
+
2 z02 + (
(z02 + (
xc x0 )2 )2
xc x0 )2
4(2 3 )
e cosh2 ,
= z
(2.32)

x1 xc 2
which involves only the z fluctuation. Combining this with the relevant part of the Gaussian
factor coming from the first line, the integral with respect to z is
4
4(2 3 )
2 4J1 cosh
2
d z [measure] exp z
z
e cosh

x1 x2 2

x1 xc 2
2
4
(2 3 )2 e2
2 4J1 cosh
d
z
[measure]
exp
z
= exp
(2.33)
J1 
x1 x2 2

x1 x2 2
to the present order of approximation. The prefactor here implies that, together with the
contribution from the 3rd term of (2.31), the total factor which is responsible for the cutoff
at large region is
2
2 (2 + 3 )
2
2 (2 3 )
exp e2
+
e
J1

x1 x2 2

x1 x2 2
2
J2 J3 2
exp 4
e .
2

x1 x2  J1
11
(2.34)
Putting all of these results together and remembering that the Jacobian and the Gaussian
integral cancel, the integral of the 3point function now takes the form
2

x1 xc (1 +2 +3 )
J12
+
J2 J3 (2)2
(1 2 3 )
2
(2.35)
d e
exp
e
J1 
x1 xc 2
in the limit of large Ji and small . This expression shows that the precise form of the
cutoff mentioned above is c with
J2 J3
2
c
.
e
(2.36)
=
J1 
x1 xc 
Thus, the 3point integral can be expressed by a Gamma function and leads to
(2 +3 1 )/2 2 +3 1
+1
2
21
(2 +3 1 ) J2 J3
2

x1 xc 
(2)
.
J1
1 + 2 + 3
J12
(2.37)
After fixing the convention for normalization suitably, this give the correct shortdistance
limit for the 3point correlator, from which we can identify the CFT coefficient by equating
the expression with
lim
0
C123
C123
,
21

x1 x2 23 
x2 x3 21 
x3 x1 22

x1 xc 21 2
(2.38)
i
3
z
d 4 x dz
z5
z2 + (
x xi )2
i=1
c(1 , 2 , 3 )
,
x3 2 +3 1 
x3 x1 3 +1 2

x1 x2 1 +2 3 
x2
c(1 , 2 , 3 )
2 12 (1 + 2 3 ) 12 (2 + 3 1 ) 12 (3 + 1 2 )
=
2
(1 )(2 )(3 )
1
(1 + 2 + 3 4) .
2
(2.39)
(2.40)
12
ln J J ,
lim (J + k) 2 exp J + k
J
2
we obtain
(2 +3 1 )/2 2 +3 1
+1
2
21
(2 +3 1 ) J2 J3
2

x1 xc 
(2)
,
J1
1 + 2 + 3
J12
(2.41)
which exactly matches (2.37).
Thus we have confirmed that the 3point correlators in the short distance limit 0
can be computed effectively as a process in the bulk occurring along a single tunneling
trajectory connecting boundary to boundary, whose amplitude essentially takes the general
form,
1 2 3
g,
1 + 2 + 3
1 2 3
2
1 + 2 + 3
J2 J3
+1 ,
g g
J1
2
(2.42)
= 2,
(2.43)
with a suitable normalization convention, apart from the usual spacetimedependent factor

x1 xc 21 .
The form (2.42) is consistent with what we have proposed as the general structure
for 3point correlators which is expected from the spacetime picture for holography
in the PPwave limit. The parameter ec is the cutoff related to the distance of
x2 ) and O3 (
x3 ) at the boundary, as (2.36). In our first original disthe operators O2 (
cussion,
the
emergence
of
this
particular form is originated from the formal integral
d exp[(1 2 3 ) ] which appears in perturbation theory along the tunneling trajectory. This is actually illdefined as it stands. To justify the expression (2.42), we had to
invoke a wavepacket picture which is rather subtle in the Euclidean tunneling as is alluded
to in [4]. According to the present argument, the illdefined integral must be replaced by
(2.35) which can be defined unambiguously by analytic continuation, due to the presence
of the natural cutoff for large region. Intuitively, the large cutoff corresponds to the
limitation of the picture using the single tunneling trajectory in discussing 3point correlators for small but nonzero . Note that if we take the limit 0 naively inside the integral,
this factor would have simply disappeared. In fact, the additional factor
J2 J3 (1 2 3 )/2
1 + 2 + 3
+1
J1
2
was missing in the previous discussions, owing to this ambiguity.
It is known that the coupling constants of supergravity modes in the bulk vanish for the
socalled extremal case with 1 2 3 = 0. However, as is well known, the extremal
correlation functions themselves do not vanish [10]. This remarkable fact corresponds to
the presence of the singular denominator in (2.42). We propose that the holographic correspondence relating the bulk 3point couplings and the CFT coefficients should obey the
13
above general relation. Of course, the computation of this section is restricted to the supergravity approximation. Below, we will argue that this should generalize to nonBPS
stringy modes too, with slight corrections. Actually, if we only consider a restricted set of
the correlators in which the numbers of impurities are conserved, as has been assumed
in the literature, the correction factor can be ignored in the large limit to the leading
order in making comparison with perturbative computation on the YangMills side, since
1 2 3 O(1/2 ) for such cases.
O I = CiI1 i2 ...ik Tr i1 i2 ik .
(3.1)
Here, C is a totally symmetric traceless tensor whose contraction is normalized as
I I
C 1 C 2 = CiI11i2 ...ik C I2 ,i1 i2 ...ik = I1 I2 ,
(3.2)
(3.3)
Since the BMN operators in the supergravity sector are essentially contained in this set
of local operators, it should be possible to derive the effective action along the tunneling
trajectory starting from an appropriate effective action for these operators. The BMN supergravity modes are a subset of (3.1), expressed by using the complex basis for the SO(2)
directions i = 5, 6 with Z = 1 (5 + i6 ):
2
k = J + k,
J , (3.4)
where CiI1 i2 ...i are now completely symmetrized traceless SO(4) tensors with unit normalk
ization (similarly as (3.2)) under contraction. The normalization constant is related to that
of the original set as
1/2
k
=
(3.5)
.
J
14
Note that in the large J limit the operators with vector excitations Di Z are regarded
essentially as the derivatives of (3.4), and hence are included in this set defined by (3.4) if
we suitably take into account the variation of the boundary coordinates x . In fact, as we
will argue later on we have to take a special care in the interpretation of vector excitations.
The bulk effective action for the SO(6) operators has been derived in Ref. [11]:
1
4N 2
1
5
d x g
(I )2 + k(k 4)I2
Sbulk =
5
2
2
(2)
I
GI I I
1
1 2 3 I1 I2 I3 ,
(3.6)
3
AI1 AI2 AI3
I ,I ,I
1
where we used Euclidean metric and assumed a particular normalization for the scalar
fields:
AI = 26k 3
k(k 1)
.
(k + 1)2
(3.7)
This is the effective action in the sense that it reproduces correlation functions through the
Witten diagrams as studied in the previous section. Therefore, the freedom of field redefinition on the bulk side and correspondingly the choice of the basis of chiral operators on the
boundary side are both fixed already. Remarkably, the set of bulk fields {I } corresponding
to the set {O I } are effectively treated as scalar fields propagating in the AdS5 background
without derivative coupling. The derivative interactions are removed by making a particular field redefinition in the derivation of this action.3 Actually, the effective action (3.6)
was obtained from the equation of motion of IIB supergravity. It was argued that the result
must be correct including the normalization of the CFT coefficients by relating them to the
Rsymmetry currents which have been known to be given exactly by the free gauge theory.
We refer the reader to [11] and the references therein for more details on this effective action. Also for a summary on the nonrenormalization properties of 2 and 3point functions
of chiral operators, see, e.g., [13] and the references therein.
The CFT coefficients C I1 I2 I3 of the above operators defined by
x1 )O I2 (
x2 )O I3 (
x3 ) =
O I1 (
C I1 I2 I3

x12 23 
x23 21 
x31 22
(3.8)
with
2 + 3 1
, etc.,
2
is related to the 3point interaction in the effective action (3.6) by
k1 k2 k3 I 1 I 2 I 3
I1 I2 I3
C C C ,
=
C
N
I I I
(k1 + 1)(k2 + 1)(k3 + 1)
GI1 I2 I3
C 1C 2C 3 =
,
7
a(k1 , k2 , k3 ) 2 ((/2)2 1)((/2)2 4)1 2 3
1 =
3 This does not mean that the interactions containing derivatives along the S 5 directions are removed.
(3.9)
(3.10)
(3.11)
= k1 + k 2 + k 3 ,
a(k1 , k2 , k3 ) =
2
15
(3.12)
k1 !k2 !k3 !
3
.
+ 2 !2(2)/2 1 !2 !3 !
(3.13)
The first expression in this list is nothing but the result of free field computation of correlators on the YangMills side, with C I1 C I2 C I3
being all the possible contractions among the
SO(6) indices of Ctensors. The second expression is its representation in terms of the bulk
quantities, in which the factor a(k1 , k2 , k3 ) arises as the coefficient relating C I1 C I2 C I3
to the overlap integral of S 5 harmonics Y I , which appears in reducing 10D theory to 5D
effective theory on AdS5 backgrounds, as
(3.14)
Y I1 Y I2 Y I3 = a(k1 , k2 , k3 ) C I1 C I2 C I3 .
S5
The definitions of the S 5 harmonics and their integrals are summarized in Appendix A, to
which we refer the reader for more details. The fact that the supergravity effective action
is exactly matched to the freefield results on the gaugetheory side is interpreted as a
consequence of nonrenormalization properties of 3point functions of chiral operators. This
is quite remarkable since the supergravity limit 0 is nothing but a strongcoupling
2 N = R 4 /( )2 on the gaugetheory side.
limit gYM
Before going to our main task of this section, let us here check that the relation between
the above effective action and the CFT coefficients is consistent with the prediction of
previous section. Taking the limit J1 (= J2 + J3 ), J2 , J3 , we find
C I1 I2 I3 =
1 2J1 + 2 9
N
3
J1 J2 J3 J1 1 1 !
GI I I .
J2 J3
1 1 2 3
J12
(3.15)
J2
J3
C 1 C 2 C 3
GI1 I2 I3 = 1 3 2J1 2 +9 J12
(3.17)
J1
J1
1 ! 2 ! 3 !
in terms of the SO(4) contractions.
16
z = z0 ( ) + z ,
(3.18)
with
x =
nz sinh + x .
(3.19)
Since the fluctuations around the trajectory can be assumed to be of order 1/ J from the
discussion of the previous section, we derive the effective metric which is correct to the
second order in the fluctuations. The result is
ds 2 = 1 + z 2 + x 2 d 2 + d z 2 + d x 2 + higherorder term
after rescaling as
z

x1 xc 
2
2 cosh
z ,

x1 xc 
x
x ,
2 cosh
sinh 2 2
z + x .
2 cosh
sinh 2
y .
2 cosh
ds 2 = 1 + y 2 d 2 + d y 2
(3.21)
(3.22)
has an SO(4) symmetry with respect to the rotation of y. We also note that, since the
order of magnitude of y is supposed to be constant in on the basis of this independent
metric, (3.20) implies that the original fluctuating coordinates x decrease as e  , when
we approach the boundary .
Using this metric, the quadratic terms of the bulk effective action (3.6) are given, to the
accuracy of second order in y, as
4N 2
1 2
4
I I + y I y I
y
y
1
d
2
(2)5
1
+ k 4) I I .
+ 1 + y 2 (J + k)(J
2
17
(3.23)
Note that we have changed the SO(6) index to the complex SO(2) SO(4) index.
The assumption that the magnitude of the fluctuating coordinates is of order 1/ J is
justified in the language of the effective action as follows. Redefine the fields by
(, y) = eJ 0 (
y )( ),
(3.24)
y) = e+J (J ) (
),
(,
y )(
0
(3.25)
1
0(J ) (
y ) = (J /)2 exp J y 2
2
(3.26)
(J )
with
being the ground state wave function for the kinetic operator for the fluctuating coordinates,
h = y2 + J 2 y 2 .
(3.27)
(3.28)
which requires the choice of signs in the definition (3.25) on the exponential. Here and in
what follows, we rewrite and z by and z without tilde again for notational brevity.
), the leading term of the free action is then of
In terms of the reduced field ( ), (
order O(J ), taking the familiar nonrelativistic form in 0 + 1 dimensions as
4N 2
+ 2k
.
(3.29)
d J
(2)5
are canceled, and the zeropoint energy of the
The order O(J 2 ) terms of the form J 2
operator (3.27) is responsible for cancelling a part of the order O(J ) term in the last term in
the Lagrangian density. Note that the term  2 is of order one and hence can be ignored
in the large J limit. Absorbing the normalization factor for the field, the free Hamiltonian
(2)5/2 1
,
2N
2J
to renormalize the quadratic terms in the standard form, the total effective action is
1
( I I I I ) + kI I I
d
2
I
1
+
(3.30)
I1 ,I2 ,I3 ( I1 I2 I3 + h.c.),
d
2
I1 ,I2 ,I3
18
1 8+J1 2/2
2
J1 J2 J3 GI1 ,I2 ,I3 ,
3N
J12
I1 ,I2 ,I3 =
(3.31)
with J1 = J2 + J3 . Comparing this result with the known relation (3.15) connecting the
CFT coefficients to the 3point coupling constant, we confirm that the relation is again
precisely the one discussed in (2.42) and (2.43). This establishes that the effective theory
along the tunneling trajectory for the BMN operators with only scalar excitations is (3.30).
Now it is very natural to extend this action by including the higher excited states
with respect to the Schrdinger operator (3.27). Since the fluctuating coordinates near the
boundary z0 0 ( ) are essentially the four vector x of the 4D base space up to an
exponential factor 
x1 xc e  , taking these excited states into account seems to corre of the BPSBMN operators. Here
spond to the inclusion of the vector excitations Di Z(Z)
we write down this extension and will give a precise discussion of this correspondence in
the next section. The decomposition (3.24) is generalized to the infinite expansion over the
complete set of excited states,
I (, y) =
(2)5/2 1 J (J )
n (
y )n ( ),
e
2N
2J
n
(3.32)
I (, y) =
(2)5/2 1 +J (J )
n (
y ) n ( ),
e
2N
2J
n
(3.33)
where
n(J ) (
y) =
4
1/4 ni /2
J
2
2
Hni J y eJ y /2
ni !
i=1
(3.34)
( I,n I,n I,n I,n ) + kI +
4
ni I,n I,n
(3.35)
i=1
I1 ,n(1) ,I2 ,n(2) ,I3 ,n(3)
1 8+J1 2/2
2
J1 J2 J3
3N
J12
(1) (2) (3)
GIn ,I ,n,I ,n I1 ,n(1) I2 ,n(2) I3 ,n(3) + h.c.
1
(3.36)
with
(1) (2) (3)
GIn ,I ,n,I ,n
1 2 3
J1
= GI1 ,I2 ,I3
J2 J3
(J )
(J )
(J )
y )n(2)2 (
y )n(3)3 (
y ).
d 4 y n(1)1 (
(3.37)
19
the present section, we do not treat the trace part for simplicity. They will be partly considered in Section 5 after
we construct the holographic string field theory.
20
, corresponding to x1 and x2 , respectively, at the boundary. The tensor Cij is usually associated with the conformal inversion x x/x2 . In our case, this is an automatic
consequence of the bulkboundary correspondence. However, (3.39) evidently leads to an
inconsistency with the orthogonality between scalar and vector excitations. In the literature
(see, e.g., [14]), a particular way out of this difficulty has been discussed without a definite
physical picture for holography. The suggested procedure recovers orthogonality only after taking the limit 
x1 x2  . Since our picture for holography must be valid for any
finite 
x1 x2 , it is not satisfactory from our standpoint.
From our viewpoint of a direct correspondence between bulk and boundary as clarified up to this point, the origin of this apparent discrepancy lies in an important difference
with respect to the SO(4) symmetry on both sides. For the bulk, we use the metric (3.22)
which characterizes the geometry close to the tunneling trajectory traversing from a fixed
boundary point to another fixed boundary point. In particular, the trajectory is asymptotically orthogonal to the boundary spacetime, and the fluctuations near the boundary have
an SO(4) symmetry under global rotations of y. The SO(4) rotations must be performed
simultaneously at all points along the trajectory, and consequently at both ends on the
boundary. It is important to recognize that this SO(4) symmetry cannot be identified with
the SO(4) isometry of the AdS metric with respect to x , as is indicated by the relation
(3.20). Since the latter isometry should correspond to the SO(4) symmetry of the boundary
theory, the rotations of y cannot be directly related to the usual SO(4) symmetry at the
boundary.
For the boundary theory, we are using the flat Euclidean metric along the boundary
space. The change of a distance caused by a small variation of points in the flat Euclidean
metric is not of second order with respect to the variations of coordinates
(
x1 + x1 x2 x2 )2
 x1 x2  2
x1 x2
2
+
= 
x1 x2  1 + 2 n
.

x1 x2 

x1 x2 
(3.41)
This is not SO(4) symmetric under the rotation of y. Note that the SO(4) transformations
of y must be done with fixed fiducial points, x1 and x2 , in such a way that the vicinities
around them are simultaneously rotated. Obviously, such an SO(4) symmetry cannot be
identified with the usual global SO(4) isometry of the original AdS metric. One might
wonder that this is then a selfcontradiction. But that is not so. As we have emphasized, the
magnitude of fluctuations around the trajectory vanishes exponentially as we approach the
boundary, and hence there is no contradiction. This is also consistent with the fact that the
tunneling trajectory is a classical solution under the restriction that the fluctuations satisfy
the Dirichlet boundary condition at z = 0. The Euclidean Smatrix, however, requires us
to blow up the vanishing fluctuations near the boundary at the vicinities of the fiducial
points by multiplying the powers of e  = eT in such a way that the SO(4) symmetry
with respect to y is kept. The process of blowingup amounts to introducing a particular
UV cutoff for the boundary theory.
Therefore, the naive identification of the vector excitations in the bulk with the derivations Di Z on the boundary is not precise. We have to check from the bulk point of view
how the small variations near the boundary are treated by deriving twopoint functions for
21
them following the approach of Section 2. Suppose that external lines are general vector
states corresponding to the operators (3.38) with two traceless symmetric tensors K L1 and
K L2 . This amounts to extracting the parts proportional to
KjL11j2 ...j y j1 y j2 y j1 ,
1
KjL 2j ...j y j1 y j2 y
j
2
1 2
2 J
j1 j2
x x
z 2
j
z
2
z + (
x x2 )2
(3.42)
J +k+
2
,
(3.43)
x i = ij 2ni nj x j .
(3.44)
Note that the use of redefined x i is required by the change of sign in the relation (3.20)
between the fluctuating coordinates with y in the limit = T (T ). Although the
4th components actually give the powers of  tanh  in converting from x/z (or x /z) to y,
they can be replaced by one in the leading singular part with respect to the regularization .
Then, the result of saddlepoint integral is
L1 L2
J2
2
(3.45)
(3.46)
1 2
(2J ) 1 !
1)

x1 x2 2(J +k+
L1 L2
2
2(J +k)

x12 g
(3.47)
with
i = (ij 2ni nj )j
(3.48)
22
being the derivatives with respect to x i . Here the subscript g for the distance 
x12  indicates
that the derivatives are computed by assuming that the small variations with respect to the
boundary positions are defined by
x1 x2
x1 x2 )2 1 2

x12 2 (
(3.49)

x1 x2 2
instead of the naive expression (3.41). Apart from the terms x12 , x22 which do not contribute owing to the traceless condition for the tensors K L1 and K L2 , this amounts to
dropping all terms which violates the SO(4) symmetry in (3.41). Strictly speaking, we
should have denoted the derivatives i , i by using different notations to indicate the situation. Effectively, our prescription is equivalent to dropping terms which violates the SO(4)
symmetry, and is consistent with the prescriptions adopted in other approaches. Practically, the following rule is valid: first we consider only the directions orthogonal to n to
obtain SO(3) symmetric answer. Then we extend the results formally to SO(4) symmetric
ones.
The variation indicated by (3.49) can be interpreted as the variation of the distances
measured using the planewave metric (3.22). The minimal distance defined by the distance
functional
2
2
d y
1 2
d y
2
(
y1 , y2 ; T ) = d 1 + y +
(3.50)
y +
d 1 +
d
2
d
with the boundary condition
y(T ) = y1 ,
y(T ) = y2
(3.51)
behaves as
y1 y2
1 2
y + y22 tanh 2T
2 1
sinh 2T
1
2T + y12 + y22 2
y1 y2 e2T + O e4T .
2
Thus the transition amplitude is
(
y1 , y2 ; T ) = 2T +
y1 ,
(J +k)(
y2 ;T )
2(J +k)T
x1 x2
12

x1 x2 2
(3.52)
J +k
(3.53)
with
x1
eT y1
,

x1 x2 
x2
eT y2
,

x1 x2 
(3.54)
which are required by the relation (3.20) for large   T . Note that we can ignore the
term 12 (
y12 + y22 ) in (3.52), which corresponds to the redefinition (3.21) and does not contribute owing to the traceless condition.
We can extend the above argument to 3point correlators with vector excitations. By an
obvious simplification of notations, we are lead to the integral
23
i=1 (2Ji )i K Li
2
j
j
dy 4 y j1 y j2 y j1 y j1 y j2 y 2 y j1 y j2 y 3 eJ1 y
2
4
J
y
1
dy e
(scalar integral with i Ji + ki + i )
= 2(1 +2 +3 )/2 J2 3 J3 2
C I1 I2 I3

x1c
(3.55)
where
1 = (2 + 3 1 )/2,
(3.56)
etc.
Taking into account the normalization factor, this leads to the CFT coefficient
J1 1 J2 2 /2 J3 3 /2
I1 I2 I3 ,L1 L2 L3
C
=
J2 J3
J1
J1
C
1 !2 !3 !
(1 1)!
(3.57)
From the viewpoint of boundary, this result can again be formulated as the consequence
of the replacements
x1 x2

x12 2 (
,
x1 xc )2 1 2

x1 xc 2
x1 x3
2
2

x13  (
(3.58)
,
x1 xc ) 1 2

x1 xc 2
2 (2)
2
x2 x3
12
22
(3.59)
1 2
2
1 2
3
directly to the general form (3.8) of the 3point correlator without vector excitations. The
prescriptions (3.58) are equivalent to (3.49) to the leading order in , while (3.59) is necessary to be consistent with the SO(4) symmetry.
Combining with the CFT coefficient of pure scalars, we obtain
1 2J1 + 2 9 J1 J2 J3 J1 1 +1 (1 + 1 )!
C I1 I2 I3 ,L1 L2 L3 =
GI1 I2 I3
N
J2 J3
1 + 1
3
J12
2 /2
3 /2
J3
1 !2 !3 ! L1 L2 L3
J2
K K K
(3.60)
J1
J1
1 !2 !3 !
in terms of the purescalar 3point coupling, or using the expression for the scalar vertex
(3.17)
24
I1 I2 I3 ,L1 L2 L3
J1 J2 J3 J1 1 +1
1
(1 + 1 )!
2
J
J
J1
2 3
1 + 1
(k2 +2 )/2
(k3 +3 )/2
J2
J3
J1
J1
1 ! 2 ! 3 !
1 !2 !3 !
1
=
N
(3.61)
in terms of the SO(4) tensors. On the other hand, the effective interaction (3.37) including
the vector excitations shows that the 3point coupling for the external lines in consideration
is given by replacing the coefficient GI1 I2 I3 for purely scalar excitations as
2 /2
3 /2
J2
J3
1 !2 !3 ! L1 L2 L3
L L L
K K K .
GI11I2 I23 3 = GI1 I2 I3
(3.62)
J1
J1
1 !2 !3 !
Comparing this with (3.60) taking into account the normalization factor in the 3point
interaction term of the effective action, we can conclude that the general form, as derived
in Section 2 for the holographic correspondence of scalar supergravity excitations, extends
to a more general case including the vector excitations,
L L L
I11I2 I23 3
L L L
CI11I2 I23 3 =
L L L
I11I2 I23 3 =
2 + 3 1
J2 J3
J1
(1 2 3 )/2
2 + 3 1
L L L
+ 1 I11I2 I23 3 ,
2
(3.63)
(3.64)
with i = ki + i = Ji + ki + i .
Supersymmetry demands that this should extend further to fermionic excitations. Since
we treat the full supersymmetric string field theory later, we do not discuss the supersymmetrization of our arguments explicitly. We only mention that to discuss orthogonality
including fermionic excitations, the spinor equivalent to the inversion tensor j1 j2 2nj1 nJ2
transforming spinor indices between two boundary points is the SO(4) gamma matrix
i i ni . In computing correlation functions, we have to extract SO(4) invariant pieces in similar manner as for bosonic excitations by taking into account the transformation of spinor
indices between two boundary points. This is effectively equivalent with the procedures
adopted in other works.
3.4. Operator representation and the corrections
On the basis of our results, we are now prepared to proceed to the construction of a full
holographic string field theory. Before that, it is convenient to first rewrite the foregoing
results using oscillator representation. By substituting the expression for the general 3point coupling, the effective action is written as
Seff = S2 + S3 ,
(3.65)
S2 =
d
I,L
S3 =
1
1
( I,L I,L I,L I,L ) + (kI + L ) I,L I,L ,
2
R
25
(3.66)
(3.67)
1 ! 2 ! 3 ! 1 !2 !3 !
where we have recovered the length dimension. The integrals over the angular momentum
J should be understood implicitly. We introduce the braket notation for the field by using
the 8dimensional oscillator algebra of ai , ai : ai 0
= 0, [ai , aj ] = ij (i = 1, 2, . . . , 8),
where the first 4 directions correspond to the SO(4) indices of the vector directions y,
and the remaining 4 directions (i = 5, . . . , 8) represents the SO(4) indices of the scalar
excitations
I,L ( )I, L
,
I ( ) ( ) =
(3.68)
I,L
1
I, L
= C iI1 i2 ...i KjL1 j2 ...j
ai ai ai aj1 aj2 aj 0
.
k
k
k!! 1 2
The free term of the action is then
1
1



+ h
sv 
S2 = d 
2
2
(3.69)
(3.70)
with
1
ai ai ,
R
4
hsv = hs + hv ,
hs =
1
aj aj .
R
8
hv =
(3.71)
j =5
i=1
The parameters familiar in string field theory literature are R 4 /J12 ( )2 = = 1/(p + )2
2 N/J 2 , g = J 2 /N . In terms of the stringlength parameter,
+
= gYM
2
(r) = pr , (r) =
1
1
Jr /R 2 . Note that 1/R since pr+ Jr /R. The supergravity limit corresponds to the
2 N )1/4 .
limit 0 with fixed R = (gYM
The interaction term is expressed in terms of the overlap state defined by
4
3
8
1 rs
rs
ai(r) n00 ai(s) +
aj (r) n00 aj (s) 0
v0
= exp
(3.72)
2
r,s=1
i=1
j =5
with the supergravity part of the familiar Neumann functions in the zeroslop limit 0
J
J
Jr
r
s
rs
1r
nrs
,
nr1
, for r, s = 1, 2 and n11
00 = n00 =
00 = 0. (3.73)
00 =
2
J1
J1
26
The unusual minus sign on the exponent is owing to our phase convention for the creation
annihilation operators. As is well known, the overlap state is the groundstate solution for
the continuity conditions satisfying
3
J2
J3
p(r) v0
= 0,
x(1) x(2) x(3) v0
= 0,
(3.74)
J1
J1
r=1
where
Jr
R
a(r) + a(r) ,
p(r) = i
a(r) a(r)
x(r) =
(3.75)
2
2R
2Jr
for all 8 directions. The ground state means that it also satisfy the locality condition of
the form
(x(1) x(2) )v0
= (x(1) x(3) )v0
= 0.
(3.76)
We find that
(1) I1 , L1 (2) I2 , L2 (3) I3 , L3 v0
1 !2 !3 ! L L L
C C C
K 1K 2K 3 .
=
1 ! 2 ! 3 !
1 !2 !3 !
(3.77)
This shows that the 3point interaction part of the effective action takes the following simple form:
1
(3)
(1)
(2) (3)  J1 J2 J3 h(2)
S3 =
(3.78)
d (1) 
s + hs hs v0
+ h.c.
2N
The integral with respect to the angular momentum under the conservation condition
J1 = J2 + J3 should be understood implicitly as before. The most important characteristic
of this expression is that the socalled prefactor involves only the energies of scalar excitations. As we have stressed in previous works, the holographic relation of our type does
not necessarily demand that the prefactor itself is the difference of the free Hamiltonian
(2)
(3)
(1)
hsv + hvs hvs . As a consequence of this remarkable property, the 3point vertices vanish whenever the scalar energies are preserved. This implies that, except for special cases
when the vector energies are simultaneously preserved, the 3point correlators vanish, as
has been explicitly exhibited in the expression of the CFT coefficient (3.61). Since we have
constraints k1 k2 + k3 and 1 2 + 3 , the conservation of both scalar and vector energies is possible only when the total energies are conserved, namely, when 1 = 2 + 3 .
Note that in this extremal cases we have finite results for the 3point correlators, owing to
the vanishing denominator in our holographic relation.
Of course, the above action for supergravity modes is valid in the limit 0 with
fixed R. The limit 0 corresponds to the imposition of the locality conditions (3.76)
for the overlap state. However, when we consider a finite , the latter conditions cannot be
preserved, as it should be for general string states which have nonzero extension. Then only
the more general continuity conditions (3.74) are satisfied. Correspondingly, the Neumann
rs
functions nrs
00 for the zero modes must be replaced by N00 , defined by
J2
J1
k2 +2
2
J3
J1
k3 +3
2
k1 !k2 !k3 !
I1
I2
I3
rs
N00
= f nrs
00
=f
rs
r1
= nr1
N00
00 =
Jr
,
J1
Jr Js
J12
27
,
11
for r, s = 1, 2 and N00
= 0,
(3.79)
where f = 1 4(1) (2) (3) K is a nontrivial function of (r) (r = 1, 2, 3). For the
definition of the quantity K, we refer the reader to Appendix B. In the supergravity limit
0, keeping R and Ji fixed, we have f 1. In the opposite large limit, it is known
that
f
(1)
R 2 J1
=
.
4(2) (3) 4 J2 J3
(3.80)
We denote the overlap state with this modification for finite by V0
.
4
3
8
1
rs
rs
V0
= exp
ai(r) N00 ai(s) +
aj (r) N00 aj (s) 0
.
2
r,s=1
(3.81)
j =5
i=1
The use of the corrected 3point vertex which is given by the same form as (3.78)
but v0
being replaced by V0
leads to a correction factor f 1 for the matrix elements,
multiplying the original local form.
On the other hand, the normalization of the CFT coefficients for the BPS operators used
in the foregoing subsections is believed to be exact to all orders in the gauge coupling =
2 N , because of the nonrenormalization property of 3point functions of chiral operators.
gYM
Indeed, as we stressed before, the CFT coefficients are nothing but the results from the
free field theory, which corresponds to the large limit. This implies that for nonzero ,
the rule of mapping between the matrix elements of the bulk 3point vertex and the CFT
coefficients must be extended as a natural ansatz to
L L L
CI11I2 I23 3 =
L L L
I11I2 I23 3
2 + 3 1
(3.82)
J J (2 +3 1 )/2
2 + 3 1
L L L
L1 L2 L3 = f 2 3
+ 1 I11I2 I23 3
I1 I2 I3
J1
2
with
L L L
J1 J2 J3 (2)
(1)
R hs + h(3)
s hs V0
,
N
(3.83)
(3.84)
and the 3point interaction term of the action is
J1 J2 J3 (2)
1
(1)
hs + h(3)
S3 =
(3.85)
d (1) (2) (3) 
s hs V0
+ h.c.
2
N
This is the main prediction of the present paper. By construction, this gives the same CFT
coefficients for the purely supergravity modes. But for higher stringy modes, the correction
factor would play an important role for impurity nonpreserving processes. The appearance
28
of the factor f with nontrivial dependence is not surprising if we recall that the origin of
the factor (J2 J3 /J1 )(2 +3 1 )/2 is the relation between the length and the cutoff with
respect to the integration, as exhibited in the integral (2.35). Our prediction
indicates that
the nonlocality of the vertex is essentially represented by a rescaling f . Finally, we
also mention that this interaction vertex is not SO(8) symmetric, contrary to a possibility
suggested in [4]. The origin of this phenomenon is the asymmetric roles played by scalar
and vector excitations from the viewpoint of boundary theory, as we have analyzed in the
previous subsection.
(4.1)
where E
is the overlap vertex including both bosonic and fermionic oscillators. The explicit form of the prefactor PSV is given in Appendix B, where all of the other necessary
definitions and formulas of string field theory are summarized, in order to make the present
paper reasonably selfcontained. It has been shown that this vertex has a nontrivial relation with the matrix of operator mixing which appears in the perturbative computation of
anomalous dimensions on the gaugetheory side. We will come back to this point later.
Another version of the 3point vertex, which was proposed [23] later in connection with
the duality relation of our type, takes the following form:
(2)
(3)
(1)
H3
D = H2 + H2 H2 E
,
(4.2)
where
H2(r) =
1
n(r) an(r) an(r) + bn(r) bn(r)
(r)  n=
with n(r) =
n2 + ((r) )2 (4.3)
are the free Hamiltonian operators. Here the stringlength parameters are chosen such that
(1) + (2) + (3) = 0 with (2) , (3) > 0 and (1) < 0 to be consistent with our conventions
used in the foregoing analyses. Note that this Hamiltonian coincides with the total free
Hamiltonian hsv , (3.71), of the previous section when it is restricted to supergravity modes
29
with the identification = 1/R. The form (4.2) is the simplest realization of the way
which was discussed in our first work [4] to obtain susycompatible interaction vertices. In
particular, it satisfies the SO(8) symmetry suggested there. The authors of [23] has shown
that this vertex gives nontrivial matrix elements in the leading order of the 1/expansion
for impurity preserving processes, following the duality relation of our type
J1 J2 J3 1
1, 2, 3H3
,
(2 + 3 1 )C123 =
(4.4)
N
when the mixing of gaugetheory operators is ignored. This duality relation itself for
impuritypreserving processes was first proposed in [26] from a viewpoint which is entirely
different from ours, and was actually abandoned later since it turned out that the relation
was not appropriate for the original purpose after correcting a sign error in some earlier
versions of works on PPwave string field theory. This relation is in fact obtained from our
holographic relation in the limit of large for arbitrary 3point processes when the number of impurities is preserved. We stress that in the light of our works, the more general
relation summarized in (3.82)(3.84) has a firm physical foundation from the holographic
principle. We should also recall that we are using the Euclidean picture of the tunneling
and hence this form of the interaction proportional to energy difference must not be thrown
away by a possible nonlinear unitary transformation: the energies are not conserved for
the Euclidean Smatrix. For a detailed discussion of our Smatrix picture, we invite the
reader to [4].
These two candidates for the 3point vertex conform to different duality relations connecting bulk and boundary. The first type H3
SV does not fit to our holographic relation.
The second type H3
D does not either, though motivated by (4.4), since for all of our arguments from Sections 2 and 3, the gaugetheory operators have to be definite conformal
operators of which 2point functions are diagonalized and 3point correlators take the standard conformal form (3.8). However, it is important here to recall that the requirement of
supersymmetry algebra to the first order in the string coupling puts only a linear constraint.
Hence, any linear combination of two possible vertices with some global
de
coefficients
pending only on string length parameters and the curvature parameter /R can
be an allowed candidate for the 3point vertex.
We can now study whether it is possible to obtain our effective action (3.85) for supergravity modes from these vertices. Let us first examine how the above two vertices behave
when they are restricted to purely supergravity modes. For purely bosonic external states,
they take the following forms:
8
4
1 i
2 i
2
2
2
XI XII
XI XII
H3
SV
Ea
2
i=5
=1
3
8
(r)
(r)
m
m (r)i (r)i
(r)i (r)i
=
a
am
a
a
Ea
(r) m
(r) m m
r=1 i=5 m=0
m=1
3
4
(r)
(r)
m
m
(r)i
(r)i
(r)i
(4.5)
a (r)i am
a
a
Ea
,
(r) m
(r) m m
r=1 i=1
m=0
m=1
30
8
4
1 i
2 i
2
2
2
XI + XII
+
XI + XII
H3
D
Ea
2
i=5
=1
3
8
(r)
(r)
m
m
(r)i
(r)i
(r)i
=
a (r)i am
+
a
a
Ea
.
(r) m
(r) m m
r=1 i=1
m=0
(4.6)
m=1
The state Ea
is the bosonic part of the overlap state. The expression XI and XII , defined
in Appendix B, contain the oscillators of only cos and sin modes, respectively, which are
(r)
(r)
represented by the creationandannihilation operator of positive (an ) and negative (an )
indices, respectively. As briefly discussed there, the factorization formula for the Neumann
matrices [17,21] allows us to derive the above simple expressions in terms of worldsheet
r .
energies m
By restricting to the zero modes, we see that neither of them separately reduces to the
effective action of the previous section. However, it is now evident that if we combine them
with equal weights we can get the desired form
1
H3
h H3
SV + H3
D
2
8
4
1 i
2 i
2
XI +
XII
Ea
2
i=5
i=1
8
3
4
(r)
(r)
m
m
(r)i
(r)i
(r)i
=
(4.7)
a (r)i am
+
a
a
Ea
,
(r) m
(r) m m
r=1
i=5 m=0
i=1 m=1
(2)
(3)
(1)
which in the zeromode sector reduces to the form (hs + hs hs )V0
, involving only
the scalar oscillators in the prefactor, as derived in the previous section. Apart from the
overall normalization fixed by comparing with (3.85), this combination is unique. Therefore, we have arrived at the uniquely possible string field 3point interaction term which is
consistent with the conclusion that we have reached by foregoing discussions.
We note that when restricted to the scalar modes the form (4.7) coincides with the one
suggested previously in [33] as the possible prefactor which is compatible with the holographic relation (4.4). Namely, they have shown that this gives the correct CFT coefficients
after taking into account the operator mixing for scalar operators with two impurities. The
result (4.7) shows that the scalar prefactor indeed consists only of the cos modes as they
have proposed, while actually the vector prefactor must consist only of the sin modes, in
order to be consistent with supersymmetry.
Thus the conclusion of this section is that the holographic string field theory is given in
the following form up to the first order in the string coupling
Sh = Sh,2 + Sh,3 ,
1
1
Sh,2 = d  
 
+ H2 
,
2
2
1
(2) (3)  J1 J2 J3 H3
h + h.c.
Sh,3 =
d (1) 
2N
(4.8)
(4.9)
(4.10)
31
2inl
1
J
Oij,n
=
e J Tr i Z l j Z J l
J N J +2 l=0
J is just the BMN operator corresponding to the ground state, with
and Ovac
.
yJ 2 n m 2
y
(5.2)
For notational brevity, in the present section, we suppress the classical part (namely, the zeroth order in ) of spacetime dependence for gaugetheory correlation functions. Note that
yJ
an;my is noting but the anomalous dimension of Om since it comes from Feynman diayJ
grams where the interaction occurs only between OnJ (x1 ) and Om (x2 ), while bn;my comes
yJ
from graphs where the interaction occurs among all three operators OnJ (x1 ), Om (x2 ) and
(1y)J
Ovac
(x3 ). This form, however, does not take the standard form of a 3point correlation
32
function of conformal operators. In other words, to this first order in , the BMN operator
J cannot be regarded as a conformal operator characterized by the standard conformal
Oij,n
transformation property.
As is well known now, this difficulty is resolved if we take into account the operator
mixing with multitrace operators. Consider a 2 2 matrix of twopoint functions of operators,
J
2inl
1
J
=
e J Tr i Z l j Z J l ,
Oij,n
J N J +2 l=0
J,y
yJ
(1y)J
:.
(5.3)
GAB =
mn
0
0
mn yz
+ g2
0
Cn;my
Cn;my
0
,
(5.5)
AB =
n2 nm
0
2
m
0
y 2 nm yz
0
+ g2
(an;my + bn;my )Cn;my
.
(5.6)
Note hat the 2point functions between a singletrace and a doubletrace operator are
obtained by taking the limit x23 1/ 0, and also that 1/ is identified with the shortdistance cutoff parameter associated with loops in the sense of Feynman diagrams.
The specific ansatz proposed in [29] in order to relate this mixing matrix to stringfield
theory vertex H3
SV is as follows. We first transform the basis for these operators by
O A = U AB OB such that
O A (0)O B (x) = AB + O( ).
(5.7)
However, this requirement alone does not determine the basis completely. To restrict the
basis further the authors demand that the transformation should be symmetric and real.
Then, the transformation matrix U can be fixed to the present order O(g2 ) as
g2
U = 1 G(1) ,
(5.8)
2
where G(1) is the O(g2 ) part of GAB . By this change of the basis, the matrix AB is
transformed to
g2 (0) (1)
,G
= (0) + g2 (1)
(5.9)
2
to the present order of approximation. The observation of [29] is that the offdiagonal
part of this matrix coincides with the matrix elements of the interaction vertex of the first
33
type, namely, H3
SV . This has been confirmed in several cases, for instance, for scalar
impurities in [14,29,30,34], and for other cases including vectors and fermions in [35,36].
For the supergravity chiral operators, the mixing matrix AB vanishes. In connection
with this, we note that in this case the matrix GAB should not be interpreted as representing operator mixing, as we will touch briefly below. This is as it should be, since the
3point functions of the chiral supergravity operators are not renormalized and hence take
the standard conformallyinvariant form even after including the higherorder effects.
We now show that this ansatz can be understood as a consequence of our holographic
string field theory, given the observation of [23]. In other words, all existing duality relations are actually compatible. Since our results are based on a clear spacetime picture
which is lacking, unfortunately, in other approaches, checking the consistency with more
formal correspondences seems to provide a useful guide for obtaining a unified viewpoint
on the holography in the planewave limit.
Let us consider a general class of three general operators O1 (x1 ), O2 (x2 ) and O3 (x3 ),
0
which have the classical conformal dimensions r satisfying the condition of degeneracy
0
0
0
1 = 2 + 3 . The number of impurities contained in O1 is thus equal to the sum of
those in O2 and O3 . Here and in what follows, we use the subscript 0
to denote the order
with respect to . The order with respect to gs will be denoted by the usual subscript (0),
(1), etc. The 3point function of them takes the following general form
O 1 (x1 )O2 (x2 )O3 (x3 )
x31 x12
0
1
1
2
2
,
= g2 C123 1 2 log(x12 ) + 3 log(x13 ) + b123 log
x23
(5.10)
0
where C123 denotes the 3point coefficient of the free gauge theory and r is the
anomalous dimension of the operator Or (xr ) to the first order in . As above, this expression in general does not conform to the standard form of the 3point correlation
functions of conformal operators. However, we can easily check that this is the most
general form which would lead to the standard form ((5.16) below) after taking into account the operator mixing. In the O( ) part on the righthand side of (5.10), the first term
1
2 log(x12 )2 comes from a class of Feynman diagrams in which the interaction occurs
0
only between O2 (x2 ) and O1 (x1 ). Similarly, the second term 3 log(x13 )2 comes from
the ones where the interaction occurs only between O3 (x3 ) and O1 (x1 ), and the third term
b123 log((x31 x12 )/x23 ) from the remaining ones, where the interaction occurs among all
three operators. Supposing that the doubletrace operator in consideration is obtained from
the product of O2 (x2 ) and O3 (x3 ) by taking the limit x23 1/, the matrices GAB and
AB in the subspace of operators O1 (0) and :O2 (x)O3 (x): takes the form
0
0
C123
1 0
,
+ g2
GAB =
(5.11)
0
0 1
C123
0
1
0
1
AB =
1
1
0
+ 3
2
1
1
0
0
(2 + 3 + b123 )C123
+ g2
(5.12)
,
1
1
0
(2 + 3 + b123 )C123
0
34
where we have taken into account the fact that the O(g20 ) part of AB gives the anomalous
dimensions of O1 and :O2 O3 :. In the case of two different scalar impurities discussed
1
1
1
0
above, 2 = m2 /y 2 , 3 = 0, 1 = n2 , b123 = bn;my = n(nm/y), and C123 = Cn;my .
Now, in order to extract the correct CFT coefficient with the operator mixing being taken
into account, we introduce the transformation matrix UAB which diagonalizes the matrix
O A (0)OB (x)
as a whole. Namely, contrary to the previous U , both of the matrices GAB
and AB are simultaneously diagonalized. It takes the form
0
D123
1 0
,
U=
+ g2
(5.13)
E123
0
0 1
where
1
D123 =
2 + 3 1 + b123
1
1
E123 =
1
2
1
3
b123
1
1
1
2
1
3
C123 ,
C123 .
(5.14)
Then the 3point function of the operators OA = UAB OB in the new basis is given by
O 1 (x1 )O2 (x2 )O3 (x3 )
= O 1 (x1 )O2 (x2 )O3 (x3 ) + g2 D12 3 :O2 O3 :(x1 )O2 (x2 )O3 (x3 ) + O g22
0
= g2 C123 + D123
1
0
0
g2 22 C123 + D123 + b123 C123 log(x12 )
1
0
0
0
+ 23 C123 + D123 + b123 C123 log(x31 ) b123 C123 log(x23 ) .
(5.15)
Here, the mixing effect does not affect the third term log(x23 ), since a correlation function which contains a double trace operator at x2 or x3 gives a O(g2 ) contribution by itself.
Taking into account the definition of D123 , we then obtain the following result regardless of
the expression of b123 , which is consistent with the canonical form of the 3point function
1
1
1
for operators with (anomalous) conformal dimensions 1 , 2 , and 3 , respectively:
O 1 (x1 )O2 (x2 )O3 (x3 )
1
1
1
= g2 C123 1 1 + 2 3 log(x12 )
1
1
1
1
1
1
+ 3 + 1 2 log(x31 ) + 2 + 3 1 log(x23 ) ,
(5.16)
0
where the true CFT coefficient C123 = C123 + D123 is expressed in terms of C123 and b123
as
0
C123 =
b123 C123
.
1 2 3
(5.17)
35
1
(5.19)
The offdiagonal matrix elements of the matrix in the particular basis which makes the
partial diagonalization are given in the form
1
0
0
off
= (2 + 3 1 )C123 + b123 C123 .
2
Using Eqs. (5.17), (5.18) and (5.19), we finally find
off
=
1
1, 2, 3H3
SV .
2
(5.20)
(5.21)
This is nothing but the claim made in [29,30], except for the overall factor 1/2. This factor can be understood from the difference of normalization. Our convention differs just by
this factor from the one adopted in the literature discussing this subject; see, for example,
(B.14) and (B.9) in [36].
It should be remarked that the above relation of the 3point vertex H3
SV with the operator mixing in perturbation theory at the boundary is intrinsically restricted to the processes
where the numbers of impurities are conserved. For this particular class of processes, our
argument clarified why the correct interaction vertex of the holographic string field theory
must be the particular combination of the two types of string interaction vertices: roughly
speaking, the part H3
D describes the bare part of the interaction of BMN operators,
while the part H3
SV describes the mixing among them. Both are necessary for describing the processes in the bulk, corresponding to a propagation of them from boundary to
boundary along the tunneling null geodesic. Note that the observation of [23] that the
string overlap vertex, the bare part of the interaction, in the large limit precisely corresponds to the freefield contraction is also quite natural for impurity preserving processes.
Through the holographic string field theory, this natural property is related to the specific
ansatz relating H3
SV to the matrix of operator mixing, which singles out the particular
basis of the gaugetheory operators.
36
We warn the reader, however, that the above intuitive interpretation on the different roles
of H3
SV and H3
D does not apply to more general processes in which the numbers of
impurities are not conserved. It is difficult to extend the above argument directly to such
cases.
6. Explicit examples
The purpose of this section is to present some concrete computations in order to confirm
our general discussions. In the present work, for simplicity we restrict ourselves to the cases
of two (conserved) impurities, for which we can utilize many results by other authors
on the gauge theory side. It is sufficient to focus on the cases of vector, mixed scalarvector, and fermionic impurities, since as we have mentioned before the case of pure scalar
impurities has been practically treated in [33] and was shown to be consistent with our
holographic relation. We are planning to study more general cases, especially the cases
where the numbers of impurities are not conserved, in a forthcoming work. In the present
section, we denote the first 4vector indices (i = 1, . . . , 4) by Greek letters (, ) for a clear
discrimination between the vector and scalar directions.
6.1. Vector impurities
Let us begin from the BMN operators with two vector impurities. The CFT coefficients
on the gauge theory side have been already computed in Ref. [14]. Nontrivial processes are
listed bellow for (2 + 3 1 )C123 :
sin2 (ny) m/y
,
y 2 n m/y
m m + vac n n :
Cvac
m m + vac n n :
Cvac
m m + vac n n :
sin2 (ny)
Cvac
,
2
y 2
(6.3)
m m + vac n n :
sin(ny) n2 + 3m2 /y 2
Cvac
,
2
y 2
n2 m2 /y 2
(6.4)
(6.1)
(6.2)
where the lefthand side represents symbolically the processes with two vector impurities
with and supposed to denote different vector indices. We have already used the fact
that the difference of the conformal dimensions, 2 + 3 1 , is given by (m2 /y 2 n2 )
with y J2 /J1 at the leading order of small = 1/((1) )2 . The mode numbers n and
m are supposed to satisfy n > 0 and m 0, respectively. The overall numerical constant
Ji
(Ji N Ji )1/2 Tr(Z Ji ) with
Cvac is the 3point function for vacuum BMN operators, Ovac
J2 + J 3 = J1 :
J1 J2 J3
y(1 y)
Cvac =
(6.5)
= g2
.
N
J1
37
Note that the normalization constant of the BMN operator has always been chosen such that
twopoint functions take the form O J (x)O J (0)
= 1/x2J . This overall factor coincides
with the corresponding factor for the 3point vertex of the holographic string field theory,
as determined in the previous section on the basis of the comparison with the supergravity
analysis.
The 3point vertex on the stringtheory side which should be compared with is the one
with the prefactor with only sin modes, since the scalar part XI2 vanishes for these cases:
XII2 E
=
(r)
rs
(r) (s)
m
(r) (s)
rs
n E
N N mn
m n + m
(r) mn
(6.6)
r,s=1 m,n=1
3
(r)
(s)
(r) (s)
m
n rs
rs
Nmn N mn
m n E
,
+
(r)
(s)
r,s=1 m,n=1
where we have expressed the formula in terms of the exponential basis defined as
0 = a0 ,
1
n = (an ian ),
2
1
n = (an + ian ),
2
(6.7)
which directly corresponds to the momentum basis of the BMN operators on the boundary.
For the first process (6.1), assuming m and n are nonzero, we obtain the following
matrix element
(1) (1) (2) (2)
n m m
123 0n
1 2
X E
2 II
(1)
(2)
m
21
1 n 12
12
21
12
Nnm N nm
+
N mn N mn
N nm
2 (1)
(2)
(1)
(2)
m
21
1 n 12
12
21
12
Nnm N nm
+
N mn N mn
+
,
N nm
2 (1)
(2)
(1)
(6.8)
(2)
where the second line comes form the case where the oscillators n and m are con(1)
(2)
tracted through the prefactor XII2 and the operators n
and m
through the overlap E
,
while the third line comes form the opposite case. The net results is
(1)
(2)
1 n
m 12
12 12
Nnm N nm
Nnm .
+
(1)
(2)
(6.9)
Using the explicit form of the Neumann coefficients in the large limit presented in (B.60),
we can confirm that this reduces in this limit to the gauge theory result (6.1), after including
the above overall factor. As for the special case m = 0, the absence of zeromode oscillators
in the prefactor XII leads to the vanishing result, which matches the gauge theory.
The second case (6.2) is related by a change m m to the first one. On the string
rs satisfies the same relation
theory side, the large behavior of Neumann functions N mn
with respect to this sign change, as exhibited in (B.60). Thus the first case ensures the
correct matching in the second case.
38
1 2
X E
2 II
(2)
22
22
11
n(1) 11
m
11
22
Nnn N nn
N mm
N mm N mm
N nn ,
(1)
(1)
(6.10)
(1)
(1)
and n
where the first term in the second line comes form the contraction of n
(2)
(2)
through the prefactor XII2 and of m and m through the overlap E
, while the second
term comes from the opposite case. Due to the property of the Neumann coefficients, the
second term vanishes, and the first term reduces to the field theory result. It is useful to
notice here that the gauge theory result for vector impurities in the case (6.3) is equal to the
scalar correspondent (, i, j ) except for the overall sign, and to recall that the scalar
case matches the duality relation (4.4) by using the vertex (1/2)XI2 E
. In addition to this,
we can easily check that the relation
2
1
(1) (1) (2) (2)
2
X
E
=
O
v
+
X
(6.11)
123
n m
m
n
I
II
2
is satisfied in the present flavorchanging process. Thus, XII2 E
can be replaced with
XI2 E
at the leading order in large limit, confirming the validity of the duality relation.
The same is true for the case of m = 0.
The last case (6.4) is the sum of the above three cases (6.1), (6.2) and (6.3) on the
gauge theory side. We can check that the same is true for the expressions on the string side,
indeed,
2
(1)
(2)
(2)
n(1)
(2)
m
1
+
2 (1)
(2)
i=5
=1
12
12 12
N mn
Nmn
+ N mn
(1)
(2)
1 n
m 12
12 12
Nmn N mn
Nmn ,
+
2 (1)
(2)
39
(6.14)
where the first part in the second line is the contribution from (XIi )2 E and the second
(6.15)
We can easily confirm that in the large limit this precisely reduces to the field theory
result (6.13).
6.3. Fermionic impurities
We first explain the convention for representing spinor impurities. We essentially follow Ref. [36]. Decompose the SU(4) Rsymmetry group as SU(4) = SO(4) U(1) =
SU(2) SU(2) U(1) with U(1) being the subgroup corresponding to large orbital angular momentum J :
4 (2, 1)+ + (1, 2) ,
where the subscript represents the U(1) charge. With this decomposition of the Rcharge
A
index, the correspondence between the fermionic fields A
and , with A and being
Rcharge and Lorentz spinor indices, and the string theory creation operators b 1 2 and
b 1 2 with SO(8) (= SO(4) SO(4) = SU(2) SU(2) SU(2) SU(2)) indices is given
by
A
(6.16)
(r,1/2 , r ,1/2 ) b1 2 , b 1 2 ,
, r ,1/2
) b 1 2 , b 1 2 .
A
(6.17)
(r ,1/2
The original SU(4) index A of A
r , 1/2) and simi is represented by the set (r, 1/2) + (
larly of A
by
its
conjugate
(r,
1/2)
+
(
r
,
1/2).
Note
that
the
range
of
indices
are r = 3, 4,
sin2 (ny)
n
,
2
n m/y
y
(6.18)
(6.19)
40
(6.20)
(6.21)
On the string theory side, the contribution from the H3
SV part in the interaction vertex
is already calculated in [36] as
1 vn n(1)11 11,m 12,m H3
SV = CN
(1)12
(2)
(2)
sin2 (ny)
,
((1) )2 2 y
(2)
(2)
(1)22
(2)
(2)
vn n(1)11 11,m 22,m H3
SV
(1)21
(2)
(2)
(6.23)
= 0,
(6.22)
(6.24)
sin2 (ny)
,
((1) )2 2 y
(6.25)
PSV = Y 4 ,
(6.27)
3
0 = b0 ,
where Y 4 is defined as Y 4 Y21 2 Y 21 2 = 12Y11 Y12 Y21 Y22 with Y21 1 = Y1 2 Y12 . See
Appendix B for the definitions of these quantities. Using the explicit form of Y 4 , we can
easily confirm that the amplitudes on the string side vanish in both case of (6.23) and (6.24),
and that (6.25) is equal to the minus of (6.22). With the definition of Y 1 2 in (B.44), the
2
(1)
(2)
stringmatrix element (6.22) is given by (G
n G
m ) , which reduces in the large limit
(r) given in (B.65).
to the righthand side of (6.22) using the asymptotic form for G
Next turning to the contribution from the H3
D part, we first find
1 vn n(1)11 11,m 11,m H3
D
(1)
(2)
2n
2m
1 12
1 12
21 2
21 2
=
+
Qnm + Qmn + Qnm Qmn
(1) (2)
4
4
(1)11
= CN
(2)
(2)
sin2 (ny)
4nm/y
,
2
2
2
((1) ) y n m2 /y 2
(6.28)
12
21
21
where we have used the relation Q12
nm = Nnm , Qmn = Nmn in the large limit,
rs
which can be easily shown by the definition of Qnm , (B.26), and the asymptotic form of
U (r) , (B.65), given in Appendix B. For other three cases (6.18), (6.20) and (6.21), we find
that the H3
D contribution gives one and the same expression
(1)
(2)
2
2n
sin2 (ny) n + m/y
2m 1 21
+
=
C
(6.29)
Qmn Q12
.
N
nm
(1)
(2) 4
((1) )2 y 2 n m/y
41
Combining these two types of contributions in each case, we obtain the matrix elements
which precisely coincide with the gauge theory results (6.18)(6.21).
7. Concluding remarks
Finally, we remark on some relevant problems left in the present paper and on possible
future directions. First of all, we have to emphasize again that our main predictions are not
restricted to the impuritypreserving processes which almost all of other works have been
limited to. In the supergravity sector, our holographic relation summarized in (3.82)(3.84)
is valid by its construction for general processes. The extremal cases of the supergravity
sector where conformal dimensions are preserved 1 = 2 + 3 are generalized by lifting
the degeneracy, owing to the higherorder effects in , to the impurity preserving processes
including string excitation modes. In the last two sections, we have confirmed that our relation is indeed valid in this case with nontrivial stringy effects. It is quite plausible that
the relation should then be naturally extended to impurity nonpreserving sectors. We can
note, for example, that the correction factor f in the holographic relation is consistent
with the fact that the CFT coefficients for such cases start from 0th order in 1/ just as the
impuritypreserving processes, while the 3point string vertices for such cases in general
start at most from the first order in 1/, because of the large behavior of the Neumann
functions. In fact, it is easy to confirm that our ansatz gives the correct results for a few simple cases. However, a systematic check of more general classes of impurity nonpreserving
processes is beyond the scope of the present paper and is left to a forthcoming work.
There are many possible directions following the present work: for instance, in connection with the question of impurity nonpreserving processes, we should investigate the
stringloop corrections. In most of the existing literature, impurity nonpreserving contributions have been ignored often without appropriate justification. It would also be interesting
to extend the discussions of Sections 2 and 3 to higherpoint correlation functions, from
both standpoints of supergravity limit and full string theory, and to see to what extent the
structure suggested in our first work [4] is realized in higher orders. Another important
problem is to derive the holographic stringfield theory in conjunction with our prescription of the holographic duality directly from the gaugetheory side. For such an attempt,
the collectivefield approach discussed in [37] seems to be suggestive.
Acknowledgements
We would like to thank H. Shimada for discussions at an early stage of the present
work. The present work is supported in part by GrantinAid for Scientific Research (No.
13135205 (Priority Areas) and No. 16340067 (B)) from the Ministry of Education, Science
and Culture.
42
(A.1)
following the convention of Ref. [11]. Using the formula for the integration over the 5dimensional unit sphere S 5 ( = 3 = the area of a unit 5sphere),
21m
1
x i1 x i2m =
(A.2)
(all possible contractions),
5
(m + 2)!
S5
we derive
I1
I2
I1 I2
Y Y = Ci1 ...ik Cj1 ...jk x i1 x ik x j1 x jk
S5
S5
= 3
and
21k
I1 I2
k! C I1 C I2 = 3 k1
,
(k + 2)!
2 (k + 1)(k + 2)
(A.3)
! ! !
2 +2 ! 1 2 3
S5
a(k1 , k2 , k3 ) C I1 C I2 C I3 ,
(A.4)
!k2 !k3 !
since in this case m = /2 and k11!
is equal to the number of same contractions occur,
2 !3 !
due to the total symmetry of the tensors C I .
Now we convert these integrals on S 5 into a Gaussian average for the SO(4) directions
by taking the large J limit. Such a calculation has previously done in Ref. [6] for some
special cases. First by expressing the S 5 harmonics in terms the S 3 harmonics Y I1 ,
iJ
(J + k)!
e
Y I = 2J /2
(A.5)
cosJ sink1 Y I1 ,
J !k!
I
Y =2
J /2
iJ
(J + k)!
e
cosJ sink1 Y I .
J !k!
(A.6)
I I
Y Y =2
S5
(J + k)!
J !k!
2
/2
d
/2
Y I Y I . (A.7)
S3
In the limit J , this can be evaluated by making a change of integration variables from
and the Cartesian coordinates x i on S 3 ( 3i=1 (x i )2 = 1) to a 4vector y 0 = , y i = x i
43
as
lim
Y I Y I = lim 2J
J
S5
= 2J
(J + k)!
2
J !k!
2
2
d 4 y eJ y y k Y I
(A.8)
I I
2
(J + k)!
3
k!
C =
2
, (A.9)
C
k
2
J !k!
2J +k1
J2
J + 12
2 J + 12
1/2
I1 I2 I3
J1 (J1 + k1 )!(J2 + k2 )!(J3 + k3 )!
Y Y Y =2
J1 !k1 !J2 !k2 !J3 !k3 !
5
S
J1
1 ! 2 ! 3 !
21+ 2 J12 J1
Equating this result with the large J1 limit of (A.4) which is equal to
1 J2 J3 1 5 I1 I2 I3
C C C ,
21+ 2 J 2 1 ! J1
3
(A.11)
we find the relation between the SO(6) and SO(4) contractions ( 1 = 1 = (k2 + k3
k1 )/2),
1
k2 /2
k3 /2
k1 !k2 !k3 ! I I I
I I I
J
J
J
1
2
3
C 1 C 2 C 3 .
C 1 C 2 C 3 = 1 !
(A.12)
J2 J3
J1
J1
1 ! 2 ! 3 !
This is used in Section 3 in order to express the 3point coupling in terms of the SO(4)
variables explicitly.
44
(r)
(r)
(r ) = x0
nr
nr
(r)
(r)
+ 2
xn cos
+ xn sin
,
(r) 
(r) 
(B.1)
n=1
(r)
1
nr
nr
(r)
(r)
(r)
pn cos
+ pn sin
(r ) =
p + 2
,
2(r)  0
(r) 
(r) 
(B.2)
n=1
(r)
(r) (r ) = 0 +
nr
nr
(r)
n(r) cos
,
2
+ n sin
(r) 
(r) 
(B.3)
n=1
1
nr
nr
(r)
(r)
(r)
(r ) =
n cos
+ n sin
+ 2
,
2(r)  0
(r) 
(r) 
(r)
(B.4)
n=1
where5
xn(r) =
(r)
a + an(r) ,
(r) n
2n
pn(r) = i
(r)
n (r)
an an(r) ,
2
(r)
1 +
(r)
(r)
0 =
1 + e((r) ) b0 1 e((r) )b0
2
2 (r) 
(r)
1
(r)
1 e((r) ) b0 + 1 + e((r) )b0
,
+
2
n(r)
(B.5)
(B.6)
n 1 + 1/2 (r)
1/2 (r)
U(r)n bn + e((r) n)U(r)n bn
=
(r)
2
2(r)  n
1 1/2 (r)
1/2 (r)
+
U(r)n bn + e((r) n)U(r)n bn
2
(n = 0),
(B.7)
(r)
n =
(r)  (r)
,
n
(B.8)
(B.9)
45
(r)
2 ,U
1 2 3 4
n2 + 2 (r)
(r)n (n (r) )/n, and ,
with i being SO(8) gamma matrices. With these Fock space basis, the free string Hamiltonian for rth string
(r)
1
=
2
2
(r) 
d 2 p (r)2 +
2
1
1
x (r) +
2 x (r)2
2
2
1
+
2
2
(r) 
d 2 (r) (r) +
1 (r)
(r) + 2(r) (r)
2
(B.10)
reduces to
H=
(r)
n (r) (r)
an an + bn(r) bn(r) .


n= (r)
(B.11)
(B.12)
r=1
where Ea
and Eb
are the bosonic and fermionic overlap vertices which satisfy
3
(r)
( )Ea = 0,
r=1
3
3
(B.13)
(B.14)
r=1
(r) ( )Eb = 0,
r=1
3
r=1
(2) (2) ,
1 = ,
(B.15)
(B.16)
(B.17)
The definitions of x(r) ( ), (r) ( ) and (r) ( ) are given in the same way. We always as+
) satisfies the relation (2) , (3) > 0, (1) < 0.
sume (r) ( p(r)
46
(B.18)
r,s=1
3
(r)1 2 rs (s)
(r) 1 2 rs (s)
Eb
= exp
bm
Qmn bn1 2 + bm
Qmn bn 1 2 vb
123 .
r,s=1 m,n=0
(B.19)
This overlap vertex is based on the ground states va
123 and vb
123 which are defined
(r)
(r)
by an va
123 = 0 and bn vb
123 = 0 for n Z. Note that Eb
is constructed on the
(r)
Fock vacuum vb
[31,32], not on the SO(8) vacuum 0
, defined as an 0
= 0 (n Z),
(r)
(r)
bn 0
= 0 (n = 0) and 0 0
= 0, on which the original fermionic interaction vertex
[1618] was constructed. As for the fermionic sector, the SO(8) spinor indices have been
decomposed as SO(8) = SO(4) SO(4) = SU(2) SU(2) SU(2) SU(2), according
to the works [19,20].
rs , N r and the fermionic Neumann coefficients
The bosonic Neumann coefficients Nmn
m
rs
r
Qmn , Qm are given by
(r ) (s )
r s
N00
(B.20)
,
= (1 4K) r s +
(1)
r1
N00
= r 1
(r )
,
(1)
(B.21)
1/2
rs
= 2(s ) s t (t ) C(r) N r m ,
Nm0
(B.22)
(B.23)
1/2
1/2
rs
= rs mn 2 C(r) C 1/2 A(r)T 1 A(s) C 1/2 C(s) mn ,
Nmn
(B.24)
rs
Nmn
= U(r) N rm U(s) mn ,
(B.25)
Qrs
mn
= e((r) )
U C N C
U(s) mn ,
(r)  (r)
Qrr
U C C N m,
m0 = r t (r ) (t )
(r)  (r) (r)
r1
Q1r
00 = Q00 =
(r )
1
,
2
(1)
(B.26)
(B.27)
(B.28)
otherwise = 0,
where n, m > 0,
r , s
Cmn = mmn ,
47
(B.29)
{2, 3}, r, s {1, 2, 3}, (1) (2) (3) , and
C(r)mn = m(r) mn ,
1
K = B T 1 B,
4
3
(B.30)
(B.31)
r=1
2 mn y(1)n+1
=
sin(my),
n2 y 2 m2
(1 y)
2 mn
A(3)
sin my,
mn =
2
n (1 y)2 m2
A(2)
mn
A(1)
mn = mn ,
Bm =
(B.32)
2
m3/2 sin(my),
y(1 y)(1)
(B.33)
1
n = (an ian ),
2
1
n = (an + ian ).
2
(B.34)
rs , is
The Neumann coefficients in terms of the exponential oscillator basis, N mn
(r) rs (s)
am
Nmn an =
m,n=
(r) rs (s)
m
Nmn n ,
(B.35)
m,n=
where
rs
rs
N 00
= N00
,
1 rs
rs
rs
rs
rs
N 0m
= N m0
= N 0m
= N m0
= N0m
,
2
1 rs
rs
rs
rs
,
= N mn
= Nmn
Nmn
N mn
2
1 rs
rs
rs
rs
N mn
.
= N mn
= Nmn
+ Nmn
2
rs s below in Appendix B.4.
We present the large behavior of N mn
(B.36)
(B.37)
B.3. Prefactors
The prefactor which was first constructed for the overlap vertex based on the SO(8)
vacuum 0
[1518] can be reformulated for the overlap vertex E
which is based on the
48
(B.38)
1 i j
K K + ij Vij K K + V
2
K 1 1 K 2 2 S1 2 (Y )S1 2 (Z) K 1 1 K 2 2 S1 2 (Y )S 1 2 (Z) , (B.39)
K J = XIJ XIIJ ,
(1 4K)1/2
Fn(r) an(r) ,
XI = i
(B.40)
(B.41)
r=1 n=0
(r)
(1 4K)1/2
Un(r) Fn(r) an ,
XII =
(B.42)
r=1 n=1
Y
1 2
(r)1 2
(1 4K)1/2
G(r)
,
n bn
(B.43)
r=1 n=0
Z
1 2
(r) 1 2
(1 4K)1/2
G(r)
,
n bn
(B.44)
r=1 n=0
with
(2)
F0 =
2
(2) (3) ,
(3)
F0 =
2
(3) (2) ,
1 1 1/2
1
U(r) C(r) CN r n
Fn(r) =
1
4K
(r)
(2)
G0
1
(2) (3) ,
=
(3)
G0
(1)
F0 = 0,
(n > 0),
1
(3) (2) ,
G(r)
U(r) C(r) C N n
n =
1 4K (r) 
(B.45)
(B.46)
(1)
(B.47)
(n > 0).
(B.48)
G0 = 0,
6 This definition differs form the one in (3.28) of [19] by a factor 2/ . Note also the difference of the total
factor of K i between here and there.
1 4
1 4 4
Y + Z4 +
Y Z
Vij ij 1 +
12
144
1
1
i
Yij2 1 + Z 4 Zij2 1 + Y 4 + Y 2 Z 2 ij ,
2
12
4
V
1 4
1 4 4
4
Y +Z +
Y Z
1
12
144
1 4
1
i 2
4
2
+ Y 2 Z 2 ,
Y 1 Z Z 1 Y
2
12
4
i
S(Y ) Y + Y 3 ,
3
49
(B.49)
(B.50)
(B.51)
with
r r
K r r K i i
Y21 1 Y1 2 Y1 2 ,
Y31 2 Y21 1 Y 1 2 ,
ij
Y 2ij Y 21 1 1 1 ,
K r r K i i r r
(r = 1, 2)
Y22 2 Y1 2 Y 1 2 ,
(B.53)
Y 4 Y21 1 Y 21 1 ,
(B.52)
Z 2ij Z 2 1 1
ij
1 1
(B.54)
,
2 2
ij
Y Z
Y 2k(i Z 2j )k . (B.55)
(r)
m (r)I (r)I (r)a (r)a
am am +
bm bm
H3
D =
(B.56)
E
m= (r)
r=1
I =1
a=1
(B.57)
1 2
K + K 2 Y 1 2 Y1 2 Z 1 2 Z 1 2 ,
4
(B.58)
where
Y 1 2 =
3
n (r) (r)1 2
,
G b
(r) n n
r=1
Z 1 2 =
3
n (r) (r) 1 2
.
G b
(r) n n
r=1
(B.59)
50
(1)m+n
22
N mn
=
,
4(1) y
33
=
N mn
23
=
N mn
1
,
4(1) (1 y)
(1)m+1
,
4(1)  y(1 y)
(1)m+n+1 sin(ny)
21
N mn
=
,
y(n m/y)
and, for m = n = 0, by
11
N 00
= 0,
12
N 00
= y,
(B.60)
31
=
N mn
(B.61)
(1)n sin(ny)
,
1 y(n m/(1 y))
(B.62)
13
N 00
= 1 y,
1
,
4(1)  y(1 y)
1
33
N 00
=
.
4(1) (1 y)
23
=
N 00
22
=
N 00
(B.63)
1
,
4(1) y
(B.64)
(r)
(r)
(1)n+1 2 sin(ny)
(1)
Gn =
,
(1) 
n
n
2(1) 
,
Un(3) =
,
Un(1) =
,
Un(2) =
2(1) y
2(1) (1 y)
n
1
,
1 4K =
4(1) y(1 y)
(B.65)
Gn = (1 K)1/2 G(r)
n .
7 Note that the definition of the Neumann vector N r here, with which F (r) and G(r) are defined, differs by
n
n
n
1/2
C(r) Ur from that of the Ref. [22].
51
(r)
n
r=1 n=0
1
an(r) an(r) E
= XI2 E
,
(r)
2
(B.66)
using the factorization formula obtained in [17,21]. Operating the annihilation operator an
on the vertex E
, the lefthand side of (B.66) can be written as
3
(r)
n
a (r) a (r) E
(r) n n
r=1 n=0
(r)
3
(r)
0
0
(r)
rs (r) (s)
N00
a0 a0 +
N rs a an(s)
=
(r)
(r) 0n 0
r,s=1
n(r)
n=1
(r)
rs (r) (s)
Nn0
an a0
n=1
n,m=1
(r)
m
rs (r) (s)
N a a
E
.
(r) mn m n
(B.67)
By the definition of Neumann matrices, the first term in the righthand side becomes
3
(r)
0
1
(r) (s)
N rs a a0 = X02 ,
(r) 00 0
2
(B.68)
r,s=1
and the sum of the second and the third terms reduces to
3
3
(r)
(r)
0
n
(s)
rs (r) (s)
N0n
a0 an +
N rs a (r) a
= X0 X+ ,
(r)
(r) n0 n 0
r,s=1 n=1
(B.69)
r,s=1 n=1
(r) (r)
1/2
(r) (r)
F0 a 0 +
Fn an
XI = i (1 4K)
r=1
r=1 n=1
X0 + X + .
(B.70)
rs
Nnm
1 1/2
rs
Nnm
U(r) C(r) CN r n U(r)
=
C(s) CN s m ,
(s)
(r)
1 4K (r) m + (s) n
(B.71)
3
3
(r)
(r)
(s)
n
1 n
rs (r) (s)
rs
rs m
Nnm
an am =
Nnm
+ Nnm
a (r) a (s)
(r)
2
(r)
(s) n m
r,s=1 n,m=1
r,s=1 n,m=1
1 2
= X+
.
2
Combining all the results above, we obtain the formula (B.66).
(B.72)
52
Noticing that the Neumann coefficient with negative Fourier modes is given by
rs
= U(r) N rs U(s) mn (m, n > 0)
Nmn
(B.73)
and the definition of XII , which has the extra iU(r) factor compared with XI , we can easily
see that the similar formula for negative modes,
3
(r)
n
r=1 n=1
1
(r) (r)
an an E
= XII2 E
,
(r)
2
(B.74)
n (r) (r)1 2
(r)
bn1 2 bn
+ bn 1 2 bn(r) 1 2 Eb
r=1 nZ (r)
= Y1 2 Y 1 2 + Z 1 2 Z 1 2 Eb
(B.75)
References
[1] D. Berenstein, J.M. Maldacena, H. Nastase, Strings in flat space and pp waves from N = 4 superYang
Mills, JHEP 0204 (2002) 013, hepth/0202021.
[2] S.S. Gubser, I.R. Klebanov, A.M. Polyakov, Gauge theory correlators from noncritical string theory, Phys.
Lett. B 428 (1998) 105, hepth/9802109;
E. Witten, Antide Sitter space and holography, Adv. Theor. Math. Phys. 2 (1998) 253, hepth/9802150.
[3] For reviews, see, e.g.,
A. Pankiewicz, Strings in plane wave backgrounds, Fortschr. Phys. 51 (2003) 1139;
J.C. Plefka, Lectures on the planewave string/gauge theory duality, hepth/0307101;
D. Sadri, M.M. SheikhJabbari, The planewave/superYangMills duality, hepth/0310119;
A.A. Tseytlin, Spinning strings and AdS/CFT duality, hepth/0311139;
R. Russo, A. Tanzini, The duality between IIB string theory on PPwave and N = 4 SYM: a status report,
hepth/0401155.
[4] S. Dobashi, H. Shimada, T. Yoneya, Holographic reformulation of string theory on AdS5 S 5 background
in the PPwave limit, Nucl. Phys. B 665 (2003) 94, hepth/0209251.
[5] T. Yoneya, What is holography in the plane wave limit of AdS(5)/SYM(4) correspondence?, hepth/0304183, expanded from the paper published in the Proceedings, Prog. Theor. Phys. 152 (2003) 108.
[6] N. Mann, J. Polchinski, ADS holography in the Penrose limit, hepth/0305230.
[7] M. Asano, Y. Sekino, T. Yoneya, PP wave holography for Dpbrane backgrounds, Nucl. Phys. B 678 (2004)
197, hepth/0308024;
M. Asano, Y. Sekino, Large N limit of SYM theories with 16 supercharges from superstrings on Dpbrane
backgrounds, hepth/0405203.
[8] Y. Sekino, T. Yoneya, Generalized AdS/CFT correspondence for matrix theory in the large N limit, Nucl.
Phys. B 570 (2000) 174, hepth/9907029;
Y. Sekino, Nucl. Phys. B 602 (2001) 147, hepth/0011122.
[9] D.Z. Freedman, S.D. Mathur, A. Matusis, L. Rastelli, Correlation functions in the CFTd /AdSd+1 correspondence, hepth/9804058.
[10] E. DHoker, D.Z. Freedman, S.D. Mathur, A. Matusis, L. Rastelli, Extremal correlators in the AdS/CFT
correspondence, hepth/9908160.
53
[11] S.M. Lee, S. Minwalla, M. Rangamani, N. Seiberg, Threepoint functions of chiral operators in D = 4,
N = 4 SYM at large N , Adv. Theor. Math. Phys. 2 (1998) 697, hepth/9806074.
[12] T. Yoneya, see http://www2.yukawa.kyotou.ac.jp/str2003/talks/yoneya.pdf.
[13] E. DHoker, J. Erdmenger, D.Z. Freedman, M. PrezVictoria, Nearextremal correlators and vanishing supergravity couplings in AdS/CFT, hepth/0003218.
[14] C.S. Chu, V.V. Khoze, G. Travaglini, BMN operators with vector impurities, Z(2) symmetry and pp waves,
JHEP 0306 (2003) 050, hepth/0303107.
[15] M. Spradlin, A. Volovich, Superstring interactions in a pp wave background, Phys. Rev. D 66 (2002) 086004,
hepth/0204146.
[16] M. Spradlin, A. Volovich, Superstring interactions in a pp wave background 2, JHEP 0301 (2003) 036,
hepth/0206073.
[17] A. Pankiewicz, More comments on superstring interactions in the pp wave background, JHEP 0209 (2002)
056, hepth/0208209.
[18] A. Pankiewicz, B. Stefanski Jr., PP wave light cone superstring field theory, Nucl. Phys. B 657 (2003) 79,
hepth/0210246.
[19] A. Pankiewicz, An alternative formulation of light cone string field theory on the plane wave, JHEP 0306
(2003) 047, hepth/0304232.
[20] A. Pankiewicz, B. Stefanski Jr., On the uniqueness of plane wave string field theory, hepth/0308062.
[21] J.H. Schwarz, Comments on superstring interactions in a plane wave background, JHEP 0209 (2002) 058,
hepth/0208179.
[22] Y.H. He, J.H. Schwarz, M. Spradlin, A. Volovich, Explicit formulas for Neumann coefficients in the plane
wave geometry, Phys. Rev. D 67 (2003) 086005, hepth/0211198;
See also: J. Lucietti, S. SchaferNameki, A. Sinha, On the planewave cubic vertex, hepth/0402185.
[23] P. Di Vecchia, J.L. Petersen, M. Petrini, R. Russo, A. Tanzini, The three string vertex and the AdS/CFT
duality in the pp wave limit, hepth/0304025.
[24] D.J. Gross, A. Mikhailov, R. Roiban, A calculation of the plane wave string Hamiltonian from N = 4 superYangMills theory, JHEP 0305 (2003) 025, hepth/0208231.
[25] C. Kristjansen, J. Plefka, G.W. Semenoff, M. Staudacher, A new double scaling limit of N = 4 superYang
Mills theory and pp wave strings, Nucl. Phys. B 643 (2002) 3, hepth/0205033.
[26] N.R. Constable, D.Z. Freedman, M. Headrick, S. Minwalla, L. Motl, A. Postnikov, W. Skiba, PP wave string
interactions from perturbative YangMills theory, JHEP 0207 (2002) 017, hepth/0205089.
[27] N. Beisert, C. Kristjansen, J. Plefka, G.W. Semenoff, M. Staudacher, BMN correlators and operator mixing
in N = 4 superYangMills theory, Nucl. Phys. B 650 (2003) 125, hepth/0208178.
[28] N.R. Constable, D.Z. Freedman, M. Headrick, S. Minwalla, Operator mixing and the BMN correspondence,
JHEP 0210 (2002) 068, hepth/0209002.
[29] J. Gomis, S. Moriyama, J. Park, SYM description of SFT Hamiltonian in a pp wave background, Nucl. Phys.
B 659 (2003) 179, hepth/0210153.
[30] J. Gomis, S. Moriyama, J. Park, SYM description of pp wave string interactions: singlet sector and arbitrary
impurities, Nucl. Phys. B 665 (2003) 49, hepth/0301250.
[31] C.S. Chu, V.V. Khoze, M. Petrini, R. Russo, A. Tanzini, A note on string interaction on the PP wave background, Class. Quantum Grav. 21 (2004) 1999, hepth/0208148.
[32] C.S. Chu, M. Petrini, R. Russo, A. Tanzini, String interactions and discrete symmetries of the PP wave
background, Class. Quantum Grav. 20 (2003) S457, hepth/0211188.
[33] C.S. Chu, V.V. Khoze, Correspondence between the three point BMN correlators and the three string vertex
on the pp wave, JHEP 0304 (2003) 014, hepth/0301036.
[34] G. Georgiou, V.V. Khoze, BMN operators with three scalar impurities and the vertex correlator duality in pp
wave, JHEP 0304 (2003) 015, hepth/0302064.
[35] G. Georgiou, V.V. Khoze, G. Travaglini, New tests of the pp wave correspondence, JHEP 0310 (2003) 049,
hepth/0306234.
[36] G. Georgiou, G. Travaglini, Fermion BMN operators, the dilatation operator of N = 4 SYM, and pp wave
string interactions, JHEP 0404 (2004) 001, hepth/0403188.
[37] R. de Mello Koch, A. Donos, A. Jevicki, J.P. Rodrigues, Derivation of string field theory from the large N
BMN limit, Phys. Rev. D 68 (2003) 065012, hepth/0305042.
Abstract
As a continuation of our previous works studying the holographic principle in the planewave
limit, we discuss the 3point correlation functions of BMN operators with bosonic excitations when
impurities are not conserved. We show that our proposal for a holographic mapping between the
conformal OPE coefficients of superYangMills theory and the 3point vertex of the holographic
string field theory is valid to the leading order in the large limit. Our results provide for the first
time a direct holographic relation for the 3point correlators of BMN operators including impurity
nonpreserving processes.
2004 Elsevier B.V. All rights reserved.
PACS: 11.25.w; 04.60.m
1. Introduction
What is the correct interpretation of holographic principle in the PPwave limit of the
AdS5 /SYM 4 correspondence has been one among several open problems since the first
original discussion [1] of the BMN limit. In the previous paper [2], we proposed a simple
direct relation between the 3point OPE coefficients of conformal BMN operators and
the bulk 3point interaction vertex of string field theory, by developing the basic ideas
Email addresses: doba@hep1.c.utokyo.ac.jp (S. Dobashi), tam@hep1.c.utokyo.ac.jp (T. Yoneya).
05503213/$ see front matter 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.nuclphysb.2004.12.013
55
presented first in Ref. [3]. In [2], we constructed the holographic string field theory which
meets requirements for the validity of the GKPWitten relation and confirmed explicitly
that it indeed reproduces the 3point OPE coefficients of N = 4 superYangMills theory
in the leading large expansion for stringy BMN operators in the impuritypreserving
sector. Furthermore, it was also clarified that two entirely different proposals for relating
3point vertices of (different versions of) string field theories to corresponding quantities
on the gaugetheory side in the impuritypreserving sector are actually compatible to each
other through our holographic map.
It is important to recall, as we have stressed there, that our holographic relation in principle should be valid for much more general impurity nonpreserving processes too, since by
construction the holographic relation is satisfied for the socalled nonextremal 3point
functions of chiral primary operators in the BMN limit. The latter functions are regarded to
be the supergravity sector of general impurity nonpreserving 3point functions. The goal
of the present paper is to confirm that our stringy holographic relation is indeed satisfied
for impurity nonpreserving cases in the leading large expansion, which is at the same
level as in the existing treatments of the 3point functions of impuritypreserving sector on
both sides.
The significance of this result seems evident, since the usual discussions in most of
other works being focused almost exclusively on the dilatation operator are practically restricted to impuritypreserving sector and therefore are difficult to extend them to impurity
nonpreserving processes.1 In contrast to this, the stringfield Hamiltonian in our approach
cannot exactly be interpreted as the dilatation operator, since it generates an infinitesimal
translation along a geodesic connecting from boundary to boundary in the Euclidean AdS
spacetime. Except for regions approaching asymptotically to the conformal boundary,
such a translation cannot be identified with dilatation. In fact, this discrepancy between
bulk Hamiltonian and dilatation operator on the boundary is, in our opinion, even more so
in the usual derivation of the Penrose limit, leading to a geodesic which is disconnected
from the boundary: there would be no asymptotic region where the Hamiltonian is related
to the dilatation, corresponding to the fact that the translation with respect to the global
time coordinate of the AdS spacetime does not coincide with dilatation on the boundary
at least within the usual geometric interpretation of the AdS spacetimes. This puzzle was
our main motivation for undertaking the present series of works [35], to which we would
like to invite the reader for further discussions of this basic issue. From our viewpoint,
the usual Penrose limit is obtained formally by a Wick rotation of the affine time parameter along the geodesics connecting boundary to boundary. Note that the relative topology
of trajectory with respect to the conformal boundary is completely changed by this Wick
rotation. This however explains why the dilatation comes about even in Minkowski formulation. In any case, given the apparent correspondence for conformal dimensions {r }
on one hand, it would be very strange if there could be no clear way of mapping the OPE
coefficients which, together with conformal dimensions, constitute the crucial data of CFT.
1 On the gaugetheory side, the matrix elements of a dilatation operator in the sense of renormalized pertur
bation theory would vanish unless the states have degenerate conformal dimensions in the lowest order. This is
borne out in computations in Refs. [911].
56
We note that, even though the matrix elements of 3point interaction vertices of string
field theory in the impurity nonpreserving sector often have additional powers in 1/,2
the OPE coefficients are not necessarily of lower order in the large expansion for large
but fixed Rcharge angular momentum J . It is also well known that a class of impurity
nonpreserving processes cannot simply be ignored in summing up intermediate states in
higherloop calculations, at least within the logic of string field theory as presently understood. Once the summation over intermediate states are involved, the limiting procedures
required in the studies of amplitudes for BMN states should be a very subtle problem, since
results would depend on the orders of various infinite sums and the limiting procedures.
We hope that our concrete results for impurity nonpreserving 3point correlators provide
a useful first step for further exploration of the AdS holography in the planewave limit
beyond the approximation of keeping only the impurity preserving sector.
In the next section, we start from summarizing briefly the holographic relation and the
holographic string field theory proposed in the previous work. In Section 3, we treat a
general case of impurity nonpreserving 3point functions where only nonsinglet (more
precisely, traceless) scalar (and bosonic) impurities corresponding to the fluctuations along
4directions among the S 5 of AdS5 S 5 are involved. In Section 4, we extend the results to
include nonsinglet vector excitations. In Section 5, we discuss a few typical cases where
singlet impurities are involved. The concluding Section 6 contains some remarks on relevant issues, especially, the uniqueness of our holographic relation and the question of
higherorder effects. Some useful formulas which are relevant for discussions in the main
text are summarized in Appendix A.
57
Under this requirement, we are led to a unique 3point vertex which should obey the
holographic relation based on the above tunneling picture. We stress that in the GKP
Witten conjecture the bulk partition function as a functional of boundary fields is fixed
uniquely with a given basis of linearized supergravity fields, independently of field redefinitions. For 3point correlators, the effect of such a field redefinition would be proportional
to the equation of motion, and integration over the interaction point in the bulk would give
vanishing result. In this sense, even though energies being now the conformal dimensions
are not conserved, the Euclidean Smatrix is very similar [3] to the ordinary Smatrix of
flat Minkowski spacetime. We will give a remark on the issue of uniqueness of our prescription in the final section of the present paper. A subtlety in the case of the extremal case
where the conformal dimensions are strictly conserved will also be clarified.
The holographic relation is summarized as
123
,
(2 + 3 1 )
J J (2 +3 1 )/2
2 + 3 1
123 = f 2 3
+ 1 123
J1
2
C123 =
with
(2.1)
(2.2)
J1 J2 J3
(2.3)
H3 h .
N
The conformal dimensions r are those in the planar limit. The CFT coefficients C123
is defined under the standard normalization of two point functions as
123 =
(1)1(2)2(3)3
x1 )O2 (
x2 ) =
O 1 (
12

x12 21
(2.4)
by
O 1 (
x1 )O2 (
x2 )O3 (
x3 ) =
C123
2
3

x12  
x23 21 
x31 22
(2.5)
with (
x12 = x1 x2 , etc.)
2 + 3 1
, etc.
(2.6)
2
The symbol H3 h is the 3point interaction vertex of the holographic string field theory
J J J
1
(2)(3) 1 2 3 H3 h + h.c.
S3 =
(2.7)
d (1)
2
N
1 =
Here the integration over the Rcharge angular momenta Jr associated with the string fields
should be implicitly understood, under the conservation condition
(r = 2, 3), (1) 
J1 = J2 + J3 . The quantity f in (2.2) depending on is responsible for the nonlocality
caused by the extended nature of strings and is given by
(r) 
(2.8)
58
where K is the wellknown expression in string field theory which is defined in terms of
various infinite matrices associated with the familiar overlap condition for string interaction. For full details of the string field theory, we refer the reader to Appendix B of the
previous paper [2] and to the references cited therein and below. The 3point interaction
vertex takes just an equalweight sum of the two previously constructed vertices which are
compatible with supersymmetry algebra:
1
H3 h H3 SV + H3 D
(2.9)
2
where H3 SV and H3 D are those proposed in [6] and in [7], respectively. In particular,
the prefactor of the part H3 D manifestly takes the form of the energy difference 21
while that of the part H3 SV is the one obtained by a natural generalization of the familiar
flatspace vertex.
Our convention for various parameters is as follows: R 4 /J12 ( )2 = = 1/(p1+ )2 =
2
gYM N/J12 , g2 = J12 /N , (r) = pr+ , (r)  = Jr /R 2 (r = 1, 2, 3). Do not confuse the
symbol (1) , etc. with the previously defined 1 , etc. as (2.6). Using these conventions, the
large limit of f is given as
f
J1
J2 J3
.
J1
4(1) 
(2.10)
We assume (1) (= (2) (3) ) < 0, (2) > 0, (3) > 0. Note that the mass parameter
becomes a meaningful curvature parameter = 1/R as we identify the lightlike momentum with the angular momentum by pr+  = Jr /R, which is the correct one for comparing
the action with flatspace form. The interaction term is of order gs forfixed R and p + as it
should be. The overall factor J1 J2 J3 /N is rewritten as 4gs ( )2 p1+ p2+ p3+ /R 5/2 in
terms of the usual string coupling constant gs . This form indicates that the effective action
in fact has a nontrivial curvature dependence.
The above normalization of the interaction vertex is fixed by matching the effective
action in the supergravity sector. The string field is normalized such that the free action
takes the standard Euclidean form
1
1
 
 + H
2  ,

S2 = d
(2.11)
2
2
1
(r)
H2 =
n(r) an(r) an(r) + bn(r) bn(r) with n(r) = n2 + ((r) )2 ,
(r)  n=
(2.12)
(r)
(r)
(r)
(r)
where (an , an )s and (bn , bn )s represent 8component bosonic and fermionic oscillators of rth string.
One of the most characteristic features of this string field theory is that for purely
bosonic string states the (total) prefactor reduces to the special form which consists of
(r)
(r)
only cos modes an (n 0) [8] and of only sin modes an (n > 0) for scalar (i = 58)
and vector (i = 14) impurities, respectively:
8
3
4
(r)
(r)
m
m (r)i (r)i
(r)i (r)i
a
am +
a
a
H3 h
(2.13)
Ea ,
(r) m
(r) m m
r=1
i=5 m=0
i=1 m=1
3
1
Ea = exp
2
59
(r) rs (s)
m
Nmn n
0(1)(2)(3)
(2.14)
r,s=1 m,n=
is the standard bosonic overlap vertex.3 Thus, Z2 (or higher SO(8) symmetry) of the free
(r) (r)
stringfield theory is completely violated by the interaction. The oscillators (n , n ,
n Z) in the exponential basis which directly corresponds to the standard convention for
the BMN operators are related to the trigonometric basis by (n > 0)
0 = a0 ,
1
n = (an ian ),
2
1
n = (an + ian ).
2
(2.15)
In our formalism, the BMN operators must have definite conformal dimensions to at
least the first order in g2 and to all orders with respect to . This means that for unprotected stringy BMN operators we have to take into account the effect of various operator
mixing including doubletrace operators [9,10] to the leading approximation in 1/N expansion in extracting the CFT coefficients C123 on the gaugetheory side. In real life, we
have to be satisfied by studies of leading 1/expansions at current stage of development,
since computation of 3point OPE coefficients at the order g2 and beyond has not been
carried out so far. Such a computation on the gaugetheory side in general requires twoloop calculations including operator mixing effects.
In [2], we have presented a general argument that the above relation must be correct for
arbitrary impuritypreserving 3point functions, by reinterpreting appropriately the previously known results [7,12] for comparison between string field theory vertices and gaugetheory calculations in the leading large expansion. We explicitly confirmed the above
relation for twoimpurity processes including vector and spinor excitations. The peculiar correction factor (f JJ2 J1 3 )(2 +3 1 )/2 ( 2 +23 1 + 1) appearing in the expression
(2.3) can be neglected to the leading order in the 1/expansion for impuritypreserving
processes. Thus, in these cases, the relation reduces to the one first conjectured in Ref. [11]
from a different viewpoint. Our previous work, however, clarified that this particular relation is valid only with our holographic string field theory vertex H3 h which takes into
account the operator mixing of gaugetheory operators by the above specific combination
of two different prefactors.
When the impurities are not conserved, the correction factor plays a crucial role, as
has already been shown by the construction in [2] for nonextremal correlators of chiral
primary operators on the basis of the GKPWitten relation. Our task is now to confirm
this for unprotected stringy BMN operators by studying the large limit. We can divide
the impurity nonpreserving 3point interactions into two classes, class I and II, respeccl
cl
cl
tively, depending on 1cl = (cl
2 + 3 1 )/2 > 0 or < 0 where r denotes the classical
conformal dimension, counting the number of fields and (spacetime) derivatives involved
3 As in [2], the minus sign on the exponential factor is due to our phase convention in defining the worldsheet
oscillators. This is different from the standard one in the literature, but is necessary for matching between bulk
and boundary.
60
in each BMN operator. Because of the SO(4) SO(4) symmetry, class II processes are
possible only when we allow singlet representations for the external line 1.
1
(J3
+ 1)N J3 +2
J3
2i aq
Tr k Z a Z J3 a e (J3 +1) ,
(3.2)
a=0
J1
1
2i an
(1)
O(i,n;j,n)
=
Tr i Z a j Z J1 a e (J1 +1) .
(J1 + 1)N J1 +2 a=0
(3.3)
The overall constants correspond to the normalization of scalar fields such that the
1 )Z(x2 ) =
free propagators are equal to (no summation over i) i (x1 )i (x2 ) = Z(x
2
1/x12  . Note that the phases associated with nonzero momenta are determined only by
relative (and oriented) distances among impurities. Here we have adopted the normalization constants and phase factors which are slightly different from those adopted in the
recent literature. In the large J limit, the difference between J and J + k 1 with k being
the number of impurities is inconsequential for our leading order computations restricted
to bosonic excitations. Therefore, in the following we will ignore this difference unless
otherwise stated explicitly and use the more familiar convention by ignoring these shift of
J , in order to save the space for mathematical expressions. However, for fermionic excitations it turns out that these shifts in the phases actually play an important role. Fermionic
excitations will not, however, be treated in the present paper and be left for a separate work.
(1)
(2)
(3)
It is obvious that the gaugetheory correlator O (i,n;j,n) O(i,m;j,m;k,p;,p) O(k,q;,q)
in the leading planar approximation must necessarily involve contractions of impurity
fields k and between 2 and 3, aside those between 1 2 and 1 3. This implies that
to the lowest nontrival order in g2 we can ignore mixing with doubletrace operators in
calculating the 3point function, since the effect of the mixing terms becomes of higher
order with respect to N1 , in contrast to the impurity preserving case where there is no
61
contraction between 2 and 3 for the mixing contribution which corresponds to the topology of a product of two cylinders (1/N ). We then find by a straightforward freefield
computation that the CFT coefficient is given by
2
1
2
sin2 (yn)
J1 J2 J3
1
2 sin (yn)
C123 =
2J1 2
=
2
2
N J J 3J
N
(n m
J12 y 2 (1 y) 2 (n m
y)
y)
1 2 3
(3.4)
with y = J2 /J1 . The first factor in the first equality comes from the overall normalization,
and the second factor is the result of freefield contractions and of summation over the
positioning of impurities. In particular, the factor 2 originates from the permutation of two
contractions between operators 2 and 3. Since the contractions between 2 and 3 can occur
in the planar limit only when the impurity fields k and are adjacent to each other, they
do not have any momentum dependence in the present largeJ limit. This is in general
true for arbitrary configurations of momenta of such impurities at least for exchanges of
bosonic fields, beyond the above special case.
Now let us turn to the corresponding calculation in string field theory. The string states
are
i(1) j (1)
(1) 0n n ,
k(3) (3)
(3) 0q q ,
(3.5)
respectively. The relevant part of the interaction vertex takes the form
H3 h P123 exp N 12 N 23 0(1),(2),(3)
where the prefactor is
(2)
P123 =
(2)
p
m
(2) (2)
(2)
(2)
(2) (2)
2 + m
+
2 + p(2) p + p p(2)
m + m m
(2)
(2)
(3)
(1)
q
n
(3)
(3)
(1)
(1)
2 + n(1) n + n n(1)
2 + q(3) q + q q(3)
(3)
(1) 
(3.6)
and
N rs
,
G f
J1
(2 + 3 1 ) 21cl 4(1) 
(3.9)
62
y(n m
y)
23
N pq
=
(1)p+1
,
4(1)  y(1 y)
(3.10)
we find that the CFT coefficient (3.4) from gaugetheory side precisely matches the string
interaction vertex with the holographic relation (2.1)(2.3).
It is not difficult to extend the above result to a more general case of scalar impurities.
In this section, we restrict ourselves to operators without any singlet representation with
respect to O(4) group of rotations of scalar directions. More precisely, the scalar impurities
i1 , i2 , . . . are contracted to polarization tensors CiI1 i2 ... which are traceless with respect
to arbitrary pair of the O(4) indices (i1 , i2 , . . .) for each conformal BMN operator in an
appropriate irreducible representation of SO(4):
CiI1 i2
I
=
O(p
1 ,p2 ,...)
J k1 N J +k
Tr i1 Z a1 i2 Z a2 e2i(a1 p2 +a2 p3 +)/J
a1 +a2 +=J
+ permutations ,
(3.11)
where k ( 1) is the number of impurities and as above we have to sum over nonequivalent
positioning of impurities.4 The traceless condition allows us to ignore the mixing of the
pairs of operators Z and Z which would be needed [13] for defining operators with definite conformal dimensions even if the mixing with doubletrace operators can be ignored. If
we consider higher orders in gYM , mixing between purely bosonic operators with antisymmetrized scalar indices (or mixed scalar and vector indices) and those involving fermion
impurities are expected to play important roles. In this paper, we ignore such complications
by restricting ourselves to the leading order effect in the large expansion on 3point functions. The symbol I of the polarization tensor indicates collectively the configurations of
4 Note that we have changed the notations slightly from Ref. [2]. For instance, k k. The convention on
the summation over permutations (for sugra modes in particular) is also changed. In the present paper, nonequivalent permutations exclude those which correspond to cyclic permutations, while in [2] they were not
excluded and hence the normalization constants had an additional power of 1/J .
63
both momenta and SO(4) representation. Its normalization is most conveniently expressed
(1) (1)
using oscillators ap , ap , etc. of strings, as
I I
(1)
(2)
1
C 2 exp
(3.12)
0(1)(2) = I1 I2 ,
(1) C
p
(2)
C I1 =
(1)
p=
I1
i1 (1) i2 (1)
(1)0Ci1 i2 ... p1 p2
etc.,
(3.13)
where the SO(4) indices on the exponential is suppressed. According to this convention,
various symmetry factors associated with the symmetry property of the SO(4) indices are
absorbed in the normalization of polarization tensors themselves. This representation naturally takes into account the summation over all allowed contractions with equal weights
corresponding to those of freegauge theory. Of course, the momenta must satisfy the level
matching condition
k
pk = 0.
(3.14)
i=1
Suppose the rth operator has kr impurities. Then, the number of contractions between
r = 2 and r = 3 is equal to
1cl = (k2 + k3 k1 )/2.
Now, in computing the gaugetheory correlators for this general type of operators, we note
the following properties, all of which already appeared in the above example.
(1) We can ignore operator mixing with doubletrace operators, since the mixing contributions are always of higher order in the 1/N expansion because of the existence of
nonzero number of 23 contractions. In the absence of the mixing with doubletrace operators, the CFT coefficients to the leading order in 1/ can be determined entirely from the
freefield contractions which give the correct leading order form of the spacetime factor
of conformal 3point functions.
(2) In the planar limit, the contractions between 2 and 3 must belong to a single group
of adjacent products of the impurity i fields in which no Z field is contained, since all Z
fields in the operators 2 and 3 must be contracted with Z fields in the operator 1 and hence
would lead to a nonplanar contribution otherwise. In other words, in the sum over ai in the
definition (3.11), such contractions are possible only when ais = ais+1 = = ais+1 1 = 0
with some s for a consecutive set of impurities (is , is+1 , . . . , is+1 1 ). See Fig. 1.
(3) By the cyclic symmetry of Tr operation and the levelmatching condition, we can
always choose s = 1. This means that we can replace the momentumdependent exponential by one for these impurities in the largeJ limit. Therefore, the 2 3 contractions do
not give any momentum dependent factor.
Using these properties, the 3point correlation function is computed as follows. First, the
overall constant factor is
cl
J1 J2 J3 cl (k1 +k2 +k3 )/2 J2 k2 /2 J3 k3 /2
N J1 +k1 +1 1
1cl !
=
1 !J1
N
J1
J1
ki 1 Ji +ki
3
N
i=1 Ji
64
Fig. 1. The dotted lines represent ZZ contractions, while real lines represent contractions of scalar fields i s.
The outer circle denotes the trace of operator product in the line 1 and two inner circles the traces of operator
products in the line 2 and 3, respectively.
k2 /2 k3 /2
cl
J2
J3
J1 J2 J3
cl
1
J1k1
1cl !J1 1 y(1 y)
N
J1
J1
(3.15)
where k2 = k2 1cl and k3 = k3 1cl represent the number of impurities in the operators
2 and 3, respectively, which contract with those of the operator 1. Note that the factor 1cl !
comes from the permutations of the ordering of impurities contracting between 2 and 3.
(1)
(2)
Secondly, each single contraction with momenta pi and pi respectively between
1 and 2 after the summation over the permutation of their positioning gives the factor
(suppressing the spacetime dependent factor 1/
x12 2 )
(2)
(1)
(1)
(1)
pi
sin ypi
pi
J1 eiypi
exp 2ia
(2)
p
J2
J1
a=0
pi(1) yi
J2
(3.16)
(2)
pi pi
(3.17)
Here the Kronecker delta of the corresponding SO(4) indices is suppressed. Similarly, each
(1)
(3)
single contraction with momenta qi and qi respectively between 1 and 3 gives
J3
a=0
(3)
(1)
(1)
q
q
J1 1 yei(1y)qi N 13(1) (3) .
exp 2ia i + i
qi qi
J3
J1
(2)
(3)
(3.18)
On the other hand, each contraction with momenta ri and ri respectively between 2
and 3 gives factor 1 (again apart from the spacetime factor) which is replaced by the
corresponding Neumann functions using the identity
(2)
1 = (1)ri +1 4(1)  y(1 y)N 23
(3.19)
(2) (3) .
ri ri
65
Collecting all these factors together, we conclude that the CFT coefficient is given as
k2 /2 k3 /2
I I I
J3
k1 J2
1cl +2cl +3cl J1 J2 J3
1
2
3
C123 = C C C (1)
J1
N
J1
J1
cl
1
cl
12
13
1
1 !J1
y(1 y)
J1 y N (1) (2)
J1 1 y N (1) (3)
12
4(1)  y(1 y)N 23
(2)
cl
cl
cl
= C I1 C I2 C I3 (1)1 +2 +3
qi qi
13
(3)
ri ri
23
pi pi
N 12(1)
(2)
pi pi
12
N 13(1)
cl
1
J1
J1 J2 J3
1 !
cl
N
4(1) 
23
N
(3)
(2) (3) ,
qi qi
13
ri ri
23
(3.20)
where rs denotes the product over the contractions and the symbol (C I1 C I2 C I3 ) denotes
the part of SO(4) contractions of polarization tensors associated with the freefield contractions. We note that the phase factors of the expressions (3.17)(3.19) cancel using the
levelmatching condition and that in terms of string notation, the following equality
I I I
cl
cl
cl
N 12(1) (2)
N 13(1) (3)
N 23
C 1 C 2 C 3 (1)1 +2 +3
(2) (3)
12
= (1) C I1 (2) C I2 (3) C I3 Ea
pi pi
13
qi qi
23
ri ri
(3.21)
is valid.
Let us next confirm that this result matches the string field theory through our holographic relation. Since we have already expressed a main part of the CFT coefficient
obtained on the gaugetheory side in terms of string states, only nontrivial part for this
task is to examine the prefactor. In the large limit, it takes the form
P123 = 0(2) 0(2) + 0(3) 0(3) 0(1) 0(1)
+
+
(2) (2)
(3) (3)
(1) (1)
m m + m
m m
m + [m m]
2
m=1
(2) (2)
(3) (3)
(1) (1)
m m + m
m m
m + [m m] .
(3.22)
m=1
Obviously, the contributions from the first and second lines is simply determined by counting the difference of the numbers of impurities in the process 2 + 3 1. Furthermore, in
the third line, the impurity preserving part corresponding to contractions 1 2 and 1 3
rs = N
rs , N
rs
rs = N
23
nm
nm
nm
cancels due to the equalities N nm
. Then, the property that N mn
is independent of the signs of momenta (m, n) shows that the contribution of the third term
is 1cl . Thus all together the total contribution of the prefactor is 21cl . But this is just
canceled by the denominator factor 1/(2 + 3 1 ) of the holographic relation. Using
66
(3.21), it is now clear that the CFT coefficient precisely satisfies our holographic relation
(2.1)(2.3) including the sign and numerical factors.
2
(
xr xs
)2
1
2 xr xs
(4.1)
in order to be consistent with the SO(4) symmetry and orthogonality of vector states. The
invariants such as ( x)2 can be ignored because of the traceless condition.
This allows us to treat the pairs of vector indices almost as those of additional virtual
scalar indices. Namely, the pair of derivatives of the spacetime factor of a free propagator
can contribute in the form
2ia( p(r) q (s) ) (r) (s) 1
2
Jr
Js
e
N rs
phase factor,
i j
2
4 ij p (r) ,q (s)

x


x

rs
rs
s
where the phase factor depending on (r, s) is the same as for the case of scalars. Thus a
pair of vector derivatives yields essentially the same factor as in the case of contractions of
scalar impurities, except for the factor 2 which is absorbed in the normalization, provided
that they act on the free propagator. This implies that the derivatives acting on 1 2
and 1 3 contractions give the same factors on both sides of gauge theory and string
field theory. In other words, the holographic relation is satisfied as it stands when vector
impurities are conserved for arbitrary impurity nonpreserving processes in which impurity
nonpreserving contractions between 2 and 3 only occur for scalar excitations.
Therefore, in the rest of this section, it is sufficient to concentrate to the case where all
vector impurities contribute to (2 3) contractions. Suppose first that we add 1cl such
derivatives for a process with nonzero scalar exchange in the 23 channel (1cl > 0). Since
cl
(2J2 )
1cl /2
(2J3 )
1
1 (2) 1cl (3) 1cl
cl
cl
(1 )!

x23 21
where the first factor comes from the normalization constant associated with the vector excitations. For notational simplicity, we suppress the SO(4) vector indices. The denominator
67
factor 1/(1cl )! is the symmetry factor to cancel the overcounting. This is equal to
cl /2 cl /2
1
1
(1cl + 1cl )
J3
1
1cl J2
J1
cl
cl
cl
2(
J1
J1
(1 ) 
x23  1 +1 )
cl (1cl + 1cl + 1)
21cl
1
cl
1
= J1 1
y(1 y)
.
cl
cl
cl
cl
cl
2(
(1 + 1) 2(1 + 1 ) 
x23  1 +1 )
As in the scalar case, we further multiply the factor
(2)
1 = (1)ri +1 4(1)  y(1 y)N 23
(2) (3)
(4.2)
ri ri
correction factor (
J1
1cl
.
cl  )
4(1)
21cl
(1cl +1cl +1)
(1cl +1) 2(1cl +1cl )
is just the
one required for modifying the energy factor in the holographic relation as
(1cl + 1)
21cl
(1cl + 1cl + 1)
2(1cl + 1cl )
(2) (2)
(3) (3)
(1) (1)
m m + m
m m
m + [m m]
2
m=1
(2) (2)
(3) (3)
(1) (1)
m m + m
m m
m + [m m] .
2
(4.3)
m=1
Then, the independence of (23) Neumann functions on the momentum (except for the
sign factor) shows that the vector contribution in this channel is zero due to cancellation
between the first and the second line in this expression. Hence, it does not contribute at
all to the modification of the 3point function. Also it is easy to check that the sign and
phase factors exactly match using level matching condition, in the same way as purely
scalar case of the previous section. Thus we can conclude that the holographic relation is
precisely satisfied.
The case with 1cl = 0 requires a separate consideration. In this case, there is no direct
contraction for scalar fields in the (23) channel. Also the correction factor before taking
derivatives is singular in the large limit, and we have to take into account mixing with
doubletrace operators. The spacetime dependence come from the order contributions
to the anomalous dimensions. Let us proceed inductively, starting from 1cl = 1. We have
to consider
1
1
1
(2) (3)
(J2 J3 )1/2 1
(2J2 )1/2 (2J3 )1/2

x23 21

x23 2
68
with 1 O(((1) )2 ). The factor 1 is precisely the necessary factor which modifies
the singular energy factor into nonsingular one corresponding to nonzero 1cl ,
(1cl + 1) (1cl + 1cl + 1) 1
(1 + 1) (1cl + 1)
1
=
= ,
21
21
21
2
2(1cl + 1cl )
with 1cl = 1 and 1cl = 0, and the power (J2 J3 )1/2 plays the same role generating the
(23) Neumann functions as in the above case with nonzero 1cl . Since there does not
occur any correction factor from the vector prefactor on the string side, this proves the
validity of the holographic relation. Note that in this special case the mixing with doubletrace operator for the operator 1 is effective on the gaugetheory side, but on the string
side the (23) Neumann functions contribute the same common factor for both H3 D
and H3 SV . Once the spacetime factor 1/
x23 2 is generated correspondingly to a single
vector exchange as above, we can apply the same argument inductively at each time of
adding one vector exchange between 2 and 3 as we have discussed for the case starting from
nonzero 1cl . This completes our argument for the holographic relation for nonsinglet
vector impurities.
69
Fig. 2. The contraction of Z and Z between 2 and 3, with 2 being the singlet operator and 3 the ground state
operator.
term is to meet this requirement by completing the vanishing SO(6) trace from the ZZ
directions.
Let us start from the case where the operator 1 and 3 are in the ground state,
1
1
O (3) =
Tr Z J1 ,
Tr Z J3 ,
O (1) =
(5.2)
J
J
J3 N 3
J1 N 1
and the operator 2 is the above singletstate operator,
J
2
2i
1
ap
(2)
J2 +1 .
Os(p;p) =
(5.3)
Tr
i Z a i Z J2 a e J2 4ZZ
2 J2 N J2 +2
a=0
On the gaugetheory side, the first term of (5.3) has no contribution in the leading order in
the perturbation theory. The second term can contribute in the leading order if and only if
(2)
one of the Z in Os(p;p) is contracted with one of Z in O (3) as in Fig. 2. The mixing with
doubletrace operator can again be ignored. Thus, apart from the spacetime factor (which
takes the correct form), the CFT coefficient is given as
J1 J2 J3 1
1
1
J3 J1 N J1 = 2
.
C123 = 4
(5.4)
J
J
J
+2
N
J2
1
3
2
J1 N J3 N
2 J2 N
On the string side, the state corresponding to the operator 2 is
(2) 0
1 (2) (2)
.
2 p,i p,i
22 =
Using the explicit form of the (22) Neumann function N mn
(1)m+n
4(1) y ,
prefactor is simply
= 2 and hence the matrix element of the interaction vertex is
given as
1
J1 J2 J3 22
J1 J2 J3
Np,p = 4
.
4
(5.5)
N
N
4(1) y
By multiplying the factor G
1
J2 J3 1
1 4(1) 
2 + 3 1
(5.6)
f
+1
,
(2 + 3 1 )
J1
2
2
J1
21cl
70
Fig. 3. The two types of contractions between 2 and 3 which are both singlet operators with two scalar impurities.
we find that the holographic relation is precisely satisfied including numerical coefficient
and sign.
This simplest example shows that the mixing of the second term (5.3) is absolutely
necessary for this agreement, and also that it is responsible to the (22) Neumann functions
on the string side. To convince the universality of this role of the Z Z mixing term, it is
useful to treat the case where both of the operators 2 and 3 are singlet states. So, now O (3)
in (5.2) is replaced by
J
3
2i
1
(3)
a
J3 a J3 aq
J
+1
3
Tr
i Z i Z
e
4ZZ
Os(q;q) =
(5.7)
.
2 J3 N J3 +2
a=0
Then, the CFT coefficient consists of two contributions with different sets of freefield
contractions as in Fig. 3,
iijj
C123 = C123
ZZ
+ C123
(5.8)
corresponding to the contributions from the contractions between scalar excitations and
between Z, Z fields, respectively. By similar calculations as above, we find
J1 J2 J3 4
J1 J2 J3 8
iijj
Z Z
C123 =
(5.9)
,
C123 =
.
N
J2 J3
N
J2 J3
On the stringfield side, we find the following matrix element of the interaction vertex,
by using selfexplanatory notations,
self
123 = 23
123 + 123
with
(5.10)
2
23 2
J1
J1 J2 J3
J1 J2 J3
1
2 Npq = 4
2
(5.11)
= 4
,
N
N
4(1)  J2 J3
2
J1
J1 J2 J3
J1 J2 J3
1
self
22
33
4Np,p Nq,q = 4
4
123 = 4
,
N
N
4(1)  J2 J3
(5.12)
23
123
71
where we have again used the property that the Neumann functions appearing here are
independent of the signs of momentum. Since the factor G is
2 + 3 1
1
J2 J3 2
1 4(1)  2
,
f
+1 2
(2 + 3 1 )
J1
2
4
J1
the results on both sides exactly match separately for two contributions between (5.8) and
(5.10). These two exercises seem sufficient to convince the general validity of our holographic relation for class I processes involving singlet representation.
5.2. Class II with scalar singlets
Let us next consider an example of class II processes which are possible with the presence of trace part in the operator 1. Consider the simplest nontrivial case with
J
1
2i
1
(1)
a
J1 a J1 ap
J
+1
1
Os(p;p) =
(5.13)
Tr
i Z i Z
e
4ZZ
,
2 J1 N J1 +2
a=0
O (2) =
1
J 2 N J2
Tr Z J2 ,
O (3) =
1
J 3 N J3
Tr Z J3 .
(5.14)
On the gaugetheory side, there is no possibility of free contraction for either term of the
operator 1. This means that the CFT coefficient vanish in the leading order at least in the
1/expansion and hence is at most of order g2 1/2 , if it does not vanish.
On the other hand, the corresponding matrix element of the 3point interaction vertex is
given as, suppressing an obvious SO(4) indices,
J1 J2 J3
(1)
(1)
(1)
0p(1) p 2 + p(1) p + p p(1)
N
2
(1) 2
1 11 (1) 2
(1)
11
11
exp N pp
p
p
N p,p
+ N p,p
p(1) p 0
2
J1 J2 J3 11
11
11
=
(5.15)
2Np,p + N p,p
.
+ N pp
N
2
The explicit form of the (11)Neumann function is
sin(my) sin(ny)
11
11
11
11
N nm
= N n,m
= N n,m
= N n,m
= (1)m+n+1
(1) 
(5.16)
in the leading order in the 1/ expansion. The correction terms for this expression is at
most of order 1/((1) )3 . Therefore, the matrix element vanishes in the leading order
and possible corrections must be at most of order ( )3 .
(1)
Let us next examine the multiplying factor,
2 +3 1
2
J2 J3
2 + 3 1
1
f
+1 .
G
(2 + 3 1 )
J1
2
72
cl
cl
cl
2 +3 1
2
is a negative integer. In
1cl = 1
which leads to the following leading behavior
G
1 J2 J3 ((1) )2
J1
((1) )2 J1 (1) 
1
=
=
.
f
J1
4(1)  p 2
p2
4p 2
(5.17)
Note that we have here taken into account the singularity of the function () 1/.
Thus we conclude that the CFT coefficient must be at most of order 1/((1) )2 in
conformity with the result from the gaugetheory side. We have again seen the crucial role
played by the factor G in relating both sides. The precise calculation of the higher order
terms for 3point correlation functions would require in general to take into account various
mixing terms involving both bosonic and fermionic impurities.
5.3. Vector singlets
The case of vectorsinglet states is more subtle. The general prescription adopted in
Section 4 for dealing with traceless vector excitations is not applicable to this case, since
in the presence of trace part there are a variety of different ways of constructing SO(4)
invariants after taking spacetime derivatives comparing the case for purely nonsinglet
processes. We consider only some simple examples.
The simplest singlet operator with vector impurities with nonzero momenta p and p
is
J
2
2i
1
Ovs(p;p) =
Tr (Dj Z)Z a (Dj Z)Z J 2a e J 1 ap
J
4 J N a=0
(5.18)
where the vector SO(4) index j is summed over from 1 to 4. The derivatives should now
be computed in the standard way. It is easy to check that this satisfies the orthonormality.
Note that we can ignore the second derivatives D 2 Z in the present approximation due to
the equation of motion.
Let us consider the simplest example of class I, in which the operator 2 is this singlet
state, while 1 and 3 are in the ground state,
J
2 2
1
2i ap
(2)
Tr (Dj Z)Z a (Dj Z)Z J2 2a e J2 1 .
Ovs(p;p) =
4 J2 N2J a=0
(5.19)
J2i
1 ap
2
= 0 (p = 0).
a=0
We conclude that the corresponding CFT coefficient must be zero in the leading order. On
the side of stringfield theory, the matrix element of the 3point vertex is
73
Fig. 4. The three types of contractions for the class II process with singlet vector operator for the line 1.
J1 J2 J3
(2)
(2)
(2)
0p(2) p 2 p(2) p p p(2)
N 2
(2) 2
1 22 (2) 2 1 22
22
(2) (2)
(5.20)
which is equal to
J1 J2 J3
22
22
22
(5.21)
2 2N p,p
+ N pp
+ N p,p
=0
N
by using explicit forms of the Neumann functions which has already been used for the
case of scalar singlet. Note that the crucial relative sign in the prefactor in the first line in
the expression (5.20), which is opposite to the corresponding scalar case. Since there is no
singularity in the factor G in the holographic relation, this shows that the both sides match.
It is easy to check that the vanishing of the CFT coefficient continues to be valid when we
increase the number of singlet vector impurities in the lines 2 and 3 on both sides.
The situation drastically changes for the class II case. Consider the simplest such
process where the operator 1 is the vector singlet state with 2 and 3 being in the ground
state:
J
1 2
2i
1
(1)
J1 2a e J1 1 ap .
Tr (Dj Z)Z a (Dj Z)Z
Ovs(p;p) =
4 J1 N J1 a=0
(5.22)
On the gaugetheory side, the 3point function consists of three different types of contributions
(1)
1(2,3)
12
13
O vs(p;p) (
(5.23)
x1 )O (2) (
x2 )O (3) (
x3 ) = F123
+ F123
+ F123
.
In the first and second terms, both of two derivatives in (5.22) acts only on (1 2)
contractions and on (1 3) contractions, respectively. In the third contribution, one of the
derivatives acts on (1 2) contractions and the other on (1 3) contractions. See Fig. 4.
We find
12
F123
=
sin2 yp
1
2
J2 J3 J12
,
2
2J
+2
2
(p) 
x12 

x13 2J3
N J1 J2 J3
(5.24)
13
F123
=
sin2 (1 y)p
1
2
J2 J3 J12
,
2
2J
(p)

x12  2 
x13 2J3 +2
N J1 J2 J3
(5.25)
74
4
J2 J3
N J1 J2 J3
sin yp sin (1 y)p
(
x12 x13 )
J12 (1)p
.
(5.26)
2
2J
(p)

x12  2 +2 
x13 2J3 +2
Summing up these three contributions, we have the expression
J1 J2 J3 sin2 yp

x23 2
J1
2
.
2
2J
+2
N
(p) 
x12  2 
x13 2J3 +2
The spacetime factor is precisely the form we should expect for a type II process with
1cl = 1, and hence the CFT coefficient is
J1 J2 J3 sin2 yp
C123 = 2
(5.27)
J1
.
N
(p)2
Let us turn to the string side. The matrix element of the interaction vertex is given by
J1 J2 J3
(1)
(1)
(1)
2
(1) 2
1 11 (1) 2
(1)
11
11
exp N pp
+ N p,p
p(1) p 0
p
p
N p,p
2
J1 J2 J3
J1 J2 J3 8
11
11
11
F123
6. Remarks
In this final section, we give a few remarks on some relevant issues related to the main
problem of the present work. We hope that this is useful for having further perspective on
our approach.
75
6.1. Uniqueness of the holographic relation and the integrability of string field theory
The holographic relation summarized in Section 2 was derived on the basis of the GKP
Witten relation and its perturbative computation in the case of 3point function using bulktoboundary propagator. The original effective action for supergravity modes on which
our derivation is based is written in terms of a particular choice of fields [16] by which
derivatives with respect to the AdS5 spacetime coordinates are completely eliminated.
This is justified since the 3point function is independent of field redefinition. To the present
order of approximation, it is sufficient to show that a wouldbe interaction term which could
be generated from the quadratic term by a field redefinition does not contribute to the 3point function. For notational simplicity, let us take the case of a single scalar field with
action
1
d 5 x g ()2 + m2 2 .
2
By a field redefinition + c 2 , a 3point interaction vertex
v3 =
gc 2 (K),
1
K = i gi + m2
g
is generated. However, since the bulkboundary propagator (z, x; y) satisfies the equation
of motion K(z, x; y) = 0 for arbitrary bulk spacetime positions (z, x), we conclude that
this vertex v3 does not contribute to the correlation function, according to the GKPWitten
prescription. Therefore, the holographic relation before the largeJ limit is unique.
Actually, in the special case of the socalled extremal correlators with 2 + 3 1 =
0, there is a wellknown subtlety. For generic configurations of U (1) Rcharge, this is
possible only for protected supergravity modes. The onshell matrix elements of 3point
interaction vertex in the bulk vanish for this case, but the 3point correlation functions do
not. In our formalism, the zero of the interaction vertex is canceled by the zero of the
denominator in relating the interaction vertex to the CFT coefficient. In this sense, the
extremal correlator has the 0/0 ambiguity. However, we can circumvent this problem in
the GKPWitten relation and in our holographic relation if we slightly shift the conformal
dimensions i i + i by analytic continuation using the generic matrix elements of
the interaction vertex such that the degeneracy 2 + 3 = 1 is lifted and, take the limit
i 0 finally after we compute the CFT coefficients.
By taking the largeJ limit, we arrived at the first order action described in Section 2
which describes the dynamics along a single tunneling trajectory. A question now arises as
to the possibility of similar field redefinitions after the largeJ limit. Here we have to be
very careful: In computing 3point correlation functions using the tunneling trajectory, we
had to introduce a cutoff near the boundary, since the approximation of a single trajectory
is violated near the boundary even if we take a short distance limit by which two among
three boundary points approach to each other up to a short distance of order . Finding the
precise prescription for this cutoff was the essence of Section 3 of our previous work [2].5
5 To avoid possible confusions, we emphasize that taking the shortdistance limit 0 here is different from
considering doubletrace operators. In our previous work [2], this limit was used for the purpose of extracting the
76
Namely, the integral with respect to the positions of interacting point along the tunneling
trajectory in the naive perturbation theory
+
d e(1 2 3 )
must be replaced by
+
J2 J3 (2)2
2
,
d e(1 2 3 ) exp f
e
J1 
x1 xc 2
(6.1)
where f plays the role of taking into account the corrections. This implies that we
cannot ignore total derivatives of the type
d (1 2 3 )
e
= (1 2 3 )e(1 2 3 )
d
inside the integral. It should be remarked that for this integral to be well defined we have to
assume an analytic continuation in 1 2 3 . Thus the shift i i + i of conformal dimensions which is necessary for the extremal case is actually a general premise for
our approach. Furthermore, the short distance limit 0 must be taken after evaluating
the integral with f JJ2 J1 3 2 = fy(1 y)J1 2 being kept fixed. In other words, the order of
two limits, large J and small , must be carefully chosen such that the integral is meaningful. Note that by overall scaling xr xr / the shortdistance limit can equivalently be
regarded as a largedistance limit in which 
x23  is fixed while x1 is sent to infinity.
The expression (6.1) is basically why we cannot impose conservation of energy for our
Euclidean Smatrix. Recall that in the argument for the uniqueness of 3point functions
before the largeJ limit, total spacetime derivatives are assumed to be vanishing. That
is justified in the case of the GKPWitten relation since the integral with respect to the
interaction point extends to the whole bulk spacetime. In contrast to this, in our tunneling picture emerging by taking the largeJ limit, we have to be very careful when total
derivatives are involved.
In connection with this, it is appropriate to reconsider the suggestion made in our
first work [3] concerning a possible integrable structure of holographic string field theory, namely, the interaction term is actually obtained from the free theory by making a
similarity transformation expressed using the CFT coefficient. In the present first order
approximation, this can equivalently be formulated by a canonical transformation
1 2
)( ),
( ) C (
( ) c ( ) = ( ) + C
2
1
) c ( ) = (
) C(
) + C 2 ( ),
(
)(
2
(6.2)
(6.3)
3point OPE coefficients. Its connection with the mixing of doubletrace operators for the impuritypreserving
sector was clarified there. For impurity nonpreserving processes, 3point functions cannot be related to the mixing matrix of a dilatation operator, since by definition we are treating operators with definite conformal dimensions
which are eigenstates of dilatation.
77
(6.4)
(6.5)
Note that we are here using a symbolic notation for brevity. It should be a trivial task to
convert those formal expressions using the braket notation for string fields. As is easily
seen, the canonical transformation, (6.2) and (6.3), generates the 3point interaction term
123 from the free Hamiltonian. This shows that the equation of motion of our holographic
string field theory is integrable provided that the CFT coefficient is well defined.
Does this imply that the threepoint functions are obtained from the free stringfield theory by the field redefinition corresponding to this canonical transformation? The answer to
this question is clearly no, since the Euclidean Smatrix is invariant under field redefinition. Recall that the equivalence with respect to the equation of motion does not in general
imply the equivalence of quantum mechanical amplitudes, because the surface terms often
play crucial roles.
Let us briefly check this using the LSZ formalism. The 3point Smatrix elements
2 )(
3 ) at the
are defined as the residue of the 3point Green functions (1 )(
poles at 1 = i1 , 2 = i2 and 3 = i3 , respectively, in the complexenergy
(i , i = 1, 2, 3) plane corresponding to external lines. If the field redefinition could generate a 3point Smatrix element from the free theory, that would be obtained from the Green
function
c (1 ) c (2 ) c (3 ) free ,
(6.6)
where we compute the Green function using the free field action for the original string
) (not for the transformed fields c and c ). Assuming that the
fields ( ) and (
Rcharge angular momenta of external states are J1 (= J2 + J3 ), J2 and J3 respectively
corresponding to our convention throughout the present work, it is sufficient to study the
following part of this Green function,
1 2
2 )(
3)
2 )(
3)
c (1 )(
C
(1 )(
free
free
2
3) .
= C (1 )(2 ) free (1 )(
free
(6.7)
Note that in this computation we do not use any partial integration with respect to the
positions of interaction, since we do not make the change of field variables explicitly for
the action itself. This expression has poles at the correct positions in the complex energy
plane for the lines 2 and 3, but not for the line 16 except for the extremal case where the
interaction vertex itself vanishes. As above, we assume the prescription i i + i for
treating the extremal correlators to remove the degeneracy. Under this definition, the Green
6 The cutoff (6.1) does not affect the positions of the leg poles.
78
Fig. 5. The diagram which contributes to the Green function of free string field after the field redefinition.
functions in general, including the extremal case, do not have correct poles corresponding
to external line 1, and hence the Smatrix elements vanish. Therefore, the free action cannot
in general generate the Smatrix elements.
When we represent the Green function in the pathintegral formalism, we would be able
to discuss the above problem by making the change of the integrated stringfields from
to the redefined ones (c , c ). Then, it would transform the free action to one with
(, )
the interaction term and with crucial surface terms. The above direct treatment without
using the change of the integration variables clearly shows that such an action generated
by the canonical transformation does give only vanishing Smatrix, due to cancellation
between local interaction terms and surface terms which are simultaneously generated by
the change of integration variables. To summarize, the conclusion of this subsection is that
our holographic relation is essentially unique, independently of possible field redefinitions,
from both viewpoints of the largeJ limit of the GKPWitten relation and of the logic of
string field theory to the present order of approximations.
In passing, we note that the above ambiguity of the extremal case in the bulk Smatrix
is essentially the same thing on the gaugetheory side as the ambiguity mentioned in [17]
with respect to the mixing of sugra BMN operators with doubletrace operators. As we
have argued in [2], however, we should not mix the doubletrace operators for sugra modes.
But if one wishes, one could mix the doubletrace operators and correspondingly make a
fieldredefinition appropriately such that one obtains the same 3point extremal correlators on both sides. The degree of arbitrariness is the same on both sides. So the existence
of this ambiguity itself may be regarded as the consistency of our Euclidean Smatrix
interpretation of holographic mapping in the PPwave limit. We do not however adopt
this viewpoint since without analytic continuation the extremal correlators cannot be computed from stringfield theory umambigously. This peculiar ambiguity is related to a well
known accidental degeneracy among multiparticle states of massless particles with strictly
collinear momenta. In any case, there is no ambiguity at all for impurity nonpreserving
cases treated in the present paper.
6.2. Higherorder effects
Extension of our ideas to higher orders is an important next problem. In the present
paper, we have only treated the leading order results with respect to 1/expansion (more
precisely, 1/(r) R 2 /J expansion) to the first order in the genusexpansion parameter 1/N . But in principle our holographic relation should be valid to all order in 1/
within the planar approximation. Actually, it is straightforward to compute next or next
79
tonext orders on the string side, since the Neumann functions are known [18,20] apart
from nonperturbative contributions. Corresponding computations on the gaugetheory
side, however, require us to obtain 3point functions at least up to 2loop orders, including all mixing effects. Once we go to such higherorders, an interesting puzzle arises. The
factor G which played crucial roles in the foregoing discussions gives corrections of the
form 2n logm (n > 0), since the large behavior (2.10) of the function f is correct
up to nonperturbative exponential corrections, according to the above references. If we
ignore the wouldbe nonperturbative terms of the Neumann functions, this implies that,
when they are not protected, 3point correlation functions are subject to these peculiar corrections which cannot be derived by the ordinary perturbation theory on the gaugetheory
side.
At present, we are not sure how this should be interpreted. A possibility may be that
the total sum of nonperturbative corrections on the Neumann functions might generate
such logarithmic terms and cancel these corrections when m = 0. For possible subtleties
of nonperturbative corrections, we refer the reader to [19] and also to [20]. Or, if such
terms with nonzero m are not canceled on the string side, nonperturbative effects (or
exact summation over the whole perturbation series) both on the gaugetheory side and on
the string side might be responsible for complete understanding of string/gauge duality. Of
course, it is also possible that our holographic relation itself may be subject to corrections.
The higher orders with respect to 1/N is also important. Since our holographic relation
is originally derived for sugra modes in the planar approximation, it is not immediately
clear whether our conjecture should be valid for nonplanar cases. In principle, we have
to examine the same limiting procedure for such cases as we have discussed for planar
3point functions. Following our philosophy, all rules of the holographic mapping should
be derived on the basis of the correspondence between correlation functions on both sides.
As for highergenus corrections, it is not evident whether the computation of loop corrections on the string side can be closed within the PPwave in the presence of the sum
over intermediate states. Recently, it has been reported [21] that the energy shift computed
from the H3 SV type string vertex disagrees with the gaugetheory prediction even at the
leading order in the first nonplanar correction, if the intermediate states are restricted to
impuritypreserving sector.
Even if it could be closed within the PPwave limit, we have to fix possible higherorder corrections for susy generators and Hamiltonians themselves on the string field side.
In our first paper [3] in this series, we have suggested that in planar (tree in the sense of
string field theory) approximation, the higherpoint vertices may be derived by a similarity transformation from free theory whose lowestorder form is nothing but the canonical
transformation discussed in the previous subsection. If we include nonplanar corrections,
these (similarity or perhaps unitary in the space of string fields) transformations generate
1/N corrections for vertices and even for quadratic kinetic terms. In order to confirm this
suggestion, it is important to investigate first the extension of our arguments for higherpoint amplitudes in the planar limit. It has been pointed out that the structure of 4point
correlation functions on the gaugetheory side suffers from some ambiguity [17] in determining the largeJ limit. From our point of view, we have to take into account the fact
that the approximation of a single tunneling trajectory can be justified only if we first take
the shortdistance limit for operator products such that the order of the distances at each
80
set of operator products at initial and final points at the boundary satisfies 2 c/J with
some finite c and finally let c 0. It is clear that the naive limit J on fixing the
spacetime configuration of operators at the boundary, which indeed leads to discontinuous results depending on the positions of operators, is not allowed. Careful analyses are
required, however, to make more precise the general structure suggested in [3] for holographic mapping for higherpoint Smatrix elements.
Note added
After completing this manuscript, we received a preliminary manuscript by H. Shimada
[22] who tried to establish the holographic relation in an approach which is slightly different from ours. His argument and result, though restricted yet to impuritypreserving sector
with scalar excitations, seem to be consistent with ours in [2] and complementary to our
argument for the uniqueness of the holographic relation.
Acknowledgements
We would like to thank H. Shimada for discussions. The present work is supported
in part by GrantinAid for Scientific Research (No. 13135205 (Priority Areas) and No.
16340067 (B)) from the Ministry of Education, Science and Culture.
(1)m+n
22
,
=
N mn
4(1) y
33
=
N mn
23
N mn
=
1
,
4(1) (1 y)
(1)m+n+1 sin(ny)
21
N mn
,
=
y(n m/y)
for (m, n) = (0, 0), and
11
N 00
= 0,
12
N 00
= y,
1
,
4(1)  y(1 y)
1
33
=
N 00
,
4(1) (1 y)
23
N 00
=
for m = 0 = n.
(1)m+1
,
4(1)  y(1 y)
(1)n sin(ny)
1 y(n m/(1 y))
13
N 00
= 1 y,
22
=
N 00
(A.1)
(A.2)
(A.3)
(A.4)
1
,
4(1) y
(A.5)
81
The asymptotic form in the large limit of f = 1 4(1) (2) (3) K is given by
f=
1
.
4(1) y(1 y)
References
[1] D. Berenstein, J.M. Maldacena, H. Nastase, Strings in flat space and pp waves from N = 4 superYang
Mills, JHEP 0204 (2002) 013, hepth/0202021.
[2] S. Dobashi, T. Yoneya, Resolving the holography in the planewave limit of AdS/CFT correspondence,
hepth/0406225.
[3] S. Dobashi, H. Shimada, T. Yoneya, Holographic reformulation of string theory on AdS5 S 5 background
in the PPwave limit, Nucl. Phys. B 665 (2003) 94, hepth/0209251.
[4] T. Yoneya, What is holography in the plane wave limit of AdS(5)/SYM(4) correspondence?, hepth/0304183,
expanded from the paper published in the Proceedings, Prog. Theor. Phys. 152 (2003) 108.
[5] M. Asano, Y. Sekino, T. Yoneya, PP wave holography for Dpbrane backgrounds, Nucl. Phys. B 678 (2004)
197, hepth/0308024;
M. Asano, Y. Sekino, Large N limit of SYM theories with 16 supercharges from superstrings on Dpbrane
backgrounds, hepth/0405203;
M. Asano, Stringy effect of the holographic correspondence for Dpbrane backgrounds, hepth/0408030.
[6] M. Spradlin, A. Volovich, Superstring interactions in a pp wave background, Phys. Rev. D 66 (2002) 086004,
hepth/0204146;
M. Spradlin, A. Volovich, Superstring interactions in a pp wave background 2, JHEP 0301 (2003) 036,
hepth/0206073;
A. Pankiewicz, More comments on superstring interactions in the pp wave background, JHEP 0209 (2002)
056, hepth/0208209;
A. Pankiewicz, B. Stefanski Jr., PP wave light cone superstring field theory, Nucl. Phys. B 657 (2003) 79,
hepth/0210246;
A. Pankiewicz, An alternative formulation of light cone string field theory on the plane wave, JHEP 0306
(2003) 047, hepth/0304232;
A. Pankiewicz, B. Stefanski Jr., On the uniqueness of plane wave string field theory, hepth/0308062.
[7] P. Di Vecchia, J.L. Petersen, M. Petrini, R. Russo, A. Tanzini, The three string vertex and the AdS/CFT
duality in the pp wave limit, hepth/0304025.
[8] C.S. Chu, V.V. Khoze, Correspondence between the three point BMN correlators and the three string vertex
on the pp wave, JHEP 0304 (2003) 014, hepth/0301036.
[9] N. Beisert, C. Kristjansen, J. Plefka, G.W. Semenoff, M. Staudacher, BMN correlators and operator mixing
in N = 4 superYangMills theory, Nucl. Phys. B 650 (2003) 125161, hepth/0208178.
[10] N.R. Constable, D.Z. Freedman, M. Headrick, S. Minwalla, Operator mixing and the BMN correspondence,
JHEP 0210 (2002) 068, hepth/0209002.
[11] N.R. Constable, D.Z. Freedman, M. Headrick, S. Minwalla, L. Motl, A. Postnikov, W. Skiba, PP wave string
interactions from perturbative YangMills theory, JHEP 0207 (2002) 017, hepth/0205089.
[12] J. Gomis, S. Moriyama, J. Park, SYM description of SFT Hamiltonian in a pp wave background, Nucl. Phys.
B 659 (2003) 179, hepth/0210153;
J. Gomis, S. Moriyama, J. Park, SYM description of pp wave string interactions: singlet sector and arbitrary
impurities, Nucl. Phys. B 665 (2003) 49, hepth/0301250.
[13] A. Parnachev, A.V. Ryzhov, Strings in the near plane wave background and AdS/CFT, JHEP 0210 (2002)
066, hepth/0208010.
[14] N. Beisert, BMN operators and superconformal symmetry, Nucl. Phys. B 659 (2003) 79, hepth/0211032.
[15] U. Grsoy, Vector operators in the BMN correspondence, JHEP 0307 (2003) 048.
[16] S.M. Lee, S. Minwalla, M. Rangamani, N. Seiberg, Three point functions of chiral operators in D = 4,
N = 4 SYM at large N , Adv. Theor. Math. Phys. 2 (1998) 697, hepth/9806074.
82
[17] C. Kristjansen, J. Plefka, G.W. Semenoff, M. Staudacher, A new double scaling limit of N = 4 superYang
Mills theory and pp wave strings, Nucl. Phys. B 643 (2002) 3, hepth/0205033.
[18] Y.H. He, J.H. Schwarz, M. Spradlin, A. Volovich, Explicit formulas for Neumann coefficients in the plane
wave geometry, Phys. Rev. D 67 (2003) 086005, hepth/0211198.
[19] I.R. Klebanov, M. Spradlin, A. Volovich, New effects in gauge theory from ppwave superstrings, Phys.
Lett. B 548 (2002) 111, hepth/0206221.
[20] J. Lucietti, S. SchaferNameki, A. Sinha, On the planewave cubic vertex, Phys. Rev. D 70 (2004) 026005,
hepth/0402185.
[21] P. Gutjahr, A. Pankiewicz, New aspects of the BMN correspondence beyond the planar limit, hepth/0407098.
[22] H. Shimada, in preparation.
Abstract
The nexttominimal supersymmetric Standard Model (NMSSM) includes a Higgs isosinglet superfield in addition to the two Higgs doublet superfields of the minimal extension. If the Higgs fields
remain weakly coupled up to the GUT scale, as naturally motivated by the concept of supersymmetry, the mixing between singlet and doublet fields is small and can be treated perturbatively. The mass
spectrum and mixing matrix of the neutralino sector can be analyzed analytically and the structure
of this 5state system is under good theoretical control. We also determine decay modes and production channels in sfermion cascade decays to these particles at the LHC and pair production in e+ e
colliders.
2005 Elsevier B.V. All rights reserved.
PACS: 12.60.Jv
1. Introduction
The minimal supersymmetric Standard Model (MSSM) [1,2] opens the path to the
analysis of supersymmetric theories. Arguments have been advanced however that suggest extensions beyond this minimal version. One wellmotivated example is the nexttominimal supersymmetric Standard Model (NMSSM) [3] in which an isosinglet Higgs
Email address: zerwas@desy.de (P.M. Zerwas).
05503213/$ see front matter 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.nuclphysb.2005.01.006
84
superfield S is introduced in addition to the two isodoublet Higgs fields H u,d incorporated
in the MSSM to generate electroweak symmetry breaking. Such an extension offers a possible solution of the problem, generating in a natural way, a value of the order of the
electroweak breaking scale v; this is achieved by identifying , apart from the O(1) coupling, with the vacuum expectation value of the scalar component S of the new isosinglet
field. [For a recent summary of this construct see Ref. [4]; a useful code has been made
available in Ref. [5].]
The superpotential of the NMSSM includes, besides the usual MSSM WY Yukawa components, an additional term, which couples the isosinglet to the two isodoublet Higgs
fields, plus the selfcoupling of the isosinglet:
H u H d ) + 1 S 3 .
W = WY + S(
3
(1)
The two parameters and are dimensionless. By demanding the Higgs fields remain
weakly interacting up to the GUT scale, the two couplings are bounded at the electroweak
scale by the inequalities , 0.7. While the scalar Higgs sector includes several soft
supersymmetry breaking parameters, the Lagrangian of the gaugino/higgsino sector is
complemented only by the familiar SU(2) and U(1) gaugino mass terms. As a result, the
parameter space of the neutralino sector is much less complex than the Higgs space.
The superpotential without the singlet selfcoupling, i.e., = 0, incorporates a Peccei
and
Quinn (PQ) symmetry: {H u (1), H d (1), S(2),
Q(1),
U (0), D(0),
L(1),
E(0)}.
Q
L are the quark and lepton SU(2) doublet superfields, while U , D and E are the up and
downquark and lepton SU(2) singlet superfields, respectively. The integer of each parenthesis indicates the PQ charge of the corresponding superfield. The
spontaneous breaking
of this symmetry by the nonzero vacuum expectation value vs / 2 of the scalar S field
gives rise to a massless Goldstone boson. However, when = 0, the mass is lifted to a
nonzero value by the selfinteraction of the S field. Still, a discrete Z3 symmetry is left
which would lead to the formation of domain walls in the early Universe. This problem
can be tamed by introducing new interactions of the inverse Planck size that, however, do
not affect the lowenergy effective NMSSM theory [6].
In contrast to the Higgs sector, masses and mixings in the chargino system are not
affected by the singlet extension. [Of course new decays such as S + or +
5 H + may be possible if allowed kinematically.]
So far the supersymmetric particle spectrum of the NMSSM has received only little
attention in the NMSSM literature, Refs.[3,79]. In this report we attempt a systematic
analytical analysis of the neutralino system. In contrast to the MSSM where exact solutions
of the mass spectrum and mixing parameters can be constructed mathematically in closed
form, this is not possible any more for the NMSSM in which the eigenvalue equation is of
5th order, not allowing closed
solutions.2 However, since the coupling between singlet and
doublet fields is weak, v/
2 O(10 ) GeV, compared with the typical supersymmetry
scale M1,2 and = vs / 2 O(103 ) GeV, a perturbative expansion of the solution gives
rise to a good approximation of the mass spectrum while the magnitude of the matrix
elements in the mixing matrix is at least qualitatively well understood. The usefulness of
a perturbative expansion has also been noticed in Ref. [9]; however, here, extending the
85
Higgs analysis in Ref. [4], we work out this approach systematically for all facets of the
NMSSM.
While plays a crucial rle in the Higgs sector, it is less crucial for the neutralino
system. The size of vs , with vs 15v to maintain a link with the electroweak scale, just
determines the singlino mass before modified by mixing effects. Once masses and mixings
are determined, the couplings of the neutralinos to the electroweak gauge bosons and to
scalar/fermionic matter particles are fixed. Decay widths and production rates of the five
neutralinos can subsequently be predicted for squark cascades at the LHC [10] and e+ e
annihilation at prospective linear colliders [11].
The report is organized as follows. In Section 2 we describe the neutralino sector of
supersymmetric models in which the pair of Higgs doublet superfields is augmented by
an additional isosinglet field. In Section 3 we show how, for a naturally expected weak
coupling, the properties of the four standard neutralinos are modified; moreover the properties of the fifth neutralino, the new singlinodominated state, are calculated. All these
spectra and mixings are predetermined analytically before the surprisingly good quality
of the weakcoupling expansion is demonstrated by comparison with numerical solutions.
In this way we achieve a satisfactory theoretical understanding of the system. In the limit
of large gaugino mass parameters M1,2 compared with the higgsino mass parameter , or
vice versa, the MSSM part can be easily diagonalized analytically and a clear and simple picture of the entire system emerges. The section is concluded by a lovely toy model
in which we set M1 = M2 and tan = 1; this set allows us to solve the system exactly,
leading to transparent closed expressions for the neutralino mass spectrum and the mixing
parameters. A sample of decay widths and production cross sections for the neutralinos
is presented in Section 4. The results are summarized in Section 5 and technical details
of the diagonalization procedure for the 5 5 neutralino mass matrix are described in the
Appendix A.
1
1
0
v
,
Hd = cos
,
S = vs / 2,
Hu = sin
(2)
v
0
2
2
the Higgshiggsino mass parameter
= vs / 2
(3)
is generated and, subsequently, the neutralino mass matrix
M X
M5 =
X T
86
M1
0
mZ c sW mZ s sW
0
mZ c cW mZ s cW
0
0
M2
s
M5 = mZ c sW mZ c cW
.
mZ s sW mZ s cW
0
c
0
(4)
This 5 5 mass matrix is constructed from the standard 4 4 MSSM neutralino mass
matrix M in the upper left corner, the mass term of the higgsino component S of the
singlet superfield S,
= 2vs / 2
(5)
and the mixing between doublets and singlet parameterized by
= v/ 2.
(6)
W 3 , H 0 , H u0 , S).
As usual, M1 and
The mass matrix M5 is defined in the group basis (B,
d
M2 are the soft SUSY breaking U(1) and SU(2) gaugino mass parameters, tan is the ratio
of the vacuum expectation values of the two neutral SU(2) Higgs doublet fields (as defined
in Eq. (2)), s = sin , c = cos , and sW , cW , tW are the sine, cosine and tangent of the
electroweak mixing angle W .
Since the neutralino mass matrix (4) is symmetric and real, it can be diagonalized by
an orthogonal matrix V 5 . The mass eigenvalues are real but not necessarily positive. They
can be mapped onto positive values by supplementing the rotation matrix to N 5 = 5 V 5
with the diagonal phase matrix ( 5 )kl = 1(i)kl in case of positive (negative) eigenvalues
so that N 5 M5 N 5 is positive diagonal. The physical neutralino states i0 [i = 1, . . . , 4]
are ordered according to ascending mass values while 50 is the predominantly singlino
state.1 They are mixtures
W 3 , H d , H u , S
[i = 1, . . . , 5],
i0 = Nij5 B,
(7)
j
of the U(1), SU(2) gauginos, the doublet higgsinos and the singlino.
The unitary matrix N 5 defines the couplings of the mass eigenstates i0 to other particles. For the neutralino production processes it is sufficient to consider the neutralino
neutralinoZ vertices
0 0
g 5 5
5 5
iL Z j L =
Ni3 Nj 3 Ni4
Nj 4 ,
2cW
0 0
g 5 5
5 5
Ni3 Nj 3 Ni4
Nj 4
iR Z j R = +
(8)
2cW
1 Note that the ordering of the masses according to ascending values is accomplished easily after the diagonalization process is finalized. For the intermediate steps it is however convenient to use the indices i = 1, 2, 3, 4
for the former MSSM type states and i = 5 for the additional state originating from the singlino field as suggested
by the structure of M5 in Eq. (4).
0
fR fR = 2gQf tW N 5 .
iL
i1
87
(9)
The coupling g is the SU(2) gauge coupling, I3 is the SU(2) isospin 3component and Qf
is the electric charge of the fermion f . In Eq. (9) the coupling to the higgsino component,
which is proportional to the fermion mass, has been neglected for light flavors. The more
involved Higgs couplings to the neutralinos are listed in detail in Section 4.
2.2. NMSSM parameter range
In contrast to the Higgs sector only two additional parameters and are introduced
in the NMSSM neutralino sector as compared to that of the MSSM including . Assuming that the fields remain weakly interacting up to the GUT scale, the two couplings are
bounded at the electroweak scale by the inequality
2 + 2 0.7.
(10)
Moreover, the renormalization group (RG) evolution of the couplings points to as
preferential target domain if the evolution starts from a random distribution of the couplings
U , U 2 at
the GUT scale [4].
While v = vu2 + vd2 = 246 GeV is fixed by the Fermi coupling GF , the parameter vs
should be expected in the same range,
vs 15v
(11)
in compliance with the arguments for introducing the NMSSM. A RG analysis of the entire
set of parameters shows that a low value of tan is favored [4]. Current experimental analyses of tan assume MSSM relations for the couplings; they are modified in the NMSSM
and the results in this extended scenario are less restrictive.
Since the size of the doubletsinglet mixing is set by = v/ 2, the mixing interaction
is expected to be small2 compared with the standard supersymmetry scales,
= vs / 2 and/or M1,2 for which values O(1 TeV) are anticipated. As a result, transparent expressions can be found by performing a systematic expansion for small mixing
between the gauginos/doublet higgsinos and the singlino, measured by the small size of
the parameter relative to the other parameters in the mass matrix.
In summary, at treelevel the NMSSM neutralino sector described above has six
free parameters which we choose as and in addition to the MSSM parameters:
{{M1 , M2 , tan , }; , }. Sometimes it is convenient to reexpress , and in
terms of , and vs . The spectrum of the NMSSM neutralino sector will now be analyzed
in detail.
2 is expected to have a lower limit from cosmological arguments; private communication with U. Ell
wanger, see also Ref. [12]. For too small a value of , i.e., very much below the typical scale v/ 2, the amount
of cold dark matter may exceed the measured value of CDM 0.25; detailed analyses are not available yet.
88
M2 mZ sW c2
M1 mZ cW c2 2
=
(13)
det(M ) M M (c s ) M m s
1
12
m2 c
M1 M2 (s c ) M12
Z
M2 = M2 ,
( )
( )
( )
2
2
M12 = M1 cW
+ M2 sW
(14)
det(M ) = M1 M2 2 2 + M12
(s2 + )m2Z .
(15)
The mixing with the singlet alters the MSSM mass eigenvalues m
i [i = 1, . . . , 4] to
O( 2 ),3 and correspondingly the singlet mass
m
5 = .
(16)
1
(V X)2i
m
i m
5
4
i
[i = 1, . . . , 4],
1
(V X)2i
m
i m
5
(17)
with the 4component vector X (0, 0, s , c ). [The eigenvalues are not necessarily ordered sequentially, and, if some of them are negative, the additional phase rotation
3 Note however that small mass differences m
i m
5  may enhance the mixing effects.
89
transforms them to positive physical masses.] Even for small mixing, the 5th eigenvalue m5
may differ significantly from the singlino mass parameter m
5 = if is small. However,
even though the relative shift may be large, the absolute shift remains small, of second
order. Trivially, the eigenvalues fulfill the spur formula
5
mi = M1 + M2 + ,
(18)
i=1
4
Vij j
(19)
j =1
linear in the mixing parameter to first approximation as expected for offdiagonal elements.
Reciprocally, the singlino component in the wave function of 50 is reduced to
1 2
i
2
4
5
V55
1
(20)
i=1
differing from unity only to second order in the mixing as expected for diagonal elements.
As long as the mixing parameter is significantly smaller than the other parameters,
we find that the approximation works remarkably well, as demonstrated in Fig. 1. As an example both the exact numerical solutions and the approximate solutions for the neutralino
masses are shown as a function of for a favored parameter set P of broken PQ symmetry, = 120 GeV with M1 = 250 GeV, M2 = 500 GeV, = 170 GeV and tan = 3. The
exact and approximate solutions agree rather well as long as is less than about 80 GeV,
as the mixing corrections are of second order in .
In Fig. 2 the exact numerical solution (solid) and the approximate solution (dashed)
are compared for the gaugino/higgsino and singlino components, {(N 5 )51 , (N 5 )53 ,
(N 5 )55 }, of the lightest singlinodominated neutralino as a function of for the same
parameter set P. Since the matrix V 5 is in general linear in the mixing term , the approximate solution differs from the exact solution already for smaller values of in the
reference point P in which m
5 = is quite close to the higgsino parameter , though the
characteristic features remain valid up to 40 GeV.
To fully exhaust the potential of our analytical method we perform the complete
NMSSM diagonalization for the two standard limits analyzed in general within the MSSM:
M1,2  and vice versa, both complemented of course by small doubletsinglet mixing
max{M1,2 , }.
3.1. Small singlino mass parameter
The first special analysis should be performed for small singlino mass parameter ,
which implies a slightly broken PQ symmetry 1 as favored by the RG flow of this
90
Fig. 1. The exact numerical solution (solid) and the approximate solution (dashed) for the masses of the five
neutralino states in the NMSSM as a function of for the parameter set P = { = 120 GeV, M1 = 250 GeV,
M2 = 500 GeV, = 170 GeV, tan = 3}. The ordering of the mass spectrum is m5 , m1 , m2 , m3 , and m4 in
increasing mass, i.e., the state 50 is the lightest neutralino for the given parameter set.
Fig. 2. The exact numerical solution (solid) and the approximate solution (dashed) for the gaugino/higgsino and
5 , N 5 , N 5 }, of the lightest singlinodominant neutralino as a function of for
singlino components, {N51
53
55
the same parameter set P as in Fig. 1.
91
coupling in grand unified theories. Due to the small doubletsinglet mixing the structure of
the original MSSM neutralinos i0 [i = 1, . . . , 4] is changed little while the properties of
the 5th neutralino 50 , the lightest for small , are determined jointly by both the singlino
parameter and the mixing parameter .
3.1.1. Large gaugino mass parameters
As a first example, we consider the case with large gaugino mass parameters, i.e.,
M1,2  mZ , .
To begin, the 4 4 diagonalization matrix V defined in Eq. (12) can be parameterized
up to second order according to standard MSSM procedure, cf. Ref. [13], as
122
122
VX
0
VG 0
.
V
(21)
VXT 122
0 VH
0
R/4
m
1
0
0
m
2
+ c
m
3
,
5
M5 M
(24)
m
4 c+
0 0 + c c+
where, in an obvious
notation, zero elements are suppressed for easier reading, and
m2Z 2
s ,
M1 W
m
2 = M2 +
m2Z 2
c ,
M2 W
92
m
3 =
M12
m2 (1 + s2 ),
2M1 M2 Z
m
4 =
M12
m2 (1 s2 ).
2M1 M2 Z
(25)
(0, 0, c , s )T
(26)
and the entire matrix V 5 can be written, up to second order, in the form
1
0
1
2
c
1 42 (1 s2 )
V 0
5
0 1
2
1 42 (1 + s2 ) c+
00
c
c+
1 22
(27)
with zeros suppressed in the upper 4 4 matrix, and antisymmetric in the offdiagonal
elements.
The rotations lead eventually to the diagonal mass matrix, of which the mass eigenvalues
to the desired order are given by
m1 M1 +
m3
m2Z 2
s ,
M1 W
m2 M2 +
m2Z 2
c ,
M2 W
2
M12
m2Z (1 + s2 ) + (1 s2 ),
2M1 M2
2
m4
2
M12
m2Z (1 s2 ) (1 + s2 ),
2M1 M2
2
m5 +
2
s2
(28)
1
0, 0, (c s ), (c + s ) ,
Vi55
2
2
i
V5i5
(0, 0, c , s )i ,
2
.
22
5
V55
1
93
(29)
sW s
cW s
sW c
cW c
,
(30)
1
2
1
VH 122
2
2 m2 /2
sW
Z
0
1
0
2 m2 /2
cW
Z
2
2
2 (1 + s2 )mZ /
,
0
1
2
2
2 (1 s2 )mZ /
,
(31)
respectively, after the higgsino submatrix has been diagonalized by the standard R/4 rotation.
These transformations diagonalize the 4 4 submatrix within the blockdiagonal matrix
5 of the same form as (24), of which the first four diagonal elements are given by
M
m
1 = M1
m
3 =+
m2Z 2
s s2 ,
W
m2Z
(1 + s2 ),
2
m
2 = M2
m2Z 2
c s2 ,
W
m
4 =
m2Z
(1 s2 ),
2
(32)
The mixing between the doublethiggsino and singlino states is then described in an
analytic form by a 4component column vector
(0, 0, c , s )T
(33)
mixing the singlino both with the gauginos and with the doublethiggsinos. The entire
matrix V 5 can be written up to second order in the form, with antisymmetric offdiagonal
94
elements,
5
V
2 c2
2 m2Z sW
2
2M12 2
2 c2
2 m2Z cW
2
2M22 2
2
2 c
22
2
2 c+
22
c+
1
0
c+
2
1 22
V
0
1
(34)
and with the same abbreviations as before in Eq. (24).
The rotations lead eventually to a diagonal mass matrix, consisting of the mass eigenvalues:
m2Z 2
m2 2
sW s2 ,
m2 M2 Z cW
s2 ,
m2
2
m3 + Z (1 + s2 ) + (1 s2 ),
2
2
m2Z
2
(1 s2 ) (1 + s2 ),
m4
2
2
2
m5 +
s2
m1 M1
(35)
with apparent reciprocity in the MSSM subsystem between gaugino and higgsino parameters in comparison with the previous case, but universal modifications from the doublet
singlet mixing.
Correspondingly, to leading order the coefficients of V 5 involving the singlino index 5
coincide with the elements of the doubletsinglet mixing matrix in Eq. (29).
3.2. Large singlino mass parameter
In the alternative extreme, the PQ symmetry is strongly broken if is large and, equivalently, the singlino mass parameter is large, i.e., , , M1,2 . This limit is not favored
by the renormalization group flow from the GUT scale down to the electroweak scale but
cannot be ruled out a priori on general grounds. The new fifth eigenstate, predominantly
composed of the singlino, would in general be the heaviest state, mixed only weakly with
the isodoublets and, as a result, coupling weakly to electroweak gauge bosons and matter
fields.
Applying the approximation method described in Appendix A and the general introduction to this section, the neutralino mass matrix can be transformed into the 4 4 and 1 1
blockdiagonal form by inserting the mixing column vector
(0, 0, s , c )T
(36)
95
in the V 5 matrix Eq. (12). Note that the mixing column vector (36) is directly proportional
to the 4component offdiagonal column vector of the mass matrix (4) unlike the column
vector (26) for a small singlino mass parameter.
From the general analysis it is apparent that the first four neutralino masses, of MSSM
type, are modified to the order 2 / through the higgsino part, as is the 5th neutralino
mass. The mass and the 55 wavefunction are approximately given by
2
(37)
2
,
22
(38)
m5 +
and
5
V55
1
(0, 0, s , c )i
(39)
sW
W 3
Z
Ha = A5 Hd = 0
Hb
Hu
0
S
S
0
sW
cW
0
0
0
0
0
1
2
1
2
0
0
1
1
2
0 B
0 W 3
0
H d .
0 H u
S
1
(40)
96
M 0 0
0
0
0 M mZ 0
0
0
M5 = A5 M5 A5 = 0 mZ 0
(41)
.
0 0 0
0 0 0
This mass matrix generates two twostate mixings between Z and H a0 , and between H b0 and
respectively. The blockdiagonal matrix can be diagonalized by the orthogonal matrix
S,
1
5
V =
(42)
Rg/ h
Rh/s
consisting of two 2 2 rotation matrices Rg/ h and Rh/s ,
cos g/ h sin g/ h
cos h/s
Rg/ h =
,
Rh/s =
sin g/ h
cos g/ h
sin h/s
sin h/s
cos h/s
(43)
2mZ
,
M (M )2 + 4m2Z
+ +
(44)
( + )2 + 42
(45)
97
Fig. 3. (a) The neutralino masses mi , mapped onto positive values, and (b) the tangent values of the mixing
angles g/ h and h/s as a function of the higgsino mass parameter for the parameter set: M = 200 GeV,
= 120 GeV, = 100 GeV.
[i, j = 1, . . . , 5]
(46)
98
are generated by schannel Z exchange, and t and uchannel eL,R exchanges. After appropriate Fierz transformations of the selectron exchange amplitudes [with the electron
mass neglected], the transition matrix element of the production process can be written as
T e+ e i0 j0 =
(47)
Q v e+ u e u i0 v j0 .
,=L,R
The transition amplitudes are built up by the sum of the products of chiral neutralino currents and chiral fermion currents. The four generalized bilinear charges Q correspond to
independent helicity amplitudes, describing the neutralino production processes for polarized electrons/positrons [13]. They are defined by the fermion and neutralino currents and
the propagators of the exchanged (s)particles as follows:
QLL = +
DZ
2
2
sW cW
2
I3 Qf sW
Zij DuL gLij ,
DZ
Qf Zij + DtR gRij ,
2
cW
DZ f
2
QLR = 2 2 I3 Qf sW
Zij + DtL gLij
,
sW cW
DZ
QRL =
(48)
with f = e in the production channel. The first term in each bilinear charge is generated
by Z exchange and the second term by selectron exchange; DZ , DtL,R and DuL,R denote
the schannel Z propagator and the t and uchannel left/righttype selectron propagators
DZ =
m2Z
s
,
+ imZ Z
D(t,u)L,R =
s
(t, u) m2
(49)
fL,R
stam variables for neutralino pair production in e+ e collisions. Finally, the matrices Zij ,
gLij and gRij are products of the neutralino diagonalization matrix elements Nij5
5 5
5 5
Zij = Ni3
Nj 3 Ni4
Nj 4 /2,
f 5
2 2
f
5
f
gLij = I3 Ni2
cW + Qf I3 Ni1
sW If3 Nj52 cW + Qf I3 Nj51 sW /sW
cW ,
5 5 2
gRij = Q2f Ni1
Nj 1 /cW .
(50)
The e+ e annihilation cross sections follow from the squares of the bilinear charges,
2 1/2
e+ e i0 j0 = Sij
2s PS
1
2
1 2i 2j + PS cos2 Q1
1/2
+ 4i j Q2 + 2PS Q3 cos d cos ,
(51)
99
polar angle of the produced neutrinos; and PS = PS (1, 2i , 2j ) denotes the familiar 2body phase space function PS (x, y, z) x 2 + y 2 + z2 2xy 2xz 2yz. The quartic
charges Qi (i = 1, 2, 3) are given by the bilinear charges as follows:
1
QRR 2 + QLL 2 + QRL 2 + QLR 2 ,
4
1
Q2 =
e QRR QRL + QLL QLR ,
2
1
Q3 = QRR 2 + QLL 2 QRL 2 QLR 2 .
4
Q1 =
(52)
+ 6m 0 m 0
e Zij2 ,
i
(53)
where PS = PS (1, m2 0 /m2 0 , m2Z /m2 0 ), with Zij defined in Eq. (50). The widths of the
j
6m m 0
e WLij WRij
(54)
i
100
Fig. 4. The production crosssections of neutralino pairs, {51} (dashed), {55} (thinsolid), {53} (dotted) and {11}
(thicksolid), in e+ e collisions with the centerofmass energy s = 500 GeV as a function of for the
parameter set P [Fig. 1] with meR = 200 GeV and meL = 250 GeV.
where PS = PS (1, m2 0 /m2 , m2W /m2 ) and the bilinear charges WL,R are defined as
j
Nj52 + ULi2
Nj53 ,
WLij = ULi1
2
WRij = URi1
Nj52 URi2
Nj54 .
2
(55)
If 2body decay channels are closed kinematically, the 3body neutralino decays, i0
j0 f f, are generated by schannel (virtual) Z exchange, and t and uchannel sfermion
exchanges. Neglecting fermion masses, the transition matrix element, cf. Ref. [15], is determined by the bilinear charges Q which are related to the bilinear charges Q introduced
for the production, by crossing symmetry as
Q = Q
(56)
u = (p 0 + pf )2 for the decays. [Neutralino decays to charginos and W bosons can be dej
scribed in the same way after obvious redefinitions of the bilinear charges.] Decay widths
and distributions depend on the quartic charges Q 1 , Q 2 and Q 3 defined analogously to
Eq. (52).
101
Fig. 5. The widths, lifetimes and flight distances [broken lines] of the decays 10 50 l + l and lR 50 l
as a function of for the parameter set P [Fig. 1]. The mass of the righthanded slepton is taken to be
ml = 200 GeV > m 0 for the 3body neutralino decays (upper panels) and to be ml = 130 GeV < m 0
R
for the 2body slepton decays (lower panels). The masses of the squarks are assumed to be mqL = 250 GeV
and mqR = 200 GeV. Right: flight distances for s = 500 GeV are shown by broken lines. The kink in the 10
lifetime and flight distance in the upper right panel is caused by accidental cancellations between sfermion and
Z exchange diagrams. [The value of the lower bound, expected from cosmological arguments on is presently
not yet known.]
(ii) At the LHC, cascade sfermion decays, f f i0 , are of great experimental interest.
The width of the sfermion 2body decay into a fermion and a neutralino follows from
g 2 1/2
PS
gfi 2 m2f m2 0 m2f ,
f f i0 =
16mf
i
(57)
where the 2phase space function PS = PS (1, m2 0 /m2 , m2f /m2 ) with f = fL or fR ; the
i
f
5
5
gfL i = 2 I3 Ni2
and gfR i = 2Qf tW Ni1
+ Qf I3 Ni1
tW
(58)
102
in obvious notation.
The reverse decays, neutralino [chargino] decays to sfermions plus fermions, i0 ff
etc, are given by the corresponding partial widths,
1/2
g 2 PS
i0 ff =
gfi 2 m2 0 + m2f m2f
32m 0
i
(59)
with the same couplings as before and PS = PS (1, m2 /m2 0 , m2f /m2 0 ). [Analogous exf
for the 2body slepton decays, respectively.4 The masses of the squarks are assumed to be
mqL = 250 GeV and mqR = 200 GeV. For small mixing the lifetimes of the second
lightest neutralino 10 and the Rsleptons lR can be quite large, giving rise potentially to
macroscopic flight paths [9]. However, cosmological bounds on must be analyzed before any (realistic) experimental conclusions can be drawn. The kink in the 10 lifetime
and flight distance in the upper right panel of Fig. 5 is caused by accidental cancellations
between sfermion and Z exchange diagrams in the decays 10 50 q q and 50 ; these accidental cancellations do not occur [to any significant degree] in the decay 10 50 l + l .
4.3. Decays to a singlino, involving Higgs bosons
Decays involving Higgs bosons can be quite different for different Higgs boson mass
spectra. Following the procedure outlined in Ref. [4] we decompose the neutral Higgs
states into real and imaginary parts as follows:
1
Hd0 = [vd S1 s + S2 c + iP1 s ],
(60)
2
1
Hu0 = [vu + S1 c + S2 s + iP1 c ],
(61)
2
1
S = [vs + S3 + iP2 ],
(62)
2
where the Goldstone states are removed by using the unitary gauge. We then further rotate
these states onto the mass eigenstates, Hi (i = 1, . . . , 3) and Ai (i = 1, 2) labeled in order
of ascending mass, by using the orthogonal rotation matrices5 O H and O A :
Hi = Sj OjHi ,
Ai = Pj OjAi .
(63)
4 In either case 0 or l is the nexttolightest SUSY particle NLSP with just one decay channel open to the
R
1
lightest SUSY particle LSP = 50 .
5 Note that the definitions of these mixings matrices differ slightly from those in Ref. [4], where the scalar
rotation matrix is defined via O T = O H and the pseudoscalar rotation is defined by a rotation through an angle
A .
103
Fig. 6. The Higgs boson mass spectrum as a function of for the parameter set P [Fig. 1] and maximal mixing.
For the purposes of example, the Higgs mass parameter MA is set to 2/ sin 2. Heavy scalar, pseudoscalar and
charged states are nearly mass degenerate: MH3 MA2 MH MA .
The resulting mass spectrum, composed of three scalars, two pseudocalars, and two
charged Higgs bosons, is shown in Fig. 6 as a function of . For the purposes of example,
we have chosen the mass parameter MA (defined to be the heavy pseudoscalar mass in the
MSSM limit) to be 2/ sin 2 567 GeV, setting the scale of the heavy Higgs bosons.
The lighter Higgs bosons consist of two scalars and one pseudoscalar. The lightest scalar
and pseudoscalar in our example are predominantly singlet states, with masses set by the
scale of .
Generally, the width of a 2body neutralino or chargino i decay to a neutralino or
chargino j and a Higgs boson k (Hk or Ak ) is given by
1/2
2
2
PS 2
mi + m2j m2k CijL k + CijRk
[ i j k ] =
16mi
+ 2 mi mj CijL k CijRk + CijLk CijRk ,
(64)
L/R
where PS = PS (1, m2j /m2i , m2k /m2i ) and the left/right couplings Cij k must be specified in each individual case; = 1 for = Hk , H , and 1 for = Ak .
(i) For the decay of a neutralino i0 to a neutralino j0 and a scalar Higgs boson Hk ,
i0 j0 Hk , the couplings are given by,
104
CijRk i0 j0 Hk
H
1 5
5
5
= g Ni2
Ni1
tW Nj53 s + Nj54 c + 2 Ni3
c Ni4
s Nj55 O1k
2
H
1 5
5
5
+ g Ni2
Ni1
tW Nj53 c + Nj54 s + 2 Ni3
s + Ni4
c Nj55 O2k
2
H
1 5 5
5 5
+ Ni3
Nj 4 Ni5
Nj 5 O3k
+ (i j ),
(65)
2
CijL k i0 j0 Hk = CijRk i0 j0 Hk ,
(66)
While the first term in each of the two square brackets in Eq. (65) are reminiscent of the
MSSM couplings i0 j0 h and i0 j0 H , respectively, the other terms are genuinely new in
origin, arising from the extra interaction terms in the NMSSM superpotential.
0 H
The widths for the kinematically allowed decays 40 1,5
1,2 are shown in
0
Fig. 7(left) as a function of . For = 0 the 5 state is decoupled from the other
neutralinos; as is switched on, the coupling, and therefore the decay widths, increase.
The decay widths for 40 50 H1 and 40 50 H2 are comparable, within an order of
magnitude, due to the large 40 mass and the near mass degeneracy of H1 and H2 . With
partial widths of order GeV, these decay modes are in the observable range of branching
ratios.
(ii) Similarly, a 2body neutralino decay to a neutralino and a pseudoscalar Higgs boson, i0 j0 Ak , follows Eq. (64) with the left/right couplings given by
CijRk i0 j0 Ak
A
1 5
5
5
= g Ni2
Ni1
tW Nj53 s + Nj54 c + 2 Ni3
c + Ni4
s Nj55 O1k
2
A
1 5 5
5 5
+ Ni3
Nj 4 Ni5
Nj 5 O2k
+ (i j ),
(67)
2
CijL k i0 j0 Ak = CijRk i0 j0 Ak .
(68)
Again, only the first term in the square brackets is similar to the MSSM coupling i0 j0 A.
0 A are shown in Fig. 7(right)
The widths for the kinematically allowed decays 40 1,5
1
as a function of for our chosen example scenario. In comparison with the scalar case,
many of the decays are kinematically disallowed, only leaving the decays of the heaviest
two neutralinos to 50 and the lightest pseudoscalar (A1 ). Note that pseudoscalar decays
are strongly suppressed compared with the scalar modes and may not be observed easily.
(iii) For completeness, we describe the decays of charginos to a neutralino and charged
Higgs boson i j0 H (i = 1, 2; j = 1, . . . , 5). These follow a similar pattern, now
with the last index of the coupling removed:
1 5
L
0
5
5
5
Cij i j H = gc Ni4
ULj 1 + Ni2
+ Ni1
tW ULj
2 s Ni5 ULj 2 ,
2
(69)
5
5
5
5
CijR i j0 H = gs Ni3
URj 1 Ni2
+ Ni1
tW URj
2 c Ni5 URj 2 .
2
(70)
105
0 H
0
0
Fig. 7. The decay widths for 40 1,5
1,2 (left) and 4 1,5 A1 (right) for the parameter set P[Fig. 1] and
maximal mixing. For the purposes of example, the Higgs mass parameter MA is set to 2/ sin 2.
However, the large mass of the charged Higgs boson means that these 2body decays are
kinematically disallowed for our specific parameter choice.
(iv) It is also possible for Higgs bosons themselves to decay into the singlinodominated
state, via the decays Hi 50 j0 , Ai 50 j0 and H 50 i , if kinematically allowed.
Clearly this is only possible for the heavier Higgs states; the lightest Higgs boson is never
heavy enough to decay in this way. The general form of the width for these decays i
j k (i = Hi , Ai , H ), is given by the crossing of Eq. (64):
2
2
PS 2
[i j k ] = Sj k
mi m2j m2k CijL k + CijRk
16mi
2 mj mk CijL k CijRk + CijLk CijRk ,
1/2
(71)
where PS = PS (1, m2j /m2i , m2j /m2i ) and Sj k = 1 or 1/2 is the usual statistical factor.
Again, = 1 for = Hk , H , and 1 for = Ak . The couplings Cij k are related to
their neutralino decay counterparts in the obvious way:
L/R
L/R
L/R
Cij k Hi j0 k0 = Ckij j0 k0 Hi ,
L/R
L/R
Cij k Ai j0 k0 = Ckij j0 k0 Ai ,
L/R
L/R
Cij H i0 j = Cij i j0 H .
(72)
(73)
(74)
Some of these decays widths are plotted in Fig. 8. Note that a significant fraction of the
Higgs boson H3 and A2 decays go into the invisible channel 50 50 only if the partial decay
width exceeds the range of 1/10 GeV. The upper left panel, showing the partial width
for the decay H3 i0 50 [i = 1, 5], has been allowed to extend down to widths of order
106
105 GeV to show the switching off of the 10 50 H3 coupling at 58 GeV. This is caused
by destructive interference between the different constituent fields in both the Higgs and
the neutralinos, and is directly analogous to the cancellations seen in Ref. [4].
107
The renormalization group flow of the parameters and from the GUT scale down to
the electroweak scale gives rise to strong upper bounds on their values at the electroweak
scale, where small is favored. The qualitative features of the neutralino masses are dependent on how strongly the PQ symmetry of the model is broken by nonzero values;
this is quite accurately described by the approximate analytical solutions.
If the PQ symmetry is slightly broken for small , the qualitative pattern for the particle
spectrum remains intact, except that the lightest singlinodominant neutralino acquires a
mass of the order of the electroweak scale. Thus the model contains four MSSMtype
heavy gaugino/higgsino dominant states and one light singlino dominant state. Since the
couplings to the Z boson can be very much reduced, the NMSSM with a slightly broken
PQ symmetry constitutes a valid scenario.
In contrast, a strongly broken PQ symmetry, though disfavored by the flow of the
couplings from the GUT scale down to the electroweak scale, could provide an extra moderately heavy neutralino state, which is only weakly coupled to the Z and (s)fermions. Such
decoupled scenarios would be more difficult to distinguish the NMSSM from the MSSM.
The methods described for real mass matrices, can readily be generalized to complex
5 5 neutralino mass matrices. In analogy to the MSSM, cf. Ref. [13], the procedure is
mathematically based upon the Takagi theorem [16] which assures that any complex symmetric matrix can be diagonalized by an orthogonal transformation. The technical details
are outlined in an addendum to the Appendix A. The system can then be solved analytically by following the same steps of approximations which have been applied to the real
mass matrices.
Acknowledgements
The work of S.Y.C. was supported in part by the Korea Research Foundation Grant
(KRF2002070C00022) and in part by KOSEF through CHEP at Kyungpook National
University.
108
where A is a 4 4 matrix incorporating elements of the order of the large SUSY scale, B
is a scalar and X is a 4component vector of order .
Performing an auxiliary orthogonal transformation O defined by the matrix,6
144 12 T
O=
(A.2)
T
1 12 T
with the mixing column vector = [A B]1 X, the mass matrix takes block diagonal
form, accurate to order 2 :
0
A + A
T
,
OHO =
(A.3)
0
B + B
where
A =
1
[A B]1 , XX T ,
2
B = X T [A B]1 X.
(A.4)
Both s are of order 2 . If A is diagonal, only the diagonal elements of A need be kept as
rediagonalization would change the mass matrix (A.16) and the orthogonal matrix (A.2)
only beyond the order considered in the systematic expansion. We note that the correction
terms satisfy the simple sum rule Tr A + B = 0.
If B is also as small as the elements of the low vector X, the mixing column vector
= A1 X and the correction terms A and B are further simplified to be
1 1
(A.5)
A , XX T
and B = X T A1 X.
2
On the contrary, if B is much larger than the other parameters, the mixing column vector
= X/B and the correction terms take the following simple form
A =
A = XX T /B,
B = X T X/B.
(A.6)
Both these approximations have been used in the derivation of all the mass and mixing
formulae discussed earlier in the report.
The diagonalization of the mass matrix M5 ,
M X
M5 =
(A.7)
5
XT m
with M being the 4 4 MSSM mass submatrix, makes use of the blockdiagonalization
method in the following way:
(1) In the first step M is diagonalized by applying the wellelaborated MSSM procedure
MD = V MV T
(A.8)
109
1
5 V X = (M m
5 )1 X.
= V : = V T MD m
(A.9)
Note that is of order quantum mechanically enhanced however if mass differences
m
i m
5  are, accidentally, small.
(3) The blockdiagonalization affects the upper left diagonalized 4 4 submatrix MD
only beyond second order and likewise the orthogonal matrix V 5 beyond the second and
first order considered, respectively, for on and offdiagonal elements. As a result, we obtain the final diagonal form of the mass matrix as
5
5T
MD
5 V M5 V
with
V5
(A.10)
144 12 (V )(V )T
(V )T
(V )
1
1 2 (V )T (V )
V
0
0
111
(A.11)
in obvious notation. While the rightmost part solves the MSSM diagonalization, the leftmost part diagonalizes the NMSSM under the assumption of small doubletsinglet mixing.
Addendum
This procedure of diagonalizing the neutralino mass matrix can readily be generalized
to complex mass matrices when the gaugino and higgsino matrix elements are no longer
assumed to be real.
The mathematical key to this generalization is the Takagi theorem [16] which assures
that any complex symmetric matrix can be diagonalized by an orthogonal transformation
of the form
N HN = diag(mi ),
(A.12)
with the unitary matrix N acting on the neutralino Majorana fields. This is widely used for
complex extensions of the MSSM mass matrix, cf. Ref. [13].
The matrix N can be decomposed, N = MU , into the unitary matrix U blockdiagonalizing the hermitian matrix H H and the diagonal phase matrix M. Writing
A2 X2
H H =
(A.13)
X2 B2
with A2 = A A + X X T , B2 = B B + X X and X2 = A X + X B, the vector 2 in
2
144 12 2 2
U=
(A.14)
2
1 12 2 2
is defined, in analogy to (A.2), by
2 = [A2 B2 ]1 X2 .
(A.15)
110
The matrix elements of the blockdiagonalized matrix can easily be derived from
0
A2 + A2
,
U H HU =
(A.16)
0
B2 + B2
where
1
[A2 B2 ]1 , X2 X2 ,
2
The matrix M is built up by the phases
e2ii = U HU ii /mi .
A2 =
B2 = X2 [A2 B2 ]1 X2 .
(A.17)
(A.18)
From this point on, the same steps can be followed as for the real mass matrix.
References
[1] P. Fayet, Phys. Lett. B 64 (1976) 159.
[2] H.P. Nilles, Phys. Rep. 110 (1984) 1;
H.E. Haber, G.L. Kane, Phys. Rep. 117 (1985) 75.
[3] P. Fayet, Nucl. Phys. B 90 (1975) 104;
M. Drees, Int. J. Mod. Phys. A 4 (1989) 3635;
J. Ellis, J.F. Gunion, H. Haber, L. Roszkowski, F. Zwirner, Phys. Rev. D 39 (1989) 844, and other references
quoted therein.
[4] D.J. Miller, R. Nevzorov, P.M. Zerwas, Nucl. Phys. B 681 (2004) 3, hepph/0304049.
[5] U. Ellwanger, J.F. Gunion, C. Hugonie, NMHDECAY, hepph/0406215.
[6] S.A. Abel, S. Sarkar, P.L. White, Nucl. Phys. B 454 (1995) 663;
S.A. Abel, Nucl. Phys. B 480 (1996) 55;
C. Panagiotakopoulos, K. Tamvakis, Phys. Lett. B 446 (1999) 224;
C. Panagiotakopoulos, K. Tamvakis, Phys. Lett. B 469 (1999) 145;
A. Dedes, C. Hugonie, S. Moretti, K. Tamvakis, Phys. Rev. D 63 (2001) 055009;
C. Panagiotakopoulos, A. Pilaftsis, Phys. Rev. D 63 (2001) 055003.
[7] M. Dine, W. Fischler, M. Srednicki, Phys. Lett. B 104 (1981) 199;
H.P. Nilles, M. Srednicki, D. Wyler, Phys. Lett. B 120 (1983) 346;
J.M. Frere, D.R. Jones, S. Raby, Nucl. Phys. B 222 (1983) 11;
J.P. Derendinger, C.A. Savoy, Nucl. Phys. B 237 (1984) 307;
A.I. Veselov, M.I. Vysotsky, K.A. TerMartirosian, Sov. Phys. JETP 63 (1986) 489, Zh. Eksp. Teor. Fiz. 90
(1986) 838 (in Russian);
R.B. Nevzorov, M.A. Trusov, J. Exp. Theor. Phys. 91 (2000) 1079, Zh. Eksp. Theor. Fiz. 91 (2000) 1251
(in Russian), hepph/0106351.
[8] F. Franke, H. Fraas, A. Bartl, Phys. Lett. B 336 (1994) 415;
U. Ellwanger, M. Rausch de Traubenberg, C.A. Savoy, Phys. Lett. B 315 (1993) 331;
U. Ellwanger, M. Rausch de Traubenberg, C.A. Savoy, Nucl. Phys. B 492 (1997) 21;
S.F. King, P.L. White, Phys. Rev. D 52 (1995) 4183;
F. Franke, H. Fraas, Z. Phys. C 72 (1996) 309;
F. Franke, H. Fraas, Int. J. Mod. Phys. A 12 (1997) 479;
B. Ananthanarayan, P.N. Pandita, Int. J. Mod. Phys. A 12 (1997) 2321;
S.P. Martin, Phys. Rev. D 62 (2000) 095008;
M. BasteroGil, C. Hugonie, S.F. King, D.P. Roy, S. Vempati, Phys. Lett. B 489 (2000) 359;
U. Ellwanger, C. Hugonie, Eur. Phys. J. C 25 (2002) 297;
F. Franke, S. Hesselbach, Phys. Lett. B 526 (2002) 370;
U. Ellwanger, J.F. Gunion, C. Hugonie, S. Moretti, hepph/0305109.
111
Abstract
We show that in string theory or supergravity with supersymmetry breaking through combined
Fterms and FayetIliopoulos Dterms the masses for charged scalars and fermions can be hierarchically split. The mass scale for the gauginos and higgsinos of the MSSM is controlled by the gravitino
mass m3/2 , as usual, while the scalars get extra contributions from the Dterms of extra Abelian
U (1) factors, which canmake them much heavier. The vanishing of the vacuum energy requires that
their masses lie below m3/2 MPl , which for m3/2 = O(TeV) sets a bound of 101013 GeV. Thus,
scalars with nonvanishing U (1) charges typically become heavy, while others remain light, producing a spectrum of scalars with masses proportional to their charges, and therefore nonuniversal.
This is a modification of the split supersymmetry scenario, but with a light gravitino. We discuss how
FayetIliopoulos terms of this size can arise in orientifold string compactifications with Dbranes.
Furthermore, within the frame work of Dterm inflation, the same vacuum energy that generates the
heavy scalar masses can be responsible for driving cosmological inflation.
2005 Elsevier B.V. All rights reserved.
PACS: 12.60.Jv; 04.65.+e; 11.25.Mj
113
1. Introduction
The conventional approach to supersymmetry breaking in models of supergravity
(SUGRA) is to assume some form of spontaneous breaking in a hidden sector, mediated
to the visible sector with contains the MSSM via gravitational interactions [1]. The overall
mass scale is then set by the gravitino mass m3/2 , and all other masses for squarks, sleptons,
gauginos and higgsinos come out roughly proportional to it, by demanding the cancellation
of the vacuum energy. Low energy supersymmetry, as required by a natural explanation of
the Higgs potential, fixes m3/2 to the electroweak scale. Now, recently it was argued [2]
that the fine tuning problem of the Higgs mass may be insignificant compared to the fine
tuning of the cosmological constant, and that an anthropic selection mechanism (see, e.g.,
[3]) may then involve actual fine tuning of MSSM parameters. The mass pattern that was
proposed under the name of split supersymmetry [4] has all the MSSM fermions at the
electroweak scale, whereas all scalars, except for the one fine tuned Higgs doublet, get
ultraheavy at a high mass scale. This scenario has attracted some attention recently [5].
More concretely, the challenge for model building is to keep the gauginos and higgsinos
light, while letting the scalars become very heavy. The motivation for this originates from
supersymmetric grand unification, even without low energy supersymmetry in the usual
sense, and the model is designed to keep the merits of gauge coupling unification as in the
MSSM. Here, we pursue the perspectives of such patterns in the mass spectrum in the context of SUGRA and string theory models, based on the paradigm of spontaneous breaking
in a hidden sector.
Given the above mentioned relations that govern gravity mediated supersymmetry
breaking, it seems very hard to achieve hierarchically split mass scales. If all masses are
proportional to m3/2 , there is no room for flexibility. There is, however, a loophole to the
argument, which has not so far been explored in the conventional approach in any depth,
probably because it quickly leads to large masses, which were thought unacceptable. It
consists of assuming not only auxiliary Fterms but also Dterms to be generated. In global
supersymmetry, this is wellknown under the label of supersymmetry breaking mediated
by an anomalous U (1), where large masses can be avoided [6,7], as will be seen later. The
mechanism basically adds a FayetIliopoulos (FI) term for an extra anomalous U (1) gauge
symmetry, independent of the Fterms which may also be present. In local supergravity, the
two contribute both to the vacuum energy, and thus get tied together at roughly the same
scale. Still, the contribution to scalar masses can be very different, since the Fterms are
mediated via gravity, while the Dterms are mediated via gauge interactions, which opens
up the possibility to have hierarchically split mass scales, splitting scalars charged under
the relevant U (1), from all other fields, i.e., gauginos, higgsinos as well as scalars not
charged under the relevant U (1)s. This is not quite along the lines of high scale supersymmetry breaking, as advocated in split supersymmetry, where the gravitino mass itself was
assumed at the high scale, and the main difficulty lies in keeping the gauginos and higgsinos lighter than m3/2 .1 Instead, we propose extra contributions to the masses of charged
1 To achieve this, it is usually assumed that gravity mediation of gaugino masses can be avoided first of all.
Furthermore, one has to find ways to suppress contributions from anomaly mediation [8]. Since the latter is not
114
scalars, which make them heavier than m3/2 , while the fermions including the gravitino
remain light.
The purpose of this paper is to study the confluence of this combined approach with
F and Dterms in supergravity and string theory models. By this we mean that we assume
that some hidden sector dynamics generates F and Dterms at some scale of supersymmetry breaking, but we do not present a full dynamical model, how this happens. Instead,
we analyze the various scenarios that can emerge in the visible sector, and in particular
identify classes of models which generically lead to hierarchically split mass scales.
1.1. FIterms in global supersymmetry
In many models of grand unification, compactification of higherdimensional supergravity or string theory, the minimal gauge symmetries that can be achieved at low energies
involve various extra Abelian U (1) gauge factors beyond the Standard Model gauge group,
i.e., the total gauge symmetry is SU(3)C SU(2)L U (1)n , where among the U (1) there
is also the hypercharge. In string theory it often happens that some of the extra factors are
actually anomalous, the anomaly being canceled by a (generalized) GreenSchwarz (GS)
mechanism, in which the gauge boson develops a Stueckelberg mass and decouples (see,
e.g., [10]). In any case, it is then permissible to add FIterms a Da to the supersymmetric
Lagrangian, one for each U (1)a . The Dterm potential in global supersymmetry is
2
g2
g2
a 2
a
i 2
D =
Qa fi  + a
VD =
(1)
2 a
2
a
a
i
and thus a > 0 leads to the formation of a condensate for at least some field fi of negative
charge Qia < 0. This breaks the gauge symmetry spontaneously, but supersymmetry can be
restored at the minimum if Da = 0.2 In the MSSM it is usually assumed that the FIterm
of the hypercharge is absent or very small, and does not play a role in the Higgs potential.
Whenever the auxiliary field obtains a nonvanishing expectation value Da = 0, supersymmetry is broken, and mass terms are generated for all the charged scalars,
m2i =
(2)
ga2 Qia Da ,
a
where it is now assumed that the charges are positive for the MSSM fields, to avoid breaking of the Standard Model gauge symmetries. This scenario can be achieved in a global
supersymmetric model with a single extra U (1)X by adding two scalars to the MSSM,
singlets under SU(3)c SU(2)L U (1)Y , but with charges 1 under U (1)X [6]. The
crucial ingredient is an interaction in the superpotential of the form
W = m + .
(3)
fully understood within string theory (see [9]) and we include the effects of gravity mediation anyway, we will
not consider anomaly mediation in the following.
2 In string theory, the FIparameter is usually a function of the moduli, = (T , T ). Therefore, turning on
a
a I I
the FIterm can correspond to a flat direction fi 2 = a in the total potential, for Qia = 1.
115
2 2
QiX fi 2 + + + X
2
(4)
+ = 0,
2
m2
= X 2 ,
gX
(5)
and
DX =
m2
,
2
gX
F + = m X + .
(6)
Gaugino masses may originate from higherdimensional operators, and are suppressed by
powers of MPl ,3
m
X
1
F + + F + m 2 .
2
MPl
MPl
(7)
cA B 3
X
+ mX B F
,
(8)
where cA and mX are two coupling constants, 3 the ChernSimons (CS) 3form, and
F X the gauge field strength of the relevant U (1)X . Now, is the imaginary part of some
3 The Planck mass M is defined so that M = 1 = (8 G)1/2 = 2.4 1018 GeV.
Pl
Pl
116
complex scalar in a chiral multiplet. For the heterotic string, the only such scalar that
participates in the GS mechanism is the dilatonaxion field S= =0 = s + i (see [11]
= ln(S + S).
g 2 tr(QX )
mX
X
.
s
192 2
(9)
In the presence of an interaction (3) the scalar fields fi charged under the U (1)X acquire
masses given by
m2i = QiX m2 O(TeV).
(10)
The remarkable feature of Eq. (10) is that it is independent of the FIparameter. Thus, the
vector boson gets a mass of the order of mX , which is close to the Planck scale, whereas
the charged sfermions and gauginos remain massless at the high scale, and get masses
of the order of the electroweak scale. This is the standard scenario of supersymmetry
breaking via an anomalous U (1) with GS mechanism.
1.3. Multiple anomalous U (1) symmetries and Dbranes
Orientifold string compactifications [13] usually involve more than one anomalous U (1)
factor. While in the heterotic string it is only the axionic partner of the dilaton that participates in the GS anomaly cancellation, now all the axionic scalars that follow from the
reduction of the RR forms from ten dimensions can do so [14]. In orientifold compactifications of type IIB strings, the relevant RR scalars originate from the twisted sectors.
The FIparameters a are then functions of the expectation values of the real parts of these
twisted scalars, instead of the dilaton s. For a special example of this class of models,
in a toroidal orbifold T6 /Z3 , it was shown, that no FIterm was generated at oneloop,
consistent with the fact that the twisted scalars vanish at the orbifold point [15]. As another class of models, orientifolds with intersecting (type IIA) or magnetized (type IIB)
Dbranes have been studied extensively in the recent past, most prominently for their very
attractive features to produce Standard Model or MSSM like gauge groups and spectra. For
these, untwisted RR scalars participate in the GS mechanism. Again the FIterm at treelevel (i.e., from a disc diagram, or the dimensional reduction of the BornInfeld action) is
proportional to the modulus that combines with the axionic scalar from the GS mechanism
into a complex scalar. The GS couplings analogous to Eq. (8) now involve many scalars
I
I
I
a
(11)
cA
B
3 +
mIa B
F
,
I,A
a,I
117
Depending on the precise form of the Khler potential, a FIterm will be generated from
this expression, that will depend on the vacuum expectation value of tI . The simplest expression would be
mIa tI .
2 a
(13)
I
It was stressed in [14] that the a given by Eq. (13) can in principle be of any size, as
opposed to the result for the heterotic string case. Another important observation is to note,
that the FIterms are not necessarily tied to anomalous gauge symmetries, but only the
nonvanishing Stueckelberg coupling mIa = 0 has to exist.4
2. FIterms in supergravity and string theory
We now examine the patterns of soft supersymmetry breaking that arise from an effective string theory Lagrangian with one or more FIterms, motivated by the appearance of
multiple U (1) factors in orientifold models, that can develop FIterms.
2.1. The vacuum energy
The degrees of freedom of the model are assumed to be given by the fields fi of the
MSSM, the extra gauge vector multiplets for the U (1)a , the moduli TI of the gravitational
sector, plus the axiondilaton S, which includes the fields that participate in the GS mechanism, producing Stueckelberg masses for the gauge bosons and FIterms. Furthermore, we
can add extra fields like the of the globally supersymmetric model, with charges a
under U (1)a . The effective scalar potential is given by the N = 1 supergravity formula
[16]
V = 4 eG GM N GM GN + 3 + VD ,
(14)
with
G = 2 K ln 6 W W ,
(15)
4 As mentioned earlier, we shall here not attempt to model the dynamics of the hidden sector in any detail,
and therefore also do not try to answer, how the FIparameters are generated dynamically. Since they are modulidependent functions, a meaningful answer would have to address the modulistabilization at the same time.
118
where indices M, N run over all fields. We define the dilaton and moduli fractions of the
vacuum energy by
1
S 2 = GS S GS GS ,
3
1
I 2 = GI I GI GI ,
3
(16)
and  2 in a similar fashion. It also turns out to be useful to introduce the following
combinations
2
y = .
m = e K/2 m,
x = + ,
m3/2 = 1 eG/2 ,
(17)
Similarly, all other fields are made dimensionless. Imposing the restriction on the model
that the vacuum energy vanishes (through fine tuning) one has
S 2 +
I 2 + + 2 +  2 +
1
2
3m23/2 MPl
g2
a
Da2 = 1,
(18)
(19)
where we ignore prefactors involving the Khler potential. As long as m3/2 O(TeV),
Eq. (19) implies roughly Da 1/2 101013 GeV, which is the usual intermediate supersymmetry breaking scale in SUGRA models. The masses that are generated by the Fterms
are given by FI /MPl O(TeV), whereas the Dterms would be able to produce much
1/2
larger mass terms proportional to Da . This means that the mass parameter in the superpotential (3), which had to be fine tuned
to the electroweak scale, can now also be assumed
as large as the intermediate scale, m m3/2 MPl . Further, we note that the scenario with
a Planck scale sized FIparameter, as is unavoidable for the heterotic string in the presence
of an anomalous U (1), is only consistent with a Planck scale sized gravitino mass. In orientifold Dbrane models, as mentioned earlier, the FIparameter can in principle be of any
value, and the problem does not occur.
In scenarios with split supersymmetry, the gravitino mass itself is not restricted to a
small value. However, gravity mediation generically leads to a contribution to gaugino and
higgsino masses which is proportional to the gravitino mass, and therefore m3/2 O(TeV)
is unavoidable in the present context. This then really puts an upper bound 101013 GeV
on the high mass scale allowed for the sleptons and squarks.
Before going into the various scenarios, let us first assemble a few general definitions
and formulas for the supergravity version of the model of supersymmetry breaking mediated by one or many anomalous U (1). For the Khler potential K we write
K = Khid (TI , TI ) + Ki (TI , TI )fi f + K+ (TI , TI ) + + + K (TI , TI ) ,
where we have now included S among the TI . The gauge kinetic functions are modulidependent,
fa = fa (TI ),
1
= (fa ),
ga2
(20)
119
but independent of . For the superpotential we assume the following factorized form
W = WMSSM (fi ) + Whid (TI ) + W + , = m + + W0 ,
(21)
where WMSSM contains the quark, lepton and Higgs fields, Whid contains the fields of
the hidden sector which break supersymmetry spontaneously by generating auxiliary field
components for FI = DI Whid , while W is still given by Eq. (3).5 With this, the total
Dterm potential VD is given by
2
g2
2
2
a
MSSM
+
Qia Ki fi 2 + a K+ + a K + a .
VD = VD
2
a
i
(22)
MSSM is the Dterm arising from the SU(2) U (1) sector, which will not be
Here VD
L
Y
important. The standard expressions for the soft breaking terms that originate from the
Fterms only, are [17]
m =
1
F I I fa ,
2(fa )
(23)
(24)
for scalar masses. For FI m3/2 MPl , both masses are of the order of m3/2 .
2.2. The simplest model
The simplest model that already displays the effects of the FIterms is given by assuming
one or more FIterms being generated by extra U (1)a gauge factors, and only including the
MSSM fields with arbitrary positive charges, but leaving out the extra fields .6 In that
case, supersymmetry is broken, and the Dterms are trivially given by
ga2 2 ga2 2
D = a .
2 a
2
Together with potential Fterms, they generate masses
m2i = m2gr,i +
ga2 Qia Ki a ,
(25)
(26)
a
5 This implies an assumption on the absence of any coupling among MSSM fields and in the superpotential, which may be problematic in the context of a concrete model. Furthermore, we also ignored any
crosscoupling in the Khler potential, where in principle the modulidependence of the various coefficients could
also involve .
6 It may sound very restrictive to allow only positive charges here, and it really would be in any reasonable
model derived from a GUT or string theory. However, there are wellknown cases in string theory compactifications, where higher order corrections in the derivative expansion of the effective action (such as the BornInfeld
Lagrangian) lead to a lifting of tachyonic negative masses in the presence of FIterms, even if some fields have
negative charge. We will come to explain this in some more detail later.
120
for all charged scalars. On the other hand, the masses of gauginos (and higgsinos) are unaffected by the Dterm, at least at leading order, and would be dictated by the Fterms
to be of order m3/2 . In a scenario with m3/2 O(TeV), and Da 101013 GeV for at
least one FIterm, this provides a hierarchical split of energy scales, however, the high
scale cannot move up all the way to the Planck scale. The most simple charge assignment would give positive charges of order one to all MSSM fields, except the Higgs,
and thus the sfermion sector of the MSSM would become very heavy and undetectable
at LHC.7 There may of course also be other interesting patterns to consider, such as different charges for the three generations, different charges for different SU(5) multiplets,
which would leave the gauge coupling unification intact, or different charges for left and
righthanded fields, which could be more easily realized in certain Dbrane models from
string theory.
2.3. The full potential with a single U (1)X
The essence of the model of [6], where an anomalous U (1) with its FIterm is responsible for supersymmetry breaking, is the scalar condensate for that cancels the
contribution of the FIparameter to the Dterm, up to small electroweak scale sized contribution, given the interaction (3) in the superpotential. Since such a condensate breaks
the gauge symmetry, one has to add extra chiral multiplets to the MSSM, neutral under the MSSM gauge symmetries. The model of [6] is within the framework of global
supersymmetry, and it is of obvious interest to study the embedding of this original model
into supergravity. So, we focus first on the case when there is just one extra U (1)X , which
develops a FIterm. The potential can be written as
2
2
) + X K+ x2 K y2 + X ,
+ 4 m(K+ + K )(xy W + x yW
2
(27)
Vhid = e K KI J DI W DJ W ,
(28)
for the contribution of the hidden sector. It is essential for the fine tuning of the vacuum
energy. The two minimization conditions are x V = y V = 0. To keep things simple, we
now also set all relevant phases to zero, i.e., we treat x, y, W as real. Then one gets the
following two equations
m23/2 2
m2
2
K+ x 2 gX
DX + 4 VF + 2 x + xy 2 (K+ + K ) + 2 K+
x
MPl
MPl
m m3/2
m2
m23/2 2
2
K y 2 gX
DX + 4 VF + 2 y + x 2 y(K+ + K ) + 2 K
y
MPl
MPl
m m3/2
121
(29)
2 D 2 , where V and V
where VF and DX are defined so that V = VF + 12 gX
F
hid are related
X
by
2
2
2
VF = Vhid + m2 MPl
x + y2 + m3/2 2 MPl
K+ x2 + K y2 3
2
(K+ + K )(xy m
3/2 + x ym
3/2 ) .
+ m MPl
(30)
Note that the redefined mass parameters are fielddependent. In the following, we also
restrict to canonically normalized fields, and set K+ = K = 1. We have illustrated the
full potential in Fig. 1, using values
4 Vhid = 3,
gX = 200,
2
6 e K W0 2 = 1,
2
2 e K m2 = 1,
2 X = 101 .
2 m2 ,
This corresponds to setting some parameters equal, dividing the total potential by MPl
3/2
and rescaling the Planck mass by many order of magnitude, so that MPl /m3/2 = 10, just
to suppress the very steep part of the potential. It turns out, that for the relevant range of
parameters x = 0 = y at the minimum.
Since the minimization of the Fterm potential basically consists of balancing terms
that scale like m2  2 , m23/2  2 or with inverse powers of MPl , with the negative term
2 , it is intuitively clear that x and y will be roughly bounded through the most
m23/2 MPl
2
dangerous term by m3/2 m1
or O(1). This is of course only valid as long as  
2
X , otherwise   gets tied to X . In any case one has that F /MPl m3/2 , which is
sufficient for light gauginos.
122
(31)
(33)
Vhid
,
2
2
m3/2 MPl
x 2 y 2 = X .
(34)
(35)
which gives
x = 0,
y 2 = X .
(36)
The values of Eq. (36) give the minimum of the potential that is in some sense analogous
to the case Eq. (32).
2.5. Multiple U (1) gauge symmetries
With multiple FIterms in the potential, it is clear that supersymmetry breaking can
occur more generically. We have seen above, that FIterms that cannot be canceled by
scalar condensates lead to large scale masses, whereas a scalar condensate with a single
FIterm was able to lower the masses to the electroweak scale m3/2 , similar to [6]. If
multiple FIterms now come accompanied by the same number of extra charged fields to
develop condensates, then the Lagrangian would just be a sum of identical copies of the
123
one of the previous sections, and nothing new is found. An interesting case arises, when
there is a mismatch, and not all the FIparameters can be absorbed by fields like , and
thus some large masses can be generated. We now analyze the situation, when there still is
only a single set of extra charged fields , that takes a nonvanishing expectation value,
but there are multiple U (1)a gauge symmetries, under which it is charged. The potential is
only modified by summing over Dterms,
4 VD =
g2
2
a
a K+ x2 a K y2 + a ,
2
a
again making a dimensionless by a 2 a . Regarding the vacuum energy cancellation, the new minimization conditions read
m23/2 2
m2
K+ x 2
a ga2 Da + 4 VF + 2 x + xy 2 (K+ + K ) + 2 K+
x
MPl
MPl
a
m m3/2
+
y (K+ x)2 + (K y)2 + 2K 3 = 0,
2
MPl
m23/2 2
m2
2
2
4
K y
a ga Da + VF + 2 y + x 2 y(K+ + K ) + 2 K
y
MPl
MPl
a
m m3/2
+
(37)
x (K+ x)2 + (K y)2 + 2K+ 3 = 0.
2
MPl
It is not necessary to study the general solution here, since the multiple FIterms already
break supersymmetry without the superpotential (3), and one can therefore set m = 0.
With K+ = K = 1 one can easily solve for x, y, finding
2
Vhid
a ga a
2
2
x2 + y2 = 2 2
(38)
,
x
y
=
2.
2
2
m3/2 MPl
MPl
a ga
Thus, in a generic situation, where not all a are equal, supersymmetry will be broken,
and Da a m3/2 MPl . In this case, the vacuum energy is given by the negative Fterm
contribution plus Dterms, and can be fine tuned even without Vhid . The Fterms are given
by F = 2 W0 m3/2 MPl . The contributions to the scalar masses now look like (with
dimensionful a )
ga2 Qia Da
ga2 Qia a + ,
m2i = m2gr,i +
(39)
a
where the terms in parentheses can be of the same order of magnitude, but the gravitymediated contributions are negligible. This realizes the split supersymmetry scenario, if
a 101013 GeV, for at least two FIparameters, but with a low gravitino mass m3/2
O(TeV).
It may also be interesting to note that at the same time all other soft breaking parameters
will get modified in the presence of FIterms and scalar field condensates. This happens
2
through the prefactor e K in the total potential. For instance, the bilinear couplings B
124
1
2
0
1/2 e 2 (x2 +y2 )+ ,
B
m3/2 e K/2 = m3/2 (S + S)
(40)
which may lead to extra suppression factors, depending on the model. Further, we note that
the Higgs mixing parameter that enters in the superpotential in the form H1 H2 , where
H1 and H2 are the two Higgs doublets of MSSM, remains essentially unaffected. This is
so because in string/supergravity scenarios one expects the term to arise from the Khler
potential using a Khler transformation after the spontaneous breaking of supersymmetry
has taken place. The Khler transformation is sensitive to the Fterm breaking and not the
Dterm. Thus, one expects the same mechanism that produces a term of electroweak
size for supergravity models to hold in this case as well.
2.6. Scenarios with partial mass hierarchies
To summarize, there are several scenarios that are possible, which would lead to different patterns in the mass spectrum: (i) There is only one FIterm, and an extra scalar field
beyond the MSSM fields which forms a vacuum expectation value. Here all the soft
scalar masses will be of electroweak size. (ii) In the case of two or more U (1)a with nonvanishing a , and the charges Qia nonzero and sufficiently generic, the scalar masses are
of the order of a , while the gauginos stay light. One can thus get a hierarchical splitting of
scalar and fermion masses. (iii) One may also achieve hybrid scenarios, when only some of
the charges Qia are nonvanishing, or such that the high mass terms just cancel out. Below
we consider two specific scenarios with partial splitting of scalars. The implications of this
partially split scenario will be very different from the usual SUGRA scenario and also from
the high scale supersymmetry scenario of split supersymmetry.
2.6.1. Model I: 2 + 1 generations
As the first model we consider the case with nonvanishing charges for the first and
second generation of squarks and sleptons, but vanishing charges for the third generation.
In this circumstance the former will develop super heavy masses and will not be accessible
at the LHC, as opposed to the third generation. The above implies that the dangerous
flavor changing neutral currents will be automatically suppressed. Some of the signals of
this scenario will be very unique, such as the decay of the gluino (g).
In SUGRA its decay
modes are as follows
ui u i j0 , di di j0 ,
di di
g u i u i ,
(41)
ui di k , di u i k+ ,
where i = 1, 2, 3 are the generational indices, j = 1, 2, 3, 4, and k = 1, 2 for neutralinos and charginos. However, for the case of the model under consideration the decay
through the first two generations is highly suppressed, and the gluino decay can only occur
through third generation squarks via the modes g bb j0 with admixtures of g bt k+ ,
g t b k , and g t t j0 , depending on the gluino mass. There will be no contribution
to the anomalous magnetic moment of the muon at the oneloop level. The following set
of phases can arise: , i , and A , where is the phase of the Higgs mixing parameter
125
, i (i = 1, 2, 3) are the phases of the SU(3)c , SU(2)L and U (1)Y gaugino masses, and
A is the phase of the common trilinear coupling for the third generation scalars (however,
it should be kept in mind that not all the phases are independent and only certain combinations enter in physical processes). Because of the superheavy nature of the first two
generations, the oneloop contributions to the electric dipole moment (edm) of the electron
and of the neutron are highly suppressed. However, there can be higher loop corrections
to the edms. Specifically, the neutron edm can get a contribution from the CP violating
dimension six operator. Unification of gauge coupling constants at the oneloop level will
remain unchanged, although at the twoloop level there will be effects from the splittings.
In this model the staus can be light and thus coannihilation of the LSP neutralinos with
the staus can occur, allowing for the possibility that the neutralino relic density could be
consistent with the current data. Finally, proton decay from dimension five operators would
not arise via dressings from the first and second generation squarks and sleptons, but can
arise from the dressings of the third generation sfermions.
2.6.2. Model II: 5 + 10 split spectrum
As another illustrative example, we consider a model where the squarks and sleptons
belonging to a 10 of SU(5) have nonvanishing charges and high scale masses, while the
squarks and sleptons belonging to a 5 have vanishing charges. In this case, the only light
scalars aside from Higgs bosons will be diC , eLi , i (i = 1, 2, 3). In this model, unlike the
case of model I, there is a oneloop supersymmetric contribution to the muon anomalous
magnetic moment g 2. The CP phases in this model consist of and i (i = 1, 2, 3).
There also is a oneloop supersymmetric contribution to the edms of the electron and of the
neutron. Further, the decays of the gluino, the chargino and the neutralino can occur only
via a smaller subset of states and thus their decay widths will be relatively smaller, though
they will still decay within the detection chamber. Finally proton decay through dimension
five operators will be highly suppressed in this model and the dominant decay will occur
through the usual vector boson interactions.
126
The gauge group for any single stack of Na Dbranes is U (Na ) = SU(Na ) U (1)a
(only sometimes an Sp(Na ) or SO(2Na ) subgroup thereof), and thus involves extra Abelian
U (1)a factors, when the Standard Model is engineered. In the first models that were constructed to reproduce the nonsupersymmetric Standard Model [20], there are four extra
U (1)a , including the anomalyfree hyper charge and gauged B L quantum numbers,
but also two extra U (1) factors which are anomalous. The relevant anomaly is actually a mixed AbeliannonAbelian anomaly, i.e., there are triangle diagrams of the type
SU(Nb )2 U (1)a nonvanishing for either SU(2) or SU(3). In particular, tr(Qa ) = 0, and
there is no gravitational anomaly. While the first models were nonsupersymmetric from
the beginning, the structure of charge assignments, given in Table 1 of [20], and the number
of U (1)a can serve as a representative example. From this it is clear that these models at
least contain two candidate U (1)a , which may develop FIterms. In generality, it is known
that, when Dbranes violate supersymmetry, this is reflected by FIterms in the effective
theory [24]. The violation of supersymmetry translates into a violation of the symmetry
on their world volume, and is geometrically captured by a violation of the socalled calibration conditions. For intersecting Dbranes models, this has a very simple geometric
interpretation. Any single brane wraps a threedimensional internal space. In the case of an
orbifold compactification it is characterized by three angles ia , i = 1, 2, 3, one for each T2
in T6 = (T2 )3 , measuring the relative angle of the Dbrane with respect to some reference
orientifold plane. The supersymmetry condition reads 1a 2a 3a = 0 mod 2 , with
some choice of signs. The FIparameter at leading order is proportional to the deviation,
a 1a 2a 3a mod 2.
(42)
The angles are modulidependent quantities, and thus the question if a FIterm is generated
cannot be finally answered without solving the moduli stabilization problem for the relevant moduli. In the mirror symmetric type IIB description with magnetized branes this is
manifest, and the relation reads
a
fa1
fa2
fa3
,
T1 + T1
T2 + T2
T3 + T3
(43)
where fai /(Ti + T ) = tan(ia ), and the Ti are the three (dimensionless) moduli, whose real
parts measure the sizes of the three T2 , while the fai are rational numbers. The natural
scale for the FIparameter is the string scale Ms = (
)1/2 , and a suppression means that
the righthand side is small numerically.
Another very important property of the string theoretic embedding of Dterms is the fact
that tachyonic masses (negative mass squared) can be lifted to positive values. Inspecting
the charge assignments, e.g., in [20], one realizes that it is not feasible to have positive
charges under the U (1)a for all the fields of the MSSM. This would naively mean that
some m2i ga2 Qia Da are negative, which would lead to a breakdown of gauge symmetry.
However, in the particular case of orbifold models, the exact string quantization can be
performed, and the mass spectrum computed for small FIparameters, without using effective field theory. It turns out that for small angles (iab = ia ib < /2), the mass of the
127
3
1 ab
i max iab ,
i
2
(44)
i=1
which, for a proper choice of signs, vanishes precisely if a = 0, consistent with the effective description. However, depending on the angles, there is a region in parameter space,
where the deviation from a = 0 only induces positive squared masses, and no tachyons
(see, e.g., [26] in this context). This comes as a surprise from the low energy point of view,
and is explained by the presence of higher order corrections in the derivative expansion of
the BornInfeld effective action, which become important for strings stretching between
intersecting branes [27]. The conclusion is, that the effective mass that follows from the
Dterm potential is corrected to positive values, and we can tolerate tachyons in the field
theoretic models.
Taking these observations together, it seems first of all possible that Dbrane models of
the type discussed have mass spectra with important contributions from FIterms. Since
multiple extra U (1)a factors exist and turn anomalous, a scenario, where the Da fields
cannot be relaxed to the electroweak scale, appears plausible. This would imply a mass
hierarchy between charged scalars and fermions. To solve the moduli stabilization problem,
and show convincingly how the D and Fterms of the desired magnitude are generated
dynamically, of course, remains an open challenge. One may also want to turn the argument
around, to conclude that within the conventional approach with low energy supersymmetry,
the potential presence of many FIterms is a great danger for these types of models, and
one has to find ways to suppress them dynamically.
Finally, we like to mention a caveat which makes the classes of Dbranes models that
we discussed rather not so good candidates to realize split supersymmetry. It is well known
that in Dbrane models in general the unification of gauge interactions really only happens
at the string scale, and not in the field theory regime. This means that the original motivation to keep the gauginos and higgsinos light, while giving up on the scalars, is upset.
The gauge kinetic functions are modulidependent fa = fa (TI ), and a unification of couplings requires some relations among moduli to hold [23,25,28], which so far can appear
accidentally in various models in the literature, but do not seem to have any independent
justification. This means, that an essential motivation for split supersymmetry, gauge unification, is only accidentally realized.
128
inflation, by undoing the fine tuning. A possible scenario is very easily illustrated along
the lines of Dterm inflation [29]. Roughly speaking, the only required modification of the
model we used so far, with the MSSM extended by one or many extra U (1)a gauge factors,
plus a pair of charged fields , is to promote the mass term (3) in the superpotential to a
dynamical field , neutral under U (1)a , which plays the role of the inflaton. We now write
W = + + W0 ,
(45)
(46)
2
2
m2 = 0 = 2 Vhid + e K0 1 + 6 W0 2 2 2 4 W0 2
ga a , (48)
2
which turns positive, when is large enough (but still well below the Planck scale). The
potential simplifies to
ga2 2
2
.
V = 0 = Vhid + 2 e K0 W0 2 2 2 3 +
2 a
a
(49)
The inflationary slowroll condition for the second derivative of the potential is [30]
2
V
 = 2 1,
(50)
V
where  1 numerically means  0.01 is acceptable. This implies that
2 2
2
MPl
a ga a
(51)
100.
2K
e
W0 2
The first derivative is then automatically also small (with a ), and inflation can
be successful. This means that the very same Dterm vacuum energy that is responsible
for the large scalar masses can drive inflation, if either the Dterms are enhanced or the
Fterm vacuum energy is sufficiently suppressed during that period. Thus, during the de
Sitter phase the relation between the Hubble parameter H and the energy density , i.e.,
2 H 2 = = 1 2 + V, shows that the Hubble expansion in this phase is
the relation 3MPl
2
dominated by the FIterm
2 2
H
3MPl
g2
a 2
a .
(52)
129
In summary, if in the phase of large the vacuum energy of the Dterms is larger than that
of the Fterms by a factor of about 100 or more, the Dterm energy can drive inflation at a
scale up to 101215 GeV, slightly above the mass scale of the heavy scalars. After the end
of the slow roll period, will eventually fall below the threshold value, and will form
condensates themselves. This can then lead to a partial relaxation of the Dterm vacuum
energy, but a more elaborate model of the hidden sector would be needed to describe
this
phase
transition
properly.
In
the
minimum,
one
may
expect
to
settle
down
to
m3/2 MPl on dimensional grounds, which is compatible with small gaugino masses.9
5. Conclusion
We have presented a model of supersymmetry breaking in the context of string and supergravity scenarios by inclusion of both F and FayetIliopoulos Dterms, arising from
extra U (1) factors in the gauge group. Such extra U (1) gauge symmetries arise quite naturally in string based models. It was shown that scalars charged under the extra U (1)
gain large masses from the FIterms, proportional to the charges of the respective scalar
fields under the extra U (1). This leads generically to nonuniversal masses
for the heavy
scalars. The cancellation of the vacuum energy puts an upper bound of m3/2 MPl on the
scalar masses, and thus also puts a bound on the FIterm X . The bound on X can be met
2.
in heterotic string models only for m3/2 close to MPl , since there X is scaled by MPl
Thus, heterotic string scenarios are not preferred from the vacuum energy constraint, when
m3/2 = O(TeV). However, m3/2 = O(TeV) and X m3/2 MPl , i.e., of size 101013 GeV,
could arise in orientifolds models which allow X of a variable modulidependent size.
The fact that the Dterm contributions to the scalar masses depend on their U (1) charges
opens up the possibility of building a new class of models with some scalars (with vanishing U (1) charges) light and others (with nonvanishing U (1) charges) heavy, while the
gauginos and higgsinos gain masses only of electroweak size. Further, the term is essentially unaffected by the FI contribution, and can be of electroweak size. We investigated
two illustrative examples of models with light and heavy scalars (model I and model II
in Section 2) and showed that they lead to significantly different phenomenologies which
could be tested at colliders and in nonaccelerator experiment. The class of models we have
discussed here are different from the high scale supersymmetry models of Ref. [2], where
all scalars and the gravitino are super heavy. However, which scalars are heavy and which
are light is now a modeldependent question. A further interesting feature is the possibility
that the vacuum energy responsible for generating heavy scalars may also drive inflation.
An analysis of how this can come about was discussed in Section 4. It would be interesting
to investigate more explicit Dbrane constructions to build models of the type advocated
here.
9 As mentioned earlier, we so far ignore here the problem of moduli stabilization, which is even more severe
in the context of inflation with Dbrane degrees of freedom, where already the correct choice of supersymmetric
coordinates is a very subtle question [31]. The FIparameters depend on the moduli fields, and it will be necessary
to stabilize the relevant fields (as, for instance, along the lines of [32]) at a scale larger that the scale of inflation,
or they would have to be considered as dynamical (see also the last reference of [29]).
130
Acknowledgements
B.K. would like to thank Angel Uranga for helpful advice. The work of B.K. was supported by the German Science Foundation (DFG) and in part by funds provided by the
US Department of Energy (D.O.E.) under cooperative research agreement #DFFC0294ER40818. The work of P.N. was supported in part by the US National Science Foundation under the grant NSFPHY0139967.
References
[1] A.H. Chamseddine, R. Arnowitt, P. Nath, Locally supersymmetric grand unification, Phys. Rev. Lett. 49
(1982) 970;
R. Barbieri, S. Ferrara, C.A. Savoy, Gauge models with spontaneously broken local supersymmetry, Phys.
Lett. B 119 (1982) 343;
P. Nath, R. Arnowitt, A.H. Chamseddine, Gauge hierarchy in supergravity GUTs, Nucl. Phys. B 227 (1983)
121;
L.J. Hall, J. Lykken, S. Weinberg, Supergravity as the messenger of supersymmetry breaking, Phys. Rev.
D 27 (1983) 2359.
[2] N. ArkaniHamed, S. Dimopoulos, Supersymmetric unification without low energy supersymmetry and signatures for finetuning at the LHC, hepth/0405159.
[3] L. Susskind, The anthropic landscape of string theory, hepth/0302219;
M.R. Douglas, The statistics of string/M theory vacua, JHEP 0305 (2003) 046, hepth/0303194;
T. Banks, M. Dine, E. Gorbatov, Is there a string theory landscape?, JHEP 0408 (2004) 058, hepth/0309170.
[4] G.F. Giudice, A. Romanino, Split supersymmetry, Nucl. Phys. B 699 (2004) 65, hepph/0406088;
N. ArkaniHamed, S. Dimopoulos, G.F. Giudice, A. Romanino, Aspects of split supersymmetry, hepph/0409232.
[5] A. Arvanitaki, C. Davis, P.W. Graham, J.G. Wacker, Oneloop predictions of the finely tuned SSM, hepph/0406034;
A. Pierce, Dark matter in the finely tuned minimal supersymmetric standard model, Phys. Rev. D 70 (2004)
075006, hepph/0406144;
S.H. Zhu, Chargino pair production at linear collider and split supersymmetry, hepph/0407072;
B. Mukhopadhyaya, S. SenGupta, Sparticle spectrum and phenomenology in split supersymmetry: some
possibilities, hepth/0407225;
W. Kilian, T. Plehn, P. Richardson, E. Schmidt, Split supersymmetry at colliders, hepph/0408088;
R. Mahbubani, Bounds on the Higgs mass in variations of split supersymmetry, hepph/0408096;
M. Binger, The Higgs boson mass at 2 loops in the finely tuned split supersymmetric Standard Model,
hepph/0408240;
J.L. Hewett, B. Lillie, M. Masip, T.G. Rizzo, Signatures of longlived gluinos in split supersymmetry,
JHEP 0409 (2004) 070, hepph/0408248;
L. Anchordoqui, H. Goldberg, C. Nunez, Probing split supersymmetry with cosmic rays, hepph/0408284;
K. Cheung, W.Y. Keung, Split supersymmetry, stable gluino, and gluinonium, hepph/0408335;
D.A. Demir, Effects of flavor violation on split supersymmetry, hepph/0410056;
R. Allahverdi, A. Jokinen, A. Mazumdar, Gravitino production from reheating in split supersymmetry, hepph/0410169;
V. Barger, C.W. Chiang, J. Jiang, T. Li, Axion models with highscale supersymmetry breaking, hepph/0410252;
J.A. Casas, J.R. Espinosa, I. Hidalgo, Implications for new physics from finetuning arguments. I: application
to SUSY and seesaw cases, hepph/0410298;
B. Bajc, G. Senjanovic, Radiative seesaw: a case for split supersymmetry, hepph/0411193.
[6] G.R. Dvali, A. Pomarol, Anomalous U(1) as a mediator of supersymmetry breaking, Phys. Rev. Lett. 77
(1996) 3728, hepph/9607383.
131
[7] P. Binetruy, E. Dudas, Gaugino condensation and the anomalous U(1), Phys. Lett. B 389 (1996) 503, hepth/9607172.
[8] L. Randall, R. Sundrum, Out of this world supersymmetry breaking, Nucl. Phys. B 557 (1999) 79, hepth/9810155;
G.F. Giudice, M.A. Luty, H. Murayama, R. Rattazzi, Gaugino mass without singlets, JHEP 9812 (1998)
027, hepph/9810442.
[9] I. Antoniadis, T.R. Taylor, Topological masses from broken supersymmetry, Nucl. Phys. B 695 (2004) 103,
hepth/0403293.
[10] M. Klein, Anomaly cancellation in D = 4, N = 1 orientifolds and linear/chiral multiplet duality, Nucl. Phys.
B 569 (2000) 362, hepth/9910143;
H. Ruegg, M. RuizAltaba, The Stueckelberg field, Int. J. Mod. Phys. A 19 (2004) 3265, hepth/0304245;
B. Krs, P. Nath, A Stueckelberg extension of the Standard Model, Phys. Lett. B 586 (2004) 366, hepph/0402047;
H. Ruegg, M. RuizAltaba, A supersymmetric Stueckelberg U(1) extension of the MSSM, hepph/0406167;
J. Louis, W. Schulgin, Massive tensor multiplets in N = 1 supersymmetry, hepth/0410149.
[11] Z. Lalak, S. Lavignac, H.P. Nilles, String dualities in the presence of anomalous U(1) symmetries, Nucl.
Phys. B 559 (1999) 48, hepth/9903160;
H.P. Nilles, Remarks on anomalous U(1) symmetries in string theory, hepph/0003102.
[12] M. Dine, N. Seiberg, E. Witten, FayetIliopoulos terms in string theory, Nucl. Phys. B 289 (1987) 589;
J.J. Atick, L.J. Dixon, A. Sen, String calculation of FayetIliopoulos D terms in arbitrary supersymmetric
compactifications, Nucl. Phys. B 292 (1987) 109.
[13] C. Angelantonj, A. Sagnotti, Open strings, Phys. Rep. 371 (2002) 1, hepth/0204089;
C. Angelantonj, A. Sagnotti, Phys. Rep. 376 (2003) 339, Erratum.
[14] L.E. Ibanez, R. Rabadan, A.M. Uranga, Anomalous U(1)s in type I and type IIB D = 4, N = 1 string vacua,
Nucl. Phys. B 542 (1999) 112, hepth/9808139.
[15] E. Poppitz, On the oneloop FayetIliopoulos term in chiral fourdimensional type I orbifolds, Nucl. Phys.
B 542 (1999) 31, hepth/9810010.
[16] A. Chamseddine, R. Arnowitt, P. Nath in [1];
E. Cremmer, S. Ferrara, L. Girardello, A. Van Proeyen, YangMills theories with local supersymmetry:
Lagrangian, transformation laws and superHiggs effect, Nucl. Phys. B 212 (1983) 413.
[17] V.S. Kaplunovsky, J. Louis, Model independent analysis of soft terms in effective supergravity and in string
theory, Phys. Lett. B 306 (1993) 269, hepth/9303040.
[18] P. Nath, T.R. Taylor, Modular invariance, soft breaking, and tan() in superstring models, Phys. Lett.
B 548 (2002) 77, hepph/0209282.
[19] R. Blumenhagen, L. Grlich, B. Krs, D. Lst, Noncommutative compactifications of type I strings on tori
with magnetic background flux, JHEP 0010 (2000) 006, hepth/0007024;
C. Angelantonj, I. Antoniadis, E. Dudas, A. Sagnotti, TypeI strings on magnetised orbifolds and brane
transmutation, Phys. Lett. B 489 (2000) 223, hepth/0007090;
G. Aldazabal, S. Franco, L.E. Ibanez, R. Rabadan, A.M. Uranga, D = 4 chiral string compactifications from
intersecting branes, J. Math. Phys. 42 (2001) 3103, hepth/0011073;
M. Cvetic, G. Shiu, A.M. Uranga, Chiral fourdimensional N = 1 supersymmetric type IIA orientifolds from
intersecting D6branes, Nucl. Phys. B 615 (2001) 3, hepth/0107166.
[20] L.E. Ibanez, F. Marchesano, R. Rabadan, Getting just the standard model at intersecting branes, JHEP 0111
(2001) 002, hepth/0105155.
[21] C. Bachas, A way to break supersymmetry, hepth/9503030;
M. Berkooz, M.R. Douglas, R.G. Leigh, Branes intersecting at angles, Nucl. Phys. B 480 (1996) 265, hepth/9606139.
[22] D. Cremades, L.E. Ibanez, F. Marchesano, SUSY quivers, intersecting branes and the modest hierarchy
problem, JHEP 0207 (2002) 009, hepth/0201205;
M. Cvetic, I. Papadimitriou, Conformal field theory couplings for intersecting Dbranes on orientifolds,
Phys. Rev. D 68 (2003) 046001, hepth/0303083;
M. Cvetic, I. Papadimitriou, Phys. Rev. D 70 (2004) 029903, Erratum;
S.A. Abel, A.W. Owen, Interactions in intersecting brane models, Nucl. Phys. B 663 (2003) 197, hepth/0303124;
132
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
P.G. Camara, L.E. Ibanez, A.M. Uranga, Fluxinduced SUSYbreaking soft terms, Nucl. Phys. B 689 (2004)
195, hepth/0311241;
D. Lst, P. Mayr, R. Richter, S. Stieberger, Scattering of gauge, matter, and moduli fields from intersecting
branes, Nucl. Phys. B 696 (2004) 205, hepth/0404134.
B. Krs, P. Nath, Effective action and soft supersymmetry breaking for intersecting Dbrane models, Nucl.
Phys. B 681 (2004) 77, hepth/0309167.
M.R. Douglas, G.W. Moore, Dbranes, quivers, and ALE instantons, hepth/9603167;
D. Cremades, L.E. Ibanez, F. Marchesano, Intersecting brane models of particle physics and the Higgs
mechanism, JHEP 0207 (2002) 022, hepth/0203160;
R. Blumenhagen, V. Braun, B. Krs, D. Lst, Orientifolds of K3 and CalabiYau manifolds with intersecting
Dbranes, JHEP 0207 (2002) 026, hepth/0206038.
I. Antoniadis, S. Dimopoulos, Splitting supersymmetry in string theory, hepth/0411032.
R. Rabadan, Branes at angles, torons, stability and supersymmetry, Nucl. Phys. B 620 (2002) 152, hepth/0107036.
A. Hashimoto, W.I. Taylor, Fluctuation spectra of tilted and intersecting Dbranes from the BornInfeld
action, Nucl. Phys. B 503 (1997) 193, hepth/9703217;
F. Denef, A. Sevrin, J. Troost, NonAbelian BornInfeld versus string theory, Nucl. Phys. B 581 (2000) 135,
hepth/0002180.
R. Blumenhagen, D. Lst, S. Stieberger, Gauge unification in supersymmetric intersecting brane worlds,
JHEP 0307 (2003) 036, hepth/0305146.
P. Binetruy, G.R. Dvali, Dterm inflation, Phys. Lett. B 388 (1996) 241, hepph/9606342;
E. Halyo, Hybrid inflation from supergravity Dterms, Phys. Lett. B 387 (1996) 43, hepph/9606423;
P. Binetruy, G. Dvali, R. Kallosh, A. Van Proeyen, FayetIliopoulos terms in supergravity and cosmology,
Class. Quantum Grav. 21 (2004) 3137, hepth/0402046.
A.R. Liddle, D.H. Lyth, Cosmological Inflation and Large Scale Structure, Cambridge Univ. Press, Cambridge, 2000.
S. Kachru, R. Kallosh, A. Linde, J. Maldacena, L. McAllister, S.P. Trivedi, Towards inflation in string theory,
JCAP 0310 (2003) 013, hepth/0308055;
J.P. Hsu, R. Kallosh, S. Prokushkin, On brane inflation with volume stabilization, JCAP 0312 (2003) 009,
hepth/0311077;
H. Firouzjahi, S.H.H. Tye, Closer towards inflation in string theory, Phys. Lett. B 584 (2004) 147, hepth/0312020;
M. Berg, M. Haack, B. Krs, Loop corrections to volume moduli and inflation in string theory, hepth/0404087;
M. Berg, M. Haack, B. Krs, On the moduli dependence of nonperturbative superpotentials in brane inflation, hepth/0409282;
H. Jockers, J. Louis, The effective action of D7branes in N = 1 CalabiYau orientifolds, hepth/0409098.
S. Kachru, R. Kallosh, A. Linde, S.P. Trivedi, De Sitter vacua in string theory, Phys. Rev. D 68 (2003)
046005, hepth/0301240.
Abstract
Auxiliary field methods in D = 2 (or 3), N = 2 supersymmetric (SUSY) nonlinear sigma models
(NL Ms) are studied. For these models auxiliary fields as Lagrange multipliers belong to a vector or
a chiral superfield, which gives a Khler quotient of complexified gauge group or a holomorphic constraint on it, respectively. Using these, NL Ms on all Hermitian symmetric spaces were formulated
previously. In this paper, we formulate new SUSY NL Ms on some ranktwo Khler coset spaces as
SUSY gauge theories with two FayetIliopoulos parameters.
2005 Elsevier B.V. All rights reserved.
PACS: 11.30.Pb; 11.30.Na; 11.10.Lm; 11.10.Kk; 11.30.Qc
1. Introduction
The auxiliary field method and the largeN method are very powerful tools to study
nonperturbative effects in a lot of theories, such as the GrossNeveu model, nonlinear
sigma models (NL Ms), gauge theories, matrix models and so on [1]. For twodimensional
(D = 2) NL Ms on coset spaces G/H , this method displays similarities with fourdimensional (D = 4) gauge theories very easily: dynamical mass generation, dynamical
symmetry breaking/restoration, dynamically induced gauge bosons, the asymptotic freeEmail addresses: nitta@physics.purdue.edu, nitta@th.phys.titech.ac.jp (M. Nitta).
1 Current address: Department of Physics, Tokyo Institute of Technology, Tokyo 1528551, Japan.
05503213/$ see front matter 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.nuclphysb.2005.01.025
134
dom and so on. Supersymmetric (SUSY) extensions are also studied for D = 2, N = 1
and N = 2 SUSY NL Ms [26]. D = 3 NL Ms are nonrenormalizable in perturbative method but renormalizable in the largeN expansion [7]. SUSY extensions of D = 3
NL Ms are also studied extensively [8,9].
As an example, the auxiliary field method for the O(N ) model can be illustrated briefly
as follows. Let gij () be the metric on S N 1 = O(N )/O(N 1) with some coordinates
i (i = 1, . . . , N 1). Then the partition function for the O(N ) model is given by
1
Z = [d] exp d D x gij () i j .
(1.1)
2
Classically O(N ) symmetry is spontaneously broken down to O(N 1). Introducing an
auxiliary field as a Lagrange multiplier, this can be rewritten as
1
Z = [d d ] exp d D x + 2 r 2
(1.2)
,
2
with = { A } (A = 1, . . . , N ) being an O(N ) vector of scalar fields. The integration over
supplies a constraint 2 = r 2 and the Lagrangian in (1.1) is rederived. On the other
hand, leaving , we can first perform the integration over dynamical fields A exactly as
the Gaussian integral. Then, the integration over can be approximated by the saddle point
in the largeN limit with taking r 2 = N/g 2 . We thus obtain the gap equation. In D = 2
gets nonzero vacuum expectation value (VEV) = 0 by solving it. Thus the O(N )
symmetry is dynamically recovered and the mass generation is found as a nonperturbative
effect contrary to masslessness in the perturbative analysis. This agrees with the Colemans
theorem which prohibits massless bosons in D = 2 [10]. In D = 3 there exist two phases
of broken and unbroken O(N ) symmetry.
Therefore, the auxiliary field formulation is crucial for a study of nonperturbative ef
fects. It is very easy to find the metric gij () solving the given constraint on linear fields :
N
for instance, for the O(N ) model, putting = (,
) and eliminating the N th component
i j
N
2
2
by = r , we obtain the metric gij () = ij + r2 2 expressed in terms of the
independent fields i . It is, however, in general difficult to find proper constraints among
linear fields to give a metric gij () of a given model. What we really want to do for a
study of nonperturbative effects is in this direction and this is the most difficulty of the
auxiliary field method. In SUSY theories, the situation becomes far more complicated as
will be explained below.
For D = 2, 3, N = 1 SUSY NL Ms, the situation is the same with bosonic case because
target spaces are Riemannian. The D = 2, N = 1 SUSY O(N ) model was investigated in
[3] and dynamical chiral symmetry breaking was found. The D = 3, N = 1 SUSY O(N )
model was discussed in [8]. D = 2, 3, N = 2 SUSY NL Ms are obtained as dimensional
reduction from D = 4, N = 1 SUSY NL Ms. For these cases target spaces must be Khler
[11,12] so that this makes the auxiliary field formulation difficult. D = 2 (D = 3), N = 2
SUSY NL Ms on Khler coset spaces G/H may have similarities with D = 4 (D = 5),
N = 2 SUSY QCD [13] as the same as bosonic theories. So pursuing similarities of these
SUSY models using nonperturbative method is very interesting.
In SUSY theories, auxiliary fields as Lagrange multipliers belong to superfields.
For these N = 2 SUSY theories, there exist vector and chiral superfields in terms of
135
Table 1
Hermitian symmetric spaces (HSS)
Type
AIII1
AIII2
BDI
CI
DIII
EIII
EVII
G/H
Complex coordinates
dimC (G/H )
)
CP N 1 = SU(NSU(N
1)U (1)
U (N )
GN,M = U (N M)U
(M)
(N 1)vector
N 1
[M (N M)]matrix
M(N M)
)
QN 2 = SO(NSO(N
2)U (1)
(N 2)vector
N 2
Sp(N )
U (N )
SO(2N )
U (N )
E6
SO(10)U (1)
E7
E6 U (1)
Symmetric (N N )matrix
Asymmetric (N N )matrix
1 N (N + 1)
2
1 N (N 1)
2
16spinor
16
27vector
27
Classification of Hermitian symmetric spaces (HSS) by Cartan, the standard complex coordinates belonging to
the representation of H and their complex dimensions are shown.
136
Table 2
Auxiliary field formulation for HSS
G/H
()
()
Superpotential
Constraints
Hosts
SO(N )
SO(N 2)U (1)
SO(2N ) Sp(N )
U (N ) , U (N )
E6
SO(10)U (1)
E7
E6 U (1)
U (1)
2 = 0
CP N 1
U (N )
tr( T J )
T J = 0
G2N,N
ij k j k = 0
CP 26
d = 0
CP 55
2N N
N N
i j k
27
27
U (1)
ij k
56
56
U (1)
Field contents, say dynamical chiral superfields () and auxiliary chiral and vector superfields () and V ,
are displayed in the first three rows. (We have given the representation of G or matrix sizes for () and ()
and gauge symmetry for V .) The superpotentials invariant under G and gauge symmetries and holomorphic
constraints obtained by the integration over () are also shown. In the last row, the host
spaces determined by
0 1
the integration over V are shown. The second rank symmetric tensor J is given by J =
1 0N with
= +1
N
(or 1) for SO(2N ) (or Sp(N )), and (d) is the E6 (E7 ) symmetric third (fourth) rank tensor whose explicit
expressions can be found in [16].
(1.5)
137
Any Khler coset space can be written as G/H = G/[Hs.s. U (1)r ] where Hs.s. is the
semisimple subgroup in H and r rank G rank Hs.s. is called the rank of this Khler
coset space [20]. Every HSS is a rank one Khler coset space, whose Lagrangian has a
U (1) or U (M) gauge symmetry and one FI parameter if formulated by the auxiliary field
method. Other rank one Khler coset spaces seem to be relatively easy to be constructed in
the auxiliary field formulation with U (1) or U (M) gauge group [16]. In this paper, we give
the auxiliary field formulation for some rank two Khler coset spaces, SU(N )/[SU(N
2) U (1)2 ] and SU(N )/[SU(N M L) SU(M) SU(L) U (1)2 ]. These models
have U (1)2 and U (M) U (L) gauge symmetries, respectively, and two FIparameters. In
addition to auxiliary vector superfields for these gauge symmetries, some auxiliary chiral
superfields are also needed even though G = SU(N ). Nonperturbative studies for these
new models become possible which will remain as a future work. We expect that U (1)2 or
U (M) U (L) gauge symmetry is dynamically induced.
This paper is organized as follows. In Section 2, we give the minimum ingredient of
SUSY nonlinear realizations needed for later discussions. Section 3 explains how to obtain compact Khler coset spaces using the superHiggs mechanism eliminating unwanted
QNG bosons. How this works in the simplest CP N and GN,M is shown. In Sections 4 and 5
we generalize these discussions to the ranktwo coset spaces SU(N )/[SU(N 2) U (1)2 ]
and SU(N )/[SU(N M L) SU(M) SU(L) U (1)2 ], respectively. Section 6 is
devoted to summary and discussions. In Appendix A, we give a review on the supersymmetric nonlinear realization with Khler G/H focusing on the case of G = SU(N ). In
Appendix B, we discuss some geometric aspects of these models, a relation with some
hyperKhler manifolds and an application to construction of a new CalabiYau metric.
138
(2.1)
with B called the Borel algebra comprising of (nonHermitian) nilpotent generators. Therefore the vacuum manifold M parameterized by massless bosons is topologically a complex
coset space GC /H , which is a noncompact Khler manifold in general. However, we note
that the isometry of this coset space should be still G but not GC because symmetry of the
Khler potential is not complexified, with being different from the superpotential.
A group element in g GC can be uniquely divided into g = h with the coset representative and h H . Here the coset representative is given by
= exp( Z)
GC
H
(2.2)
(2.4)
1
dimC M = NP + NM = (NNG + NQNG ).
2
(2.5)
139
If there are no QNG bosons (NQNG = NM = 0) the manifold becomes compact, whereas
if there exists at least one QNG bosons (NQNG = NM 1) the manifold becomes noncompact. For noncompact cases, NG bosons parameterize a compact homogeneous submanifold G/H embedded into the total manifold M GC /H [34]. For both cases, the
isometry of M is G but not GC although GC acts on M transitively.
Even if the same NG bosons appear for the same G/H , NQNG depends on the underlying linear model. Two extreme or natural cases can be considered: the case of the maximal
NQNG equal to NNG (NP = 0) and the case of NQNG = 0 (NM = 0).
(1) If there exist only Mtype multiplets without Ptype multiplets (NP = 0, NQNG =
NNG ), it is called the maximal realization (or fullydoubling), which corresponds
to GC /H = GC /H C T (G/H ).4 Some sufficient conditions for maximal realizations are known [29]: it occurs if a symmetry G is broken down to H by VEVs of
linear fields (1) which belong to a real representation of G or (2) with G/H a symmetric space. In both cases, absence of gauge fields is assumed.
(2) On the other hand, when there exist only Ptype multiplets without Mtype multiplets
(NM = NQNG = 0), it is called the pure realization, which corresponds to a compact
homogeneous Khler manifold (Khler coset) G/H . In this case GC /H G/H holds.
The Khler metric on arbitrary Khler coset G/H was constructed by Borel [20]. All
Khler coset spaces G/H were classified by using the painted Dynkin diagrams [23].
Their Khler potentials were given by Itoh, Kugo and Kunitomo (IKK) [24]. For pure
realizations a nogo theorem due to Lerche and Shore is known [29,30] (see also [16,
32,34]): there must appear at least one QNG bosons and therefore pure realizations
cannot be realized if a symmetry G is broken by VEVs of linear fields and if there are
no gauge symmetries.
When we reformulate NL Ms by auxiliary fields, dynamical fields must be embedded
into fields in some linear representations of G. Therefore the LercheShore theorem is
the most difficulty for the auxiliary field formulation of Khler coset spaces G/H . As
discussed in this paper in detail, this theorem can be avoided introducing appropriate gauge
symmetry.
[37].
140
one QNG bosons, with preserving SUSY.5 Hence, one V can eliminate one noncompact
direction with one compact direction. If we can eliminate all noncompact directions of
QNG bosons, a compact manifold of the pure realization can be realized. Such consideration for the CP N model and the GN,M model was given in [16,22] and then applied to
HSS [16]. We briefly review the cases of CP N and GN,M in this section, with some new
consideration on the geometry of noncompact manifolds before gauging.
3.1. Auxiliary field formulation for CP N
First, let us discuss CP N 1 . Prepare linear fields N of SU(N ). The system has an
additional phase symmetry U (1)D , given by = ei , and the total global symmetry becomes G = U (N ) = SU(N ) U (1)D . When the fields develop the VEV it can be
transformed by G into = (0, . . . , 0, v)T with v real positive. By this VEV, G is spontaneously broken down to its subgroup H = U (N 1). There appear NNG = dim G/H =
2N 1 NG bosons.
To discuss the whole massless bosons including the QNG bosons, we consider the complexification of the groups. The complex unbroken and broken generators are
0
P
.
..
C .
.
H = U(N 1) . ,
(3.1)
G C H = 0N 1
,
0
P
B
0 0 M
B 0
where B denote generators in a Borel subalgebra B in Eq. (2.1). Here, M and P denote one
Hermitian and N 1 nonHermitian broken generators, generating M and Ptype chiral
superfields, respectively (NM = 1, NP = N 1). So the numbers of NG and QNG bosons
are NNG = 2N 1 and NQNG = 1, respectively. The number of the QNG bosons coincides
can be
with the number of Ginvariant 2 as was discussed in [36]. The manifold M
locally written as
R
M
SU(N )
U (N )
R
U (N 1)
SU(N 1)
(3.2)
eV eV = eV ei+i ,
(3.3)
with (x, , ) a gauge parameter of a chiral superfield. Note that this gauge symmetry is
complexified to U (1)C = GL(1, C) because the scalar component of is a complex scalar
5 When a gauge field in V absorbs an NG boson in a Ptype multiplet, SUSY is spontaneously broken [39].
field. The simplest invariant Lagrangian for these matter contents is written as
L = d 4 eV cV ,
141
(3.4)
with c a real positive parameter called the FayetIliopoulos (FI) parameter.6 If we eliminate
V by its equation of motion eV c = 0, we obtain
4
L = d c log + 1 = d 4 c log ,
(3.6)
where the second term has disappeared under the superspace integral. There still exist the
gauge symmetry (3.3) for matter fields in this Lagrangian which can be fixed as T =
( T , 1) using U (1)C . Then the nonlinear Lagrangian in terms of independent fields is
obtained as
L = d 4 c log 1 + .
(3.7)
We thus have obtained the Khler potential for the FubiniStudy metric on CP N 1 .
This construction of CP N is known as the Khler quotient method [51,52]: CP N 1
M/U
(1)C .
Note that can be rewritten as = v with v = (0, . . . , 0, 1)T and = exp 0N1
=
0 0
1N1
being the same as (A.7). Comparing the coset generator in and (3.1), we find
0 1
that one QNG boson is absorbed, with one NG boson, by the U (1) gauge field V and that
a pure realization occurs.
Although we have used the classical equation of motion to eliminate V here, we can
show that this holds in the quantum level using the path integral formalism [17,18].
3.2. Auxiliary field formulation for GN,M
This can be generalized to nonAbelian gauge symmetry to construct GN,M . This case
is a little bit complicated due to nonuniqueness of vacua. Let be an (N M)matrix
chiral superfield, on which the global symmetry G = SU(N ) SU(M) U (1) as
g
= gL gR ,
(3.8)
L=
d 4 f eV cV ,
(3.5)
with an arbitrary function f . However, we can show that we get the same Lagrangian (3.7) below when V is
eliminated [17].
142
because the Ginvariants det and tr( )n (n > M) are not independent with these
due to the CayleyHamilton equation for M by M matrices. The Ginvariants (3.9) de to be M. The most general invariant
termine the cohomogeneity of the manifold M
Lagrangian is
L = d 4 f X 1 , X 2 , . . . , XM
(3.10)
with an arbitrary function f .
Using G symmetry, generic VEVs can be transformed into the form of
0(N M)M
v
0
1
Vgeneric = =
,
..
.
0
vM
(3.11)
i1
Mi
tion by U (1)D . Hence the number of NG bosons is NNG = dim(G/Hg ) = 2MN M. The
M constants vi in (3.11) are (VEVs of) QNG bosons parametrizing noncompact directions
SU(N ) U (M)
SU(N ) SU(M)
RM
,
SU(N M) U (1)M
SU(N M) U (1)M1
(3.12)
and so it is cohomogeneity M. In the last equality, the overall phase rotation is cancelled.
When some vi s coincide, unbroken symmetry is enhanced. For instance, when two of
them coincide vi = vj , the unbroken symmetry is H = SU(N M) U (2) U (1)M2 .
So the number of NG bosons NNG = dim(G/H ) = 2MN 3 is less than the generic
points. Some NG bosons have changed into QNG bosons with total number of massless
bosons unchanged. This phenomenon was found by Shore [34] and was named the SUSY
vacuum alignment. The number of QNG bosons changes from a point to a point, but
the minimum number of QNG bosons realized at the generic points is bounded below by
the number of the Ginvariants composed of fundamental fields as found in [36]. So it
determines the cohomogeneity of the manifold.
At the most symmetric point v1 = v2 = = vM v, the VEV becomes
0(N M)M
Vsymmetric = v
(3.13)
.
1M
143
(3.14)
def
posite phase rotation by U (1)D with fixing the M M unit matrix in the VEV (3.13). The
number of NG bosons is NNG = dim(G/H0 ) = 2MN M 2 .
At this point complex unbroken and broken generators become
SU(N M)C 0(N M)M
H =
, AMM U(1)C ,
BM(N M)
AMM
P(N M)M
0N M
, 0M U(1)C
G C H =
(3.15)
M.
0M(N M) MMM
Here B represents the matrix of the Borel generators and M and P represent the matrices
of M and Ptype broken generators, respectively, and U (1) in G C H is generated by Q1
itself which is also Mtype. So NP = M(N M) and NM = M 2 hold. The numbers of NG
and QNG bosons are NNG = 2NP + NQNG = 2MN M 2 and NQNG = M 2 , respectively.
Here NNG is the least and NQNG is the most. The manifold looks like
RM 2
M
SU(N ) U (M)
,
SU(N M) U (M)
(3.16)
eV eV = ei eV ei ,
(3.17)
144
(3.21)
can be used to eliminate V . Solving this equation as V = log( /c) and substituting
this back into the original Lagrangian (3.19), we obtain
L = d 4 c log det ,
(3.22)
where a constant has disappeared under the superspace integral. This still has the
GL(M, C) gauge symmetry (3.17) for matter fields. We can fix this gauge degree of freedom as
=
(3.23)
1M
with an N M by M matrix of chiral superfields. Therefore, we obtain
L = d 4 c log det 1M + .
(3.24)
This is the Khler potential for the Grassmann manifold GN,M . In terms of Khler quotient
(M)C .
we can write GN,M M/U
0
NM
Note that in (3.23) can be rewritten as = Vgauge using = exp 0M(NM)
0M =
1NM
Eq. (3.15), we conclude that M 2 QNG bosons at the most symmetric point are absorbed,
with the same number of NG bosons, by the U (M) gauge fields V and that a pure realization occurs. The less symmetric points do not contribute to the resultant manifold.
The U (M) gauge symmetry can be replaced with the U (N M) gauge symmetry with
the same SU(N ) global symmetry considering an N by N M matrix, due to the duality
GN,M GN,N M .
4. Auxiliary field formulation of SU(N)/[SU(N 2) U (1)2 ]
Generalizing the discussion in CP N , we formulate the ranktwo Khler coset space
SU(N )/[SU(N 2) U (1)2 ] by the auxiliary field method. In the first subsection, we
construct a noncompact Khler manifold putting a holomorphic constraint by an auxiliary
chiral superfield without gauging. Then in the second subsection, we obtain a compact
manifold by gauging U (1)2 part of the isometry of the noncompact manifold introducing
auxiliary vector superfields with two FIparameters.
7 The most general invariant Lagrangian is
L=
2
M
d 4 f tr eV , tr eV , . . . , tr eV
c tr V
(3.20)
with an arbitrary function f . However, we think that the resulting Lagrangian after eliminating V coincides with
the simplest case although we do not have a proof.
145
(4.2)
U (1)2
symmetry,
i i = eii i
(4.3)
v2 = 2 = ( 0, . . . , 0, u, 0)T ,
(4.5)
N 2
with v and u real positive. By these VEVs, G = SU(N ) U (1)2 is spontaneously broken
down to H = SU(N 2) U (1)2 . Its generators are given by
0 0
. .
SU(N 2) .. ..
U(1) U(1),
H=
(4.6)
0 0
0 0 0 0
0 0 0 0
with two U (1)s given by generators
Q1 diag( 1, . . . , 1, 2 N, 0),
N 2
Q2 diag( 1, . . . , 1, 0, 2 N ),
(4.7)
N 2
8 (g 1 )T is not equivalent to g when we consider complex extension of the group for the symmetry of the
superpotential. We should define the transformation law of the antifundamental representation by the former.
9 The constraint = a 2 , with a being a real constant, was discussed in [36]. In this case, one of U (1)
1
2
symmetries (4.3) is explicitly broken.
146
in SU(N ) accompanied with the 2  and 1 phase rotations (4.3) with opposite angles,
with fixing v2 and v1 , respectively. The number of NG bosons is NNG = dim(G/H ) =
4N 4.
To discuss QNG bosons, we consider the action of the complexification of G. The complex unbroken and broken generators are
B 0
. .
SU(N 2)C .. ..
U(1)C U(1)C ,
H =
B 0
0 0 0 0
B
B B 0
..
0N 2
.
C
C
U(1)C
G H =
M U(1)M ,
0 P
P P 0 P
0 0 0 0
0
..
.
(4.8)
where B denote generators in a Borel subalgebra, P denote nonHermitian broken generators Ptype chiral superfields and two U(1)C
M s are given by Q1 and Q2 defined in
Eq. (4.7), both of which generate Mtype chiral superfields. The numbers of the P and the
Mtype superfields are NP = 2N 3 and NM = 2, respectively. Therefore, there appear
NNG = 4N 4 NG bosons, whose number coincides with the dimension of G/H , and
NQNG = 2 QNG bosons. The number of the QNG bosons coincides with the number of the
G invariants, 1 2 and 2 2 , as discussed in [36]. The manifold can be locally written as
R2
M
SU(N ) U (1)2
SU(N )
R2
2
SU(N 2)
SU(N 2) U (1)
(4.9)
with i (x, , ) being a chiral superfield as a gauge parameter of U (1)i . We thus obtain the
invariant Lagrangian for the auxiliary field formulation of SU(N )/[SU(N 2) U (1)2 ]
as
V
4
V2
2
1
L = d e 1 1 + e 2 2 c1 V1 c2 V2 +
d 1 2 + c.c. .
(4.11)
147
To confirm if this really gives a compact Khler coset space, we now eliminate auxiliary
superfields , V1 and V2 by their equations of motion. The integration over gives the
holomorphic constraint
1 2 = + + = 0,
in which we have used the notation
1 = ,
2 =
,
(4.12)
(4.13)
with and N 2 vectors and the rests singlets. Introducing a new chiral superfield
(x, , ), we rewrite this constraint as
2 = 2,
2 = 2.
(4.14)
If we eliminate from these equations, we get the constraint (4.12) again. Instead, we
solve and by other fields:
1
1
+
,
=
.
=
(4.15)
2
The integration over Vi gives eVi i i ci = 0, which can be solved as
Vi = log i i /ci .
(4.16)
Substituting these solutions and (4.15) back into the Lagrangian (4.11), we obtain
L=
d 4
2
ci log i i
i=1
1
2
4
2
2
= d c1 log  +  + +
2
2
1
,
2
(4.17)
where additional constant has disappeared under the integration over the superspace. Since
this still has gauge invariance (4.10) of U (1)2 for matter fields, we can take a gauge of
= = 1:
2
4
2
L = d c1 log 1 +  + +
2
2
.
+ c2 log 1 + 2 +
(4.18)
2
This is the Khler potential of SU(N )/[SU(N 2) U (1)2 ] with a suitable complex
structure [31] (see Eq. (A.17) in Appendix A).
148
To see the relation with the noncompact case, we note the coset representative is given
by
1N 2 0
= eZ = T 1 + 12
0 0
1
0N 2 0
with Z = T 0 ,
0 0 0
(4.19)
where Z represent complex broken generators, and = {, , }. Using this representative, the superfields, after solving the constraint and fixing a gauge, can be written as
T
1 = v1 ,
(4.20)
2 = 1 v2 ,
where vi are VEVs given in (4.5) with u = v = 1. By comparing the generators (4.8) and
(4.19), it is now obvious that two Mtype superfields are eliminated by gauging of U (1)2 .
From the structure of generators (4.19), we find that this manifold has one of two complex
structures, called II , for SU(N )/[SU(N 2) U (1)2 ] [25].
If we forget the superpotential term in the Lagrangian (4.11), it becomes the auxiliary field formulation for CP N 1 CP N 1 . Therefore SU(N )/[SU(N 2) U (1)2 ] is
embedded into CP N 1 CP N 1 by a holomorphic constraint 1 2 = 0 so that it is
algebraic.
g1 U (M),
g2 U (L).
= g2 ,
(5.1)
T = 0LM .
(5.2)
...,
M
XM tr
,
2
Y2 tr ,
Y1 tr ,
...,
L
YL tr
,
149
(5.3)
so that this manifold is cohomogeneity M + L.10 The most general invariant Lagrangian is
L=
d 4 f (X1 , . . . , XM , Y1 , . . . , YL ) +
d 2 tr T + c.c. ,
(5.4)
0(N M)M
v
0
1
V1 = =
,
..
.
0
vM
u1
0
..
0LM ,
(5.5)
uL
(5.6)
so that it is cohomogeneity M + L.
At the most symmetric points, the VEVs become
0
V1 = = v (N M)M
1M
,
(5.7)
10 The invariants made by traces of products M and N are probably not independent of these
invariants although we do not have a proof. It is plausible considering the relation with VEVs as below.
150
XLL
0LM
H = 0L(N ML)
, XLL , YMM
0M(N ML)
0ML
YMM
U(1) U(1),
(5.8)
with X and Y generators of SU(L) and SU(M), respectively. Here the two U (1)s are
given by
Q1 diag( L, . . . , L, N + M + L, . . . , N + M + L, 0, . . . , 0 ),
N ML
Q2 diag( M, . . . , M , 0, . . . , 0, N + M + L, . . . , N + M + L ),
N ML
(5.9)
in SU(N ) combined with phase rotations with opposite angles by U (1)1 and U (1)2 , respectively.
Complex unbroken and broken generators become like
H = 0L(N ML)
XLL
0LM
, XLL , YMM
BM(N ML)
BML
YMM ,
U(1)C U(1)C ,
MLL
PLM
G C H = PL(N ML)
, 0LL , 0MM
0M(N ML)
0ML
MMM
C
U(1)C
M U(1)M ,
(5.10)
respectively, where each subscript denotes the size of each block, and two U (1)s in G C
H are generated by Q1 or Q2 in Eq. (5.9), both of which are Mtypes. The numbers of the
M and Ptype superfields are NM = L2 + M 2 and NP = M(N M) + L(N M L),
respectively. There appear NNG = 2(N M + N L L) M 2 L2 NG bosons and NQNG =
(L2 + M 2 ) QNG bosons. The manifold looks like
RM 2 +L2
M
(5.11)
= ei2 ,
151
= ei1 ei2 ,
T
(5.12)
where 1 (x, , ) and 2 (x, , ) are chiral superfields of gauge parameters, taking values
in U(M) and U(L), respectively. Then the Lagrangian can be written as
L = d 4 tr eV1 + tr eV2 c1 tr V1 c2 tr V2
+
d 2 tr T + c.c. .
(5.13)
We decompose the matrix chiral superfields into submatrices, like
I A
I
= A ,
=
,
AB
A
(5.14)
where a new index I , which runs from 1 to N M L, has been introduced. In terms of
these decompositions, the integration over A yields the LM holomorphic constraints
T
T
T
A = I
I A +
A + B
BA = 0,
(5.15)
in which the summation over repeated indices is implied. Introducing a new L M matrix
chiral superfield, A (x, , ), these constraints can be rewritten as
T
T
2
A I
I A = 2A ,
T
T
2B
BA I
I A = 2A .
(5.16)
Elimination of from these equations gives the original constraints (5.15). Instead, we
express and by other fields in the region such that det = 0 and det = 0 hold:
1
1
= 1T + T ,
(5.17)
= 1T T ,
2
2
where we have used the matrix notation.
The equations of motion for V1 and V2 read eV1 c1 1M = 0 and eV2 c2 1L =
0, respectively. These equations can be solved, to give
,
V2 = log det
.
V1 = log det
(5.18)
c1
c2
Substituting these solutions back into the Lagrangian (5.13), we obtain
L = d 4 c1 log det + c2 log det
= d 4 c1 log det + + + c2 log det + + .
(5.19)
152
By substituting (5.17) into this Lagrangian and taking a gauge fixing of = 1M and =
1L , we obtain
1
1
L = d 4 c1 log det 1M + + + + T
2
2
1
1 T
.
+ c2 log det 1L + +
(5.20)
2
2
We thus have obtained the Khler potential of SU(N )/[SU(N M L) U (M) U (L)]
153
Second, let us discuss ranktwo coset spaces G/H with other groups G. For rankone case it was possible to reduce G = SU(N ) to other groups G G imposing G invariant Fterm (holomorphic) constraints by auxiliary chiral superfields which give an
embedding for the whole coset G /H G/H , as in the Lagrangian (1.5) for QN 2 =
SO(N )/[SO(N 2) U (1)] CP N 1 . In principle, it should work also for ranktwo
cases but naive attempts failed. This remains for a future work.
Third, we would like to discuss higherrank coset spaces. The LercheShore theorem
[29,30] implies that any Khler G/H needs a gauge group to be formulated by linear
fields. It is natural for a gauge group to include the U (1)r factor with r FIparameters for
a rankr coset space. Moreover we should introduce at least r irreducible representations
of G and put suitable Fterm constraints among them which are needed to fix all GC invariants composed of them. However the similar problem for SU(N )/[SU(N 2)
U (1)2 ] with I occurs and we are unable to achieve this in the present time.
Perturbatively, dynamics of D = 2, N = 2 SUSY NL Ms have very different features
according to their first Chern classes c1 (M) on the target manifold M. CalabiYau manifolds M have vanishing first Chern classes c1 (M) = 0. D = 2, N = 2 SUSY NL Ms on
these manifolds are conjectured to be finite to all orders and are considered as models of
superstring theory [41]. On the other hand, all Khler coset spaces have positive first Chern
classes: c1 (G/H ) > 0. For all NL Ms on M with c1 (M) > 0, it is conjectured that these
models are asymptotically free and have the mass gap like D = 4 QCD.
Nonperturbative analyses for these features were discussed extensively using the mirror
symmetry [42]. Exact beta functions for D = 2 NL Ms on Hermitian symmetric spaces
were derived using the instanton method [43]. D = 2, 3, N = 2 SUSY NL Ms were also
discussed using the Wilsonian renormalization group (WRG) [44], in which they used
the socalled Khler normal coordinates [45] to derive the WRG equation. In particular,
D = 2, 3, N = 2 SUSY NL Ms on the KhlerEinstein manifolds including Khler coset
spaces were discussed.
Combined with these several methods, we expect that the largeN method plays an
important role to reveal nonperturbative aspects of SUSY NL Ms and their similarity
with D = 4, 5 QCD.
Acknowledgements
The author thanks Kiyoshi Higashijima and Makoto Tsuzuki for a collaboration in early
stages of this work. His work is supported by the US Department of Energy under grant
DEFG0291ER40681 (task B).
U
(1)
i
i=1
i=1 ni = N and r( 1) called the
154
SU(n1 )
SU(n2 )
H=
..
U(1) U(1) .
(A.1)
SU(nr+1 )
All offdiagonal blocks are zero matrices and r U (1) generators Q ( = 1, . . . , r) are
given by
Q = diag( n+1 , . . . , n+1 , 0, . . . , 0, n1 , . . . , n1 0, . . . , 0 ).
n1
=2 n
n+1
(A.2)
r+1
=+2 n
r
c Q
(A.3)
=1
with c R. Complex linear combination I CI XI (CI C) of the real coset generators
XI G H can be divided into Bi H and Zi G C H according to positive and negative Y charges, respectively: [Y, Bi ] +Bi and [Y, Zi ] Zi . Note that the generators
in H carry zero Y charges and therefore all generators in H carry nonnegative Y charges.
Thus, rank one coset spaces have one complex structure. Rank (more than) two coset spaces
have (more than) two inequivalent complex structures. The complex coset representative
can be defined by = exp( Z) GC /H with i NG chiral multiplets whose scalar
components are both genuine NG bosons. There exists homomorphism GC /H G/H for
pure realizations because there are no QNG bosons.
Once a complex structure is provided we can construct a Ginvariant Khler potential
on G/H . There exist r projection operators ( = 1, . . . , r) satisfying the conditions
H = H ,
2 = ,
(A.4)
r
c log det ,
(A.5)
=1
155
H = SU(N 1)
0
..
.
U(1)C ,
0
G C H =
0N 1
P
..
. ,
(A.6)
0 0 0
B 0
respectively, with U(1)C generated by Q1 . Then the complex coset representative is given
by
1N 1
0N 1
,
=
Z=
(A.7)
0 0
0 1
with an (N 1)vector of SU(N 1). The projection operator is = diag( 0, . . . , 0, 1)
N 1
(A.8)
SU(N M)C 0(N M)M
U(1)C ,
BM(N M) SU(M)C
P(N M)M
0N M
,
G C H =
0M(N M)
0M
H =
(A.9)
1N M
0N M
,
=
Z=
(A.10)
0M(N M) 0M
0M(N M) 1M
with an N M by M matrix. The projection operator is = diag( 0, . . . , 0, 1, . . . , 1 )
N M
(A.11)
156
Q2 diag( 1, . . . , 1, 0, N + 2).
(A.12)
N 2
Two complex structures I and II denoted in [25] are represented, for instance, by YI
Q2 and YII Q1 Q2 , respectively. According to these Y charges, complex unbroken
and broken generators are given by
P P
0 0
.. ..
. .
0N 2
. .
SU(N 2)C .. ..
C
C
P P
2U(1) ,
HI =
G H I =
0 0
,
B B 0 0
0
0
0
P
0 0 0 0
B B 0
for I and
0
..
SU(N 2)C
.
H II =
2U(1)C ,
B
0
0 0 0 0
B B B 0
B
..
.
P
..
0N 2
.
C
G H II =
,
0
P
P P 0 P
0 0 0 0
0
..
.
for II . For both cases two U (1) generators are given by Eq. (A.12). The coset representative for I is calculated as
1N 2 + 12
0N 2
( Z)I = 0 0 ,
(A.13)
I = 0 1
,
0 0 0
0 0
1
with and belonging to N 2 of SU(N 2). The coset representative for II is
0N 2 0
( Z)II = T 0 ,
0 0 0
1N 2 0
II = T 1 + 12
0 0
1
(A.14)
(2 )I = diag( 0, . . . , 0, 1, 1),
(A.15)
(A.16)
N 2
N 2
N 2
N 2
157
Q2 diag( M, . . . , M , 0, . . . , 0, N + M + L, . . . , N + M + L ).
N ML
(A.18)
As the same with the last example, the Y charge can be taken, for instance, as YI Q2 or
YII MQ1 LQ2 for the complex structure I or II , respectively. Complex unbroken
and broken generators are given by
C
SU(L)C
0LM
H I = BL(N ML)
2U(1) ,
BM(N ML)
BML
SU(M)C
0L
PLM
G H I = 0L(N ML)
(A.19)
,
0M(N ML)
0ML
0M
for I and
C
SU(L)C
0LM
H II = 0L(N ML)
2U(1) ,
BM(N ML)
BML
SU(M)C
0L
PLM
G H II = PL(N ML)
,
0M(N ML)
0ML
0M
11 We have chosen these charges in a different way from those in Refs. [22,48].
(A.20)
158
0N ML
1N ML + 12
,
0
0L ,
I =
( Z)I =
(A.21)
0
1L
0
0 0M
0
0
1M
1N ML 0
0N ML 0
( Z)II = T
II = T
0L ,
1L + 12 T , (A.22)
0
0 0M
0
0
1M
for I and II , respectively. The projection operators are given by
(1 )I = diag( 0, . . . , 0, 0, . . . , 0, 1, . . . , 1 ),
N ML
(2 )I = diag( 0, . . . , 0, 1, . . . , 1, 1, . . . , 1 ).
N ML
(A.23)
for I and
(1 )II = diag( 0, . . . , 0, 0, . . . , 0, 1, . . . , 1 ),
N ML
(2 )II = diag( 1, . . . , 1, 0, . . . , 0, 1, . . . , 1 ),
N ML
(A.24)
for II . Using these projection operators, the Khler potentials are calculated as
1
1
KI = c1 log det 1M + + + +
2
2
1L +
+ ( + 12 )
+ c2 log det
1
1
1
+ ( + 2 ) 1M + + ( + 2 )( + 2 )
1
1 T
T
1N ML +
+ ( + 12 T )
+ c2 log det
1 T
1
+ ( + 2 )
,
1M + + ( + 2 )( + 12 T )
,
(A.25)
for I and II , respectively.
The second one should coincide with the Khler potential (5.20) up to a holomorphic
coordinate transformation and a Khler transformation, but we are unable to show their
equivalence.
159
d 4 eV 1 1 + eV 2 2 cV +
d 2 1 2 + c.c. ,
(B.1)
d tr eV + tr T eV c tr V +
4
T
d tr + c.c. ,
2
(B.2)
with c c2 c1 . As the same as the above discussions, this is the HK quotient construction
for the LindstrmRocek metric on T GN,M [51,57]. (Putting = 0 in the Lagrangian
(B.2) we obtain the Lagrangian (1.4) for GN,M , and therefore we find this bundle structure.) Thus SU(N )/[SU(N 2M) U (M)2 ] is a U (M) Khler quotient of T GN,M , and
the latter is a CM bundle over the former.
Before closing this appendix, we discuss possibility to construct a new CalabiYau (CY)
metric of cohomogeneity one [4649]. If we freeze out V1 + V2 in SU(N )/[SU(N 2)
U (1)2 ] starting from the most general Lagrangian, we obtain
L=
d 4 eV 1 1 + f eV 2 2 cV +
d 2 1 2 + c.c. ,
(B.3)
160
the Khler structure. If we freeze out aV1 + bV2 (with a, b R) in SU(N )/[SU(N 2)
U (1)2 ], we get
V
qV
4
2
L = d e 1 1 + f e 2 2 cV +
(B.4)
d 1 2 + c.c. ,
with q a/b being the relative charge of the remained U (1) gauge for 1 and 2 , and
c c1 + qc2 . This is the Lagrangian suggested in [49] ((A.2) in Appendix A) as a generalization of the construction of a CY metric using matter coupling in the CP N model.
Integrating V and we obtain
q
L = d 4 log 1 1 + h 1 1
(B.5)
2 2
with the constraint 1 2 = 0 and some function h related with f . In the case of q = 0, the
CY metric was obtained in [49] solving the Ricciflat condition for h. The case of q = 1
corresponds to the HK Calabi metric. If we solve it for general q, we will be able to obtain
the most general CY metric on the line bundle over SU(N )/[SU(N 2) U (1)2 ], if it
exists, which is cohomogeneity one and can be locally written as
R
SU(N )
,
SU(N 2) U (1)
(B.6)
where freedom to embed U (1) H into SU(N ) corresponds to q. The holonomy structure
for the case of N = 3, R SU(3)/U (1), was discussed in detail in [58] including Spin(7)
holonomy.
References
[1] A.M. Polyakov, Gauge Fields and Strings, Harwood Academic, Reading, UK, 1987;
S. Coleman, Aspects of Symmetry, Cambridge Univ. Press, Cambridge, 1985;
E. Abdalla, M.C.B. Abdalla, K.D. Rothe, NonPerturbative Methods in 2Dimensional Quantum Field Theory, World Scientific, Singapore, 1991;
K. Higashijima, Prog. Theor. Phys. Suppl. 104 (1991) 1.
[2] A. DAdda, M. Lscher, P. Di Vecchia, Nucl. Phys. B 146 (1978) 63.
[3] E. Witten, Phys. Rev. D 16 (1977) 2991;
O. Alvarez, Phys. Rev. D 17 (1978) 1123.
[4] E. Witten, Nucl. Phys. B 149 (1979) 285.
[5] A. DAdda, P. Di Vecchia, M. Lscher, Nucl. Phys. B 152 (1979) 125.
[6] K. Higashijima, T. Kimura, M. Nitta, M. Tsuzuki, Prog. Theor. Phys. 105 (2001) 261, hepth/0010272.
[7] I.Ya. Arefeva, Ann. Phys. 117 (1979) 393;
I.Ya. Arefeva, S.I. Azakov, Nucl. Phys. B 162 (1980) 298;
B. Rosenstein, B.J. Warr, S.H. Park, Nucl. Phys. B 336 (1990) 435.
[8] V.G. Koures, K.T. Mahanthappa, Phys. Rev. D 43 (1991) 3428;
V.G. Koures, K.T. Mahanthappa, Phys. Rev. D 45 (1992) 580.
[9] M. Ciuchini, J.A. Gracey, Nucl. Phys. B 454 (1995) 103, hepth/9508176;
T. Inami, Y. Saito, M. Yamamoto, Prog. Theor. Phys. 103 (2000) 1283, hepth/0003013;
T. Inami, Y. Saito, M. Yamamoto, Phys. Lett. B 495 (2000) 245, hepth/0008195;
T. Inami, Y. Saito, M. Yamamoto, Mod. Phys. Lett. A 16 (2001) 1643, hepth/0106204.
[10] S. Coleman, Commun. Math. Phys. 31 (1973) 259.
[11] B. Zumino, Phys. Lett. 87B (1979) 203.
161
[12] J. Wess, J. Bagger, Supersymmetry and Supergravity, second ed., Princeton Univ. Press, Princeton, 1992.
[13] N. Seiberg, E. Witten, Nucl. Phys. B 426 (1994) 19, hepth/9407087;
N. Seiberg, E. Witten, Nucl. Phys. B 431 (1994) 484, hepth/9408099;
P.C. Argyres, M.R. Plesser, N. Seiberg, Nucl. Phys. B 471 (1996) 159, hepth/9603042.
[14] S. Aoyama, Nuovo Cimento A 57 (1980) 176.
[15] J.S. Schwinger, Phys. Rev. 125 (1962) 397.
[16] K. Higashijima, M. Nitta, Prog. Theor. Phys. 103 (2000) 635, hepth/9911139.
[17] K. Higashijima, M. Nitta, Prog. Theor. Phys. 103 (2000) 833, hepth/9911225.
[18] K. Higashijima, M. Nitta, in: H. Suganuma, et al. (Eds.), Proceedings of Confinement 2000, World Scientific,
Singapore, 2001, pp. 279286, hepth/0006025;
K. Higashijima, M. Nitta, in: C.S. Lim, et al. (Eds.), Proceedings of ICHEP2000, World Scientific, Singapore, 2001, pp. 13681370, hepth/0008240.
[19] K. Higashijima, E. Itou, M. Tsuzuki, in preparation.
[20] A. Borel, Proc. Natl. Acad. Sci. 40 (1954) 1147.
[21] M. Bando, T. Kuramoto, T. Maskawa, S. Uehara, Phys. Lett. B 138 (1984) 94;
M. Bando, T. Kuramoto, T. Maskawa, S. Uehara, Prog. Theor. Phys. 72 (1984) 313;
M. Bando, T. Kuramoto, T. Maskawa, S. Uehara, Prog. Theor. Phys. 72 (1984) 1207.
[22] T. Kugo, Soryushiron Kenkyu (Kyoto) 95 (1997) C56;
T. Kugo, in: J. Nishimura, K. Yamawaki (Eds.), Proceedings of 1996 International Workshop on
Perspectives of Strong Coupling Gauge Theories, SCGT 96, World Scientific, Singapore, 1996,
http://www.phys.nagoyau.ac.jp/Scgt/proc/.
[23] M. Bordemann, M. Forger, H. Rmer, Commun. Math. Phys. 102 (1986) 605.
[24] K. Itoh, T. Kugo, H. Kunitomo, Nucl. Phys. B 263 (1986) 295;
K. Itoh, T. Kugo, H. Kunitomo, Prog. Theor. Phys. 75 (1986) 386.
[25] W. Buchmller, O. Napoly, Phys. Lett. B 163 (1985) 161.
[26] M. Bando, T. Kugo, K. Yamawaki, Phys. Rep. 164 (1988) 217.
[27] S. Aoyama, Nucl. Phys. B 578 (2000) 449, hepth/0001160.
[28] A.J. Buras, W. Slominski, Nucl. Phys. B 223 (1983) 157.
[29] W. Lerche, Nucl. Phys. B 238 (1984) 582.
[30] G.M. Shore, Nucl. Phys. B 248 (1984) 123.
[31] W. Buchmller, U. Ellwanger, Phys. Lett. B 166 (1985) 325.
[32] U. Ellwanger, Nucl. Phys. B 281 (1987) 489;
U. Ellwanger, Fortschr. Phys. 36 (1988) 881.
[33] W. Buchmller, W. Lerche, Ann. Phys. 175 (1987) 159.
[34] A.C. Kotcheff, G.M. Shore, Int. J. Mod. Phys. A 4 (1989) 4391;
G.M. Shore, Nucl. Phys. B 320 (1989) 202;
G.M. Shore, Nucl. Phys. B 334 (1990) 172.
[35] K. Higashijima, M. Nitta, K. Ohta, N. Ohta, Prog. Theor. Phys. 98 (1997) 1165, hepth/9706219.
[36] M. Nitta, Int. J. Mod. Phys. A 14 (1999) 2397, hepth/9805038.
[37] K. Furuta, T. Inami, H. Nakajima, M. Nitta, Prog. Theor. Phys. 106 (2001) 851, hepth/0106183.
[38] T. Kugo, I. Ojima, T. Yanagida, Phys. Lett. B 135 (1984) 402;
W. Lerche, Nucl. Phys. B 246 (1984) 475.
[39] J. Bagger, E. Witten, Phys. Lett. B 118 (1982) 103.
[40] T. Itoh, P. Oh, C. Ryou, Phys. Rev. D 64 (2001) 045005, hepth/0101041.
[41] L. AlvarezGaum, Nucl. Phys. B 184 (1981) 180;
C.M. Hull, Nucl. Phys. B 260 (1985) 182;
L. AlvarezGaum, P. Ginsparg, Commun. Math. Phys. 102 (1985) 311.
[42] K. Hori, C. Vafa, Mirror symmetry, hepth/0002222.
[43] V.A. Novikov, M.A. Shifman, A.I. Vainshtein, V.I. Zakharov, Phys. Lett. B 139 (1984) 389;
A.Y. Morozov, A.M. Perelomov, M.A. Shifman, Nucl. Phys. B 248 (1984) 279.
[44] K. Higashijima, E. Itou, Prog. Theor. Phys. 108 (2002) 737, hepth/0205036;
K. Higashijima, E. Itou, Prog. Theor. Phys. 109 (2003) 751, hepth/0302090;
K. Higashijima, E. Itou, Prog. Theor. Phys. 110 (2003) 563, hepth/0304194.
162
[45] K. Higashijima, M. Nitta, Prog. Theor. Phys. 105 (2001) 243, hepth/0006027;
K. Higashijima, E. Itou, M. Nitta, Prog. Theor. Phys. 108 (2002) 185, hepth/0203081.
[46] K. Higashijima, T. Kimura, M. Nitta, Phys. Lett. B 515 (2001) 421, hepth/0104184.
[47] K. Higashijima, T. Kimura, M. Nitta, Phys. Lett. B 518 (2001) 301, hepth/0107100;
K. Higashijima, T. Kimura, M. Nitta, Nucl. Phys. B 623 (2002) 133, hepth/0108084;
K. Higashijima, T. Kimura, M. Nitta, Ann. Phys. 296 (2002) 347, hepth/0110216.
[48] K. Higashijima, T. Kimura, M. Nitta, Nucl. Phys. B 645 (2002) 438, hepth/0202064;
K. Higashijima, T. Kimura, M. Nitta, in: S. Bentvelsen, et al. (Eds.), Proceedings of the 31st International Conference on High Energy Physics, ICHEP2002, Elsevier, Amsterdam, 2002, pp. 867869, hepth/0210034.
[49] M. Nitta, Noncompact CalabiYau metrics from nonlinear realizations, hepth/0309004.
[50] L. AlvarezGaum, D.Z. Freedman, Commun. Math. Phys. 80 (1981) 443.
[51] U. Lindstrm, M. Rocek, Nucl. Phys. B 222 (1983) 285.
[52] N.J. Hitchin, A. Karlhede, U. Lindstrm, M. Rocek, Commun. Math. Phys. 108 (1987) 535.
[53] E. Calabi, Ann. Sci. cole Norm. Sup. 12 (1979) 269.
[54] T.L. Curtright, D.Z. Freedman, Phys. Lett. B 90 (1980) 71;
L. AlvarezGaum, D.Z. Freedman, Phys. Lett. B 94 (1980) 171;
M. Rocek, P.K. Townsend, Phys. Lett. B 96 (1980) 72.
[55] M. Arai, M. Naganuma, M. Nitta, N. Sakai, Nucl. Phys. B 652 (2003) 35, hepth/0211103;
M. Arai, M. Naganuma, M. Nitta, N. Sakai, BPS wall in N = 2 SUSY nonlinear sigma model with Eguchi
Hanson manifold, in: Garden of Quantain Honor of Hiroshi Ezawa, World Scientific, Singapore, 2003,
pp. 299325, hepth/0302028.
[56] A. Dancer, A. Swann, J. Geom. Phys. 21 (1997) 218.
[57] M. Arai, M. Nitta, N. Sakai, Vacua of massive hyperKhler sigma models of nonAbelian quotient, hepth/0307274, Prog. Theor. Phys., in press.
[58] H. Kanno, Y. Yasui, J. Geom. Phys. 43 (2002) 293, hepth/0108226;
H. Kanno, Y. Yasui, J. Geom. Phys. 43 (2002) 310, hepth/0111198.
Abstract
We study gauged linear sigma models for noncompact CalabiYau manifolds described as a line
bundle on a hypersurface in a projective space. This gauge theory has a unique phase if the Fayet
Iliopoulos parameter is positive, while there exist two distinct phases if the parameter is negative.
We find four massless effective theories in the infrared limit, which are related to each other under
the CalabiYau/LandauGinzburg correspondence and the topology change. In the Tdual theory, on
the other hand, we obtain two types of exact massless effective theories: one is the sigma model on
a newly obtained CalabiYau geometry as a mirror dual, while the other is given by a Landau
Ginzburg theory with a negative power term, indicating N = 2 superconformal field theory on
SL(2, R)/U (1). We argue that the effective theories in the original gauged linear sigma model are
exactly realized as N = 2 Liouville theories coupled to welldefined LandauGinzburg minimal
models.
2005 Elsevier B.V. All rights reserved.
PACS: 11.15.Ex; 11.25.Hf; 11.25.Pm
1. Introduction
Field theory in twodimensional spacetime is one of the most powerful tools for analyzing dynamical phenomena in particle physics. It has been studied as a toy model of
Email address: tetsuji@post.kek.jp (T. Kimura).
05503213/$ see front matter 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.nuclphysb.2005.01.029
164
low energy effective theory including symmetry breaking mechanism. Nonlinear sigma
model (NLSM) on the projective space is a typical example to investigate chiral symmetry
breaking [1,2]. String worldsheet theory is also described as a twodimensional conformal
field theory (CFT). Conformal invariance in the worldsheet theory gives a set of equations
of motion of spacetime physics [3]. When we discuss string theory on curved spacetime,
supersymmetric NLSMs play significant roles.
Coupled to a gauge field, twodimensional field theory has been applied to more complicated physics. The gauge field plays a key role in the compactification of the target
space via the Higgs mechanism. This theory is also useful to study mathematical problems such as Morse theory [4] and mirror symmetry [5]. Many people have constructed
twodimensional supersymmetric gauge theories in order to understand various kinds of
physical and mathematical structures.
Higashijima and Nitta formulated supersymmetric NLSMs on hermitian symmetric
spaces (HSSs), which are specific Khler manifolds, starting from supersymmetric gauge
theories with four supercharges [6]. By using this, we constructed Khler metrics on complex line bundles over compact EinsteinKhler manifolds [7,8]. These noncompact Khler
manifolds have vanishing Ricci tensors, and hence are CalabiYau (CY) manifolds [9,10].
In attempt to investigate the gauge/gravity duality in string theory [1113] on these CY
manifolds, however, it is indispensable to understand global aspects such as cohomology
classes.
On the other hand, it is a wellknown fact that the gauged linear sigma model (GLSM)
is useful to investigate worldsheet string theories on toric varieties [14]. In this framework,
we can understand some global properties of the CY manifolds. GLSM includes at least
two kinds of SCFTs in the infrared (IR) limit. One is a supersymmetric NLSM on CY
manifold and the other is an N = 2 LandauGinzburg (LG) theory. We can read the cohomology ring of the CY manifold from the chiral ring derived from the LG superpotential
[15]. Furthermore, its Tdual theory provides us with mirror descriptions of the original
geometry [16,17].
Since each HSS is constructed as a submanifold of a complex projective space or of
a Grassmannian, we can, in general, construct the GLSMs for the line bundles on HSSs.
Investigating LG theories in the IR limit of the GLSMs, we will be able to understand
cohomology rings of the noncompact CY manifolds. Unfortunately, however, it is difficult
to embed the sigma models on HSSs into the GLSMs. Thus we study the GLSM for a line
bundle of homogeneous hypersurface in the projective space whose Ricci tensor vanishes
[18]. This noncompact manifold is represented as O(N + ) bundle on CPN 1 [], which
can be seen as a toy model of the line bundles on HSSs.
In this paper we will find the following theories in the IR limit: NLSMs on CY manifolds, orbifolded LG theories, gauged WessZuminoWitten (WZW) models on coset
SL(2, R)/U (1) and Liouville theories. They appear as N = 2 SCFTs. The former two
theories give unitary conformal field theory. Under the conformal invariance, the sigma
model and the LG theory become appropriate SCFTs from differential geometric and
algebrogeometric points of view, respectively. They often emerge when we analyze superstring theory on compact manifolds [19,20]. The latter two theories are slightly different.
These theories appear in string theory on noncompact curved spacetime such as a twodimensional black hole [21,22]. They are also utilized in noncritical string theory and
165
matrix model [23,24]. It is quite important to study all four SCFTs simultaneously when
we consider string theories on noncompact CY manifolds. Thus we study the GLSM for
noncompact CY manifolds including all the above four theories in the low energy limit.
This paper is organized as follows. In Section 2 we study the GLSM for line bundles
and discuss how massless effective theories appear in some specific limits. In this analysis
we find that there exist two distinct phases in the negative FI parameter region. This phenomenon newly appears, while other wellknown GLSMs do not include this. In Section 3
we discuss the Tdual of the GLSM. We obtain two types of exact effective theories, the
sigma models on newly constructed mirror CY geometries and the LG theories with negative power. There we discuss the exact effective theories in the original GLSM. We devote
Section 4 to the summary and discussions. In Appendix A we introduce conventions of
N = 2 supersymmetry in twodimensional spacetime. In Appendix B we review a definition of weighted projective space. Finally, we briefly introduce the linear dilaton CFT and
discuss an interpretation of LG superpotential with a negative power term in Appendix C.
This argument is useful to understand the LG theories in Section 3.
V V = V + i( ),
a a = e+2iQa a ,
= ,
where Qa is a U (1) charge of the chiral superfield a . For convenience, we restrict these
charges to integers: Qa Z. The complexified gauge parameters are described by a chiral
, ), respectively: D = 0, D = 0.
and an antichiral superfields (x, , ) and (x,
By using superfields, we construct a supersymmetric gauge invariant Lagrangian:
1
+
LGLSM = d4 2
a e2Qa V a
e
a
1
2
2
+
d W () + c.c. +
d WGLSM (a ) + c.c. ,
2
where we assume that all chiral superfields have nonzero U (1) charges Qa = 0 because
a neutral chiral superfield is completely free from the system. The abelian gauge coupling
constant e, which appears in front of the kinetic term of , has mass dimension one. There
exist two types of superpotentials. One is a superpotential written as WGLSM (a ). This is a
holomorphic function of chiral superfields a . The other is called a twisted superpotential
W () described as
W () = t,
t = r i,
166
where t is a complex parameter defined by the FayetIliopoulos (FI) parameter r and the
thetaangle . We also refer t to the (complexified) FI parameter.
We are interested in supersymmetric low energy effective theories. Thus we need to
study the potential energy density U() described by the scalar component fields of superfields:
e2
Fa 2 + U (),
U() = D2 +
(2.1a)
2
a
1
D=r
Qa a 2 ,
2
e
a
2
Q2a a 2 ,
U () := 2 
D :=
Fa =
WGLSM (),
a
(2.1b)
M := (a ) Cn D = Fa = U = 0 U (1),
where n is the number of scalar component fields in the GLSM. The dividing U (1) group
indicates the abelian gauge symmetry. Since we consider N = (2, 2) supersymmetric theories, the manifold M becomes a Khler manifold [25] where the FI parameter denotes
the scale of M.
Under a generic configuration for chiral superfields a of charges Qa , the FI parameter
r receives a renormalization via wavefunction renormalizations of a . Thus the bare FI
parameter r0 is related to the renormalized one rR under the following equation:
UV
,
r0 = rR +
(2.2)
Qa log
a
where UV and are the ultraviolet cutoff and the scale parameter, respectively. Thus we
observe that the scale of M changes under the renormalization group (RG) flow. Studying
the function of the FI parameter derived from (2.2), we find whether the effective theories
expanded on M are asymptotically free or not. In particular, if we impose
Qa = 0,
(2.3)
a
the FI parameter does not receive the renormalization. Thus there appears a nontrivial
conformal field theory in the IR limit. From the geometric point of view, the sum a Qa
is equal to the first Chern class c1 (M) of the vacuum manifold M. If the condition (2.3)
is satisfied on M, this manifold becomes a CY manifold. Thus we refer (2.3) to the CY
condition. In attempt to study CY manifolds, we impose this on the GLSM.
We usually study how massless effective theories are realized on the supersymmetric
vacuum in M. Recall that in twodimensional field theory the massless modes are not
167
welldefined because of the IR divergence in their twopoint functions. The Colemans theorem on nonexistence of NambuGoldstone modes [26] is closely related to this difficulty.
In order to avoid this problem, we assume that there exists an IR cutoff parameter. Furthermore we take the large volume limit r when we consider a NLSM whose target
space is the vacuum manifold M. In this limit the FI parameter r of GLSM can be related
to the coupling constant g of the NLSM:
r=
1
.
g2
This means that the large volume limit r is the weak coupling limit g 0.
Next we consider fluctuation fields around the vacuum. It is so complicated to analyze
massless/massive fluctuation modes that we perform here a general calculation. Let us
decompose scalar component fields a into three kinds of variables:
a = a + a + a ,
(2.4)
d2 x (a + a ) = 0,
where a , a and a mean the vacuum expectation values (VEVs), the fluctuation modes
tangent and nontangent to the vacuum manifold M, respectively. They satisfy the following relations:
F ()VEV := F 0,
(2.5a)
F ()
()
:=
a
0,
F
(2.5b)
VEV
a
a
F ()
()
F
(2.5c)
:=
a
= 0.
VEV
a
a
Note that F () are the set of functions given by (2.1b): F = {D, Fa , U }. The symbols
and denote holomorphic variations with respect to the complex variables a . Of course
the VEVs a satisfy Eq. (2.5a). Eq. (2.5b) provides that the first order variations of F ()
with respect to the fluctuation modes a vanish. This is nothing but the definition that a
move only tangent to the vacuum manifold M. The third equation (2.5c) means that the
other fluctuation modes a do not propagate tangent to M. Substituting (2.4) and (2.5) into
the potential energy density U() described by (2.1), we investigate the behaviors of the
low energy effective theories. If Eqs. (2.5a) and (2.5b) furnish nontrivial relations among
the fluctuation modes a , these modes constitute a supersymmetric NLSM whose target
space is M. However, if these equations are trivially satisfied, a are free from constraints
and propagate on a flat space with potential energy. Then we find that a field theory appears
described by a superpotential of fluctuation fields such as a LG superpotential WLG .
2.2. Field configuration and supersymmetric vacuum manifold
Now we are ready to analyze massless low energy effective theories in the GLSM for
O(N + ) bundle on CPN 1 []. We consider a U (1) gauge theory with N + 2 chiral
168
S1
1
...
...
SN
1
P1 P2
N +
(2.6)
(2.7)
By definition, the numbers N and are positive integers: , N Z>0 . We assume that these
two integers satisfy 1 N 1 and 2 N . The sum of all charges Qa vanishes (2.3)
in order to obtain nontrivial SCFTs on the CY manifold.
Now we consider the potential energy density and look for supersymmetric vacua. Imposing the WessZumino gauge, we write down the bosonic part of the potential energy
density U():
U() =
N
2
e2 2
p1 i G (s)2 + U (),
D + G (s) +
2
(2.8a)
i=1
D=r
N
i=1
U () = 2 
N
(2.8b)
(2.8c)
i=1
Imposing zero on them, we obtain the supersymmetric vacuum manifold M. Since the Lagrangian has N = (2, 2) supersymmetry and the single U (1) gauge symmetry, the vacuum
manifold becomes a Khler quotient space:
M = (a ) CN +3 D = G = p1 i G = U = 0 U (1).
(2.9)
In attempt to study effective theories, we choose a point on M as a vacuum and give VEVs
of scalar component fields: a a . Then we expand the fluctuation modes around the
vacuum. In general, the structure of M is different for r > 0, r = 0 and r < 0 and there
appear various phases in the GLSM. The phase living in the r > 0 region is referred to the
CY phase, and the phase in r < 0 is called to the orbifold phase. A singularity of the
model emerges in the phase at r = 0. Thus we sometimes call this the singularity phase.
We will treat these three cases separately. We comment that in each phase the vacuum
manifold is reduced from the original M. We often refer the reduced vacuum manifold to
Mr M. The VEVs of the respective phases can be set only in Mr .
2.3. CalabiYau phase
In this subsection we analyze the CY phase r > 0. In this phase, D = 0 requires
some si cannot be zero and therefore must vanish. If we assume p1 = 0, the equations
169
G (s) = i G (s) = 0 with the condition (2.7) imply that all si must vanish. However, this
is inconsistent with D = 0. Thus p1 must be zero. The variable p2 is free as long as the
condition D = 0 is satisfied. Owing to these, the vacuum manifold M is reduced to MCY
defined by
2
N
1
+
si i G s +
(s + s )i1 (s + s )ik i1 ik G s
k!
i=1
k=2
i1 ,,ik
2
1
1
+ p 1 2
(s + s )j1 (s + s )jk i j1 jk G s
i G s +
k!
j1 ,...,jk
i=1
k=2
N
2
2
2
2
2
2
si + si + si + p 1  + (N ) p2 + p 2 + p 2 .
+ 2 
N
i=1
Fluctuation modes si and p 2 remain massless and move only tangent to MCY because they
VEV = G
VEV = 0. The variation (p
1 i G )VEV = 0 indicates p 1 = 0.
are subject to D
The modes , p 1 , si and p 2 have mass m2 = O(e2 r). The gauge field vm also acquires mass
of order O(e2 r) by the Higgs mechanism. The fermionic superpartners behave in the same
way as the scalar component fields because of preserving supersymmetry. In the IR limit
e and the large volume limit r , the massive modes decouple from the system.
Thus we obtain
N = (2, 2) supersymmetric NLSM on MCY
(2.11)
170
Table 1
Classification of O(N + ) bundle on CPN 1 []
Degree
=1
2N 1
as a massless effective theory. Notice that this description is only applicable in the large
volume limit because the NLSM is welldefined in the weak coupling limit from the viewpoint of perturbation theory. This effective theory becomes singular if we take the limit
r +0 because the decoupled massive modes becomes massless. This phenomenon also
appears in the SeibergWitten theory [27,28], the black hole condensation [29,30], and
so on.
Let us make a comment on the target space MCY . By definition, the number means the
degree of the vanishing polynomial G (s) = 0, which gives a hypersurface in the projective
space CPN 1 . We can see that if = 1, G=1 (s) = 0 gives a linear constraint with respect
to the homogeneous coordinates si and the hypersurface CPN 1 [ = 1] is reduced to (N
2)dimensional projective space CPN 2 . This reduction does not occur if 2 N 1.
Here we summarize the shape of the target space MCY in Table 1.
Although the = 1 case has been already analyzed in the original paper [14], the other
cases 2 N 1 are the new ones which have not been analyzed.
2.4. Orbifold phase
Here we consider the negative FI parameter region r < 0. In this region the total vacuum
manifold (2.9) is restricted to a subspace defined by
1
2
Mr<0 := (p1 , p2 ) C D = 0, r < 0 U (1),
(2.13b)
2
N +1
M
(2.13c)
D = G = 0, r < 0 U (1).
:= (p2 ; si ) C
r<0
In the former subspace the condition (2.7) is trivially satisfied whereas in the latter subspace
it is satisfied nontrivially. Both of the two subspace include a specific region p1 = si = 0.
171
2
2
i G (s + s )2
+ G (s + s ) + p1 + p 1 + p 1
+ 2 2
2
2
si + si 2 + 2 p1 + p 1 + p 1 + (N )2 p2 + p 2 + p 2 ,
where p1 and p2 are VEVs of scalar components of P1 and P2 , respectively. They live
in the weighted projective space (2.13b). Because the VEVs of si are all zero, the U (1)
gauge symmetry is spontaneously broken to Z , where is the great common number
between and N : = GCM{, N }. This potential energy density provides that all
fluctuation modes si and si appear as linearly combined forms such as si + si , which do
not acquire any mass terms. The modes p 1 and p 2 remain massless and move tangent to
the subspace (2.13b). The other fluctuation modes acquire mass of order m2 = O(e2 r).
Thus, in the IR limit e , all the massive modes are decoupled from the system. Thus
we obtain the following massless effective theory:
N = (2, 2) supersymmetric NLSM on WCP1,N
coupled to LG theory with WLG = p1 + P1 G (S) Z ,
(2.14)
where P1 and Si are massless chiral superfields. Note that the sigma model sector also contains the Z orbifold symmetry coming from the property of WCP1,N . As is well known
that the term p1 G (S) forms an ordinary LG superpotential. Thus in the IR limit we can
interpret that this term is marginal and flows to the N = (2, 2) minimal model. The second
term P1 G (S) is somewhat mysterious. Since this term has not any isolated singularities
we might not obtain welldefined unitary CFT. This difficulty causes the noncompactness
of the manifold MCY which appears in the CY phase.
There are two specific points in the subspace WCP1,N . One is the point p2 = 0 and
the other is p1 = 0. In the former point the gauge symmetry is enhanced to Z . Furthermore, the mode p 1 disappears and p 2 becomes massless, which combines with a massless
fluctuation modes p 2 linearly. This combined mode is free from any constraints. The
other massless modes si + si in (2.14) remain massless and are also free from constraints.
Thus in the IR limit and the large volume limit, the massless effective theory becomes an
N = (2, 2) supersymmetric theory as
CFT on C1 LG theory with WLG = p1 G (S) Z .
(2.15)
172
LG theory with WLG = P1 G (S) on CN +1 ZN ,
(2.16)
which consists of N + 1 massless chiral superfields such as P1 and Si . This theory is not a
welldefined LG theory because the superpotential WLG has no isolated singularities. We
interpret the defect of isolated singularities as a noncompactness of the manifold MCY in
the CY phase via CY/LG correspondence (if this correspondence is satisfied in the case
of sigma models on noncompact CY manifolds). This property prevents from calculating
a chiral ring of this model in the same way as unitary LG minimal models describing
compact CY manifolds [15].
Here we study massless effective theories on the subspace M2r<0 defined in (2.13c).
As mentioned before, there are nontrivial constraints in M2r<0 . Thus, as we shall see, the
effective theories are also under these constraints. In the same way as discussed before,
we choose one point in the subspace M2r<0 and make all the scalar fields fluctuate around
it. Then we write down the expanded potential energy density (2.8) in terms of VEVs and
fluctuation modes a , a and a :
e2
2 Re
U=
si si + (N )p 2 p 2
2
i
2
si + si 2 + p 1 2 + (N )p 2 + p 2 2
2
1
+
si i G s +
(s + s )i1 (s + s )ik i1 ik G s
k!
i
k=2
i1 ,...,ik
2
1
1
2
+ p 1 
(s + s )j1 (s + s )jk i j1 jk G s
i G s +
k!
i
j1 ,...,jk
k=1
si + si + si 2 + 2 p 1 2 + (N )2 p2 + p 2 + p 2 2 .
+ 2 2
i
173
This potential energy density indicates the following: the fluctuation modes si , p 1 and p 2
are massive; si and p 2 move tangent to M2r<0 . Thus, taking e and r , we
obtain
N = (2, 2) supersymmetric NLSM on M2r<0 .
(2.17)
In this theory there exist P2 and Si as massless chiral superfields, which move tangent to
M2r<0 . Notice that in general points in M2r<0 the U (1) gauge symmetry is completely
broken because of the existence of nonzero VEVs si . However, taking si = p1 = 0
and p2 = 0 in the subspace M2r<0 , we find that the gauge symmetry is partially restored
to ZN .
Even though the vacuum manifolds M1r<0 and M2r<0 are connected on p1 = si = 0,
the effective theories given by (2.16) and (2.17) are quite different from each other. The
reason is that while the subspace M1r<0 is free from constraints G = p1 i G = 0, in the
subspace M2r<0 these constraints are still valid on the region p1 = si = 0. On account of
the existence of these constraints, a phase transition occurs when the theory moves from
one to the other. Thus we conclude that a new phase appears on the subspace M2r<0 , which
has not been discovered in wellknown GLSMs such as the models for O(N ) bundle on
CPN 1 , for CPN 1 [N], for resolved conifold, and so on. We refer this phase to the 3rd
phase. Here we refer the phase on M1r<0 to the orbifold phase, as usual.
2.4.2. Effective theories of = 2
Let us consider the orbifold phase of = 2. In the same way as the previous analysis,
the constraints G = p1 i G = 0 and the property (2.7) decompose the manifold Morbifold
into two subspaces:
Morbifold =2 = M1r<0 M2r<0 ,
M1r<0 := (p1 , p2 ) C2 D = 0, r < 0 U (1) WCP12,N 2 ,
(2.18a)
(2.18b)
(2.18c)
These two subspaces are glued in the region given by p1 = si = 0. Although this situation
is same as to the case of 3 N 1, the appearing massless effective theories are quite
different.
Here let us analyze the effective theories on the subspace M1r<0 = WCP12,N 2 . We
choose a point in this subspace as a supersymmetric vacuum and take VEVs of all scalar
fields. Then we make all scalar fields fluctuate around the VEVs. Fluctuation modes p 1
and p 2 are subject to the constraints such that they move only tangent to WCP12,N 2 . The
fluctuation modes si have no degrees of freedom because of the variation of the constraint
p1 i G2 = 0. (In the case of (2.13b), the equations G = 0 and p1 i G = 0 are trivially
satisfied in WCP1,N . These variations are also trivial. However, the case of = 2 is
quite different. By definition, some i j G2 must have nonzero values. Thus even though
the above equations are trivially satisfied in the subspace WCP12,N 2 , their variations give
nontrivial constraints on the fluctuation modes.) Under these conditions we write down
the potential energy density (2.8) in terms of VEVs a and fluctuation modes a and a :
174
U=
e2
2 Re 2p 1 p 1 + (N 2)p 2 p 2
2
2
2
2
2
2
2
i G2 (s )2
+ G2 (s ) + p1 + p 1 + p 1
+ 2 2
2
2
si 2 + 4p1 + p 1 + p 1 + (N 2)2 p2 + p 2 + p 2 .
This function denotes the following: the fluctuation modes si , p 1 and p 2 acquire masses
m2 = O(e2 r); the modes p 1 and p 2 remain massless and move tangent to WCP12,N 2 .
Thus taking e and r , we obtain the massless effective theory described by
N = (2, 2) supersymmetric NLSM on WCP12,N 2 .
(2.19)
This sigma model has Z orbifold symmetry coming from the property of
where = GCM{2, N 2}. This effective theory does not include massless LG theory. The
reason is that the degree two polynomial G2 generates mass terms such as
p1 2 i i G2 2 . (See, for example, [31].)
Now we consider the effective theory on two specific points in WCP12,N 2 like (2.15)
and (2.16). Expanding the theory on the one point (p1 , p2 ) = (p1 , 0), the gauge symmetry
is partially restored to Z2 . Thus we obtain the effective theory on this specific point as
WCP12,N 2 ,
(2.20)
Note that this theory can possess the LG theory with a quadratic superpotential WLG =
p1 G2 (S), which gives massive modes of Si .
The effective theory drastically changes if we expand the theory on another point
(p1 , p2 ) = (0, p2 ) in WCP12,N 2 . On this point, the broken gauge symmetry is enhanced to
ZN 2 and the fluctuation modes si become massless with being free from any constraints.
Both p 1 and p 1 are massless and linearly combined in the potential energy density. The remaining field p 2 becomes zero because there exists a nontrivial variation of the constraint
D = 0. Summarizing these results, we find that the following massless effective theory
appears in the limit e, r :
N = (2, 2) LG theory with WLG = P1 G2 (S) on CN +1 ZN 2 .
(2.21)
Although this superpotential also has no isolated singularities, this theory should describe
a nontrivial SCFT. We shall return here in later discussions.
We next study the massless effective theories on the subspace M2r<0 defined in (2.18c).
The potential energy density is obtained as
e2
2 Re
U=
si si + (N 2)p 2 p 2
2
i
2
si + si 2 + 2p 1 2 + (N 2)p 2 + p 2 2
175
1
2
+
si i G2 s +
(s + s )i (s + s )j i j G2 s
2!
i
i,j
2
2
+ p 1 
(s + s )j i j G2 s
i G2 s +
i
si + si + si 2 + 4p 1 2 + (N 2)2 p2 + p 2 + p 2 2
+ 2 2
i
under the following constraints on fluctuation modes: the fluctuations si and p 2 move tangent to M2r<0 ; the other tangent mode p 1 is zero; the fluctuations a are all massive of
m2 = O(e2 r). Thus the effective theory expanded around generic points in M2r<0 becomes
N = (2, 2) supersymmetric NLSM on M2r<0
(2.22)
in the IR and large volume limit: e, r . The U (1) gauge symmetry is completely
broken if some si = 0 exist. On the other hand, if we expand the theory on a specific
point p1 = si = 0, the gauge symmetry is partially restored to ZN 2 .
So far we have studied the effective theories on all regions of the vacuum manifold
Morbifold of = 2. From the same reason discussed in the case of 3 N 1, there
exists a phase transition between the theories (2.21) and (2.22) because of the nontrivial
constraint coming from the variation of the equation G2 = 0. Thus we find that the GLSM
for the O(N + 2) bundle on CPN 1 [2] also includes two phases in the negative FI parameter region. The phase on (2.19) is called the orbifold phase, and we refer the phase on
(2.22) to the 3rd phase.
Here we illustrate the relation among the phases in the GLSM schematically in Fig. 1.
In the large volume limit r , these three effective theories (2.11), (2.14) and
(2.17) become welldefined. In later discussions we shall consider how these effective theories deform in the small FI parameter limit r 0. There we must consider the singular
phase [14].
Fig. 1. Various phases in GLSM for O(N + ) bundle on CPN 1 [] with 2 N 1. The axes with
thin/thick lines represent the vacuum space coordinates in the positive/negative FI parameter regions, respectively.
176
2
i G1 s 2
+ G1 (s ) + p 1 2
i
2
2
2
2
si + si + si + p 1  + (N 1) p2 + p 2 + p 2 .
+ 2 
i
This indicates that the modes si and p 2 remain massless whereas the modes si , p 1 and p 2
become massive. Thus the following massless effective theory appears in the limit e, r
:
N = (2, 2) supersymmetric NLSM on M2r<0 ,
(2.24)
where the U (1) gauge symmetry is completely broken because si = 0. While if we set
the VEVs to si = 0, the broken U (1) gauge symmetry is enhanced to ZN 1 . In addition
the modes si become massless and are combined with the tangent modes si , which are still
under constraint G1 = 0. The mode p 2 becomes zero, which is derived from the variation
of D = 0. In this specific point we can see that the space M2r<0 is deformed to CN 1 /ZN 1
and the effective theory are
N = (2, 2) SCFT on CN 1 /ZN 1 .
(2.25)
The two effective theories (2.24) and (2.25) are smoothly connected without any phase
transitions coming from the variations of constraints. Thus we find that in the = 1 case
there exists only one phase in the negative FI parameter region. We refer this to the orbifold
phase, as usual. Here we illustrate the schematic relation between the CY phase and the
orbifold phase in Fig. 2.
Note that the GLSM for O(N + 1) bundle on CPN 1 [1] is completely the same as
the one for O(N + 1) bundle on CPN 2 . This is because the hypersurface CPN 1 [1] is
nothing but CPN 2 . Thus the vacuum structure and phases are also equal to each other.
177
= 0,
(2.26b)
which is satisfied only on r = 0. Although the first solution (2.26a) appears in each GLSM,
the second solution (2.26b) does not satisfy the supersymmetric vacuum condition U() =
0 in some GLSMs, for example, the GLSM for CPN 1 [N ].
178
In the Coulomb branch, we can set that a = 0 and is free. This solution appears
only when the FI parameter vanishes. If we choose to be zero, i.e., all the VEVs of
scalar fields vanish a = 0, the Coulomb branch connects with the Higgs branch.
Let us consider a massless effective theory in the Coulomb branch. Since the scalar field
has mass dimension one, we take the VEV to be very large. Owing to this, all chiral
superfields Si , P1 and P2 acquire very large masses via U () in (2.8). Taking
and integrating out all massive fields1 , we obtain the following effective Lagrangian:
1
+
d2 W eff () + c.c. ,
Leff = d4 Keff (, )
2
Qa
1
Qa log
W eff () = t
a
= t log() (N ) log(N + ) .
Note that the twisted superpotential was deformed by the quantum effects coming from the
integration of massive fields. This effective Lagrangian presents the asymptotic form of the
potential energy density
Ueff ( ) =
2
eeff
e2
W eff ( )2 = eff t log() (N ) log(N + )2 .
2
2
(2.27)
In order that the effective theory remains supersymmetric, the potential energy density
must be zero in a specific value of . Notice that if the complexified FI parameter is given
by
t = log() + (N ) log(N + ),
(2.28)
the potential energy density becomes always zero. If it happens, the effective theory does
not have any mass gap and becomes singular as a twodimensional field theory. Thus (2.28)
is the quantum singular point of the GLSM. In the classical point of view, the value t = 0
looks like a singular point in the theory. Integrating out the massive fields, we find that
the singular point moves to (2.28). The massless effective theories in Coulomb and Higgs
branches are connected with each other avoiding this singular point.
2.5.2. CY/LG correspondence and topology change
As mentioned before, the massless effective theories are only valid if we take the FI
parameter to be infinitely large r . In this limit the effective theories are (partly)
described by the NLSMs. However, if we change the FI parameter to be small, the NLSM
representations are no longer welldefined and must be deformed. This phenomenon has
been already studied in [14] as following: if the FI parameter goes to zero r 0, the effective theory on the CY phase moves to the theory on the Coulomb branch in the singularity
phase avoiding the singular point. Furthermore, the effective theory connects to the LG
theory in the orbifold phase when r .
1 Here we can integrate out because the superpotential W
a
GLSM does not contribute to the deformation of
the effective twisted superpotential W eff (). For the precise derivation, see Chapter 15 in [17].
179
The above phenomenon suggests that, rather than the LG theory being equivalent to the
sigma model on the CY manifold, they are two different phases of the same system, i.e., the
system of the single GLSM. Thus the CY/LG correspondence [3238] can be read from
the phase transition. In fact, it has been proved that the sigma model on CP4 [5] and the Z5 orbifolded LG theory with WLG = G5 (S), which are equivalent to each other, appear as the
distinct phases in the single GLSM. Furthermore, the topology change is also understood
in the framework of the phase transition of the GLSM. The flop of the resolved conifold
O(1) O(1) CP1 is a typical example [14].
Now let us apply the above discussions to the GLSM for O(N + ) bundle on
CPN 1 []. For example, we consider the relations among the various phases of 3
N 1, where we have found three phases: the CY phase on MCY , the orbifold phase on
M1r<0 and the 3rd phase on M2r<0 . Furthermore, we have found four effective theories.
Let us discuss the relations among them:
The effective theory on the CY phase (2.11) and the theory on the 3rd phase (2.17)
are related to each other via a topology change, because the defining equations of the
target spaces are equal except for the sign of the FI parameter. Furthermore, these two
effective theories are both sigma models, which do not include the potential theory
sectors such as a LG theory.
(2.16) and (2.17) are connected at the point p1 = si = 0 in Morbifold . Since the former is the sigma model and the latter is the LG theory, there exists a phase transition
between these two theories, which are equivalent to each other by the CY/LG correspondence.
Both the theories (2.15) and (2.16) are on the weighted projective space and are included in the theory (2.14).
The LG theory (2.15) is equivalent to the sigma model (2.11) by the CY/LG correspondence.
Notice that the sigma model (2.11) is not related to (2.16) directly, because the theory
(2.16) has already been connected to (2.17) while the CY/LG correspondence connects
between two theories by onetoone. These connections are realized through the singularity
phase. Even though we wrote down the connections only from the qualitative point of view,
we can acquire nontrivial relations among the effective theories as illustrated in Fig. 3.
In the case of compact CY manifolds, we have already understood that the local rings
in the LG theory are identified with the chiral rings of the SCFT and these chiral rings are
related to the harmonic forms on such manifolds [15]. However, we have no proof that this
relation is also satisfied in the case of noncompact CY manifolds. Thus we must investigate
the spectra of the above effective theories as a future problem.
As discussed before, we have obtained various massless effective theories by decomposing all massive modes. Thus they are just approximate descriptions which must be
deformed if we can exactly integrate out massive modes. In the next section we will study
the Tdual theory of the GLSM [16]. This formulation is so powerful to obtain the exact effective theories. Analyzing them exact theories we will reinvestigate the massless effective
theories in the original GLSM.
180
Fig. 3. The relation among various phases around the singularity: a conjecture.
3. Tdual theory
In this section we consider Tdual of GLSMs. It is quite significant to study it because
we can obtain exact descriptions of the low energy effective theories. Furthermore, they
will also indicate how the exact effective theories are realized in the original GLSM. In
fact, in the original GLSM, we obtained just approximate effective theories. There we did
not perform integratingout but just decomposed all massive modes because it is generally
impossible to integrate them out. In the model proposed in [16], we calculate a function
which is directly related to the partition function. Thus we will obtain exact effective theories as quantum field theory.
3.1. General construction
Here we briefly review the Tduality of a generic GLSM without any superpotentials
[16]. We start from the following Lagrangian in twodimensional worldsheet:
1
1
2Qa V +Ba
L = d 2 +
(Ya + Ya )Ba
e
2
e
a
1
d2 (t) + c.c. ,
+
2
(3.1)
where Ya are twisted chiral superfields whose imaginary parts are periodic of period 2 .
We incorporate real superfields Ba as auxiliary fields.
Integrating out twisted chiral superfields Ya , we obtain D + D Ba = D+ D Ba = 0,
whose solutions are written in terms of chiral superfields a and a such as Ba = a + a .
When we substitute them into the Lagrangian (3.1), a GLSM Lagrangian appears:
1
+
a e2Qa V a
d4 2
e
a
1
2
d (t) + c.c.
+
2
LGLSM ,
181
L B
a
a =a +
(3.2)
where we rewrote a := ea . On the other hand, when we first integrate out Ba in the
original Lagrangian (3.1), we obtain
Ya + Ya
.
Ba = 2Qa V + log
(3.3)
2
Let us insert these solutions into (3.1). By using a deformation
1
1
d4 Qa V Ya = Qa d2 D + D V Ya = Qa d2 Ya ,
2
2
we find that a Lagrangian of twisted chiral superfields appears:
1
1
(Ya + Ya ) log(Ya + Ya )
LT = d 4 2
2
e
a
1
d2 W + c.c. ,
+
2
W =
Qa Ya t +
eYa .
a
(3.4a)
(3.4b)
This Lagrangian is Tdual of the gauge theory (3.2). Notice that the twisted superpotential
W is corrected by instanton effects where the instantons
are the vortices of the gauge
theory. In attempt to analyze a model satisfying a Qa = 0, the scale parameter is
omitted by field redefinitions. Relations between chiral superfields a in (3.2) and twisted
chiral superfields Ya in (3.4) are
2 a e2Qa V a = Ya + Ya .
(3.5)
We can see that the shift symmetry Ya Ya + 2i comes from the U (1) rotation symmetry
on a . In the IR limit e , becomes nondynamical and generates a following
constraint from W :
Qa Ya = t,
a
(3.6)
where W is defined in (3.4b). When we consider low energy effective theories of the theory
(3.4), we take the gauge coupling constant to be infinity e . In this limit is no
182
longer dynamical and becomes just an auxiliary field. Thus the function (3.6) is rewritten
by integratingout of :
=
dYa
Qa Ya t exp
eYa .
a
Via suitable field redefinitions, we can read a LG theory of twisted chiral superfields. Moreover, we also obtain a period integral of a mirror pair of the manifold which appeared in
the effective theories in the original GLSM. Thus we often refer the function (3.6) to the
period integral.
Suppose the theory is topologically Atwisted [39]. In the topologically Atwisted theory, twisted chiral superfields are only valid while the other fields such as chiral superfields
and real superfields are all BRSTexact. Due to this the Lagrangian is reduced only to the
twisted superpotential and the partition function is obtained as the integral of weight eW .
This is nothing but the period integral defined in (3.6). Thus as far as considering the Atwisted sector, the effective theories derived from this function are exact.
Unfortunately, no one knows an exact formulation of a Tdual Lagrangian LT of a
GLSM with a generic superpotential WGLSM . This is partly because the above formulation
is only powerful when we consider Tduality of topologically Atwisted GLSMs. As mentioned above, any deformations of WGLSM are BRSTexact in the topological Atwisted
theory. However, even though in the Atheories, we can analyze Tdual theories of specific
GLSMs with superpotentials of type WGLSM = P G (S), where G (S) is a homogeneous
polynomial of degree with respect to chiral superfields S. In order to do this, let us deform
the above period integral (3.6) to
dYa () exp(W ).
= d
(3.7)
a
This function can be derived through the discussions of Cecotti and Vafa [40], and Morrison and Ronen Plesser [41]. Here we omit a precise derivation. Please see it in [16].
In this formulation we also take the IR limit e and integrate out the superfield ,
because we want to obtain the mirror dual descriptions of the effective theories of the
original GLSM with superpotential. However, the factor in (3.7) prevents from exact
integratingout of . Thus we need to replace this factor to other variable which does not
disturb the integration. If we wish to obtain the LG description we replace the factor to the
differential with respect to the FI parameter such as t . On the other hand when
we derive the mirror geometry we replace this to the differential operator of an appropriate
twisted chiral superfield derived from the negatively charged chiral superfield, for example,
YP , where YP is the twisted chiral superfield of the chiral superfield P of charge
. The resulting geometry has a Z type orbifold symmetry. For example, we start from
the GLSM for quintic hypersurface CP4 [5]. Performing Tduality and taking the IR limit,
we obtain not only the mirror dual geometry CP4 [5]/(Z5 )3 but also its LG description [16].
This procedure is so powerful that we develop it in order to obtain the mirror descriptions
of the noncompact CY manifolds. In Sections 3.3 and 3.4 we will study how to obtain LG
theories defined by the twisted superpotential and mirror dual geometries, respectively. We
can also obtain another geometry with a different orbifold symmetry if we replace to
the differential of other suitable twisted chiral superfield.
183
S1
1
Y1
. . . SN
... 1
. . . YN
P1
YP1
P2
N +
YP2
(3.8)
The twisted chiral superfields Ya are periodic variables Ya Ya + 2i. They are defined
from the chiral superfields a via (3.5). As we have already discussed, the twisted superpotential W and the period integral , given by the followings, play key roles:
W =
N
Yi YP1 (N )YP2 t +
i=1
d
N
N
(3.9a)
i=1
(3.9b)
i=1
Let us take the IR limit e in order to consider the low energy effective theories.
It is clear that the dynamics of is frozen and this superfield becomes just an auxiliary
superfield. Thus we must replace the factor in the period integral (3.9b) to appropriate
variables.
3.3. Mirror LandauGinzburg descriptions
In this subsection we will derive LG theories with orbifold symmetries. In order to do
this, we change the variable in the period integral (3.9b) to
.
t
This replacing can be easily performed because of the existence of the term ( a Qa Ya
t) in . Then we integrate out the superfield and obtain
=
N
Yi YP1 (N )YP2 t
i=1
YP1
YP2
Yi
.
e
e
e
exp
(3.10)
Next let us solve the function in this function. We note that there are two ways to solve
it. One is to write the variable YP1 in terms of Yi and YP2 . The other is to solve YP2 by Yi
and YP1 . Both two solutions give consistent LG theories with orbifold symmetries.
184
Performing the tderivative in (3.10) after the substitution of this solution, we obtain
= et/
N
1Y
N
e i dYi e YP2 dYP2
i=1
1 N
YP2
YP2
Yi
t/
Yi
exp
e
e
e
e
e
.
i
It is clear that the integral measure is not canonical. Transforming the variables into
Xi := eYi and XP 2 := e(N )YP2 , we obtain the following period integral with a canonical
measure up to an overall constant:2
N
=
dXi dXP2 exp
Xi XP2N et/ X1 XN XP2 .
(3.11)
i
i=1
Since Ya are periodic with respect to the shifts of their imaginary parts Ya Ya + 2i, the
new variables Xi and XP2 are symmetric under the following phase shifts:
Xi i Xi ,
XP2 P2 XP2 ,
i = P2N = 1 2 N P2 = 1. (3.12)
We can read from (3.11) and (3.12) that the following orbifolded LG theory appears:
N
Xi + XP2N + et/ X1 XN XP2 (Z )N .
(3.13)
W =
i=1
This theory is still illdefined from the minimal model point of view. Even though the terms
of positive powers such as Xi are welldefined and they consist of N = 2 LG minimal
model, there exists a term XP2N , which does not generate any critical points at finite XP2 .
However, there is an interpretation to avoid this difficulty. Recall a discussion on the linear
dilaton CFT and the Liouville theory [42,43]. (We prepare a brief review in Appendix C.)
Based on this argument, we can interpret the negative power term corresponds to Z0k in
(C.4), which gives an N = 2 SCFT on the coset SL(2, R)k /U (1) at level k assigned by
.
N
This assignment is correct because the conformal
weights ra in Appendix C are all
ra = 1/, where n + 1 = N . Thus we obtain r = a ra 1 = N/ 1 1/k, which
gives the above equation. This theory is given as an N = 2 KazamaSuzuki model on
k=
185
the coset SL(2, R)k /U (1) at level k [44], which is the gauged WZW model on the twodimensional Euclidean black hole [21]. Furthermore, this theory is exactly equivalent to
N = 2 Liouville theory of background charge Q2 = 2/k via Tduality [22,43]. We will
continue to argue in later discussions.
3.3.2. Solution two: ZN orbifolded LG theory
In the same analogy of the previous discussion,3 we study the theory of the second
solution
N
1
Yi + YP1 ,
YP2 =
t
N
i=1
which comes from the function in the period integral (3.10). Substituting this into (3.10),
we find
N
1 Yi
=
e N dYi e N YP1 dYP1
i=1
t
1
N
Yi N
YP1
YP1
Yi
N
.
exp
e
e
e
e
e
i
XiN
t
N
N
X
+X
+ e N X1 XN XP1 (ZN )N .
W N =
(3.14)
i
P1
i=1
Applying the discussions in Appendix C to the negative power term in the superpotential
W N , we find that the theory is also described by the welldefined LG theory with an
orbifold symmetry coupled to N = 2 KazamaSuzuki model on the coset SL(2, R)k /U (1)
at level k, which is given by
N
2
=k = 2,
Q
where Q is the charge of equivalent N = 2 Liouville theory.
3.4. Mirror geometry descriptions
In the previous subsection we found two orbifolded LG theories as exact effective theories. They are obtained by solving the twisted chiral superfields YP1 and YP2 , respectively.
Next, we will read geometric informations from the same period integral (3.9b). Here we
will also obtain two solutions which are related to the LG theories. The derivation procedure is so complicated that we try to imitate the method discussed in Section 7.3 of [16]
3 From now on we omit overall constant factors which appear in the period integral.
186
and we develop detailed calculations, explicitly. In order to obtain the geometric informations in the IR limit, we integrate out the superfield in the period integral (3.9b) after the
replacement of in (3.9b) to other variables, as we performed before.
3.4.1. Z orbifolded geometry
Let us study how to obtain the geometry with Z type orbifold symmetry. Replacing
in the period integral (3.9b) to
,
YP1
N
dYi eYP1 dYP1 dYP2
i=1
YP1
YP2
Yi
.
Yi YP1 (N )YP2 t exp
e
e
e
(3.15)
eYa =: P1 Ua
eYb =: P2 Ub
for a = 1, . . . , ,
for b = + 1, . . . , N.
exp P1
Ua + 1 P2
Ub + 1
a=1
b=+1
dUi
dP2 du dv log
=
Ui + t
Ua + 1
Ui
a
i
i
Ub + 1 uv
exp P2
b
dUi
du dv log
=
Ui + t
Ua + 1
Ub + 1 uv ,
Ui
a
i
i
b
(3.16)
where we introduced new variables u and v taking values in C and used a following equation
1
= du dv exp(P2 uv).
P2
187
It is obvious that the resulting function (3.16) still includes a noncanonical integral measure. Thus we perform further redefinitions such as
Ua =: et/
Za
,
Z1 ZN
Ub =: Zb .
Note that the period integral (3.16) is invariant under the following transformations acting
on the new variables Zi :
Za a Za ,
Zb b Zb ,
a = b = 1 N = 1,
where is an arbitrary number taking in C . The i come from the shift symmetry of the
original variables Yi Yi + 2i. Combining these transformations we find that the period
integral has C (Z )N 2 symmetries. Substituting Zi into (3.16), we obtain
N
N
1
t/
dZi du dv
Za + e Z1 ZN
Zb + 1 uv ,
=
vol(C )
i=1
a=1
b=+1
Za + Z1 Z ,
G(Zb ; u, v) :=
a=1
N
Zb + 1 uv,
(3.17a)
(3.17b)
b=+1
:= et/ Z+1 ZN .
(3.17c)
N
i=1
188
(Z )N 2 orbifolded geometry. Here we replace to the differential with respect to YP2 ,
which is dual of the chiral superfield P2 of charge (N ):
.
N YP2
N
dYi dYP1 eYP2 dYP2
i=1
Yi YP1 (N )YP2 t
exp
eYi eYP1 eYP2 .
(3.18)
Let us perform the following redefinitions of the variables Yi , YP1 and YP2 :
eYP1 =: P1 ,
eYP2 =: P2 ,
eYa =: P1 Ua
eYb =: P2 Ub
for a = 1, . . . , ,
for b = + 1, . . . , N.
Substituting the redefined variables into (3.18) and introducing auxiliary variables u and
v in order to integrate out P1 completely, we obtain
N
N
dUi
du dv log
Ui + t
=
Ui
i=1
i=1
N
(3.19)
Ub + 1
Ua + 1 uv .
b=+1
a=1
The integral measure still remains noncanonical. We next introduce further redefinitions
of Ui :
Ua =: ZaN ,
Ub =: et/(N)
ZbN
.
Z1 ZN
189
We can see that the map from Zi to Ui is onetoone modulo the C (ZN )N 2 action
given by
Za a Za ,
Zb b Zb ,
aN = bN = 1 N = 1,
where takes value in C . On account of the above redefinitions and symmetries we find
that the period integral is rewritten as
N
1
N
dZi du dv
Za + 1 uv
=
vol(C )
i=1
a=1
N
N
t/(N)
Zb
+e
Z1 ZN ,
b=+1
N = (Zi ; u, v) CN +2 F(Za ; u, v) = 0, {G(Zi ) = 0}/C (ZN )N 2 ,
M
(3.20a)
N
ZaN + 1 uv,
G(Zi ) :=
ZbN + Z+1 ZN ,
F(Za ; u, v) :=
a=1
:= et/(N) Z1 Z .
b=+1
(3.20b)
(3.20c)
190
is deformed to the LG theory coupled to the Liouville theory as an exact quantum theory.
This is equivalent to (3.13) via Tduality. On account of the CY/LG correspondence, (3.13)
and sigma model on (3.17) are identical with each other. Finally the original CY manifold
(2.10) and (3.17) are mirror dual with each other. Notice that the CY manifold MCY is
also deformed because the Liouville theory indicates that the dilaton field propagates on
the target space [45,46]. Of course, we find that there are the same relations among effective
theories (2.16), (2.17), (3.14) and (3.20).
Let us consider the case = 1. As discussed before, the GLSM has only two massless effective theories (2.11) and (2.25). In addition, the subspace CP1 [] in (3.17) is
illdefined if = 1 and then the LG description (3.13) is also illdefined. Thus the Tdual
theory has only two descriptions (3.14) and (3.20) in the IR limit. This situation is consistent with the result in [16], where the GLSM for O(N ) bundle on CPN 1 and its Tdual
was discussed.
191
etry group. These symmetries give the information of the complex structures of not only
the base spaces HSSs but also the noncompact CY manifolds. However, the Tdual theory
[16] is only valid when we consider the GLSM without a superpotential or with a superpotential given simply by a homogeneous polynomial such as WGLSM = P G (S). Even
though the polynomial G (S) has an additional symmetry, the period integral (3.6) or (3.7)
cannot recognize the existence of this additional symmetry. Thus the Tdual theory does
not map all structures of the CY M to the mirror geometry completely. For example, we
can argue the sigma model on the resolved conifold and its mirror dual in the framework
of GLSM and its Tdual, however, we have not even understood any correct descriptions
for the deformed conifold represented by the GLSM. Therefore, if we wish to obtain the
correct Tdual theories of the sigma models on such noncompact CY manifolds, we must
improve the formulation so that it may recognize the complex structure of the manifold. It
is quite significant to solve this problem in order to understand mirror symmetry for more
general noncompact CY manifolds.
Acknowledgements
The author would like to thank Hiroyuki Fuji, Kentaro Hori, Takahiro Masuda, Shunya
Mizoguchi, Toshio Nakatsu, Kazutoshi Ohta, Takuya Okuda, Makoto Sakaguchi, Hitoshi
Sato, Dan Tomino and Atsushi Yamaguchi for valuable comments. The author also thank
Debashis Ghoshal for correspondence. The author would also like to express the gratitude
to Yukawa Institute for Theoretical Physics (YITP) for the hospitality during the authors
stay. This work was supported in part by the JSPS Research Fellowships for Young Scientists (#1503926).
Appendix A. Conventions
In this appendix we will write down the notation and convention which are modified from the ones defined in WessBaggers book [51]. In [51] supersymmetric field
192
theory is defined in fourdimensional spacetime. However, in this paper we discuss supersymmetric field theory in two dimensions. Thus let us first perform the dimensional
reduction. The coordinates in two dimensions (x 0 , x 1 ) are related to the fourdimensional
ones (y 0 , y 1 , y 2 , y 3 ):
0 1
0 3
x ,x y ,y .
Note that we perform the dimensional reduction for y 1  and y 2 directions. Next we redefine
the irreducible representation for spinors. Weyl spinors in four dimensions becomes
Dirac spinors. For convenience, we define the Dirac spinor indices in two dimensions as
[14]:
1 2
+
, = , ,
(1 , 2 ) = ( , + ),
= + ,
+ = ,
12 = 21 = 1 + = + = 1,
+
+
,
= , = ( + , ).
Under the above convention, the super differential operators D are also changed as follows:
D =
i (0 1 ),
D = + i (0 1 ).
= + + (v0 + v1 ) + (v0 v1 ) 2 + 2 +
2i + + + + + 2i + + + + 2 + + D.
193
Note that we consider only U (1) gauge theories in this paper, where v0 and v1 are components of a U (1) gauge potential, are gaugino fields as Dirac spinors and D is a real
auxiliary field. The complex fields and are coming from the dimensionally reduced
components of fourdimensional U (1) gauge potential. In general we set this superfield to
be dimensionless.
Twisted chiral superfield. A twisted chiral superfields is also an irreducible superfield
in two dimensions. The definition is D + Y = D Y = 0. Expanding a twisted chiral superfield Y in terms of { , }, we obtain
= i 2 + + i 2 + 2 + (D iF01 )
i (0 1 ) i + + (0 + 1 ) + 2 + (0 1 ) +
+ 2 + + (0 + 1 ) + + 02 12 .
This is also a twisted chiral superfield D + = D = 0. This superfield is gaugeinvariant under the U (1) gauge transformation.
Here let us define integral measures of the fermionic coordinates and in the
superspace:
1
d2 := d + d ,
2
1
d2 := d + d ,
2
1
d4 := d + d d + d .
4
194
First let us review onedimensional (ordinary) projective space CP1 . We prepare a twodimensional complex plane without the origin such as W = C2 {0}. The coordinates on
this plane are described as (z1 , z2 ). The projective space CP1 is defined as a space whose
coordinate is given by the ratio of the complex variables z1 and z2 . (The complex variables
are called homogeneous coordinates in the projective space.) Under this definition, the two
points (z1 , z2 ) and (z1 , z2 ) in W plane are identified with each other:
(z1 , z2 ) (z1 , z2 ),
(B.1)
2n = 10 d,
195
coupling region:
Rd1,1 R M = Rd1,1 R S 1 M/U (1),
where M is a compact and nonsingular manifold. The real line R is parameterized by .
We can define CFT on each subspace. On the flat space Rd1,1 we can define N = 1
SCFT whose central charge is
3
cd = d.
2
(C.1)
(C.2)
If
weights ra such as r
we introduce a new variable r with respect to the conformal
2 = 2r .
r
1,
we
can
express
the
background
charge
Q
to
Q
a a
Here let us combine the above discussion with the conjectures proposed by Muhki and
Vafa [52], Ghoshal and Vafa [53], and Ooguri and Vafa [54], where they insisted that
an N = 2 SCFT on the noncompact space R M can be given formally by the LG
superpotential
W = Z0k + F (Za ),
(C.4)
1
2
= 2.
r
Q
(C.5)
This formulation is useful to describe the sigma model on deformed conifold [53]. The first
term in the superpotential appears to be illdefined from the LG minimal model point of
view. The corresponding potential does not have a minimum at the finite value of Z0 . The
196
topological LG model with such a superpotential has already been studied by Ghoshal
and Mukhi [55], and Hanany, Oz and Ronen Plesser [56] in order to investigate twodimensional string theory. Moreover, in general, k is not an integer, which makes (C.4)
nonsingle valued. Thus, it was proposed that this first term can be interpreted as an N = 2
SCFT on the coset SL(2, R)/U (1) at level k. From the geometric point of view, this coset
space corresponds to a semiinfinite cigar, and in the IR limit this geometry deforms to
the twodimensional Euclidean black hole [21]. This SCFT on the coset had been believed
to be isomorphic to the Liouville theory in the sense of SCFT. They are related by strongweak coupling duality on the worldsheet: the theory (C.4) can be valid as a Liouville theory
in the large Q limit, while this can be seen as a coset SCFT in the large k limit (k = 2/Q2 ).
Finally, it has been proved that the N = 2 SCFT on the coset SL(2, R)k /U (1) is exactly
equivalent (or Tdual) to the N = 2 Liouville theory to each other in any values of k > 0
[22]. This equivalence was also proved by Tong in the framework of twodimensional domain wall physics in threedimensional theory [57].
To summarize, we find that the string theory on a singular CY manifold X 2n can be
holographic dual to string theory as a product theory of the N = 2 SCFT on the coset
SL(2, R)k /U (1) and the N = 2 LG minimal model on M/U (1). The coset SCFT sector
is also equivalent to the N = 2 Liouville theory on R S 1 .
References
[1] A. DAdda, P. Di Vecchia, M. Lscher, Confinement and chiral symmetry breaking in CPN 1 model with
quarks, Nucl. Phys. B 152 (1979) 125.
[2] D.J. Gross, A. Neveu, Dynamical symmetry breaking in asymptotically free field theories, Phys. Rev. D 10
(1974) 3235.
[3] D. Friedan, E.J. Martinec, S.H. Shenker, Conformal invariance, supersymmetry and string theory, Nucl.
Phys. B 271 (1986) 93.
[4] E. Witten, Topological sigma models, Commun. Math. Phys. 118 (1988) 411.
[5] B.R. Greene, M. Ronen Plesser, Duality in CalabiYau moduli space, Nucl. Phys. B 338 (1990) 15.
[6] K. Higashijima, M. Nitta, Supersymmetric nonlinear sigma models as gauge theories, Prog. Theor. Phys. 103
(2000) 635, hepth/9911139.
[7] K. Higashijima, T. Kimura, M. Nitta, Gaugetheoretical construction of noncompact CalabiYau manifolds,
Ann. Phys. 296 (2002) 347, hepth/0110216.
[8] K. Higashijima, T. Kimura, M. Nitta, CalabiYau manifolds of cohomogeneity one as complex line bundles,
Nucl. Phys. B 645 (2002) 438, hepth/0202064.
[9] E. Calabi, Mtriques khlriennes et fibrs holomorphes, Ann. Scient. col. Norm. Sup. 12 (1979) 269.
[10] S.T. Yau, Calabis conjecture and some new results in algebraic geometry, Proc. Nat. Acad. Sci. 74 (1977)
1798.
[11] I.R. Klebanov, E. Witten, Superconformal field theory on threebranes at a CalabiYau singularity, Nucl.
Phys. B 536 (1998) 199, hepth/9807080.
[12] I.R. Klebanov, M.J. Strassler, Supergravity and a confining gauge theory: duality cascades and SBresolution of naked singularities, JHEP 0008 (2000) 052, hepth/0007191.
[13] M. Cvetic, G.W. Gibbons, H. L, C.N. Pope, Ricciflat metrics, harmonic forms and brane resolutions,
Commun. Math. Phys. 232 (2003) 457, hepth/0012011.
[14] E. Witten, Phases of N = 2 theories in two dimensions, Nucl. Phys. B 403 (1993) 159, hepth/9301042.
[15] W. Lerche, C. Vafa, N.P. Warner, Chiral rings in N = 2 superconformal theories, Nucl. Phys. B 324 (1989)
427.
[16] K. Hori, C. Vafa, Mirror symmetry, hepth/0002222.
197
[17] K. Hori, S. Katz, A. Klemm, R. Pandharipande, R. Thomas, C. Vafa, R. Vakil, E. Zaslow, Mirror Symmetry,
Clay Mathematics Monographs, vol. 1, American Mathematical Society, Providence, 2003.
[18] T. Kimura, Towards mirror symmetry on noncompact CalabiYau manifolds, contribution to the Proceedings of the 12th International Conference on Supersymmetry and Unification of Fundamental Interactions
(SUSY 2004), hepth/0409003.
[19] D. Gepner, Exactly solvable string compactifications on manifolds of SU(N ) holonomy, Phys. Lett. B 199
(1987) 380.
[20] D. Gepner, Spacetime supersymmetry in compactified string theory and superconformal models, Nucl.
Phys. B 296 (1988) 757.
[21] E. Witten, String theory and black holes, Phys. Rev. D 44 (1991) 314.
[22] K. Hori, A. Kapustin, Duality of the fermionic 2d black hole and N = 2 Liouville theory as mirror symmetry, JHEP 0108 (2001) 045, hepth/0104202.
[23] P.H. Ginsparg, G.W. Moore, Lectures on 2D gravity and 2D string theory, Lectures given at TASI Summer
School (1992), hepth/9304011.
[24] Y. Nakayama, Liouville field theory: a decade after the revolution, Int. J. Mod. Phys. A 19 (2004) 2771,
hepth/0402009.
[25] B. Zumino, Supersymmetry and Khler manifolds, Phys. Lett. B 87 (1979) 203.
[26] S.R. Coleman, There are no Goldstone bosons in two dimensions, Commun. Math. Phys. 31 (1973) 259.
[27] N. Seiberg, E. Witten, Electricmagnetic duality, monopole condensation, and confinement in N = 2 supersymmetric YangMills theory, Nucl. Phys. B 426 (1994) 19, hepth/9407087;
N. Seiberg, E. Witten, Nucl. Phys. B 430 (1994) 485, Erratum.
[28] N. Seiberg, E. Witten, Monopoles, duality and chiral symmetry breaking in N = 2 supersymmetric QCD,
Nucl. Phys. B 431 (1994) 484, hepth/9408099.
[29] A. Strominger, Massless black holes and conifolds in string theory, Nucl. Phys. B 451 (1995) 96, hepth/9504090.
[30] B.R. Greene, D.R. Morrison, A. Strominger, Black hole condensation and the unification of string vacua,
Nucl. Phys. B 451 (1995) 109, hepth/9504145.
[31] N.P. Warner, Lectures on N = 2 superconformal theories and singularity theory, in: M. Green, et al. (Eds.),
Superstrings 89, World Scientific, Singapore, 1990.
[32] E.J. Martinec, Algebraic geometry and effective Lagrangians, Phys. Lett. B 217 (1989) 431.
[33] C. Vafa, String vacua and orbifoldized LG models, Mod. Phys. Lett. A 4 (1989) 1169.
[34] C. Vafa, N.P. Warner, Catastrophes and the classification of conformal theories, Phys. Lett. B 218 (1989) 51.
[35] B.R. Greene, C. Vafa, N.P. Warner, CalabiYau manifolds and renormalization group flows, Nucl. Phys.
B 324 (1989) 371.
[36] E.J. Martinec, Criticality, catastrophes and compactifications, in: L. Brink, et al. (Eds.), Physics and Mathematics of Strings, V.G. Knizhnik Memorial Volume, EFI, Chicago, 1989.
[37] K. Intriligator, C. Vafa, LandauGinzburg orbifolds, Nucl. Phys. B 339 (1990) 95.
[38] S. Cecotti, N = 2 LandauGinzburg vs. CalabiYau models: nonperturbative aspects, Int. J. Mod. Phys.
A 6 (1991) 1749.
[39] E. Witten, Mirror manifolds and topological field theory, hepth/9112056.
[40] S. Cecotti, C. Vafa, Topological and antitopological fusion, Nucl. Phys. B 367 (1991) 359.
[41] D.R. Morrison, M. Ronen Plesser, Summing the instantons: quantum cohomology and mirror symmetry in
toric varieties, Nucl. Phys. B 440 (1995) 279, hepth/9412236.
[42] A. Giveon, D. Kutasov, O. Pelc, Holography for noncritical superstrings, JHEP 9910 (1999) 035, hepth/9907178.
[43] A. Giveon, D. Kutasov, Little string theory in a double scaling limit, JHEP 9910 (1999) 034, hepth/9909110.
[44] Y. Kazama, H. Suzuki, New N = 2 superconformal field theories and superstring compactification, Nucl.
Phys. B 321 (1989) 232.
[45] N. Nakayama, K. Sugiyama, Construction of supergravity backgrounds with a dilaton field, hepth/0411143.
[46] M. Nitta, Conformal sigma models with anomalous dimensions and Ricci solitons, hepth/0411149.
[47] T. Eguchi, Y. Sugawara, Modular invariance in superstring on CalabiYau nfold with ADE singularity,
Nucl. Phys. B 577 (2000) 3, hepth/0002100.
[48] S. Mizoguchi, Modular invariant critical superstrings on fourdimensional Minkowski space twodimensional black hole, JHEP 0004 (2000) 014, hepth/0003053.
198
[49] J.M. Maldacena, G.W. Moore, N. Seiberg, Geometrical interpretation of Dbranes in gauged WZW models,
JHEP 0107 (2001) 046, hepth/0105038.
[50] K. Hori, A. Kapustin, Worldsheet descriptions of wrapped NS fivebranes, JHEP 0211 (2002) 038, hepth/0203147.
[51] J. Wess, J. Bagger, Supersymmetry and Supergravity, second ed., Princeton Univ. Press, Princeton, NJ, 1992.
[52] S. Mukhi, C. Vafa, Twodimensional black hole as a topological coset model of c = 1 string theory, Nucl.
Phys. B 407 (1993) 667, hepth/9301083.
[53] D. Ghoshal, C. Vafa, c = 1 string as the topological theory of the conifold, Nucl. Phys. B 453 (1995) 121,
hepth/9506122.
[54] H. Ooguri, C. Vafa, Twodimensional black hole and singularities of CY manifolds, Nucl. Phys. B 463
(1996) 55, hepth/9511164.
[55] D. Ghoshal, S. Mukhi, Topological LandauGinzburg model of twodimensional string theory, Nucl. Phys.
B 425 (1994) 173, hepth/9312189.
[56] A. Hanany, Y. Oz, M. Ronen Plesser, Topological LandauGinzburg formulation and integrable structure of
2d string theory, Nucl. Phys. B 425 (1994) 150, hepth/9401030.
[57] D. Tong, Mirror mirror on the wall: on twodimensional black holes and Liouville theory, JHEP 0304 (2003)
031, hepth/0303151.
Abstract
We study weakly coupled SU(N ) N = 4 superYangMills theory on R S 3 at infinite N , which
has interesting thermodynamics, including a Hagedorn transition, even at zero YangMills coupling.
We calculate the exact oneloop partition function below the Hagedorn temperature. Our calculation
employs the representation of the oneloop dilatation operator as a spin chain Hamiltonian acting on
neighboring sites and a generalization of Plyas counting of necklaces (gaugeinvariant operators)
to include necklaces with a pendant (an operator which acts on neighboring beads). We find that
the oneloop correction to the Hagedorn temperature is ln TH = +/8 2 .
2005 Elsevier B.V. All rights reserved.
PACS: 11.15.q
1. Introduction
The past several years have witnessed a tremendous amount of effort invested into the
careful study of N = 4 superYangMills (SYM) theory at large N from a number of complementary approaches. One motivation for much of this work is the fact that this theory
is believed to provide, via the AdS/CFT correspondence, the simplest context in which we
might hope to understand how to solve large N gauge theories in four dimensions. Optimistically, the apparent integrability [1] of the string theory dual to N = 4 SYM theory
suggests that it might be possible to calculate all physical quantities (at least at infinite N )
Email address: spradlin@kitp.ucsb.edu (M. Spradlin).
05503213/$ see front matter 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.nuclphysb.2005.01.007
200
D2 (xk )
k
P D2 (xk , xm ) ,
+
ln Tr eH 1loop = 2
(1)
1 z(xk )
4
k=1
m=1
TH () = TH (0) 1 +
(2)
+ , TH (0) 0.380
8 2
(measured in units where the radius of the S 3 is one), which is presumably another example
of an interesting physical quantity which we might hope to one day calculate as an exact
function of .
We begin in Section 2 by reviewing how the free partition function may be calculated by
first assembling the partition function for the elementary fields into a partition function for
singletrace operators, and then into the full partition function for multitrace operators.
We discuss the general structure of the oneloop correction to the partition function and
comment on the role of operator mixing and 1/N effects. In Section 3 we rephrase the
problem of counting gauge theory operators into the language of spin chains and introduce
a generalization of Plyas necklace problem to what we call necklaces with a pendant.
such as a spin chain Hamiltonian, inserted
These are necklaces with some local operator O,
at neighboring beads on the necklace. We derive a general formula expressing a partition
function for such necklaces in terms of some basic quantities O(x) and P O(w, y)
In Section 4 we apply this machinery to various familiar
which are easily obtained from O.
subsectors (SU(2), SO(6), and SL(2)) of the N = 4 SYM theory. This section is mostly
a warmup for Section 5, where we calculate D2 (x) and P D2 (w, y) for the oneloop
201
dilatation operator D2 of the full N = 4 SYM theory. Finally in Section 6 we present our
final results for the oneloop corrections to the single and multitrace partition functions
of N = 4 SYM.
2. The N = 4 SYM partition function
We begin in this section with a discussion of weakly coupled N = 4 SYM theory on
R S 3 following [15,17]. Since there is more than one way to calculate the free partition
function, we review here that derivation which best sets the stage for our calculation of the
oneloop correction in the following sections.
2.1. Initial considerations
In general, the partition function is given by
Z() = Tr eH ,
where is the inverse temperature and H is the Hamiltonian of the theory on
convenient to introduce the bookkeeping parameter
x = e = e1/T ,
(3)
R S3.
It is
(4)
202
traces removed, since antisymmetric derivatives can be replaced by field strengths and
traces of derivatives give terms which are zero by the equations of motion. Following
Polyakov [19] we refer to such objects as the letters of the N = 4 alphabet. We will
use A to denote the collection of letters.
We define d(A) of a letter A to be its engineering dimension, so d( I ) = 1, d(a ) =
3/2, d(F ) = 2, and each covariant derivative adds one to the dimension. The enumeration of letters weighted by dimension gives rise to the following elementary partition
function for the N = 4 alphabet [15,17],
2x(3 x )
d(A)
x
=
.
z(x) =
(6)
(1 x )3
AA
This function has a power series which converges for all temperatures 0 x < 1 and can
be understood as follows:
3/2
z(x) =
6x + 16x
+
6( I )
16(a )
+ .
2
30x
5/2
+ 48x
+
70x 3
24(D I )+6(F )
48(D a )
54(D D I )+16(D F )
(7)
There are only 48 D a instead of 64 because 16 components are set to zero by the Dirac
equation. Similarly there are only 54 components of D D I because D 2 I is zero. Finally, 4 of the 24 components of D F are zero by the equations of motion and another 4
are equal to zero by the Bianchi identity.
In what follows we will frequently need to know various partition functions with the
fermion number operator (1)F inserted. These factors can be easily dealt with by making
use of the fact that bosonic and fermionic operators respectively have integer or halfinteger
dimensions at tree level. To this end we introduce the quantity
= e2i ,
(8)
2x(3 + x )
2id(A) d(A)
z(x) =
e
x
=
= 6x 16x 3/2 +
(1 + x )3
AA
(1)F (A) x d(A) .
=
(9)
AA
203
Tr[A1 A2 AL ] is the overall cyclic invariance of the trace. The enumeration of such
singletrace operators is identical to the combinatorial problem of counting the number
of distinct necklaces composed of a collection of different types of beads. The solution to
this problem, which we review in the next section, may be expressed in terms of z(x) as
Z(x) = z(x)
(n)
n=1
ln 1 z n+1 x n ,
(11)
where (n) is Eulers totient function which counts the number of integers less than n
which are relatively prime to n. The first term z(x) in (11) is present simply to subtract
away traces of a single letter Tr[A], since these vanish automatically in the SU(N ) theory.
Plugging (6) into (11) gives an expansion which goes as follows:
Z(x) =
2
21x
21(Tr[ I J ])
5/2
+ 96x
96(Tr[ I a ])
+.
3
376x
76(Tr[ I J K ])+144(Tr[ I D
J ])+36(Tr[ I F
(12)
])+120(Tr[a b ])
Two comments are in order. The first is that the result (11) is only valid in the N = limit
of the SU(N ) gauge theory, since we allow arbitrarily high powers of the individual letters.
If N were finite, then trace identities would allow a single trace of more than N letters to
be reexpressed in terms of highertrace operators, indicating that the basis of operators we
are using would be overcomplete.
The second observation is that the power series expansion of (11) converges only for
0 x < xH , where
xH = 7 4 3 0.072.
(13)
The divergence of the partition function at this value, which corresponds to the temperature
TH ( = 0) =
1
1
=
0.380
ln xH ln(7 + 4 3 )
(14)
(measured in units of R 1 ) has been argued to be the gauge theory dual of the Hagedorn
transition in string theory [15,17]. Hagedornlike behavior in other free large N systems
was also observed in [20].
2.4. From singletrace to multitrace operators
Having determined the partition function Z(x) for singletrace operators, it is an easy
combinatoric problem to calculate from this the partition function for an arbitrary number
of traces, since the only constraint is that traces should be treated as indistinguishable
bosons or fermions. The result is
Z(n+1 x n )
Z(x) = exp
(15)
.
n
n=1
204
The partition function Zk (x) for ktrace operators can be extracted by inserting y n into the
sum and then reading off the coefficient of y k in the expansion of the exponential.
From the results reviewed above we can see that the complete partition function of
N = 4 SU(N ) gauge theory at infinite N and zero YangMills coupling is given by the
formula
z(n+1 x n )
1
Z(x) = exp
(16)
.
n
1 z(n+1 x n )
n=1
n=1
The exponential term in (16) is (one over) the partition function for gauge group U (1) and
the infinite product is the partition function for gauge group U (N ), so (16) expresses the
expected fact that
ZSU(N ) =
ZU (N )
.
ZU (1)
(17)
In fact, it is interesting to note that although this analysis has been done at infinite N , the
result (16) remains correct to all orders in 1/N (though certainly not at finite N ). This
is true because the treelevel dilatation operator D0 obviously does not receive any 1/N
corrections, and trace relations are nonperturbative in 1/N .
2.5. Turning on the YangMills coupling
In (16) we have reviewed the complete partition function for SU(N ) N = 4 SYM theory
on R S 3 at infinite N and zero YangMills coupling. The goal of this paper is to calculate
the firstorder correction to this result when we turn on the YangMills coupling gYM . The
2 N as
dilatation operator D can be expanded in powers of the t Hooft parameter = gYM
D2 + ,
4 2
so to first order in the partition function (5) is given by
D = D0 +
ln x D
D + D +
= Tr x D0 +
Tr x 0 D2 + .
Z(x, ) = Tr x 0 4 2 2
4 2
Therefore we need to calculate the trace of the oneloop dilatation operator,
Z (1) (x) = Tr x D0 D2 .
(18)
(19)
(20)
Our calculation will proceed by first calculating the oneloop partition function in the
singletrace sector and then assembling together the result for multitrace operators as in
the previous subsection.
2.6. Operator mixing and 1/N
The treelevel dilatation operator D0 is diagonal in the trace basis, since the engineering
dimension of a ktrace operator is obviously just the sum of the engineering dimensions of
the k individual operators. At one loop this is no longer true. Instead we have a formula of
205
the form
D2 k =
k
i=1
1
1
Tr[O1 ] D2 Tr[Oi ] Tr[Ok ] + k 1 + k + 1,
N
N
(21)
for a generic ktrace operator k = Tr[O1 ] Tr[Ok ]. At first glance one might be tempted
to disregard the O(1/N) terms since we are working at infinite N . However it is well
known that there are classes of ktrace operators whose oneloop anomalous dimensions
receive nonzero contributions from mixing with k 1trace
operators [2123]. This occurs
generically for BMN operators, which consist of L N letters. For such operators the
1/N suppression is overwhelmed by the growth of the number of nonplanar diagrams [22].
Therefore we will not use N = as a justification to omit the last two terms in (21).
Fortunately we have an even better excuse, which is simply that these terms are nondiagonal in the trace basis and therefore do not contribute to the quantity (20) that we
are computing. The situation becomes more complicated starting at two loops, where the
correction to the partition function involves Tr[x D0 D4 ] and Tr[x D0 D22 ], both of which
have some diagonal terms of order 1/N 2 which cannot necessarily be dropped. (In some
sense this like an order of limits problem, like the one discussed in [24].) Of course, if one
focuses on calculating the planar partition function (as opposed to the large N partition
function), then none of these issues arise and one can ignore the 1/N terms in (21) from
the beginning. Subtleties in the 1/N expansion have been noted in [25].
Note that we have implicitly been using a scalar product on the space of operators which
is diagonal in the trace basis,
kl kl .
(22)
The N = 4 SYM theory provides, via the stateoperator correspondence, a natural inner
product Skl = kl on the space of operators which is diagonal in the trace basis at infinite
N but receives nondiagonal, tracemixing corrections beginning at O(1/N ). This operator
mixing does not concern us here since the trace
1
kx D0 D2 lSkl
,
Tr x D0 D2 =
(23)
k,l
is completely independent of the scalar product S. Therefore we are free to choose the
most convenient inner product Skl = kl . A nontrivial inner product S appeared in studies
of 1/N corrections to the BMN correspondence, such as [26,27] where matrix elements
of the Hamiltonian were compared between gauge theory and string theory. The utility of
ignoring the gauge theory inner product for the purpose of calculating basisindependent
quantities was emphasized in [28] in the context of calculating eigenvalues of D2 in the
BMN sector.
3. Plya necklaces
In this section we develop the machinery which will reduce the calculation of the oneloop partition function
Z (1) (x) = Tr x D0 D2
(24)
206
L
d(Ai ).
(25)
i=1
The analysis of this section will be general, but of course for the desired application to
N = 4 SYM theory we remember in the back of our mind that we will take A to be the
N = 4 alphabet, and d(A) will be the engineering dimension of the letter A.
A central result of Plyas counting theory is that the necklace partition function
Z(x) =
(26)
x d(N ) ,
N
(n)
n=1
ln 1 z x n ,
(27)
(28)
AA
The formula (27) is valid when the beads are all bosons. The generalization to include
fermions is straightforward and will be presented below.
3.2. Necklaces with a pendant
Instead of reviewing the elementary derivation of the result (27), we will consider a
useful generalization from which we will recover (27) as a special case. To describe the
generalization that we are interested in, it is useful to think of a necklace of length L as a
spin chain of length L, where on each site of the spin chain the spin vector takes values in
the set A. Of course we have the constraint that only cyclically invariant spin configurations
correspond to necklaces. Therefore, we can recast the calculation of (26) into spin chain
207
x d(N ) =
TrL Px d ,
(29)
L=1
where TrL denotes the trace over spin chains of length L and P denotes the projection
operator onto the subspace of cyclically invariant spin configurations. The projector can be
written explicitly as
1
1 + T + T 2 + + T L1 ,
(30)
L
where T is the translation operator which sends site i on the chain to site i + 1 and satisfies
T L = 1.
The generalization of TrL [Px d ] that we would like to consider is to the case
TrL Px d O ,
(31)
P=
where O is any homogeneous operator which commutes with d and which acts only on
two neighboring sites at a time, so that it may be decomposed into the form
O =
L
O i,i+1 ,
(32)
i=1
Spin chain Hamiltonians are of course prime examples of such operators, and eventually
we will apply the present machinery to the case O = D2 , but our analysis will continue
to be as general as possible. Since the operator O only connects two neighboring sites at
a time, but can slide all the way around the necklace, we can think of O as a pendant
hanging from two adjacent beads on the necklace.
Given any such operator O, an obvious quantity of interest is the expectation value
x d(A1 )+d(A2 ) A1 A2 OA1 A2 .
O(x) = TrAA x d O =
(33)
A1 ,A2 A
We will see that knowing O(x) is almost, but not quite enough information to allow for
the calculation of (31). The other quantity that we will need to know is the permuted trace
w d(A1 ) y d(A2 ) A1 A2 OA2 A1 , (34)
P O(w, y) = TrAA P w d1 y d2 O =
A1 ,A2 A
where P is the permutation operator on A A and the two sites are counted with different
variables w and y. Let us now see how to reduce the calculation of (31) to the calculation
of these two quantities.
Since P projects onto cyclically invariant states anyway, the sum over i in (32) is actually redundant. We can affix the pendant to sites (i, i + 1) = (1, 2), and the sum in (32) just
gives a factor of L which cancels the 1/L in (30) giving
d L1
TrL T k x d O 12 .
TrL Px O =
k=0
(35)
208
L1
k=0 A1 ,...,AL
L
Ai ,Ai+k . (36)
i=3
Upon contemplating the formula (36), it is clear that after we sum over A3 , . . . , AL ,
only two possible structures can emerge,
(i)
A1 A2 OA1 A2
(37)
depending on whether k and L are such that the Kronecker delta functions in (36) end up
connecting site 1 + k to site 1 or to site 2. For example, k = 0 clearly gives the former
structure while k = 1 clearly gives the latter. But for general k and L, what is the criterion
which tells us which of the two possibilities (37) we get?
Let us start with site 1 + k and follow it around the necklace using n of the Kronecker
delta functions,
1 + k 1 + 2k 1 + 3k 1 + (n + 1)k.
(38)
Since there are only L 2 delta functions in total, we have 0 n L 2. Now let m =
(L, k) be the greatest common divisor of L and k. If m > 1, then by step n = L/m 1 at
the latest, site 1 + k will have been connected to site
1 + k 1 + (n + 1)k = 1 + (k/m)L = 1 mod L.
(39)
On the other hand, if k and L are relatively prime (m = 1), then there is no n < L such
that 1 + (n + 1)k = 1 mod L, so 1 + k cannot get connected to site 1 and hence must be
connected to site 2. We conclude that we get the trace structure of type (ii) if and only if k
and L are relatively prime, and we get structure (i) otherwise. Let us therefore study these
cases separately.
3.3. Case (i): k and L have a common factor
As a warmup exercise let us consider the case L = 15, k = 6, so that m = (15, 6) = 3.
Then as we sum over sites 3, 4, . . . , L, the delta functions in (36) sew together the beads of
the necklace as follows:
7 13 4 10 1,
8 14 5 11 2,
9 15 6 12 3 9.
(40)
Each line in this formula can be though of as a strand of the necklace. So the necklace
with L = 15 and k = 6 is composed of m = (15, 6) = 3 strands, and the Kronecker delta
functions in (36) force all of the beads on any given strand to be the same.
209
The strands starting with 1 + k and 2 + k end up respectively at sites 1 and 2, confirming
our earlier analysis. The third line in (40) denotes the following contribution to (36):
x d(A3 )+d(A9 )+d(A15 )+d(A6 )+d(A12 ) A3 ,A9 A9 ,A15 A15 ,A6 A6 ,A12 A12 ,A3
A3 ,A9 ,A15 ,A6 ,A12
x 5d(A3 ) = z x 5 .
(41)
A3
The first two strands in (40) involve the sites 1 and 2 where the pendant is attached, and it
is not hard to see that they end up contributing a factor of
(42)
x 5d(A1 )+5d(A2 ) A1 A2 OA1 A2 = O x 5 ,
A1 ,A2
using the definition (33). Combining these results, we find that the total contribution to the
sum (36) for the case L = 15 and k = 6 is
z x5 O x5 .
(43)
The generalization of this analysis is straightforward. A necklace with general L and k
will have m = (L, k) different strands, each with L/m beads. Two of those strands (like
the first two in (40) will involve the sites 1 and 2 and give rise to
L/m
.
O x
(44)
The remaining m 2 strands (note that we are assuming here that m 2), like the last line
in (40), give a factor of
L/m m2
.
z x
(45)
Combining (44) and (45), we conclude that the total contribution to (36) from all k such
that m = (k, L) > 1 is
L1
(46)
k=0
m=(k,L)>1
(47)
confirming our general analysis that site 1 + k gets connected to 2, and site 2 + k gets
connected to 1, giving trace structure (ii). The first line indicates a contribution of
A6 ,A11 A11 ,A2 x d(A6 )+d(A11 )+d(A2 ) = x 3d(A2 ) ,
(48)
A6 ,A11
210
while the second line similarly denotes a contribution of x 11d(A1 ) . Putting everything together, we find that for L = 14 and k = 5, (36) reduces to
x 11d(A1 )+3d(A2 ) A1 A2 OA2 A1 = P O x 11 , x 3 ,
(49)
A1 ,A2
P O x Ln(L,k) , x n(L,k) ,
(50)
k=0
(k,L)=1
P O x Lk , x k .
(51)
k=0
(k,L)=1
L1
1 L
P O x Lk , x k z x L
O x
,
(52)
k=0
(k,L)=1
where in the first term we omitted the constraint m > 1 from (46) at the expense of subtracting off the extra terms on the second line. Now we can trade the sum over k in the first
term of (52) for a sum over divisors a of L, to write
L/a2 a
O x
(a) z x a
TrL Px d O =
aL
L1
k=0
(k,L)=1
1 L
P O x Lk , x k z x L
O x
.
(53)
211
At this step let us pause for a moment to explain how to recover the Plya formula
(27) as promised. To this end we need to consider the special case where the operator
O is proportional to the identity matrix, and specifically O = 1/L. This is the correct
normalization which gives rise to O = 1 when plugged into the sum over sites in (32). For
O = 1/L we easily find
1
1
O(x) = z(x)2 ,
P O(w, y) = z(wy).
L
L
The second term in (53) therefore drops out, leaving just
1
L/a
TrL Px d =
(a) z x a
.
L
(54)
(55)
aL
L=1
(n)
ln 1 z x n ,
TrL Px d =
n
(56)
n=1
L=2
1
O(x)
O(x n )
+
TrL Px d O =
(n)
n
z(x)
z(x ) 1 z(x n )
n=1
L1
L=2
k=0
(k,L)=1
O(x L )
P O x Lk , x k
.
z(x L )
(57)
The first term is present to subtract off the part of the second term which would correspond
to L = 1, which we omit since it cannot support a pendant (and moreover is irrelevant
in the SU(N ) gauge theory). A final step is to simplify (57) by changing the summation
variable from n to L and combining everything into the main result
TrL Px d O
L=2
O(L+1 x L )
Lk+1 Lk k+1 k
+ L=1 P O
.
x
, x
1 z(L+1 x L )
(58)
L=1 k:(k,L)=1
212
The first term in (58) for L = 1 is precisely what one would obtain by making the crude
estimate that the only effect of the projection P onto cyclically invariant states is to insert
a factor of 1/L. The rest of (58) is the detailed correction to this approximation.
In all the cases relevant to N = 4 SYM theory that we study below, we will see that
the second term in (58) is a very small correction in the sense that its contribution to
the coefficient of x n is negligible for large n. In particular, we will find that O(x) and
P O(w, y) converge for all temperatures so that the large temperature behavior of (58) is
dominated by the pole 1/(1 z(x)) in the first term.
4. Examples
We can gain some insight into the formula (58) by applying it to some subsectors of the
gauge theory. The implication of the results presented in this section for the thermodynamics of N = 4 SYM theory is unclear since there is no sense in which the sectors decouple
from each other at finite temperature (we do not consider here the addition of chemical
potentials for various charges). However, we believe this section is a useful warmup exercise for the more complicated analysis which follows. Moreover, the results given here
for the traces of the SU(2), SO(6) and SL(2) spin chain Hamiltonians may be of interest
from the point of view of integrability in those sectors. An additional subsector which is
of independent interest is the SU(24) subsector, which at one loop is isomorphic to the
t Hooft large N limit of the planewave matrix model [10]. This subsector is considered
in [29].
4.1. The SU(2) subsector
This subsector consists of all operators of the form
Tr[XXZZZXZZ XXZ],
(59)
where X and Z are two holomorphic scalar fields. The alphabet for this sector is
A = {X, Z}, the dimension formula is d(A) = 1, and the elementary partition function
is z(x) = 2x. In the natural basis {XX, XZ, ZX, ZZ} for A A, the matrix elements of
the permutation operator P and the oneloop Hamiltonian D2 are simply
1 0 0 0
1
0 0 1 0
P =
D2 = (1 P ).
(60)
,
0 1 0 0
2
0 0 0 1
We immediately find
D2 (x) = x 2 ,
(61)
P D2 (w, y) = wy.
Plugging these into the main formula (58) gives, after some simplification, the following
formula for the trace of the Hamiltonian in the SU(2) sector:
1 3x n
.
(n)x n
Tr x D0 D2 = x
(62)
1 2x n
n=1
213
As a check, we used a computer to calculate the trace of the SU(2) spin chain Hamiltonian
for all chains of length L 26 and successfully matched the coefficients in the expansion
of (62) up to order x 26 .
4.2. The SO(6) subsector
This subsector consists of all operators built only out of scalar fields with no derivatives,
Tr I1 IL ,
(63)
Ii = {1, . . . , 6}.
The alphabet is A = { 1 , . . . , 6 }, the dimension formula is d(A) = 1, and the elementary
partition function is z(x) = 6x. In the natural basis I1 I2 = I1 I2 for A A, the matrix
elements of D2 and P are [5]
1
1
1
I1 I2 D2 J1 J2 = I1 I2 J1 J2 + I1 J1 I2 J2 I1 J2 I2 J1 ,
4
2
2
I1 I2 P J1 J2 = I1 J2 I2 J1 .
(64)
27
P D2 (w, y) = wy,
2
D
27
3
n 9 65x
0
Tr x D2 = x
(n)x
2
2
1 6x n
(65)
(66)
n=1
for the trace of the SO(6) Hamiltonian. As a check, we used a computer to calculate the
trace of the SO(6) spin chain Hamiltonian for all chains of length L 11 and successfully
matched the coefficients in the expansion of (66) up to order x 11 .
4.3. The SL(2) subsector
This subsector consists of all operators of the form
Tr D i1 Z D iL Z , in {0, 1, . . .},
(67)
AA
x d(A) =
i=0
x i+1 =
x
.
1x
In the basis i1 i2 = D i1Z D i2Z for A A the matrix elements of D2 are [30]
i1 =j1
1
,
i1 i2 D2 j1 j2 = i1 +i2 ,j1 +j2 i1 ,j1 h(j1 ) + h(j2 )
2
i1 j1 
(68)
(69)
214
j
1
n=1
h(0) = 0.
(70)
(71)
x2
ln(1 x),
(1 x)2
1 wy
(1 w)(1 y)
P D2 (w, y) =
ln
.
2 1 wy
(1 wy)2
(72)
After some simplification, we find for the trace of the SL(2) Hamiltonian the result
Tr x D0 D2 =
(n)
n=1
L=1
xn
ln 1 x n
n
1 2x
xL
1 xL
L
ln 1 x k .
(73)
k=1
(k,L)=1
As a check, we used a computer to calculate the trace of the SL(2) spin chain Hamiltonian
for all chains with total dimension D0 19 and successfully matched the coefficients in
the expansion of (73) up to order x 19 .
(75)
needed to invoke (58) for the full PSL(44) spin chain Hamiltonian D2 .
The calculation of D2 (x) is greatly facilitated by making use of the PSL(44) symmetry of N = 4 SYM theory. Since the dilatation operator D2 commutes with this symmetry,
the action of D2 on an arbitrary state can be decomposed into its action on irreducible
representations of PSL(44). We therefore begin with a discussion of the relevant representations. Unfortunately, the operator w D0(1) y D0(2) does not commute with the twoletter
PSL(44) Casimir, so the calculation of P D2 (w, y) will prove more difficult.
215
(76)
under SL(4) flavor rotations and the SL(2) SL(2) Lorentz algebra. The singleton representation is frequently denoted VF , although we shall continue to refer to it as A for
consistency.
Since the oneloop dilatation operator D2 only acts on two letters at a time and com2
mutes with the twoletter Casimir J(12)
of PSL(44), it is sufficient to consider the decomposition of the product of two copies of the singleton representation into irreducible
representations of PSL(44). The decomposition is
AA=
Vj ,
(77)
j =0
[1, 0, 1](0,0)
for j 2.
(78)
2 2
A = B[0,1,0](0,0)
,
Vj = C 1,1
1 1
2 2
V0 = B[0,2,0](0,0)
,
[0,0,0]( 12 j 1, 12 j 1)
1 1
4 4
V1 = B[1,0,1](0,0)
,
and
for j 2.
(79)
In linguistics, a group of two successive letters whose phonetic value is a single sound,
such as ng in Yang or th in theory, is called a digraph, so we can think of the Vj as the
digraphs of N = 4 YangMills theory.
It is straightforward to count the primary states in Vj , weighted in the usual way by
x D0 . For j = 0 and j 2 the results can be read off respectively from (6.13) and (5.45)
of [31], or from Tables 7 and 8 of [18]. We did not immediately find the primary content of
V1 in the literature, but the derivation thereof is straightforward and the result is presented
in Appendix A. The results for all Vj can be summarized in the expressions
4
V0 (x) = 4x 2 1 + x (5 x),
7
Vj (x) = x j 1 + x j 1 + (j + 2) x j 1 + 5 x (j + 2)x , j 1.
(80)
Setting x = 1 counts the total number of primaries in Vj , which is
+ 1) for any
j 0.
Note that for j = 0, 1 some powers of x in Vj (x) have negative coefficients. This may
be thought of as a bookkeeping device (explained in [31]) which allows for easily keeping
28 (2j
216
track of fields which are eliminated by the requirement of imposing equations of motion
or conservation laws. One consequence of this is that the full partition function for the
module Vj (including descendants) may be calculated naively with derivatives treated as
if they acted freely, without worrying about equations of motion or conservation laws. The
partition function for the module Vj is therefore simply
Vj (x)
,
TrVj x D0 = TrAA Pj x D0 =
(1 x)4
(81)
Pj = 1.
(82)
j =0
TrAA Pj x D0 = TrAA x D0 = z(x)2 ,
(83)
j =0
j
1
n=1
h(0) 0,
(84)
h(j )Pj .
(85)
j =0
Vj (x)
h(j )
.
D2 (x) = TrAA x D0 D2 =
(1 x)4
(86)
j =0
(1 + x )2
2
D2 (x) =
1 4 x + x ln(1 x) x 1 8 x + 2x .
(1 x )6
(87)
217
(where Tr denotes the projection onto singlets of the cyclic group ZL ). Then PSL(44)
invariance guarantees that in each resulting module VI the dilatation operator D2 is proportional to the identity operator, with some calculable eigenvalue hI . The desired trace
would then be given by
hI Tr PI x D0 ,
Tr x D0 D2 =
(89)
L=2 I CL
1
(1 + w )2 (1 + y )2
=
2
(1 w ) (1 y )2 (1 + wy )3 ( w + y )2
(1 w)(1 y) (1 wy )( w + y )2
1
p2 (w, y)
wy p1 (w, y) + ln
2
(1 w)(1 y)
(1 wy)2
1 w (1 + wy )3
1( w y)
p3 (w, y) ,
(90)
ln
2( w+ y)
1 y (1 w)(1 y)
in terms of the three polynomials
218
(91)
and
p3 (w, y) = y 2 p2 (1/y, w) = w 2 p2 (y, 1/w).
(92)
Vj (x)
P D2 (x, x) = TrAA P x D0 D2 =
TrVj P x D0 D2 =
(1)j h(j )
,
(1 x)4
j =0
j =0
(93)
where we used the fact that the permutation operator P acts as
in Vj [33,34]. If we
now substitute the expressions for Vj (x) from (80) and perform the sum over j , we obtain
P D2 (x, x)
1
(1 + x )3
=
x 1 7x 1/2 + x + x 3/2 6x 2 + 2x 5/2
4
3
(1 x ) (1 + x)
1 7x 1/2 + 15x 25x 3/2 + 25x 2 15x 3/2 + 7x 3 x 7/2 ln(1 + x) .
(94)
Encouragingly, this expression agrees precisely with that obtained by setting w = y = x in
(90).
Now we begin in earnest the calculation of (90). The first step is to use the matrix
elements of D2 in a GL(44) oscillator basis, as presented in [30], to write down a sum
which gives (90):
(1)j
P D2 (w, y) =
4
4j
p1 !p2 !
4
4j
(1)
j F1 j F2 j k!(k + 1)!
j
2
1
1
1
1 si + pi Fi (1 + si )(1 + pi ),
2
2
2
(95)
i=1
(96)
219
1 n
.
c(n, 0, 0) = h
2 2
(97)
The detailed derivation of (95), which is not entirely straightforward, is presented in Appendix B. The next several steps of the calculation will be shown schematically. The
skeptical reader is free to verify that the power series expansion of (95) agrees with that of
(90) to any desired order.
After several manipulations similar to the ones in Appendix B, Eq. (95) can be cast into
the form
P D2 (w, y) =
(1)n (n + 1)(n + 3)w 1+n/2 y n/2
n=0
(1 + y )2
n+1 1+n/2 2 ln(1 y)
,
Pn (y) + 1 + (1) y
(98)
y
(1 y )2
where Pn (y) is a polynomial in y whose highest term is O(y n+1 ). We did not obtain
an explicit formula for Pn (y), though presumably it could be reverse engineered from the
final answer (90). Instead, we use a trick by investigating the quantity
(1 y ) (1 w ) 2
.
Q(w, y) = P D2 (w, y)
(99)
(1 + y ) (1 + w )
1 w 2
=
dm w m/2 , dm = m,0 + 4m(1)m
1+ w
(100)
m=1
to write
Q(w, y) =
dm (1)n (n + 1)(n + 3)w 1+n/2+m/2 y n/2 Pn (y) + Wn (y) ,
m,n=0
where
Wn (y) = y
n/2
(101)
2 ln(1 y)
1 y 2
.
1 + (1)n+1 y 1+n/2
1+ y
y
(102)
The trick is now to break the power series expansion of Q(w, y) into the upper diagonal
terms (where the power of y is greater than the power of w), the diagonal terms, and the
lower diagonal terms. Since Q(w, y) is a symmetric function, the lower diagonal terms
will be known once the upper diagonal terms are known. Furthermore, the diagonal terms
can be extracted from (94), so all we have left to calculate are the upper diagonal terms.
But since Pn is a polynomial whose highestorder term is y n+1 , we see from (101) that it
220
never contributes to the upper diagonal. Therefore, for purposes of computing the upper
diagonal we are free to omit the Pn (y) term in (101), and make the replacement
Wn (y) Wn (y)y k : k>1+n/2+m/2 ,
(103)
where the notation means that we write out Wn (y) as a series in y and throw away all
powers of y less than or equal to 1 + n/2 + n/2. This finally gives a formula for Q(w, y)
which is amenable to a calculation in M ATHEMATICA. In this manner, we obtain after a
tedious but straightforward calculation the result (90).
(1)
(106)
221
(a factor of 1/n is canceled by ln x n = n ln x) where we recall that the tree level result
Z (0) (x) is written in (16).
Now let us plug the result from (58) into (106). The term proportional to P D2 gives
the sum
P D2 n(Lk)+1 x n(Lk) , nk+1 x nk
n=1 L=2 k:(k,L)=1
P D2 b+1 x b , a+1 x a .
(107)
a,b=1
To see why this equation is true, pick any positive a and b, and look at the lefthand side to
see how many times (if any) the term P D2 (a+1 x a , b+1 x b ) appears. This is equivalent
to asking how many solutions, for given a and b, there are to the equations
nk = a,
nL = a + b,
(108)
for L 2 and k: (k, L) = 1. The answer is that there is always precisely one solution:
n = (a, b), L = (a + b)/n and k = a/n. Certainly if n were not the greatest common
divisor of a and b but some smaller common divisor, then (108) would still give solutions
for k and L, but these would not satisfy the constraint (k, L) = 1.
Next we plug the D2 (x) terms from (58) into (106), which gives the sum
D2 (nL+1 x nL )
D2 (k+1 x k )
=
(L)
(L)
1 z(nL+1 x nL )
1 z(k+1 x k )
k=1 Lk
n=1 L=1
D2 (k+1 x k )
.
k
=
1 z(k+1 x k )
(109)
k=1
Combining (107) and (109) into (106) gives the final result
D2 (k+1 x k )
k+1 k m+1 m
ln x (0)
(1)
+
P D2 x ,
,
Z (x)
k
x
Z (x) =
4 2
1 z(k+1 x k )
k=1
k,m=1
(110)
as advertised already in (1), for the oneloop partition function of N = 4 SYM theory on
R S 3 , expressed in terms of the free partition function Z (0) , the elementary partition
function (6), and the traces (87) and (90).
6.3. Oneloop Hagedorn temperature
The partition function Z(x) has a simple pole at the Hagedorn temperature
c
Z(x)
(111)
,
xH x
where c is an irrelevant overall numerical coefficient. To compute the oneloop correction
xH to the Hagedorn temperature, we simply expand
c
c
xH
(112)
=
1
+ .
xH + xH x xH x
xH x
222
When we compare this to (110) and recall that xH is such that z(xH ) = 1, we note that
only the k = 1 term in the sum of D2 (x k ) contributes to the double pole at the Hagedorn
temperature.1 Reading off the residue of this pole, we find
2 ln xH
ln x D2 (x)
=
D
(x
)
.
xH = lim (xH x)
(113)
x
H
2
H
xxH
3
4 2 1 z(x)
4 2
Remarkably, we find from (87) that D2 (xH ) = 3/4, which gives
ln xH
xH
=
,
xH
8 2
and hence
TH
1 xH
=
=
.
TH
ln xH xH
8 2
The oneloop Hagedorn temperature is therefore
1
+ , TH (0) =
TH = TH (0) 1 +
.
2
8
ln(7 + 4 3 )
(114)
(115)
(116)
It is encouraging that the sign is positive, consistent with the simple guess that the Hagedorn temperature is a monotonically increasing, smooth function of from TH = T0 at zero
coupling to the AdS/CFT prediction that TH 1/4 at strong coupling.
7. Discussion
In this paper we have presented, in (1), the oneloop correction to the partition function
of SU(N ), N = 4 SYM theory on R S 3 at infinite N and below the Hagedorn temperature.
Several recent papers including [17,3537] address the thermodynamics and phase transition structure of weakly coupled gauge theories with the goal of smoothly connecting
onto the strong coupling predictions implied by the AdS/CFT correspondence. String theory in AdS5 S 5 has both a Hagedorn transition [14] at TH 1/4 and a HawkingPage
transition [38] involving the nucleation of AdS5 black holes at THP = 3/2 . The authors
of [17] lay out the ways in which the phase diagram of the weakly coupled theory can
be matched onto these strong coupling predictions. Various qualitatively different phase
diagrams can in part be distinguished by the sign of a particular coefficient in the effective action for the Polyakov loop U , which is the order parameter for the phase transition.
The computation of this coefficient requires a threeloop calculation in thermal YangMills
theory on S 3 which is in progress [39].
One of the motivations for the present work was the desire to provide an independent
check of some pieces of the calculation of [39] from a completely orthogonal starting
1 It is possible that the second term in (110) develops a pole at the Hagedorn temperature after evaluating the
sum, in which case it would add a finite correction to the oneloop Hagedorn temperature. However, this does
not happen in any of the subsectors that we studied, and numerical evidence suggests that it does not occur here
either.
223
Acknowledgements
It is a pleasure to thank O. Aharony, S. Minwalla, H. Reall and R. Roiban for useful
discussions, correspondence, and comments on the manuscript. We are grateful to J. Plefka
for pointing out a flaw in our original proof of Eq. (51). This research was supported in
part by the National Science Foundation under Grant No. PHY9907949.
4 4
module V1 , using the notation of [31] (where it is referred to as B[1,0,1](0,0)
).
224
(A.1)
With the help of the SL(4) SL(2) SL(2) dimension formula
dim[k, p, q](j1 ,j2 ) =
1
(k + p + q + 3)(k + p + 2)(p + q + 2)(k + 1)
12
(p + 1)(q + 1)(2j1 + 1)(2j2 + 1)
(A.2)
we can immediately read off the partition function for primary states in V1 ,
V1 (x) = 15x 2 + 96x 5/2 + 252x 3 + 336x 7/2 + 210x 4 + 0x 9/2 84x 5
48x 11/2 9x 6 ,
(A.3)
225
(B.2)
1
1
D0 = 1 + a a + b b .
(B.3)
2
2
To consider two letters A A we simply have two copies the above algebra, indexed
by a subscript (i) {(1), (2)}. A general state in A(i) will be labeled by its oscillator
1 , a 2 , b1 , b2 , c1 , c2 , c3 , c4 ) N . In this basis we have
occupation numbers (a(i)
(i)
(i) (i) (i) (i) (i) (i) (i)
1 1
1 1
1 1
2
2
2
3
4
a(i) + a(i)
+ b(i) + b(i)
c(i) + c(i)
,
+ c(i)
+ c(i)
2
2
2
1 1
1 1
2
2
+ b(i) + b(i)
.
+ a(i)
D0(i) = 1 + a(i)
(B.4)
2
2
Then to calculate a trace TrA(i) literally means that we perform a sum of the form
C(i) = 1
A(i)
(C(i) ) =
N(i)
1
1 ,a 2 ,b1 ,b2 =0
a(i)
(i) (i) (i)
(C(i) )
(B.5)
over all possible oscillator numbers, subject to the physical state constraint. As a check, it
is straightforward to confirm using this formula and (B.4) that
D
2x(3 x )
TrA x D0 =
(B.6)
x 0=
3 ,
(1 x )
A
s1 , . . . , sn ; {Ii } = A
(B.7)
I1 (s1 ) AIn (sn ) 0,
where si {1, 2} indicates on which site the oscillator acts. The dilatation operator does
not change the type or number of elementary GL(44) oscillators, but can only cause them
to hop from site 1 to 2 or vice versa according to the rule given in [30]:
226
D2(12) s1 , . . . , sn ; {Ii }
c(n, n12 , n21 )(C(1) )(C(2) )s1
, . . . , sn
; {Ii } ,
=
(B.8)
s1 ,...,sn =1,2
where n12 , n21 count the number of oscillators hopping from site 1 to 2 or vice versa and the
coefficients c(n, n12 , n21 ) are given in (97). In what follows we will consider a generalized
operator of the form (B.8),
n12 n21
q n q12
q21 s1 , . . . , sn
; {Ii } ,
Qs1 , . . . , sn ; {Ii } =
(B.9)
s1
,...,sn
=1,2
n12 n21
with matrix elements q n q12
q21 instead of c(n, n12 , n21 ).
(B.10)
i=1
with
n = a(1) + a(2) ,
si = {1, . . . , 1, 2, . . . , 2}.
a(1)
(B.11)
a(2)
2
s1
,...,sn
=1
n12 n21
q n q12
q21 a(1) , a(2) P
n
i=1
a(s
) 0,
(B.12)
where Q is defined in (B.9). The function h counts the number of ways that the initial state
a(1) , a(2) is mapped to itself, up to a permutation P , under the action of an operator of the
form (B.8), weighted according to the powers of q12 and q21 which tell us the number of
oscillators that have hopped from 1 to 2 or vice versa. An elementary combinatoric analysis
reveals that
a(1) a(2) a(1) +a(2) a(1) a a(2) a
q
q12
q21
ha(1) ,a(2) (q, q12 , q21 ) =
a
a
a=0
= (qq12 )a(1) (qq21 )a(2) F a(1) , a(2) , 1, (q12 q21 )1 , (B.13)
where F is the hypergeometric function.
A similar analysis can be done for the case of a single fermionic oscillator c. For a
general state c(1) , c(2) we find
c(1) c(2)
.
gc(1) ,c(2) (q, q(12) , q(21 ) = (qq12 )c(1) (qq21 )c(2) 1
(B.14)
q12 q21
227
Since fermionic oscillators only have occupation numbers which are 0 or 1, this formula
encodes the four cases
g0,0 = 1,
g0,1 = qq21 ,
g1,0 = qq12 ,
(B.15)
For g0,0 there are no oscillators, so there is no hopping possible. For g0,1 we start with one
oscillator on site 2, which moves to site 1 giving a factor of q21 . In the final case, g1,1 we
have one fermionic oscillator on each site. They can either stay where they are, or they can
flip, accounting for the two terms in g1,1 .
For a system with multiple oscillators we simply multiply together the appropriate partition functions (B.13) and (B.14) for the individual oscillators. For GL(44) this gives
PN(1) ,N(2) (q, q12 , q21 ) = N(1) , N(2) P QN(1) , N(2)
2
2
4
a
a
.
=
h a(1) , a(2)
h b(1) , b(2)
g c(1)
, c(2)
=1
=1
(B.16)
a=1
The quantity
R(w, y; q, q12 , q21 ) = TrAA P w D0(1) y D0(2) Q
(C(1) )(C(2) )w D0(1) y D0(2) PN(1) ,N(2) (q, q12 , q21 )
=
N(1) ,N(2)
(B.17)
then represents a trace over A A which counts all of the possible hoppings between
a state N(1) , N(2) and its permutation N(2) , N(1) , weighted appropriately by w to the
power of the dimension of site 1, times y to the power of the dimension of site 2, times q
to the power of the total number of oscillators on both sites, times q12 to the power of the
number of oscillators hopping from site 1 to 2, times q21 to the power of the number of
oscillators hopping from site 2 to 1. Concretely, we obtain from (B.5), (B.13) and (B.14)
the formula
R(w, y; q, q12 , q21 )
1
(C(1) )(C(2) )
F a(1)
, a(2)
, 1, (q12 q21 )1
1
q12 q21
a=1
=1
2
F b(1)
, b(2)
, 1, (q12 q21 )1
(B.18)
=1
where
n(i) =
2
=1
a(i)
+
2
=1
b(i)
+
4
a=1
a
c(i)
(B.19)
228
s(i)
1 2
f a(i) , a(i) =
f (t(i) , s(i) t(i) ).
(B.20)
s(i) =0 t(i) =0
1 ,a 2 =0
a(i)
(i)
(For the b oscillators we will use s and t as the new summation variables.) After this
substitution, the t variables only appear in the arguments of the hypergeometric functions.
The sum over t(1) and t(2) can be done with the help of the identity
s(1) s(2)
t(1) =0 t(2) =0
(B.21)
The sum over fermionic occupation numbers can be similarly simplified with the formula
4
4
1
4
a
a
a a
1 c(1)
f
c(1) ,
c(2)
c(2) z
1 ,c2 ,c3 ,c4 ,c1 ,c2 ,c3 ,c4 =0
c(1)
(1) (1) (1) (2) (2) (2) (2)
4
4
F(1) =0 F(2) =0
a=1
a=1
a=1
4
4j
4j
j j 4
.
f (F(1) , F(2) )
(1) z
j F(1) j F(2) j
(B.22)
j =0
4
(1)j zj
4
4j
4j
j F(1) j F(2) j
2
1
1
1
(B.23)
i=1
where
z=
1
,
q12 q21
(B.24)
Of course, we are not particularly interested in the quantity R(w, y; q, q12 , q21 ). It is
simply an auxiliary quantity which counts the possible ways for the operators to hop between sites. We are interested in the trace of (B.8) instead of the trace of (B.9), which we
can calculate via the replacement
P D2 (w, y) = R(w, y; q, q12 , q21 )q n q n12 q n21 c(n,n12 ,n21 ) ,
(B.25)
12
21
229
where the notation means simply that we expand (B.23) in powers of q, q12 and q21 and
then make the substitution indicated for the various powers.
In order to proceed, we expose the powers of q12 and q21 in (B.23) by using the identity
F (s(1) , s(2) , 2, z)F (s(1) , s(2) , 2, z)
s(1) !s(2) !
F(1 k, k, s(1) , s(2) ; 2, 1 k + s(1) , 1 k + s(2) ; 1)
=
zk
k!(k + 1)!
k=0
(B.26)
where F is the regularized hypergeometric function. Combining (B.25), (B.26) and (B.23)
finally leads to the formula (95) for the desired trace, after some simplification of the
notation.
References
[1] I. Bena, J. Polchinski, R. Roiban, Hidden symmetries of the AdS5 S 5 superstring, Phys. Rev. D 69 (2004)
046002, hepth/0305116.
[2] J.K. Erickson, G.W. Semenoff, K. Zarembo, Wilson loops in N = 4 supersymmetric YangMills theory,
Nucl. Phys. B 582 (2000) 155, hepth/0003055.
[3] N. Drukker, D.J. Gross, An exact prediction of N = 4 SUSYM theory for string theory, J. Math. Phys. 42
(2001) 2896, hepth/0010274.
[4] S.S. Gubser, I.R. Klebanov, A.M. Polyakov, A semiclassical limit of the gauge/string correspondence, Nucl.
Phys. B 636 (2002) 99, hepth/0204051.
[5] J.A. Minahan, K. Zarembo, The Betheansatz for N = 4 superYangMills, JHEP 0303 (2003) 013, hepth/0212208.
[6] N. Beisert, M. Staudacher, The N = 4 SYM integrable superspin chain, Nucl. Phys. B 670 (2003) 439,
hepth/0307042.
[7] A.A. Tseytlin, Spinning strings and AdS/CFT duality, hepth/0311139.
[8] A.A. Tseytlin, Semiclassical strings in AdS5 S 5 and scalar operators in N = 4 SYM theory, hepth/0407218.
[9] N. Beisert, The dilatation operator of N = 4 superYangMills theory and integrability, hepth/0407277.
[10] D. Berenstein, J.M. Maldacena, H. Nastase, Strings in flat space and pp waves from N = 4 superYang
Mills, JHEP 0204 (2002) 013, hepth/0202021.
[11] A. Santambrogio, D. Zanon, Exact anomalous dimensions of N = 4 YangMills operators with large R
charge, Phys. Lett. B 545 (2002) 425, hepth/0206079.
[12] A.V. Ryzhov, A.A. Tseytlin, Towards the exact dilatation operator of N = 4 superYangMills theory, Nucl.
Phys. B 698 (2004) 132, hepth/0404215.
[13] G. Arutyunov, S. Frolov, M. Staudacher, Bethe ansatz for quantum strings, JHEP 0410 (2004) 016, hepth/0406256.
[14] E. Witten, Antide Sitter space, thermal phase transition, and confinement in gauge theories, Adv. Theor.
Math. Phys. 2 (1998) 505, hepth/9803131.
[15] B. Sundborg, The Hagedorn transition, deconfinement and N = 4 SYM theory, Nucl. Phys. B 573 (2000)
349, hepth/9908001.
[16] B. Sundborg, Stringy gravity, interacting tensionless strings and massless higher spins, Nucl. Phys. B (Proc.
Suppl.) 102 (2001) 113, hepth/0103247.
[17] O. Aharony, J. Marsano, S. Minwalla, K. Papadodimas, M. Van Raamsdonk, The Hagedorn/deconfinement
phase transition in weakly coupled large N gauge theories, hepth/0310285.
[18] M. Bianchi, J.F. Morales, H. Samtleben, On stringy AdS5 S 5 and higher spin holography, JHEP 0307
(2003) 062, hepth/0305052.
[19] A.M. Polyakov, Gauge fields and spacetime, Int. J. Mod. Phys. A 17S1 (2002) 119, hepth/0110196.
[20] M.B. Halpern, On the large N limit of conformal field theory, Ann. Phys. 303 (2003) 321, hepth/0208150.
230
[21] N. Beisert, C. Kristjansen, J. Plefka, G.W. Semenoff, M. Staudacher, BMN correlators and operator mixing
in N = 4 superYangMills theory, Nucl. Phys. B 650 (2003) 125, hepth/0208178.
[22] N.R. Constable, D.Z. Freedman, M. Headrick, S. Minwalla, Operator mixing and the BMN correspondence,
JHEP 0210 (2002) 068, hepth/0209002.
[23] N.R. Constable, D.Z. Freedman, M. Headrick, S. Minwalla, L. Motl, A. Postnikov, W. Skiba, PPwave string
interactions from perturbative YangMills theory, JHEP 0207 (2002) 017, hepth/0205089.
[24] I.R. Klebanov, M. Spradlin, A. Volovich, New effects in gauge theory from ppwave superstrings, Phys.
Lett. B 548 (2002) 111, hepth/0206221.
[25] N. Beisert, C. Kristjansen, M. Staudacher, The dilatation operator of N = 4 superYangMills theory, Nucl.
Phys. B 664 (2003) 131, hepth/0303060.
[26] D.J. Gross, A. Mikhailov, R. Roiban, A calculation of the plane wave string Hamiltonian from N = 4
superYangMills theory, JHEP 0305 (2003) 025, hepth/0208231.
[27] J. Pearson, M. Spradlin, D. Vaman, H. Verlinde, A. Volovich, Tracing the string: BMN correspondence at
finite J 2 /N , JHEP 0305 (2003) 022, hepth/0210102.
[28] N. Beisert, C. Kristjansen, J. Plefka, M. Staudacher, BMN gauge theory as a quantum mechanical system,
Phys. Lett. B 558 (2003) 229, hepth/0212269.
[29] M. Spradlin, M. Van Raamsdonk, A. Volovich, Twoloop partition function in the planar planewave matrix
model, Phys. Lett. B 603 (2004) 239, hepth/0409178.
[30] N. Beisert, The complete oneloop dilatation operator of N = 4 superYangMills theory, Nucl. Phys. B 676
(2004) 3, hepth/0307015.
[31] F.A. Dolan, H. Osborn, On short and semishort representations for fourdimensional superconformal symmetry, Ann. Phys. 307 (2003) 41, hepth/0209056.
[32] N. Beisert, M. Bianchi, J.F. Morales, H. Samtleben, Higher spin symmetry and N = 4 SYM, JHEP 0407
(2004) 058, hepth/0405057.
[33] L. Dolan, C.R. Nappi, E. Witten, A relation between approaches to integrability in superconformal Yang
Mills theory, JHEP 0310 (2003) 017, hepth/0308089.
[34] L. Dolan, C.R. Nappi, E. Witten, Yangian symmetry in D = 4 superconformal YangMills theory, hepth/0401243.
[35] L. Fidkowski, S. Shenker, Dbrane instability as a large N phase transition, hepth/0406086.
[36] H. Liu, Fine structure of Hagedorn transitions, hepth/0408001.
[37] O. Aharony, J. Marsano, S. Minwalla, T. Wiseman, Black holeblack string phase transitions in thermal
(1 + 1)dimensional supersymmetric YangMills theory on a circle, Class. Quantum Grav. 21 (2004) 5169,
hepth/0406210.
[38] S.W. Hawking, D.N. Page, Thermodynamics of black holes in antide Sitter space, Commun. Math. Phys. 87
(1983) 577.
[39] O.Aharony, J. Marsano, S. Minwalla, K. Papadodimas, M. Van Raamsdonk, in preparation.
Abstract
TypeIIB supergravity in ten dimensions admits two consistent Z2 truncations. After the insertion
of D9branes, one of them leads to the lowenergy action of typeI string theory, and it can be performed in two different ways, in correspondence with the fact that there are two different consistent
tendimensional typeI string theories, namely, the SO(32) superstring and the USp(32) model, in
which supersymmetry is broken on the D9branes. We derive here the same results for typeIIA theory compactified on a circle in the presence of D8branes. We also analyze the symmetric action
for a brane charged with respect to the Sdual of the RR 10form of typeIIB, and we find that the
tension of such an object has to scale like gS2 in the string frame. We give an argument to explain
why this result is in disagreement with the one obtained using Weyl rescaling of the brane action, and
we argue that this brane can only be consistently introduced if the other Z2 truncation of typeIIB
is performed. Moreover, we find that one can include a 10form in typeIIA supersymmetry algebra,
and also in this case the corresponding symmetric brane has a tension scaling like gS2 in the string
frame.
2005 Elsevier B.V. All rights reserved.
PACS: 11.25.w; 11.30.Pb
1. Introduction
TypeII string theories in the nonperturbative regime contain in their spectrum BPS
Dbranes, that are charged states with respect to RR fields [1], and are defined as hyEmail address: f.riccioni@damtp.cam.ac.uk (F. Riccioni).
05503213/$ see front matter 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.nuclphysb.2005.01.034
232
persurfaces on which open strings end [2]. In the lowenergy effective action, these states
appear as 1/2supersymmetric solitonic solutions carrying electric or magnetic charge with
respect to the RR fields of typeIIA and typeIIB supergravities. The effective action describing the massless modes of a Dbrane is characterized by a DiracBornInfeld (DBI)
term and a WessZumino (WZ) term. In [3,4] it was shown that the effective action describing the massless open string states at string tree level is the DBI action in the approximation
in which one neglects derivatives of the field strength, while the coupling to the RR fields
is contained in the WZ term. The relative coefficient of the DBI and WZ terms is fixed,
since the tension and the RRcharge of the brane are related by the BPS condition.
The method for constructing actions for supersymmetric Dbranes is known in the literature [59]. These actions are obtained embedding the worldvolume of the brane in
superspace. The fermionic superspace coordinate becomes consequently a fermion on the
brane. Apparently, this seems to imply that the brane breaks all the supersymmetries, since
the fermion plays the role of the goldstino field. The solution of this apparent paradox is
the fact that the brane action possesses an additional local fermionic symmetry, known as
symmetry [10,11], whose role is to decouple half of the fermions in the brane action.
After gauge fixing half of the supersymmetries become linearly realized, while the other
half are still nonlinearly realized, la VolkovAkulov [12]. Therefore, symmetry is a
basic ingredient in the construction of brane actions, and it is the worldvolume remnant of
the BPScondition.
Spacetimefilling Dbranes characterize the vacua of typeI models. TypeI string theory is obtained from typeIIB through an orientifold projection [13] that removes the states
that are odd under orientation reversal of the string. From a target space point of view, this
can be pictured in terms of orientifold planes, and the appearance of tadpoles corresponds
in this picture to nonvanishing tension and charge of the Oplane. Tadpole cancellation
typically requires the introduction of an open sector, and this corresponds to Dbranes. In
ten dimensions, one can thus introduce an O plane (with negative tension and negative
charge) and 32 D9branes, with a resulting gauge group SO(32) [14]. The cancellation
of the overall tension and charge of the configuration corresponds to the cancellation of
dilaton and RR tadpoles [15,16]. The resulting theory is N = 1 supersymmetric, and the
massless spectrum contains the gravity multiplet from the closed sector and an SO(32)
YangMills multiplet from the open sector. There is actually a second possibility, corresponding to a change of sign of the tension and the charge of the orientifold plane, so that
RR tadpole cancellation requires the addition of 32 antiD9 branes, with a resulting gauge
group USp(32) [17]. The overall tension of the configuration does not vanish, so that the
resulting theory has a dilaton tadpole. Nevertheless, the theory is anomalyfree, as a consequence of the vanishing of the RR tadpole [16,18]. The spectrum is not supersymmetric,
and more precisely the closed sector is not modified, still describing at the massless level
the N = 1 gravity multiplet, while the massless fermions in the open sector are not in
the adjoint but in the antisymmetric representation of USp(32), so that supersymmetry is
broken on the brane [19]. The gravitino couplings can then only be consistent if supersymmetry is nonlinearly realized in the open sector. Since the antisymmetric representation of
symplectic groups is reducible, the massless spectrum contains a spinor that is an USp(32)
singlet, and this spinor is the goldstino of the nonlinearly realized supersymmetry [20].
The presence of the NS tadpole is a manifestation of the fact that the theory has been ex
233
panded around the wrong vacuum, and an analysis of this problem, addressed long time
ago in [21], has been recently performed in [22].
From the point of view of the lowenergy effective action, the closed sector of typeI
strings is obtained performing a consistent Z2 truncation of the typeIIB theory, while the
open sector corresponds to the first order in the lowenergy expansion of the D9brane
action in a typeI background. It is then natural to ask what is the fate of symmetry in
this background. The result is that there are two possibilities of performing this truncation
[23], and in a flat background, with all bulk fields put to zero, the D9brane action reduces
in one case to the VolkovAkulov (VA) action [12], and in the other case to a constant. In
[24] these results were extended to a generic background, showing that also in the curved
case there are two possibilities of performing the truncation. In one case one gets a dilaton
tadpole and a RR tadpole plus goldstino couplings, while in the other case the goldstino
couplings vanish and one is left with a dilaton and a RR tadpole. This result is equivalent
to the string result: the two different truncations correspond to the two different choices of
the relative sign of tension and charge of orientifold plane and D9branes. The first case
corresponds to the nonsupersymmetric one, in which the orientifold plane and the Dbrane
have both positive tension, and in the case of 32 coincident D9branes it gives rise to the
lowenergy action [20,25] of the USp(32) model. The second case, in which the goldstino
disappears, corresponds to an orientifold plane with negative tension, and in the case of 32
coincident D9branes it gives rise to the lowenergy action of the supersymmetric SO(32)
superstring. In other words, the supersymmetric truncation projects out the spinor that
would not be projected out fixing the symmetry gauge. As a result, in general one expects
that only linearly realized supersymmetry survives. The nonsupersymmetric truncation
does the opposite, namely, it projects out the spinor that would have been projected out
by symmetry, so that no symmetry is left in the truncated theory, and in the resulting
action supersymmetry is only nonlinearly realized, i.e., completely broken.
If one wants to generalize these results to lowerdimensional cases, the first possibility is to consider the Tdual of this configuration, that is a typeI orientifold of typeIIA
compactified on a circle [2]. In this case the orientifold projection has fixed points on the
circle, corresponding to the positions of the Oplanes. In [26] the lowenergy action for
a D8brane located at one of the fixed points was constructed in the typeI background,
without including the fermionic fields. In this paper we want to apply the techniques used
in [24] to this case, in order to obtain the lowenergy brane + bulk action up to four Fermi
terms, for a generic 9dimensional background. We will see that the results are in complete
agreement with Tduality, since also in this case one has two possible consistent truncations, leading in one case to a supersymmetric model, and in the other case to a model in
which supersymmetry is nonlinearly realized on the brane.
Sduality is a symmetry of typeIIB string theory mapping weak coupling to strong
coupling [27]. On the other hand, typeI string theory is related in ten dimensions by a
strongweak coupling Sduality to the heterotic SO(32) theory [28]. In this respect, it is
interesting to study the behavior of the O9D9 system of [23,24] under Sduality, and
whether the result can be related to the lowenergy action of the heterotic theory. TypeIIB
supersymmetry algebra includes two 10forms [23,29]. One of them is the RR 10form
that couples to D9branes, while the other couples to other spacetimefilling branes, called
NS9branes in [29]. TypeIIB supergravity also admits an additional Z2 truncation, re
234
moving all the RR fields. This truncation was conjectured in [29] to be the Sdual of the
orientifold projection, and consequently the introduction of 32 NS9branes was conjectured to give origin to the SO(32) heterotic string after performing this projection [29,30].
Under Sduality, the DBI part of the D9brane action acquires a dilaton factor e4 in the
string frame, and this led to conjecture that the tension of the NS9branes is proportional to
gS4 [29,31]. We will argue in this paper that this is actually not the case. We will prove that
symmetry requires a dilation factor e2 in front of the DBI term of an NS9brane action, and thus a tension proportional to gS2 . This seems to be inconsistent with the analysis
of [31], since starting from a D9brane action and performing an Sduality transformation
one should end up with a symmetric action. The solution of this paradox is that in the
presence of NS and RR 10forms Sduality is no longer a symmetry of the typeIIB algebra.
This does not mean that Sduality symmetry of typeIIB is actually broken, since introducing spacetimefilling branes is only consistent after performing a truncation. We will also
analyze the typeIIA case, since typeIIA supersymmetry algebra can be extended including an NS 10form, and the resulting supersymmetric NS9brane action is symmetric if
the tension scales like e2 in the string frame.
The paper is organized as follows. In Section 2 we review some known results about
superDbranes. In Section 3 we discuss the typeI truncation of typeIIA compactified on
a circle. We make use of the democratic formulation of the theory [26], in which both
the RR forms and their magnetic duals appear as independent fields in the supersymmetry
algebra, and duality relations between electric and magnetic field strengths are imposed as
constraints (the same formulation was introduced in [23] for the typeIIB case). We show
that the two possible truncations lead in one case to a supersymmetric model, and in the
other case to a model in which supersymmetry is spontaneously broken on the D8brane.
In Section 4 we discuss Sduality of typeIIB in the presence of NS and RR 10forms.
First of all, one realizes that these two fields, besides transforming as a doublet under Sduality, acquire a dilaton dependence e2 . Moreover, an additional constraint has to be
imposed for Sduality to be a symmetry. In other words, Sduality is broken in the presence
of spacetimefilling branes. Section 5 is devoted to the study of the supersymmetric NS9brane in both IIA and IIB. We show that symmetry implies that the tension of the NS9brane in the string frame scales like e2 . In analogy to the D9 [24] and the D8 cases, one
can perform a truncation to show that half of the fermions decouple from the spectrum. In
this case the spectrum is projected by means of a heterotic truncation. Finally, Section 6
contains the conclusions.
235
the Dbranes of typeIIB, we write down the IIB algebra in the democratic formulation, in
which all the forms and their magnetic duals appear in the algebra. Following the notations
of [23], the supersymmetry transformations of the IIB bulk fields are
e a = a ,
1
1
1
= D H 3 + e
G(2n+1) 1 ...2n+1 Pn ,
8
16
(2n + 1)! 1 ...2n+1
5
n=0
(2)
B
= 2 3 [ ] ,
(10)
B1 ...10 = e2 3 (10[1 ...9 10 ]
C(2n)
1 ...2n
1 ...10 ),
1
2n ]
= (2n)e Pn [1 ...2n1 2n ]
2(2n)
(2n2)
=
n=0
1
= ,
(2.1)
2
where Pn is 1 for n odd and i2 for n even. We are neglecting terms cubic in the fermions
in the case of the transformations of the spinors. We have introduced the field strengths for
the RR fields and their duals, related by duality according to the relations
G(7) = G(3) ,
G(9) = G(1) ,
G(5) = G(5) .
(2.2)
An advantage of this formulation is that all the ChernSimons terms in the supergravity
Lagrangian are hidden in the definitions of the field strengths and their magnetic duals.
The matching between bosonic and fermionic degrees of freedom is of course restored
only once these duality relations are imposed.1 The field strengths are defined through the
relations
H = dB,
(2.3)
(2n1)
C (2n) = dRR
(10)
B (10) = dN S ,
(2n3)
RR
H,
(2.4)
so that the field strengths are gauge invariant. The dilaton dependence in the variations of
the forms shows that the algebra of Eq. (2.1) is expressed in the string frame. Moreover,
it is important to observe that two 10forms are present in the algebra. We stress again
that, even though these forms do not have any dynamics since they do not have any field
1 This is a generalization of what is typically done for the selfdual 5form field strength, when one writes a
Lagrangian for an ordinary 5form, and imposes selfduality as a constraint on the equations of motion.
236
strength, they are associated to spacetimefilling branes, whose presence is consistent only
after one performs a suitable projection of the spectrum.
The general idea is to describe supersymmetric Dbranes through the embedding of
a bosonic brane in superspace [5,7,8]. We thus introduce the worldvolume fields as the
supercoordinates
Z M i = x i , I i
(2.5)
defining the position of the brane in superspace. Here i are the worldvolume coordinates (i = 0, . . . , 9 for a 9brane), while = 0, . . . , 9 is a spacetime vector index and
= 1, . . . , 32 a spinor index, and I = 1, 2. The Majorana spinors I are both lefthanded.2
We denote with V i ( ) the Abelian worldvolume vector. The bulk superfields are denoted
with
(2n)
, EM A , BMN , BM1 ...M10 , CM
(2.6)
, n = 0, . . . , 5,
1 ...M2n
and the brane action is
S = SDBI + SWZ =
d 10 e det(g + F) +
CeF ,
M10
(2.7)
M10
where
Fij = Fij + Bij ,
(2.8)
and one defines the pullback of the bulk fields on the worldvolume according to
gij = Ei a Ej b ab ,
Bij = i Z M j Z N BMN
(2.9)
and
C=
5
(1)n C (2n) ,
C (2n) =
n=0
1
(2n)
dZ M1 dZ M2n CM1 ...M2n .
(2n)!
(2.10)
In a flat space background [6,9] these expressions have a simpler form, since from the
(global) supersymmetry transformations of the supercoordinates,
1
,
2
one derives a supersymmetry invariant object
= ,
i = i x +
x =
(2.11)
1
i ,
2
(2.12)
that is the flat space analogous of i Z M EM a . Consequently, the pullback of the metric
becomes
gij = i j = i x j x + (i x j ) + ,
2 In the IIA case the two chiral spinors I are substituted with a single nonchiral Majorana spinor.
(2.13)
237
where we neglect higher terms in the fermions. Analogously, the pullback of the NS
2form is
3 j ] + ,
Bij = [i x
(2.14)
(2.15)
The brane action (2.7) is then supersymmetric, provided that one chooses the supersymmetry transformation for the worldvolume vector Vi to be
1
1
Vi = i 3 j Fj i
(2.16)
2
2
up to a gauge transformation.
The action (2.7) is invariant under worldvolume general coordinate transformations,
and one can then choose a static (or Monge) gauge, in which the coordinates i are
identified with x i , i = 0, . . . , p, where p + 1 is the spacetime dimension of the brane.
A supersymmetry variation then induces a compensating general coordinate transformation, and the resulting variation for is
1 i
i .
(2.17)
2
The other xs in this gauge become worldvolume scalars, whose supersymmetry transformations is
1
1
a = a i i a , a = p + 1, . . . , 10.
(2.18)
2
2
Focusing again on the flat space limit, one can recognize in Eq. (2.17) the Volkov
Akulov (VA) transformations [12]. We will concentrate in the following on spacefilling
9branes, so that the target spacetime matrices can be identified with the worldvolume
matrices, and the spacetime index is the same as the worldvolume index i. The commutator of two transformations (2.17) is a translation,
[1 , 2 ] = 2 1 ,
(2.19)
=
(2.20)
ea = L ea ,
with L the Lie derivative with respect
(2.21)
to3
1
= ( ).
2
3 The parameter should not be confused with the worldvolume coordinates.
(2.22)
238
The action of supersymmetry on e is thus a general coordinate transformation, with a parameter depending on , and therefore
L = det e
(2.23)
is clearly an invariant Lagrangian. Using the same technique, for a generic field A that
transforms under supersymmetry as
A = L A,
(2.24)
me
m ,
(2.25)
This is what happens in the brane action (2.7) in the Monge gauge in a flat space background, since the pullback of the metric of Eq. (2.13) in the Monge gauge equals the VA
metric g . Moreover, the second term in the variation of V i is a general coordinate transformation with the same parameter plus an additional gauge transformation, while the
first term combines with the variation of the pullback of the NS form of Eq. (2.14) in such
a way that F transforms covariantly. Finally, the pullbacks of the RR forms in Eq. (2.15)
are such that the WZ term in the brane action transforms as a total derivative.
It is then natural to generalize this VA construction to D9branes in a generic background. One must construct from the bulk fields quantities whose supersymmetry variations are general coordinate transformations with the parameter plus additional gauge
transformations [20,25]. Supersymmetry guarantees that this way of constructing the D9brane action coincides with the superspace construction of [7,8].4 For instance, from the
supersymmetry variation of one defines
1 n3
1
1
(2n1) i1 ...i2n1
ij k 3 + e
G
Pn ,
= + Hij k
2
48
16
(2n 1)! i1 ...i2n1
n=1
(2.26)
whose supersymmetry transformation is a general coordinate transformation with the correct parameter i given in (2.22), up to higher order Fermi terms. With the same technique,
one can construct all the other hatted fields [20,25], so that the resulting D9brane action
in a generic typeIIB background is
F,
S = SDBI + SWZ =
(2.27)
d 10 e det(g + F) +
Ce
6
M10
M10
where
Fij = Fij + B ij .
(2.28)
We come now to a brief discussion of the degrees of freedom that the action (2.7) propagates. If all the fermions were dynamical this would lead to a complete spontaneous
4 See [32] for a similar construction in the case of a generic pbrane.
239
n = 1, 3, 5,
(10)
= 0,
(1 1 )f = 0,
(2.31)
where we have denoted with f the gravitino and the dilatino. The surviving bosonic fields
are thus the dilaton, the metric, the RR 2form and its dual, and the RR 10form, while the
two different signs in the projection of the fermions indicate that there are two possible
typeI truncations.
The truncation on the D9brane action was performed in [23] in flat space, and generalized in [24] to an arbitrary background. We review here the results. The brane fields are
projected according to
Vi = 0,
(1 1 ) = 0.
(2.32)
The lower sign choice leads to no surviving symmetry, since it projects out the spinor
components that would have been put to zero using symmetry before the truncation,
while the upper sign choice leads to no leftover components of . This last choice, then,
5 See [34] for a similar analysis.
240
corresponding to the vanishing of all the terms containing the goldstino, results in a supersymmetric typeI spectrum. Actually, in the case of a single D9brane, there are no
remaining worldvolume degrees of freedom after the truncation, but the generalization to
a stuck of branes would result in a spectrum in which supersymmetry is linearly realized,
and the goldstino is projected out. The resulting action contains a dilaton tadpole and a RR
tadpole, that in the SO(32) string are both canceled against the orientifold plane contribution. The other choice, instead, corresponds to the curved generalization of the VA action.
The resulting spectrum breaks supersymmetry in the brane sector [19], or more precisely
N = 1 supersymmetry is nonlinearly realized on the brane. The brane action again contains a dilaton tadpole and a RR tadpole, but in this case, a suitable orientifold projection
only cancels the brane RR charge, and a dilaton tadpole remains [17].
The typeIIB supersymmetry algebra in D = 10 also admits an alternative Z2 truncation, projecting out all the RR fields and acting as (1 3 )f = 0 on the fermions, and for
this reason called heterotic truncation [23]. We will show in Section 5 how this truncation
can be consistently implemented on spacetimefilling branes electrically charged with respect to B (10) , after a discussion about Sduality of typeIIB carried out in Section 4. First,
in the next section, we are going to describe the Tduals of these results, i.e., the typeI
truncation of IIA in the presence of D8branes.
3. TypeI truncation of IIA
After reduction to D = 9, Tduality relates the system described in the previous section
to typeIIA theory compactified on a dual circle, and the corresponding truncation is in this
case the lowenergy manifestation of the typeI orientifold projection. In this section we
want to discuss this truncation in the presence of D8branes. Since D8branes are charged
with respect to the RR 9form, whose field strength is dual to a cosmological constant,
the massive Romans IIA supergravity [35] is the bulk lowenergy theory describing this
system [28]. In has been shown in [26] that both massless and massive 10dimensional IIA
supergravities can be described in terms of the same supersymmetry algebra, once the Romans cosmological constant is treated as a dynamical 0form dual to the RR 10form field
strength. Again, it is convenient to work in the democratic formulation [23,26], treating all
the RRforms and their magnetic duals as independent, and imposing the duality relations
as constraints. The resulting supersymmetry algebra is
e a = a ,
1
1
1
G(2n) 1 ...2n (11 )n ,
= D + H 11 + e
8
16
(2n)! 1 ...2n
5
n=0
(2)
B
= 2 11 [ ] ,
C(2n1)
1 ...2n1
= (2n 1)e
(11 ) [1 ...2n2
n
1
2n1 ]
]
2(2n 1) 2n1
(2n3)
B2n2 2n1 ] ,
+ (n 1)(2n 1)C[
1 ...2n3
241
1
1 5 2n (2n)
H 11 + e
G
(11 )n 1 ...2n ,
12
8
(2n)! 1 ...2n
5
=
n=0
1
= .
2
The RR field strengths are defined as
(3.1)
(2)
(3.2)
(2)
eB ,
(3.3)
The field equations of typeIIA supergravity obtained in this formulation are supersymmetric only after these duality relations are imposed. This algebra can be naturally extended to
include a 10form, whose supersymmetry transformation is
= e2 (10 [1 ...9 10 ] + 1 ...m10 ).
B(10)
1 ...10
(3.4)
(3.5)
but we will show that it does not correspond to any spacetimefilling symmetric brane.
Consequently, we will only consider (3.4) as a natural extension of the typeIIA supersymmetry algebra.
We now continue reviewing the results of [26] concerning the possible consistent Z2
truncations of the algebra of Eqs. (3.1) and (3.4). In 10 dimensions, only a single truncation
is available, projecting out all the RR fields, and acting on the fermions as
{ , , } 11 { , , }.
(3.6)
We will construct in Section 5 the resulting symmetric spacetimefilling brane, for which
this truncation is consistent in the way described in the previous section. It will turn out
that this brane is electrically charged with respect to the 10form B (10) .
If we compactify the theory on a circle S 1 , the resulting theory admits another Z2 truncation, acting on the compactified coordinate as
x 9 x 9 ,
(3.7)
thus acting as an orbifold projection, being the lowenergy manifestation of the orientifold
projection generated by introducing two orientifold 8planes at the fixed points. If we only
consider spacetime indices in the uncompactified directions, the projection acts on the
fields according to
(2)
(2)
g , , B
g , , B
,
C(2n1)
()n+1 C(2n1)
,
1 ...2n1
1 ...2n1
{ , , } 9 { , , }.
(3.8)
242
Any index in the 9direction corresponds to an additional minus sign with respect to these
projection rules, and consequently the 10form B (10) of Eq. (3.4) (having an index in the
9direction) is consistently projected out.6 In order to make the analogy with the IIB case
in 10 dimensions manifest, we define the 10dimensional matrices as
= 2 ,
9 = 1 1 ,
11 = 1 3
(3.9)
(1 1 ) = 0.
(3.10)
We now want to consider the introduction of D8branes, and we will only take into
account the case in which a single D8brane is located at one of the two fixed points of the
orientifold projection. Consistency requires that the truncation acts on the worldvolume
fields as7
V i = 0,
(1 1 ) = 0.
(3.11)
det g + C (9) .
e
(3.12)
M9
M9
The supersymmetrization of this action is obtained in the Monge gauge following the same
arguments of the previous section. Taking into account only the terms that are relevant after
the truncation, and neglecting higher order Fermi fields, we thus define the hatted fields
1
= + + ,
2
g = g + 2 ( ) + ( D) + ,
1
C (9)1 ...9 = C(9)1 ...9 9e [1 ...8 11 9 ] e 1 ...9 11
2
9
e [1 ...8 11 D9 ] + ,
(3.13)
2
whose supersymmetry transformation has the form of a dependent general coordinate
transformation (plus an additional gauge transformation in the case of the 9form). Expressing then the brane action in terms of these hatted fields, it turns out that if one chooses
the upper sign in the projection of the fermions, all the terms containing the goldstino
disappear in the action, while the lower sign choice leads to an action of the VA type for .
We interpret this result in the same way as we did for the IIB case in 10 dimensions. The
6 In the case of B (10) defined in Eq. (3.5), consistency would require that this form survive the projection.
7 The other worldvolume field, the scalar x 9 , is of course projected out because of Eq. (3.7).
243
upper sign choice corresponds to a supersymmetric spectrum. Again, just as in the case of
a single D9brane, for a single D8brane there are no remaining worldvolume degrees of
freedom after the truncation, but the generalization to a stuck of branes would result in a
spectrum in which supersymmetry is linearly realized, and the goldstino is projected out.
The resulting action contains a dilaton tadpole and a RR tadpole, that in consistent supersymmetric orientifold models are both canceled against the orientifold plane contribution.
The other choice, instead, corresponds to the case in which the brane and the orientifold
plane have both positive tension. Consequently, N = 1 supersymmetry is nonlinearly realized on the brane, and a suitable orientifold projection only cancels the brane RR charge,
while a dilaton tadpole remains. The fact that this result is in agreement with the IIB result
of Refs. [23,24] is a manifestation of Tduality [36,38].
4. Sduality of typeIIB
TypeIIB superstring theory is conjectured to be invariant under SL(2, Z) transformations [27], a discrete subgroup of the isometry group SL(2, R) of typeIIB supergravity
[39]. This group acts on the complex scalar
= C0 + ie
(4.1)
as
where
a
c
a + b
,
c + d
b
d
(4.2)
SL(2, R),
while the 2forms B (2) and C (2) transform as a doublet. The matrix
0 1
S=
1 0
(4.3)
(4.4)
(4.5)
244
=
10e
,
C(10)
1 [1 ...9
10 ]
]
1 ...10
20 10
1
1
= H 3 e G(3)
,
1
12
12
1
= ,
(4.6)
2
again neglecting higher order Fermi terms in the transformations of the fermions.
Our strategy will be to derive the transformations of the fields under Sduality requiring
that the supersymmetry algebra is preserved. We already know that
,
e a e/2 e a .
(4.7)
We now obtain the transformations of , and e requiring that they are consistent with
Eq. (4.7), i.e., imposing that the supersymmetry variation of the Stransformed fields is
still Eq. (4.6), up to other local symmetry transformations of the typeIIB theory. We know
from the supersymmetry transformation of that and must acquire the same dilaton
dependence. Moreover, we expect that all the fermions undergo an overall SL(2, R) rotation determined by a 2 2 unitary matrix . Finally, the transformation of the gravitino
can contain a term proportional to . Hence, imposing that the transformed vielbein has
the correct supersymmetry variation, one gets
1
(4.8)
e/4 e/4 .
4
It turns out that the supersymmetry transformation of the vielbein is mapped to itself plus
an additional local Lorentz transformation of parameter
e/4 ,
1
ab = e/2 ab .
(4.9)
4
We neglect this term when we study the Stransformations of the supersymmetry variations
of the fermions, since they would lead to cubic Fermi terms. The Sduality transformation
of is straightforwardly obtained imposing that the transformed dilaton varies according
245
(4.10)
C (2) B (2)
B (2)
and
C (2)
form an SL(2, R)
(4.11)
if
1 3 = 1 ,
whose solution is8
= ei2 /4 =
1 1 = 3 ,
1/2 1/ 2
.
1/ 2 1/ 2
(4.12)
(4.13)
Implementing these transformations on the supersymmetry variation of C (4) , one then obtains
(2)
(2)
(4.14)
(4.15)
The correctness of the transformations (4.8) and (4.10) is finally proven by showing that
the supersymmetry variation of the transformed Fermi fields is consistent with Eq. (4.6).
Following the same arguments, one can now determine the Sduality transformations
of the two 10forms B (10) and C (10) . One expects these fields to transform as a doublet,
but the surprising result is that the first requirement one has to make to impose Sduality
is that the transformation of this doublet must have a nontrivial dilaton dependence. More
precisely, the only possibly consistent transformation is
B (10) e2 C (10) ,
C (10) e2 B (10) .
(4.16)
This still does not guarantee that Sduality is preserved, and in fact the additional variation
of in (4.16) is canceled only once one imposes the additional constraints9
C(10)
( ) = e ( 1 1 ...10 ),
1 ...10
B(10)
( ) = e2 ( 3 1 ...10 ).
1 ...10
(4.17)
After performing the typeI truncation of Eq. (2.31), in which B (10) is projected out, the
first constraint becomes
1 1 ...10
C1 ...10 = e det g.
(4.18)
10!
8 The inverse choice for corresponds to a sign change in the transformations of C (2) and B (2) .
9 As a consistency check, one can show that these constraints are related by an Sduality transformation.
246
A similar result holds for the second constraint, after performing the heterotic truncation,
in which all the RR fields are projected out and the spinors are projected according to
(1 3 )f = 0.
(4.19)
As we will see in the next section, a possible interpretation of this result is that Sduality
is actually broken in the presence of spacetimefilling branes. The constraint of Eq. (4.17)
can be justified only in the truncated theory, and this is in agreement with the fact that only
in the truncated theory the presence of spacetimefilling branes can be consistent.
We will see in the next section that if one tries to construct the Sdual of a supersymmetric D9brane using Eqs. (4.7), (4.8), (4.10), (4.11) and (4.16), the action one gets is
no longer symmetric, and the constraints of Eq. (4.17) have to be imposed to restore
symmetry. This means that the breakdown of symmetry is consistent with the breakdown of Sduality. The brane action obtained in this way has a DBI term proportional to
e4 and a WZ term proportional to e2 . As we are going to prove, it turns out that, in
the untruncated theory, a symmetric action for a spacetimefilling brane charged with
respect to B (10) has an e2 dilaton dependence in the DBI term, and no dilaton factor in
the WZ term.
(5.1)
The same projection applies to the spinor in the brane sector, since transforms under
Sduality like , while the worldvolume vector V i is projected out, again in agreement
with Eqs. (4.8) and (4.12). We will assume that the 10forms B (10) and C (10) transform
according to Eq. (4.16), keeping in mind that this is consistent only if the constraints (4.17)
are satisfied. We will see that the action we end up with is consistently symmetric only
if these constraints are satisfied.
After the projection, the Sdual of the action (2.7) becomes
10
4
S=
(5.2)
d e
det g +
e2 B (10) .
M10
M10
247
The supersymmetrization of this action is then worked out in the same way as the D9 and
D8 cases. One first constructs the hatted fields
1
= + + ,
2
g = g + 2 ( ) + ( D) + ,
B (10)
= B(10)
+ 10e2 [1 ...9 3 10 ] e2 1 ...10 3
1 ...10
1 ...10
+ 5e2 [1 ...9 3 D10 ] + ,
and then writes the supersymmetric action
S=
d 10 e4 det g +
e2 B (10) .
M10
(5.3)
(5.4)
M10
If this procedure preserved symmetry, it would be expected that one of the two truncations (the one with the upper sign choice in Eq. (5.1), as one would get using Eq. (4.12))
leaded to a brane action in which all the goldstino terms disappear. This is actually not the
case, since for the upper sign choice a term proportional to
B(10)
( ) + e2 ( 1 ...10 )
1 ...10
(5.5)
survives. This term vanishes if the second constraint of Eq. (4.17) is imposed. This is
not surprising, since only if this constraint is valid the Sduality transformations can be
performed. Thus, the picture that emerges is that the breakdown of Sduality is in agreement with the breakdown of symmetry, and the constraint of Eq. (4.17) provides a
restoration of both. On the other hand, the constraint (4.17) leads to a vanishing action
for the NS9brane, and this could simply mean that such an object does not exist. We
will comment about this in the conclusions. The lower sign choice in (5.1), again analogously to the Dbrane case, corresponds to a VAtype action for , after Eq. (4.17) is
imposed.
Let us consider now the action
S=
(5.6)
d 10 e2 det g +
B (10) .
M10
M10
In this case, using Eqs. (5.3), one obtains that the upper sign choice in Eq. (5.1) leads to
an action with no goldstino, while the lower sign choice leads to a VA action, and in both
cases no constraint is required. This means that the untruncated action is symmetric,
and thus we argue that this is the correct action for an NS9brane. More precisely, the
complete action would result from rescaling the Sdual of the Lagrangian of Eq. (2.7), and
in order to compute this, one should know how C (8) and C (6) transform under Sduality.
This analysis is currently under investigation, and we expect that it would shed some light
on the problem of studying the Sdual of a D7brane as well. Anyway, we do not expect
the results of this section to be altered by the inclusion of additional terms, since we do not
see how the constraint of Eq. (4.17) can be removed modifying B (10) by the inclusion of
other bulk fields.
248
One could also discuss the Sdual of this picture, starting from the action of Eq. (5.6),
and then performing an Sduality transformation. The result is that one ends up with the
action
S=
(5.7)
d 10 e3 det g +
e2 C (10) .
M10
M10
= 11 .
(5.8)
(5.9)
Vi
M10
is supersymmetric choosing the upper sign in Eq. (5.8), while supersymmetry is spontaneously broken if one chooses the lower sign. It can be shown that, after an S 1 reduction,
Tduality relates this action with the one of Eq. (5.6).
6. Conclusions
The starting point of this paper was a continuation of [24], where the results of [23]
were generalized to a curved background, showing that the possible typeI truncations of
typeIIB are in correspondence with the possible consistent typeI strings in D = 10. We
showed here that the same results apply to the D = 9 truncations of typeIIA, in accordance
with Tduality. We then proceeded constructing the symmetric spacetimefilling branes
that are charged with respect to the NS 10forms of typeIIB and typeIIA.
In [29] it was argued that Sduality of typeIIB implies the existence of NS9branes, that
together with the D9branes form an SL(2, Z) doublet. From the standard Weylrescaling
argument, it turns out that the tension of these branes appears to scale like 1/gS4 in the
string frame. Here we have argued that the actual tension of these branes scales like 1/gS2 ,
249
like the other solitonic NS objects, namely NS5branes. The solution of the paradox is
that the doublet of NS and RR 10form potentials does not transform covariantly under
SL(2, Z). It is therefore not possible to derive the action for an NS9brane performing a
Weyl rescaling. In [29,30] it was also conjectured that Sduality implies the existence of a
dual of the orientifold projection, and the SO(32) heterotic theory should result from this
projection, after the introduction of 32 NS9branes. This projection would naturally act
like 3 on the fermion doublets, since in the heterotic theory the fermions come only from
the left sector. We do not expect this truncation to be an ordinary Z2 orbifold, since it is
well know that a Z2 orbifold of typeIIB gives rise to typeIIA. In any case, if there is a
way of deriving the heterotic SO(32) theory from typeIIB, we expect that the twisted
sector of the projection would correspond to inserting symmetric branes, that would
therefore have the structure of Eq. (5.6). In any case, should a brane interpretation of the
heterotic theory be possible, a natural question would arise, namely what is the heterotic
string equivalent of brane supersymmetry breaking.10
Similar arguments hold for the IIA case. Since the typeIIA superstring and the E8 E8
heterotic theory have both an Mtheory origin [40,41], it has also been conjectured [29] that
the E8 E8 heterotic theory can arise in ten dimensions from a projection of typeIIA,
that would result in the lowenergy action in a Z2 truncation removing the RR fields. We
emphasize again that if this projection exists, it cannot act as a Z2 orbifold of typeIIA,
since such an orbifold gives rise to typeIIB. The twisted sector of the heterotic theory
would result in this case form the insertion of NS9branes. Starting from typeIIB and using
Weylrescaling arguments, it has been argued that Tduality would imply that the tension of
this branes, apart from having an e4 dilaton dependance, is proportional to R 3 , where R
is the radius of the isometry direction [31], and this would mean that they are not defined
in 10 uncompactified dimensions. The NS9brane, as well as the D8brane, would then
result from a 9brane in Mtheory whose effective action and target space solution [42]
can be written only if the 11dimensional supergravity has an isometry, and thus cannot be
covariant in 11 dimensions. Again, the symmetric spacetimefilling brane we obtained
in this paper has instead an e2 dilaton dependance, and it is related by Tduality to the
typeIIB NSbrane of Eq. (5.6). This different scaling with respect to the one of [31] implies
that this NS9brane and the D8brane cannot have a common Mtheory origin. Since the
fieldstrength of a 10form in 11 dimensions would be dual to a cosmological constant,
this result is basically rephrasing the fact that no cosmological constant can be included
in 11dimensional supergravity [43], and Romans IIA supergravity cannot be obtained
by dimensional reduction from 11 dimensions. After compactification on a 2torus Mtheory is related to typeIIB by Tduality, and thus this picture is the Tdual analogous
of the typeIIB picture, where Sduality is broken by the presence of spacetimefilling
branes.
Finally, it would be interesting to see if the Sduality rules of this paper can be used
to understand the strong coupling behavior of the D7branes. In [44] it was shown that
D7branes of typeIIB belong to a triplet of 7branes. One could then determine a supersymmetric effective action for these branes requiring symmetry, and relate them to the
10 I am grateful to E. Dudas for discussions about this point.
250
Acknowledgements
I am grateful to E. Bergshoeff, M. Bianchi, E. Dudas, M. Green and G. Pradisi for
discussions. This work is supported by a European Commission Marie Curie Postdoctoral
Fellowship, Contract MEIFCT2003500308.
References
[1] J. Polchinski, Dirichletbranes and RamondRamond charges, Phys. Rev. Lett. 75 (1995) 4724, hepth/9510017.
[2] J. Dai, R.G. Leigh, J. Polchinski, New connections between string theories, Mod. Phys. Lett. A 4 (1989)
2073.
[3] E.S. Fradkin, A.A. Tseytlin, Nonlinear electrodynamics from quantized strings, Phys. Lett. B 163 (1985)
123.
[4] A. Abouelsaood, C.G. Callan, C.R. Nappi, S.A. Yost, Open strings in background gauge fields, Nucl. Phys.
B 280 (1987) 599.
[5] M. Cederwall, A. von Gussich, B.E. Nilsson, A. Westerberg, The Dirichlet superthreebrane in tendimensional type IIB supergravity, Nucl. Phys. B 490 (1997) 163, hepth/9610148.
[6] M. Aganagic, C. Popescu, J.H. Schwarz, Dbrane actions with local kappa symmetry, Phys. Lett. B 393
(1997) 311, hepth/9610249.
[7] M. Cederwall, A. von Gussich, B.E. Nilsson, P. Sundell, A. Westerberg, The Dirichlet superpbranes in
tendimensional type IIA and IIB supergravity, Nucl. Phys. B 490 (1997) 179, hepth/9611159.
[8] E. Bergshoeff, P.K. Townsend, SuperDbranes, Nucl. Phys. B 49 (1997) 145, hepth/9611173.
[9] M. Aganagic, C. Popescu, J.H. Schwarz, Gaugeinvariant and gaugefixed Dbrane actions, Nucl. Phys.
B 495 (1997) 99, hepth/9612080.
[10] W. Siegel, Hidden local supersymmetry in the supersymmetric particle action, Phys. Lett. B 128 (1983) 397.
[11] M.B. Green, J.H. Schwarz, Covariant description of superstrings, Phys. Lett. B 136 (1984) 367.
[12] D.V. Volkov, V.P. Akulov, Is the neutrino a goldstone particle?, Phys. Lett. B 46 (1973) 109;
U. Lindstrom, M. Rocek, Constrained local superfields, Phys. Rev. D 19 (1979) 2300;
S. Samuel, J. Wess, A superfield formulation of the nonlinear realization of supersymmetry and its coupling
to supergravity, Nucl. Phys. B 221 (1983) 153;
S. Samuel, J. Wess, Realistic model building with the AkulovVolkov superfield and supergravity, Nucl.
Phys. B 226 (1983) 289;
S. Samuel, J. Wess, Secret supersymmetry, Nucl. Phys. B 233 (1984) 488;
J. Bagger, A. Galperin, Matter couplings in partially broken extended supersymmetry, Phys. Lett. B 336
(1994) 25, hepth/9406217.
[13] A. Sagnotti, in: G. Mack, et al. (Eds.), NonPerturbative Quantum Field Theory, in: Cargese 87, Pergamon,
Elmsford, 1988, p. 521;
A. Sagnotti, Open strings and their symmetry groups, hepth/0208020;
G. Pradisi, A. Sagnotti, Open string orbifolds, Phys. Lett. B 216 (1989) 59;
M. Bianchi, A. Sagnotti, On the systematics of open string theories, Phys. Lett. B 247 (1990) 517;
M. Bianchi, A. Sagnotti, Twist symmetry and open string Wilson lines, Nucl. Phys. B 361 (1991) 519;
M. Bianchi, G. Pradisi, A. Sagnotti, Toroidal compactification and symmetry breaking in open string theories, Nucl. Phys. B 376 (1992) 365.
[14] M.B. Green, J.H. Schwarz, Anomaly cancellation in supersymmetric D = 10 gauge theory and superstring
theory, Phys. Lett. B 149 (1984) 117;
M.B. Green, J.H. Schwarz, Infinity cancellations in SO(32) superstring theory, Phys. Lett. B 151 (1985) 21.
251
[15] N. Ohta, Cancellation of dilaton tadpoles and two loop finiteness in SO(32) type I superstring, Phys. Rev.
Lett. 59 (1987) 176.
[16] J. Polchinski, Y. Cai, Consistency of open superstring theories, Nucl. Phys. B 296 (1988) 91.
system and the USp(32) string theory, Prog. Theor.
[17] S. Sugimoto, Anomaly cancellations in type I D9D9
Phys. 102 (1999) 685, hepth/9905159.
[18] M. Bianchi, J.F. Morales, Anomalies and tadpoles, JHEP 0003 (2000) 030, hepth/0002149] .
[19] I. Antoniadis, E. Dudas, A. Sagnotti, Brane supersymmetry breaking, Phys. Lett. B 464 (1999) 38, hepth/9908023.
[20] E. Dudas, J. Mourad, Consistent gravitino couplings in nonsupersymmetric strings, Phys. Lett. B 514 (2001)
173, hepth/0012071.
[21] W. Fischler, L. Susskind, Dilaton tadpoles, string condensates and scale invariance, Phys. Lett. B 171 (1986)
383;
W. Fischler, L. Susskind, Dilaton tadpoles, string condensates and scale invariance. 2, Phys. Lett. B 173
(1986) 262.
[22] E. Dudas, G. Pradisi, M. Nicolosi, A. Sagnotti, On tadpoles and vacuum redefinitions in string theory, hepth/0410101.
[23] E. Bergshoeff, M. de Roo, B. Janssen, T. Ortin, The superD9brane and its truncations, Nucl. Phys. B 550
(1999) 289, hepth/9901055.
[24] F. Riccioni, Truncations of the D9brane action and typeI strings, Phys. Lett. B 560 (2003) 223, hepth/0301021.
[25] G. Pradisi, F. Riccioni, Geometric couplings and brane supersymmetry breaking, Nucl. Phys. B 615 (2001)
33, hepth/0107090.
[26] E. Bergshoeff, R. Kallosh, T. Ortin, D. Roest, A. Van Proeyen, New formulations of D = 10 supersymmetry
and D8O8 domain walls, Class. Quantum Grav. 18 (2001) 3359, hepth/0103233.
[27] C.M. Hull, P.K. Townsend, Unity of superstring dualities, Nucl. Phys. B 438 (1995) 109, hepth/9410167.
[28] J. Polchinski, E. Witten, Evidence for heterotictype I string duality, Nucl. Phys. B 460 (1996) 525, hepth/9510169.
[29] C.M. Hull, Gravitational duality, branes and charges, Nucl. Phys. B 509 (1998) 216, hepth/9705162.
[30] C.M. Hull, The nonperturbative SO(32) heterotic string, Phys. Lett. B 462 (1999) 271, hepth/9812210.
[31] E. Bergshoeff, E. Eyras, R. Halbersma, J.P. van der Schaar, C.M. Hull, Y. Lozano, Spacetimefilling branes
and strings with sixteen supercharges, Nucl. Phys. B 564 (2000) 29, hepth/9812224.
[32] I.A. Bandos, J.A. de Azcarraga, J.M. Izquierdo, J. Lukierski, An action for supergravity interacting with
superpbrane sources, Phys. Rev. D 65 (2002) 021901, hepth/0104209;
I.A. Bandos, J.A. de Azcarraga, J.M. Izquierdo, J. Lukierski, D = 4 supergravity dynamically coupled to a
massless superparticle in a superfield Lagrangian approach, hepth/0207139;
I.A. Bandos, J.A. de Azcarraga, J.M. Izquierdo, J. Lukierski, On dynamical supergravity interacting with
superpbrane sources, hepth/0211065.
[33] E.A. Bergshoeff, M. de Roo, A. Sevrin, NonAbelian BornInfeld and kappasymmetry, J. Math. Phys. 42
(2001) 2872, hepth/0011018.
[34] J. Gomis, K. Kamimura, P.K. Townsend, Nonrelativistic superbranes, hepth/0409219.
[35] L.J. Romans, Massive N = 2a supergravity in ten dimensions, Phys. Lett. B 169 (1986) 374.
[36] E. Bergshoeff, M. De Roo, Dbranes and Tduality, Phys. Lett. B 380 (1996) 265, hepth/9603123.
[37] M.B. Green, C.M. Hull, P.K. Townsend, Dbrane WessZumino actions, Tduality and the cosmological
constant, Phys. Lett. B 382 (1996) 65, hepth/9604119.
[38] E. Bergshoeff, C.M. Hull, T. Ortin, Duality in the type II superstring effective action, Nucl. Phys. B 451
(1995) 547, hepth/9504081.
[39] J.H. Schwarz, Covariant field equations of chiral N = 2 D = 10 supergravity, Nucl. Phys. B 226 (1983) 269.
[40] E. Witten, String theory dynamics in various dimensions, Nucl. Phys. B 443 (1995) 85, hepth/9503124.
[41] P. Horava, E. Witten, Heterotic and type I string dynamics from eleven dimensions, Nucl. Phys. B 460 (1996)
506, hepth/9510209.
[42] E. Bergshoeff, J.P. van der Schaar, On M9branes, Class. Quantum Grav. 16 (1999) 23, hepth/9806069.
[43] K. Bautier, S. Deser, M. Henneaux, D. Seminara, No cosmological D = 11 supergravity, Phys. Lett. B 406
(1997) 49, hepth/9704131.
252
[44] P. Meessen, T. Ortin, An Sl(2, Z) multiplet of ninedimensional type II supergravity theories, Nucl. Phys.
B 541 (1999) 195, hepth/9806120.
[45] G.W. Gibbons, M.B. Green, M.J. Perry, Instantons and sevenbranes in type IIB superstring theory, Phys.
Lett. B 370 (1996) 37, hepth/9511080.
[46] M.B. Einhorn, L.A. Pando Zayas, On sevenbrane and instanton solutions of type IIB, Nucl. Phys. B 582
(2000) 216, hepth/0003072.
[47] E. Bergshoeff, U. Gran, D. Roest, Type IIB sevenbrane solutions from ninedimensional domain walls,
Class. Quantum Grav. 19 (2002) 4207, hepth/0203202.
Abstract
We derive the nonperturbative effect in type 0B string theory, which is defined by taking the
double scaling limit of a onematrix model with a twocut eigenvalue distribution. However, the
string equation thus derived cannot determine the nonperturbative effect completely, at least without
specifying unknown boundary conditions. The nonperturbative contribution to the free energy comes
from instantons in such models. We determine by direct computation in the matrix model an overall
factor of the instanton contribution, which cannot be determined by the string equation itself. We
prove that it is universal in the sense that it is independent of the detailed structure of potentials in
the matrix model. It turns out to be a purely imaginary number and therefore can be interpreted as a
quantity related to instability of the Dbrane in type 0 string theory. We also comment on a relation
between our result and boundary conditions for the string equation.
2005 Elsevier B.V. All rights reserved.
1. Introduction
The nonperturbative effect in string theory can be studied using matrix models. In particular, the noncritical string theory, which is a simplified model of string theory, is exactly
solvable via matrix models [1]. The string equation, which can be derived from them,
contains the nonperturbative effect of the noncritical string theory [2]. On the other hand,
Email address: ymatsuo@gauge.scphys.kyotou.ac.jp (Y. Matsuo).
05503213/$ see front matter 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.nuclphysb.2005.01.002
254
study of the Liouville theory [3] enables us to find the effect of the Dbrane, which is the
nonperturbative effect of string theory and can be identified with the effect that appears in
the string equation. In [4], we have shown that the string equation does not describe the
nonperturbative effect completely, at least in c = 0 noncritical string theory. To obtain the
whole nonperturbative effect, it is necessary to study the matrix model directly.
Recently, a matrix model that corresponds to the noncritical string with worldsheet supersymmetry has been proposed [5,6]. For the c = 0 noncritical string, which is described
as twodimensional pure supergravity on the worldsheet [7], we consider the double scaling limit around the GrossWitten transition [8]. This critical point can be found in the
onematrix model with a twocut eigenvalue distribution. In string theoretical interpretation, the onematrix model with two cuts corresponds to the NSR string theory of type 0B.
This matrix model can be solved with the string equation. However, as in the case of c = 0
string theory, it does not contain the nonperturbative effect completely.
In this paper we study the nonperturbative effect of type 0B string theory by analyzing
the matrix model with a twocut eigenvalue distribution. We compute the effect of instantons directly in the matrix model, which corresponds to the Dbrane in the string theory.
The result is summarized as follows.
From the string equation, the nonperturbative effect in the free energy is obtained as
2 3/2
C
,
= 3/4 exp t
(1.1)
3
t
where C cannot be determined from the string equation by itself without specifying unknown boundary conditions. From the direct computation using the matrix model, we can
determine the constant C as
i
C= .
(1.2)
4
Moreover, it is shown that this value is universal, namely, it does not depend on the detailed structure in the potential of the matrix model. Because it is purely imaginary, this
nonperturbative effect is related to the instability of the Dbrane.
The paper is organized as follows. In Section 2, we identify the instantons in the matrix
model and the contribution from instantons to the free energy. In Section 3, we compute
the effect of the instantons using the method of orthogonal polynomials. In Section 4, we
take the double scaling limit and consider the universal behavior of the effect of instantons.
In Section 5, we present the conclusions. Appendices AC show the details of calculations.
255
extended to the case of twocut distribution. In this case, the effective potential has two
separated minima. Half of the eigenvalues are in one of these minima and the other half
are in the other minimum except that a single eigenvalue is at the maximum.
In the onematrix model, the partition function is given by
Z = d eN tr V () .
(2.1)
Here, is an N N Hermitian matrix. Hereafter we consider the case where the potential
V (x) is invariant under x x and thus the eigenvalue distribution of has this Z2
symmetry.
Diagonalizing the matrix , the partition function can be expressed as
Z=
(2.2)
di 2 () eN i V (i ) .
i
i=1
i=1
dx det(x )2 N 1 eN V (x)
(2.3b)
dx eN Veff (x) .
(2.3c)
Here, the subscript N 1 indicates that the quantities concerned are those in the (N
1) (N 1) matrix model, and the expectation value O is defined as
1
O =
(2.4)
d O eN tr V () .
Z
In the largeN limit, the system of an (N 1) (N 1) matrix is the same as the system
of an N N matrix. Hence, we can use the standard N N matrix model to calculate
these expectation values.
The effective potential Veff (x) defined above can be expressed in terms of connected
diagrams. After some algebra, we obtain
2
1
N Veff (x) = N V (x) 2 tr log(x ) c
(2.5)
2 tr log(x ) c .
2
Here, the subscript c indicates the connected part. In the largeN limit, the first and
second terms are of order N and third term is of order N 0 . If we restrict ourselves to the
leading order of N , the terms other than the first two can be neglected. Using the resolvent1
1
1
1
= V (x) V 2 (x) + p(x) ,
R(x) =
(2.6)
tr
N x
2
1 Here the branch of the square root is chosen so that R(x) 1/x as x .
256
(1)
1 (2)
V + ,
N eff
1
2 tr log(x )
N
x
x
= V (x) 2 Re dx R(x ) = Re dx V 2 (x ) + p(x).
Veff = V (x)
(2.7a)
(2.7b)
(2.7c)
The resolvent R(x) has the cut on the real axis. If x is on the cut, the effective potential
becomes constant and the eigenvalue density (x) takes a nonzero value. Interpreting this
in physical terms, we are considering the N th eigenvalue x and the other N 1 eigenvalues are those of whose distribution is expressed by (x). In the cut, where the N 1
eigenvalues are distributed, the forces from them acting on the N th eigenvalue cancel each
other. The effective potential Veff (x) has is at the minimum over the entire cut. From the
standpoint of the original system of the N N matrix including the N th eigenvalue, it
is natural that the eigenvalue density (x) should not change due to the N th eigenvalue
at the leading order of N ; that is, the back reaction is at the subleading order. Hence, in
the integration with respect to x in (2.3), most of the contribution comes from the region
where x is inside the cut. In the case of one cut, the integration over this region gives the
partition function of the N N matrix system. To extend this to type 0B string theory, we
should consider the case of two cuts. In the case of two cuts, there are two minima of the
effective potential. Because the potential under consideration is Z2 symmetric, these two
minima should be symmetric under x x and give the same contribution to the partition
function. Hence, we can deal with these two regions together as inside the cut. There is
another nonzero contribution from the region where x lies outside the cut. Because there
is a maximum of the effective potential, we should take this into account. If the N th eigenvalue x lies outside the cut, the eigenvalue density of the N N matrix system differs
from that of the (N 1) (N 1) matrix system. It can be identified as the instanton
characterized by the configuration where x is located at the local maximum of the effective potential. The resolvent is related to the disk amplitude of the Liouville field theory
with the boundary condition corresponding to the FZZTbrane [10]. The effect of the instanton comes from the (integrated) resolvent with some fixed value of the cosmological
constant corresponding to the local maximum of the effective potential. It is related to the
ZZbrane [9].
Now, we divide the integration region into two parts, namely, inside the cut and outside
the cut, as
dx det(x )2 eN V (x)
Z = ZN 1
inside the cut
+ ZN 1
dx det(x )2 eN V (x) .
(2.8)
The second term comes from the integration outside the cut, and is identified as the contribution of the instanton. So far, we have restricted ourselves to the N th eigenvalue x and
257
identified the instanton as the configuration where x lies outside the cut. However, there
are N eigenvalues and all other N 1 eigenvalues can be possibly outside the cut as well.
Hence, there is an ninstanton sector; that is, n eigenvalues lie outside the cut. The partition
function can be expressed the sum of those in the ninstanton sector as
Z = Z (0inst) + Z (1inst) + Z (1inst) + ,
Z (0inst) =
di 2 () eN i V (i )
i
(0inst)
= ZN 1
(0inst) N V (x)
dx det(x )2
e
,
(2.9a)
(2.9b)
(1inst)
(0inst)
= N ZN 1
(0inst) N V (x)
dx det(x )2
e
.
(2.9c)
Here, the superscript (ninst) indicates the ninstanton sector. In the partition function
Z (0inst) and the expectation value O(0inst) in the 0instanton sector, all eigenvalues are
inside the cut. Hereafter we will omit the superscript (0inst) in the expectation value.
The factor N in front of the 1instanton sector partition function (2.9c) reflects the number
of ways of specifying an eigenvalue that lies outside the cut. If we consider all ninstanton
sectors, neglecting interaction between instantons, which is valid in the largeN limit, the
partition function can be expressed in terms of those of the 0instanton and the 1instanton
sector
Z (1inst)
(0inst) +
F
(0inst)
1 + (0inst) + = eF
,
e =Z=Z
(2.10a)
Z
dx det(x )2 eN V (x)
Z (1inst)
= (0inst) = N outside the cut
.
2 N V (x)
Z
inside the cut dx det(x ) e
(2.10b)
The additional term is the contribution from instantons. We regard this term as the
chemical potential of the instanton. This effect corresponds to the ZZbrane in the noncritical string theory.
258
Our expectation here is that these divergences cancel the overall N in (2.10b) to make
the chemical potential finite. In order to confirm this, we have to retain the N dependence
in the computation. For this purpose, it is appropriate to use the method of orthogonal
polynomials, because it is available for any N .
We begin with definitions and properties of the orthogonal polynomials. The partition
function of the onematrix model can be expressed in terms of orthogonal polynomial
Pn (x) as
Z=
(3.1)
di det Pn (n ) det Pm (m ) eN i V (i ) .
i
nn
mm
Here, the orthogonal polynomial Pn (x) = x n + O(x n1 ) satisfies the orthogonality condition
N
1
fnN n .
(3.3)
n=1
(3.4a)
Pn (x) = Pnm Pm (x) = N V (X) nm Pm (x).
(3.4b)
fn = fn = f ( ) (n: even),
(3.5)
fn = f( ) (n: odd).
In this limit, a summation over the index n can be approximated by an integration. If we
want to calculate up to the nexttoleading order in 1/N , we should use the EulerMclaurin
259
N
fn =
2
1+ N1
N
d f( ) +
2
0 N1
1
d f( ).
(3.6)
(3.9)
(3.10)
This formula enables us to evaluate the chemical potential. Substituting this formula into
(2.10b), the chemical potential can be expressed in terms of the orthogonal polynomials,
which can be determined by (3.4), and we will obtain the definite value of the chemical
potential.
When we evaluate det(x )2 , the result will be different depending on whether
x is inside the cut or outside the cut. This difference can be described as follows. The
orthogonal polynomial Pn can be expressed as the expectation value in the n n matrix
system as in (3.7), and thus if x is inside the cut of this system, Pn (x) is oscillating rapidly,
otherwise Pn (x) is monotonic as x n . As n increases, the oscillatory region becomes wider.
When x is inside the cut, we define n (x) as the minimum value of n such that x is
inside this
oscillatory region of Pn (x). The normalized orthogonal polynomials n (x) =
Pn (x)/ hn take values of the same order for n n (x), while those for n < n (x) n (x)
become small so that the contributions from the last n (x) terms containing Pn (x) (n <
n (x)) in (3.10) are damped exponentially. Hence, we can neglect the term of Pn with
n < n (x) in (3.10). On the other hand, if x is outside the cut, Pn (x) for all n are not
oscillating but monotonic with respect to x as x n and we cannot neglect the latter terms
with Pn (x) for n < n (x). Thus, in the computation of (3.10), we should consider these
two cases, namely, inside the cut and outside the cut, separately.
260
First, we consider the case in which x is outside the cut. In this case, the largest contribution to det(x )2 comes from the first term in (3.10) and contributions from latter terms
become smaller like a geometric series. Using the ratio of the orthogonal polynomials
ekn
Pn (x)
,
Pn1 (x)
(3.11)
n
1
1+O
.
PN2 (x) rN e2kN
DN =
N
(3.12)
(3.13)
n=0
In the case of two cuts, however, we should distinguish a quantity with even n and odd n.
Hence, we approximate this up to O(1/N ) by a sum of two geometric series as
DN
n
PN2 1 + rN e2kN rN rN e2(kN +kN )
(3.14a)
n=0
(3.14b)
n
PN2 1 + rN e2kN rN rN e2(kN +kN )
(3.15a)
n=0
(3.15b)
1 (1)
k + ,
N n
(3.16a)
(0)
1 2
x + rn rn
2x
(0)
1 2
x rn + rn
2x
ekn =
ekn =
2
x 2 rn rn 4rn rn ,
(3.16b)
2
x 2 rn rn 4rn rn ,
(3.16c)
2
1
kn(1) + kn(1) = log x 2 rn rn 4rn rn ,
2
(3.16d)
(0)
where the double sign is understood to be + for x 2 > rn + rn + 2 rn rn so that ekn x as
(0)
x , and for x 2 < rn + rn 2 rn rn so that ekn does not diverge at x = 0. Using
d k (0) ( ) + k (0) ( ) = 2
x
dx R(x).
261
(3.17a)
(3.17b)
Here, the sign distinguishes the cases of even N and odd N . The relation between
W (x) and resolvent R(x) can be checked with a concrete potential V (x), or in the double
scaling limit. We will examine this point later.
Next, we consider the case where x is inside the cut. In this case, the contribution from
Pn (x) for n < n (x) is exponentially suppressed. Hence, DN = det(x )2 can be
approximated as
DN =
N
Pn2 (x)
N n1
rN m .
(3.18)
m=0
n=n (x)
In the largeN limit, the summation over n can be replaced by an integration. By definition,
x is in the oscillatory region of Pn (x) with n n (x). Because its frequency is of the order
N as shown in Appendix A, in order to calculate Pn (x)2 in the largeN limit it is sufficient
to take the average of this oscillation, which amounts to dividing the amplitude of Pn (x)2
by 2. In the case of two cuts, generally, we should distinguish a quantity for even
n and a
quantity for odd n. However, after we average this oscillation, n (x) = Pn (x)/ hn can be
approximated by a continuous function of = n/N whether n is even or odd. Finally we
should take the extra factor 2 into account. This factor comes from a relative normalization
between the Pn (x) inside the cut and Pn (x) outside the cut as in the continuity formula in
the usual WKB approximation [4]. Using the asymptotic formula for n (x) inside the cut
given in Appendix A, (3.18) takes a value up to O(1/N ) in the exponent
DN =
N
hN n2 (x)
n=n
= 2e
N Re W (x)
N n1
rN m
(3.19a)
m=0
1
d
rN N
x
4r ( )r ( ) (x 2 r ( ) r ( ))2
= 2N(x) rN eN (V (x)+W0 ) ,
(3.19b)
(3.19c)
where = n /N and W (x) is again related to the resolvent as in (3.17b). Because the
real part of the resolvent becomes V (x) inside the cut, Re W (x) = V (x) + W0 , where
(0)
W0 is a constant that determines the origin of Veff (x). Note that because Re W (x) does
not depend on = n/N for n n , as shown in Appendix A, we can put it outside the
integration with respect to . We have also used the fact that the integration in the second
line can be identified with the eigenvalue density (x), which is shown in Appendix B.
262
Now, we are ready to take the ratio of the partition functions inside and outside the cut,
and obtain the chemical potential of the instanton. This can be expressed as
dxdet(x )2 eN V (x)
.
= N outside the cut
(2.10b)
2 N V (x)
inside the cut dxdet(x ) e
Most of the contribution from outside the cut comes from the saddle point of the effective
potential in the largeN limit. In our case, there is a maximum of the effective potential at
x = 0. This maximum always exists when we consider the case where there are two cuts
that are symmetric under x x. We consider only the instanton that is located at this
maximum. Using the method of steepest descent, we obtain the chemical potential with an
overall factor of O(N 0 ) as
rN + rN + rN rN 
=i
1
N (R (0) 2 V (0)) 4 rN rN rN 
0
dx 2R(x) V (x) ,
exp N
(3.20)
a
where a corresponds to W0 and can be chosen as any point in the left one of the two cuts. As
(0)
mentioned above, it indeed determines the origin of Veff (x). If there is another maximum
or minimum of the effective potential, we should take it into account in the contribution
from outside the cut. Under the double scaling limit, however, there can be no contribution
from such a saddle point. Hence, it is sufficient to consider the instanton at the maximum
of x = 0. We will elucidate this point further in the next section.
(4.1)
[gV (X)]n,n1 .
263
variable for odd n can be obtained by interchanging r and r , which is nothing but F (r , r ).
At the critical point, the two cuts merge and r and r take the same value rc . This means
that r = r = rc for n = N at the critical point. Hence, we obtain gc = F (rc , rc ). Expanding
(4.1) around rc , it can be expressed as
c r ) A(r
c r )
F (r , r ) = gc A(r
c r )2 B(r
c r )2 C(rc r )(rc r ).
B(r
(4.2)
Case 3: A = A and B = B.
The condition that is relevant to the twocut case is case 2. In this case, if g gc , a solution
of (4.1) with real rs exists that can be rewritten as
r = rc gc g ,
(4.3a)
r = rc + gc g .
(4.3b)
For g > gc , we cannot take this condition. In this case, the condition r = r will be
satisfied, which is the case of one cut. Using the definition in (4.3), a solution of (4.1) in
this case can be expressed as (see Appendix C)
r = rc
2
(gc g ).
4rc
(4.4)
First, we evaluate the free energy without the contribution from the instanton. In the
GrossWitten phase transition comprising the third order, the third derivative of the free
energy in general has a discontinuity, which is universal in the sense that it is not affected by
details of the potential. Thus, in order to extract the universal part of the free energy, we will
compare the free energy of each phaseone cut and two cutsand pick up a discontinuity of
(the third derivative of) the free energy, and which is the only contribution to the universal
part. For the free energy in the twocut phase, from (3.3) we obtain
N2
F=
2
1
d (1 ) log r ( ) + log r ( )
(4.5a)
2 2
d (1 )
(gc g )
rc
(4.5b)
N2
1
0
N2
12
grc
2
(gc g)3 ,
(4.5c)
264
F = N2
(4.6a)
2 2
d (1 )
(gc g )
4rc
1
2
0
N2
0
2 2
d (1 )
(gc g )
rc
N2
24
grc
(4.6b)
2
(gc g)3 .
(4.6c)
Here, we have omitted the terms that do not contribute to the discontinuity of the free energy. In computation of the free energy in the onecut phase, we have noticed the following
fact: in the onecut phase, in general there is only one cut constructed from N eigenvalues.
However, if there were n < n0 = Ngc /g eigenvalues, the cut would split into two. Hence,
for < 0 n0 /N , we should take the twocut solution (4.3a), even in the onecut phase.
Comparing (4.5c) and (4.6c), we can see that the third derivative of the free energy indeed
has a discontinuity.
Second, we consider the contribution from the instanton in a background with a fixed
number of instantons. In the oneinstanton background, an instanton is located at the top
of the effective potential. The contribution to the free energy from this instanton can be
obtained by the height of the potential barrier. The effective potential can be obtained from
the resolvent that is expressed near the critical point as
1
R(x) = x
2
rc
C = 2 ,
g
1
d k (0) ( ) + k (0) ( )
C x a 2 x 2 ,
(4.7a)
a2 =
2
(gc g),
rc
(4.7b)
where we have dropped the contribution from the potential V (x), which is nonuniversal.
Then, the height of the potential barrier is
0
2N
a
2
dx R(x) = N
(gc g)3/2 .
3 grc
(4.8)
This is the contribution to the free energy from the instanton in the oneinstanton background. In the ninstanton background, the contribution of the instantons at the leading
order is multiplied by n.
265
Third, we evaluate the chemical potential of the instanton. From (2.10b), (4.3), and
(4.8), we obtain
grc
2
i
3/2
.
N
exp
(g
g)
=
(4.9)
c
3 grc
4 N (gc g)3/4
Now, we are ready to consider the double scaling limit. Two limits, N and g
gc , are related as
2/3
3
N =a ,
(4.10)
(gc g) = a 2 t, a 0.
grc
There is an ambiguity in the definition of t. Here, we define t so that the discontinuity of
the free energy is described as F (t) = t/8. Using this definition, the free energy in our
calculation and that in the string equation coincide. The string equation in this definition
takes the form of the Painlev II equation,
th = h3 2t2 h.
(4.11)
h(t) = t +
(4.12)
(4.13)
in perturbative expansion. The solution in the onecut phase is h(t) = 0. Thus, the discontinuity of the free energy is consistent with ours.
In this double scaling limit, the chemical potential of the instanton becomes
2 3/2
i
.
= 3/4 exp t
(4.14)
3
4 t
As shown above, the value does not depend on the details of the potential and hence it is
a universal quantity. Moreover, it agrees with the nonperturbative effect derived from the
string equation. This can be seen as follows. If there are two solutions of the string equation
with the same asymptotic expansion for large t
1
h(t) = t t 5/2 + ,
(4.15)
4
then the difference between them due to the nonperturbative effect is given by
hinst (t) = Const t 1/4 e2/3t
3/2
(4.16)
From this, the nonperturbative effect of the free energy can be obtained as
Finst = Const t 3/4 e2/3t
3/2
(4.17)
which is in accord with our chemical potential of the instanton. This justifies our identification of the instanton in the matrix model as the nonperturbative effect of string theory.
However, the overall normalization factor cannot be determined from the string equation
itself. We have determined it from direct computations in the matrix model and have shown
that it is universal.
266
5. Conclusions
In this paper, we have investigated in full detail the nonperturbative effect in type 0B
string theory, which is defined by taking the double scaling limit of the onematrix model
with a twocut eigenvalue distribution. We have computed the contribution from the instanton to the free energy directly in the matrix model. In the double scaling limit, the
chemical potential of the instanton does not depend on the details of the potential, and is
a universal quantity. It takes exactly the same form as the nonperturbative effect derived
from the string equation. Moreover, by computation via the matrix model keeping N finite,
we have fixed the overall factor of the chemical potential, which cannot be determined by
the string equation itself. In [4], it is shown that in (bosonic) c = 0 noncritical string theory,
only the closed string, or the string equation, does not describe the nonperturbative effect
completely, and that the matrix model is more fundamental, capturing the nonperturbative
effect correctly. Here, we have found that this is also the case with type 0B, or c = 0,
noncritical string theory.
It is worth noting a crucial difference between c = 0 and c = 0 string theory. In the
case of c = 0 string theory, the potential of the matrix model is unbounded from below.
Therefore, the vacuum with one cut is unstable and the imaginary part of the free energy
obtained in [4] reflects this instability. On the other hand, in the case of c = 0 string theory
defined by a matrix model with a twocut eigenvalue distribution, the potential is a doublewell type and bounded from below. Thus, the matrix integration in the definition of the
partition function or the free energy as in (2.1) is well defined and gives a real number. In
fact, in the double scaling limit, the integration outside the cut in (2.10b) becomes
0
rc
t 2/3(t 2 )3/2
d
e
.
2
t 2
(5.1)
Therefore, the contribution from the instanton to the free energy is given by an integration
of a real function, but it diverges. This divergence can be attributed
to the largeN limit,
in the double
where some eigenvalues are at the edge of the cut = t. Moreover,
scaling limit, some of the eigenvalues are pushed out of the cut t, so that
the edge
of the eigenvalue distribution is smeared. In this sense, the boundary at = t between
the outside and inside of the cut becomes subtle, and it becomes somewhat artificial in the
double scaling limit to divide the integration region into these two regions. Because this
divergence originates from only a part of the eigenvalues that spread into the outside of
the cut, it is less divergent compared to the integration over the inside of the cut given as
the denominator in (2.10b), which is of order N as shown in (3.19). Therefore, it may be
possible to change the definition of inside the cut slightly and then renormalize the above
divergence into the contribution in the denominator in (3.19). In that case, it is important
to confirm that the chemical potential is still universal irrespective of a slight change in the
definition of the inside of the cut. However, at least as long as the string coupling constant
gs is sufficiently small,2 we can take advantage of the saddle point method to evaluate the
2 g can be restored on dimensional grounds as t 3/2 t 3/2 /g .
s
s
267
above integration. Thus, the above integration is dominated by the saddle point 0 and
if we can choose the contour of as the whole imaginary axis, the integration becomes
finite and gives an imaginary number as in (4.14). The chemical potential obtained in this
way is meaningful at least for small gs . In fact, the instanton in the matrix model we have
considered corresponds to the Dbrane known as a ZZbrane [9]. This can be checked
by noting that it gives an open boundary to the worldsheet, or more quantitatively, by
comparing the disk amplitude in the fixed instanton background to that in the ZZbrane
background computed in the super Liouville theory as done in [4]. In type 0B string theory,
the ZZbrane is unstable. The chemical potential that we have computed as above is a
purely imaginary number. It can be considered to reflect the fact that the Dbrane under
consideration is unstable. Although c = 0 string theory does not have the time direction
and the instanton does not decay, if we consider the additional time direction associated
with the energy of the system, the instanton will decay due to its instability. The chemical
potential of the imaginary number can be considered to show this instability. In this sense,
the chemical potential is analogous with the energy of a statistical system, where there
is no time direction, but the energy gives the probability that the system is realized in the
ensemble. Likewise, because the instanton corresponds to the ZZbrane in c = 0 noncritical
string theory, the chemical potential is related to the decay rate of the ZZbrane. Note that
it is not necessary for our result to agree with the computations in [5] of the decay rate of
the ZZbrane (Dparticle) in c = 1 noncritical string theory, because the time in the target
space is a priori not the same as the extra time direction associated with the free energy
mentioned above.
The most important problem remaining unsolved is to identify the boundary conditions
of the string equation. Note that the matrix model from which the string equation is derived has no ambiguity. Therefore, it is natural to expect that the matrix model specifies
some boundary conditions, by which we can determine the nonperturbative effect precisely. Because the string equation is a differential equation of the second order, there are
two boundary conditions to be specified. It can be readily seen that the asymptotic behavior
(4.15) needs finetuning of one parameter, and therefore there is one boundary condition
remaining. It is not clear in general whether the contribution from the instanton to the
free energy, or the nonperturbative effect in the free energy, has something to do with this
boundary condition. In particular, if gs is sufficiently small, the nonperturbative effect is
exponentially small and it is impossible in general to determine it as an ambiguity to be
added to a perturbative series of the free energy, which is an asymptotic expansion. However, if we concentrate on the process where each term of a perturbative series is completely
zero, we can identify the nonperturbative effect unambiguously like the tunneling effect.
The chemical potential we have computed can be regarded as such an example. That is, it
is purely imaginary and the free energy cannot have an imaginary part perturbatively. We
therefore conclude that by defining the contour of the integration with respect to the eigenvalue as the imaginary axis, we can compute the chemical potential of the instanton, and
this will give a boundary condition for the imaginary part of a complexified string equation.
However, we have not yet fixed a boundary condition for the real part of the string equation. Note here that in order to identify the nonperturbative ambiguity, we should specify a
choice of the contour in general as mentioned in [12].
268
Acknowledgements
The authors would like to thank M. Hanada, N. Ishibashi, and T. Matsuo for their fruitful discussions. This work is supported in part by the GrantinAid for Scientific Research
(14540254) and the GrantinAid for the 21st Century COE Center for Diversity and
Universality in Physics from the Ministry of Education, Culture, Sports, Science and
Technology (MEXT) of Japan. The work of T.K. is supported in part by the Special Postdoctoral Researchers Program.
(A.1)
Here we have used the fact that sn is identically zero due to the symmetry with respect to
x x. In the case of two cuts, we should distinguish quantities for even n and odd n as
follows:
x = ekn+1 + rn ekn
x=e
kn+1
+ rn e
kn
(A.3a)
(A.3b)
(0)
x = ekn + rn ekn
(A.5a)
(0)
x = ekn + rn ekn
269
(A.5b)
(0)
(0)
0 = kn(1) + kn(0) ekn rn kn(1) ekn
(0)
(0)
0 = kn(1) + kn(0) ekn rn kn(1) ekn
(A.6a)
(A.6b)
2
(0)
1 2
kn
e =
(A.7a)
x + rn rn x 2 rn rn 4rn rn ,
2x
2
1 2
(0)
ekn =
(A.7b)
x rn + rn x 2 rn rn 4rn rn ,
2x
(0)
where the sign of
is + for x 2 > rn + rn + 2 rn rn so that ekn x as x , and
for x 2 < rn + rn 2 rn rn not to diverge at x = 0. Moreover, from (A.6),
2
1
kn(1) + kn(1) = log x 2 rn rn 4rn rn .
2
By using these ks, the orthogonal polynomials can thus be expressed as
n
Pn (x) = exp
kn .
(A.8)
(A.9)
m=1
N
Pn (x) = exp
2
1+ N1
)+ N
d k(
2
N
Pn (x) = exp
2
1
)
d k(
)+ N
d k(
2
1+ N1
)
d k(
(A.10b)
0+ N1
(A.10a)
0+ N1
1
d k (0) ( ) + k (0) ( ) .
(A.11a)
(A.11b)
Here,
is
for even n and kn for odd n.
If x is in the oscillatory region of Pn (x), where (x 2 rn rn )2 4rn rn < 0, kn is no
longer real, but Pn (x) should be real. We note here that Pn (x) is a solution of a set of
linear differential equations (3.4) and we have obtained its solution in a classically allowed
(0)
kn
(0)
kn
(0)
270
region, (A.11b) via the WKB approximation. Therefore, in order to find Pn in the oscillatory region which is an analog of a forbidden region, we invoke the continuity condition
[4]. That is, it is sufficient to take a linear combination of two complex solutions of (3.4a)
to obtain the real solution, which is nothing but the real part of the naive expression in
(A.11b) with an appropriate phase factor. We thus obtain Pn (x) in the oscillatory region as
1/4
x 2 rn
Pn (x) = 2
4rn rn (x 2 rn rn )2
n
n
N
1 (0)
N
sin Im
+ kn + . (A.12)
exp
Re W0 x,
W0 x,
2
N
2
N
2
Here, we have used
1
1
Re kn(0) = log rn ,
(A.13)
Re kn(0) = log rn ,
2
2
in the oscillatory region. The additional factor 2 comes from a relative normalization to the
solution for the allowed region (x 2 rn rn )2 4rn rn > 0 in the continuity condition, and
is a constant phase factor. The quartic root is defined as real and positive, and the overall
sign is included in .
n
n
1
Wr x,
= W0 x,
d log r ( ) + log r ( ) .
N
N
2
(A.14b)
n
N
= nN
1
1
d k (0) ( ) + k (0) ( ) log rn log rn ,
2
2
(A.15)
hence Re Wr (x, ) does not depend on . Thus, we can set Wr (x, ) = Wr (x, 1) for
n /N . As seen in (A.12), the frequency of the oscillation of Pn (x) is of order N . Therefore,
if we take the average of this oscillation for Pn2 (x), it simply gives 1/2. We thus obtain
x r0
2
exp N Re Kr (x, 1) .
n (x)
2
(A.16)
4rn rn (x 2 rn rn )2
Here, the branch of the square root is chosen to be real and positive. This expression is
convenient for calculating DN because of the independence of at order N .
271
1
1
= x K(x) = x d k (0) ( ) + k (0) ( ) .
(B.3b)
2
2
0
The resolvent inside the cut can be obtained by analytic continuation, and it is consistent
with the choice of the double sign in the definition of k (0) ( ) and k (0) ( ) given in (3.16b)
and (3.16c). We can evaluate the integration in (3.19) by using this relation. In the case of
two cuts, we have
2x
,
x k (0) ( ) + k (0) ( ) =
(B.4)
(x 2 rn rn )2 4rn rn
which is again understood as an analytic function. Thus, the eigenvalue density can be
obtained as
1
i 1
1
i 1
tr
tr
(x) =
(B.5a)
2 N x + i
2 N x i
1
1
x
= Re d
(B.5b)
.
4r ( )r ( ) (x 2 r ( ) r ( ))2
Here, the square root is again defined as an analytic function. The sign of the square root
is positive for positive x. Note that (x) is positive even for negative x because the square
root also becomes negative there. Thus, the integration in (3.19) gives the eigenvalue density.
The relation between the effective potential and the resolvent (B.3) can be checked by
a direct computation near the critical point, where it can be expressed as
x
,
x k (0) ( ) + k (0) ( )
(B.6a)
2
b( ) a ( ) x 2
272
a( )2
2
(gc g ),
rc
b2 ( )
4rc .
(B.6b)
Here, a = a(1) and b = b(1) are identified with the endpoints of the cut. Integrating this
equation with respect to x, the effective potential near the critical point can be obtained
as
1
1 (0)
Veff (x) = x
2
2
rc
C = 2 ,
g
1
) + k(
)
C x a 2 x 2 ,
d k(
(B.7a)
a2 =
2
(gc g),
rc
(B.7b)
(C.1)
1/2
r = rc B + B C
(C.2)
gc g .
The double sign indicates + for r and for r , because r vanishes for = 0. In the onecut
phase, r = r = r , and then r near the critical point is
r = rc
1
(gc g ).
2A
B,
and C can be expressed in terms of F (r , r ) as
A, B,
F (r , r )
F (r , r )
A=
=
,
r r =r =rc
r r =r =rc
1 2 F (r , r )
B =
,
2 r 2 r =r =rc
1 2 F (r , r )
B =
,
2 r 2 r =r =rc
2 F (r , r )
C=
.
r r r =r =rc
(C.3)
(C.4a)
(C.4b)
(C.4c)
(C.4d)
273
Here, we use the fact that F (r , r ) originates from [gV (X)]n,n1 , and that it can be expressed in terms of V (x). Because the potential we consider is even, V (x) has only the
odd power of x and can be expressed as gV (x) = xU (x 2 ). Using this, F (r , r ) is expressed
as
r
r
r
U z+
z+
,
F (r , r ) = dz z +
(C.5)
z
z
z
for even n, and whereas odd n we should use F (r , r ). The condition A = A is not trivial,
but comes from (4.1). Using (C.4), (4.1) expressed as
r
dz
r
z+
.
0 = (r r )
(C.6)
U z+
z
z
z
To satisfy this condition with r = r , we need
r
r
dz
U z+
z+
.
0 = G(r , r ) =
z
z
z
(C.7)
(C.8)
This is the relation between r in the onecut phase and in the twocut phase.
References
[1] E. Brezin, V.A. Kazakov, Exactly solvable field theories of closed strings, Phys. Lett. B 236 (1990) 144;
M.R. Douglas, S.H. Shenker, Strings in less than onedimension, Nucl. Phys. B 335 (1990) 635;
D.J. Gross, A.A. Migdal, Nonperturbative twodimensional quantum gravity, Phys. Rev. Lett. 64 (1990)
127;
D.J. Gross, A.A. Migdal, A nonperturbative treatment of twodimensional quantum gravity, Nucl. Phys.
B 340 (1990) 333.
[2] F. David, Phases of the large N matrix model mnd nonperturbative effects in 2D gravity, Nucl. Phys. B 348
(1991) 507;
F. David, Nonperturbative effects in matrix models and vacua of twodimensional gravity, Phys. Lett. B 302
(1993) 403, hepth/9212106;
B. Eynard, J. ZinnJustin, Large order behavior of 2D gravity coupled to d < 1 matter, Phys. Lett. B 302
(1993) 396, hepth/9301004.
[3] V.G. Knizhnik, A.M. Polyakov, A.B. Zamolodchikov, Fractal structure of 2dquantum gravity, Mod. Phys.
Lett. A 3 (1988) 819;
F. David, Conformal field theories coupled to 2D gravity in the conformal gauge, Mod. Phys. Lett. A 3
(1988) 1651;
J. Distler, H. Kawai, Conformal field theory and 2D quantum gravity or whos afraid of Joseph Liouville?,
Nucl. Phys. B 321 (1989) 509.
[4] M. Hanada, M. Hayakawa, N. Ishibashi, H. Kawai, T. Kuroki, Y. Matsuo, T. Tada, Loops versus matrices:
the nonperturbative aspects of noncritical string, Prog. Theor. Phys. 112 (2004) 131, hepth/0405076.
[5] M.R. Douglas, I.R. Klebanov, D. Kutasov, J. Maldacena, E. Martinec, N. Seiberg, A new hat for the c = 1
matrix model, hepth/0307195.
[6] I.R. Klebanov, J. Maldacena, N. Seiberg, Unitary and complex matrix models as 1d type 0 strings, hepth/0309168.
274
Abstract
We compute the complete gauge and chiral superheavy mass spectrum and couplings of the minimal SUSY GUT (based on the 21012612610 irreps as the Higgs system) by decomposing SO(10)
labels in terms of PatiSalam subgroup labels. The spectra are sensitive functions of the single
complex parameter that controls MSGUT symmetry breaking. We scan for the dependence of the
threshold corrections to the Weinberg angle and unification scale as functions of this parameter. We
find that for generic values of the GUT scale parameters the modifications are within 10% of the one
loop values and can be much smaller for significant regions of the parameter space. This shows that
contrary to longstanding conjectures, high precision calculations are not futile but rather necessary
and feasible in the MSGUT. The couplings of the matter supermultiplets are made explicit and used
to identify the channels for exotic (B = 0) processes and to write down the associated bare d = 5
operators (some of both are novel). The mass formulae for all matter fermions are derived. This sets
the stage for a comprehensive RG based phenomenological analysis of the MSGUT.
2005 Elsevier B.V. All rights reserved.
PACS: 12.10.Dm; 12.10.Kt; 12.60.i; 12.60.Jv
1. Introduction
The supersymmetric SO(10) GUT based on the 126, 126, 210 Higgs multiplets [14]
has, of late, enjoyed a much delayed bloom of interest motivated by its economy and preEmail address: aulakh@pu.ac.in (C.S. Aulakh).
05503213/$ see front matter 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.nuclphysb.2005.01.008
276
dictivity. Besides the traditional virtues of SO(10) this is the minimal renormalizable model
which has shown itself capable of matching the observed fermion spectra, including the
prima facie GUT repellent feature of maximal mixing in the neutrino sector [57]. Beyond
the traditional scenario of perturbative unification of couplings due to the RG flow between
MS and MX it also offers strong indications that the gauge coupling becomes strong above
the GUT scale. We have argued [8,9] that this necessarily leads to dynamical symmetry
breaking of the GUT symmetry at a scale U (just above the perturbative unification scale
MU 1016 GeV). Utilizing the quasiexact supersymmetry at the GUT scale we made
plausible [9] a scenario in which U is calculably determined by only the low energy data
and structural features of the theory (such as the gauge symmetry group, supersymmetry
and the very restricted Higgs multiplets available to generate fermion massesparticularly
neutrino massesin a renormalizable theory). This scenario offers interesting possibilities
of a novel picture of elementarity and dual unification characterized by a new fundamental
length scale 1
U characterizing the hearts of quarks [8,9].
The MSGUT is thus the focus of multifaceted interest and a detailed phenomenological
analysis of the theory in terms of the structure dictated by its GUT scale minimality is thus
called for. However such an analysis has been delayed by the computational difficulty of
obtaining the GUT scale spectra and couplings and the effective Lagrangian describing the
normal and exotic features (baryon and lepton violation, etc.) of the GUT derived MSSM
(i.e., extended by the leading (d = 5) exotic operators of the theory). The spectra and
couplings are necessary to analyse threshold corrections to the gauge couplings near the
GUT scale and are also a crucial input into deriving the Lagrangian for exotic processes
and parameters mediated by GUT scale massive fermions. In [11] we presented techniques
for computing the decomposition of SO(10) invariants in terms of the unitary labels of its
maximal (PatiSalam) subgroup SU(4) SU(2)L SU(2)R . Once this decomposition is
performed the computation of the complete spectrum and couplings is quite easy and the
long standing vagueness regarding the Clebsches that arise can finally be banished. This
allowed us to present, by way of illustration of the power of our method, the two most
important mass matrices (4 4 and 5 5 respectively) affecting electroweak symmetry
breaking, fermion masses and nucleon decay: namely those for the MSSM type Higgs
SU(2)L doublets and baryon number violation mediating SU(3)c colour triplets that mix
with the doublets and triplets of the fermion mass (FM) Higgs (10, 126). Moreover since
our methods allow computation of the actual couplings of Higgs to spinors we could also
obtain the d = 5 operators for baryon violation generated via exchange of triplet higgsinos
contained not only in the traditional (6, 1, 1) submultiplets (of the 10 or those in the 126
[12]) which had been noticed to provide a connection between neutrino masses and proton
decay, but also in other channels arising from the exchange of colour triplets contained
in (10, 1, 3)126 submultiplets involved in neutrino mass generation [11]. A more complete
calculation of these spectra and effective Lagrangians and an initial estimate of their effects
is the subject of this paper.
While the calculations presented were in their final stages we were collaborating and
cross checking with another parallel calculation [13] of mass spectra using a different [10]
method which has since been published. Moreover another group [14,15] has also recently
published a calculation (using the same methods as [13]) of spectra and baryon decay effective potentials recently. As far as computation of chiral superfield spectra are concerned
277
our results coincide (upto normalization and phase conventions) with those of [13]. However both our results diverge [11,13] in certain details from the chiral spectra given in [14].
Moreover as already noted by us in (an update to) [11] we also disagreed with the results
of [14] regarding the higgsino channels available for baryon decay in this model. We found
[11] that [14] obtained couplings between the 126 multiplet and matter in the spinorial 16
representation which were in contradiction with ours [11] not only as regards the numerical
coefficients but also in the heavy Higgs channels to which matter fields couple in a baryon
number violating way. In the revised version of [14], i.e., [15] this defect has apparently
been corrected at least modulo disagreement on values of Clebsches. We shall try to settle
these questions by tracing the reasons for the continuing discrepancy in explicit detail and
confirm our previous assertions. We have also analyzed the gauge Dirac multiplet structure
arising from the superHiggs effect and the masses and vevs responsible for the type I and
type II mechanisms [16,17] of neutrino mass generation.
We emphasize that our method allows computation, not only of spectra but also of the
couplings of all the multiplets in the theory (whether they are renormalizable or heavyexchange induced effective couplings) without any ambiguity. Moreover our results are
obtained by an analytic tensorial reprocessing of labels of fields in the Lagrangian. This
approach might thus find preferment with field theorists in comparison with the more
restricted capabilities of the approach of [10], which, so far, has not proved capable of
generating all the Clebsches of the SO(10) theory and which relies on an explicit multiplet representative and computer based approach which is tedious to connect to the unitary
group tensor methods so familiar to particle theorists.
It has long been held by some that SO(10) GUTs specially SUSY SO(10) GUTs,
are essentially selfcontradictory [18] due to the apparently enormous threshold effects
that might arise due to the large number of superheavy Higgs residuals in these theories.
Thus the authors of [18] speak of the futility of highprecision calculations in SO(10).
However these assertions have never been tested against any actual computations of mass
spectra of a SUSY SO(10) GUT and are only worst case estimates assuming that no cancellations occur. However we expect that cancellations will generically occur since the
leptoquark mass has no reason to lie at the edge of the mass spectrum nor are the coefficients all of the same sign. With the computed spectra now available, the threshold effects
on observable quantities such as Sin2 w , MX , etc., become computable in terms of the
relatively small number [4] of GUT scale parameters of the MSGUT. In fact [4] a single
complex parameter controls the solutions of the cubic equation in terms of which all the
superheavy vevs are defined. We have performed a preliminary scan of the parameter space
of the MSGUT to see what is the typical size of these corrections. We find the striking result that such threshold corrections are generically 10% of the 1loop results [2326]
that underpin the GUT scenarios viability. Moreover, for significant and possibly interestingly restricted regions of parameter space these corrections can be much smaller, i.e., as
small as 0.5% of the one loop results. Thus far from indicating futility our results indicate
that a thoroughgoing investigation of the compatibility of low energy precision data with
the threshold corrections may significantly constrain the parameter space of the MSGUT.
In any case we show that inclusion of threshold corrections is necessary and not futile.
Such an analysis is now in progress [27].
278
In Section 2 we present a brief review of the principal features of the minimal SUSY
SO(10) theory [14] and compute the gauge supermultiplet masses. In Section 3 we provide the PS reduction of the SO(10) Higgs superpotential. From this we computed the
chiral fermion mass terms and thus the supermultiplet spectra which we discuss here and
list in Appendix A. In Section 4 we compute the threshold corrections to the 1loop values
of the unification scale MX and sin2 W (MS ). In Section 5 we present the couplings of
the matter fields to FM Higgs fields in the superpotential as well as their couplings to the
gauginos of the SO(10) model. This permits us to identify the possible channels for baryon
violation in the low energy theory via exchange of higgsinos or gauginos and compute
the relevant effective Lagrangians. Using the associated mass matrices we write down the
d = 5 effective Lagrangians for baryon and lepton number violation which arise via exchange of superheavy fermions. In Section 6 we discuss the mass formulae for the matter
fermions in this model. The Majorana mass terms of the left and right handed neutrinos
and the SU(2)L triplet microvev responsible the type II mechanism for neutrino mass is
calculated along with the charged fermion mass matrices. In a final section we discuss our
conclusions and results and plans for further investigations using the results derived here.
M
WAM = MH Hi2 + ij kl ij kl + ij kl klmn mnij + ij klm ij klm
2
4!
4!
5!
1
+ ij kl ij mno klmno + Hi j klm ( ij klm + ij klm )
4!
4!
(1)
and
1
(5)
f T C i i5 B i1 ,...,i5 .
(2)
5! AB A 2 1
Our notations and conventions for spinors can be found in [11]. The Yukawa couplings
are complex symmetric 3 3 matrices and one of them, say f can thus be
hAB , fAB
diagonalized (by an orthogonal transform Uf U T using a unitary matrix U which leaves the
= F
matter kinetic terms invariant) to a real positive diagonal form fAB
A AB , thus leaving
15 residual real parameters in WFM . In addition the 7 complex parameters in WAM can be
WFM = hAB AT C2 i B Hi +
(5)
279
reduced to 10 real ones by absorbing 4 phases by Higgs field redefinitions. Then together
with the gauge coupling one has in all exactly 26 nonsoft parameters [4]. Coincidentally,
MSSM also has 26 nonsoft couplings consisting of 9 quark and charged lepton masses,
3 Majorana neutrino masses 3 quark (CKM) and 3 lepton (PMNS) mixing angles and
1 quark but 3 lepton CP phases together with 3 gauge couplings and a parameter. Thus
we see that the 15 parameters of WFM must be essentially responsible for the 22 parameters
describing fermion masses and mixings in the MSSM.
The kinetic terms are given by covariantizing in the standard way the global SO(10)
invariant Dterms
1
1
1 The relations between the quantities of [13] (denoted by primes) and ours are = / 2,
= /
(also for vevs) = 2 , = 2 , = 2, M = 2M.
280
this theory was laid and the spectrum of MSSM like SU(2)L doublets and SU(3)c baryon
decay triplets first computed.
We denote quantum numbers w.r.t. the SM gauge group by enclosing them in square
brackets while those with respect to the PS group are denoted by round brackets. We have
adopted the rule that any PS submultiplet of an SO(10) field is always denoted by the
same symbol as its parent field, its identity being established by the indices it carries or
by additional sub/superscripts ((a), (s), , L, R) denoting (anti)symmetry or (anti)self
duality, if necessary. On the other hand, since one often encounters several chiral MSSM
multiplets of the same type arising from different SO(10) Higgs multiplets we will also
introduce a naming convention using roman letters for these multiplets. If we need to denote the scalar component of a chiral superfield we use a tilde over the superfield symbol
and sometimes use a superscript F to denote fermionic components of chiral superfields
while gauginos are denoted by . Our notation for indices is as follows: the real indices
of the vector representation of SO(10) are denoted by i, j = 1, . . . , 10. The real vector
index of the upper left block embedding (i.e., the embedding specified by the breakup of
the vector multiplet 10 = 6 + 4) of SO(6) in SO(10) are denoted a, b = 1, 2, . . . , 6 and
of the lower right block embedding of SO(4) in SO(10) by ,
= 7, 8, 9, 10. These in 2,
3,
4,
5,
6
dices are complexified via a unitary transformation and denoted by a,
b = 1,
1 , 2,
2 , 3,
3 where 1 1,
2 1 , etc. Similarly we denote the complexified
,
= 1,
9,
10
Using this complexification we showed [11] how
0.
versions of ,
by ,
= 7 , 8,
all SO(6) SO(4) subinvariants of SO(10) tensor invariants could be systematically converted to SU(4) SU(2)L SU(2)R invariants whose indices are as follows: the indices of
2).
Finally the index of
= 1,
the doublet of SU(2)L (SU(2)R ) are denoted , = 1, 2 (,
the fundamental 4plet of SU(4) is denoted by a (lower) , = 1, 2, 3, 4 and its upperleft
block SU(3) subgroup indices are ,
= 1, 2, 3. The corresponding indices on the 4 are
carried as superscripts. These doublets and quartets correspond to the chiral spinor representations of the SO(4) and SO(6) subgroups of SO(10). Details of the spinorial invariant
decomposition techniques may be found in [11]. The component of the SU(4) adjoint in
(15)
the direction of the GellMann generator i is labeled with a superscript (15) or (B L).
2
Thus the PS decomposition of our SO(10) multiplets is
2, 2),
+ , (a)
(6, 2, 2) + , (s)
(10, 2, 2) + (s)
(10,
(4)
= + = 126
(s)
L
(10, 1, 3) + (s)
(10, 3, 1) + (6, 1, 1) + , (15, 2, 2),
=
(5)
= = 126
(s) (10, 3, 1) + (6, 1, 1) + (15, 2, 2),
= R(s)
(10, 1, 3) +
(6)
(7)
(a)
(a)
1, 2) + F (4, 2, 1),
A = 16 = 16+ = (4+ , 2+ ) + (4 , 2 ) = F (4,
(8)
F (4, 2, 1) = (Q
, L ),
with
Q=
U
,
D
L=
1, 2) = Q
, L
F (4,
,
e
=
Q
d
,
u
L =
281
(9)
e
.
(10)
The GUT scale vevs that break the gauge symmetry down to the SM symmetry are [1,2]
a
(15, 1, 1) 210 : abcd
= abcdef ef ,
(11)
2
where [ef ] = Diag(2 , 2 , 2 ), 2 = i2 . One dualizes
ab
1
abcdef cdef .
4!
(12)
ia
ia
Diag(I3 , 3)
,
=
2
2
(15, 1, 3) 210 :
ab
= ab ,
(13)
(14)
1 2 = i (R) 0 ,
2
(1, 1, 1) 210 :
= p ,
(15)
(16)
(10, 1, 3) 126 :
(R) 441 1
= ,
1 3 5 8 0
= = i 44(+)
2
(17)
(10, 1, 3) 126 :
(R)44 44
2 4 6 7 9
= = i () = 22 .
2
(18)
(19)
(20)
(21)
(22)
(23)
282
The vanishing of the Dterms of the SO(10) gauge sector potential imposes only the
condition
  =  .
(24)
(25)
where =
and the other vevs can be expressed in terms of values of the variable x
which solve Eq. (25). This parametrization of the MSGUT ssb problem [4] is of great help
computationally and clearly exhibits the crucial importance of the , i.e., of the ratio M/m.
The important role played by a similar ratio in the other renormalizable SO(10) GUT based
on the 45, 54, 126, 126 representations has already been noted [20].
When we measure vevs or masses in units of m
= m/ we will put a tilde over the sym
bol. We also define the additional dimensionless parameters = / and M H = MH /m.
Then the dimensionless vevs are = x [4] and
M
m
a =
(x 2 + 2x 1)
,
(1 x)
p =
x(5x 2 1)
,
(1 x)2
2 x(1 3x)(1 + x 2 )
.
(1 x)2
(26)
The solutions of the cubic equation (25) are generically complex. We will therefore
nowhere assume hermiticity for our mass matrices, preferring to leave them undiagonalized
for eventual numerical diagonalization so that all our results are applicable in the general
case. We will not generate arrays of expressions in terms of the variable x, although it is
easy to do so since, practically speaking, the substitutions are now handled via a computer
anyway.
We conclude this section with a description of the superHiggs effect for the breaking
SO(10) SU(3) SU(2) U (1)Y which is achieved by the above superheavy vevs. As
is well known, as a consequence of gauge symmetry breaking, each massive gauge boson forms a massive supermultiplet together with its longitudinal Goldstone pseudo scalar
(and its real scalar partner) as the 4 bosonic degrees of freedom. Its gaugino and the chiral
fermion superpartner of the Goldstone scalar pair make up one Dirac fermion superpartner also with 4 degrees of freedom. This is the socalled Dirac or massive vector gauge
supermultiplet. These gauge boson/gaugino masses are the most fundamental thresholds
of the GUT and it is appropriate to begin with a discussion of their values for this model.
In Section 4 we follow [21,22] to compute the threshold corrections using the spectra we
compute. Then a Dirac gauge coset multiplet in representation R of the MSSM gives rise to
a RG mass threshold above which the gauge and chiral components of the Dirac multiplets
separately contribute 3Ti (R) and Ti (R), respectively, to the beta function coefficients of
the individual MSSM couplings (including the U (1)Y coupling!).
It is easiest to keep track of the gaugino masses and mixings. The combination of chiral
fermions that forms a Dirac fermion together with a gaugino must, for consistency, be a
zero mode of the mass matrix arising from the superpotential and this makes it easy to
disentangle the gauge spectrum even in the case of complex vevs and parameters. For the
symmetry breaking to the MSSM the gauginos of the coset SO(10)/G321 lying in the PS
283
representations (6, 2, 2) (1, 1, 3) plus the triplets and antitriplets in (15, 1, 1) (i.e., 33
Dirac multiplets in all) obtain a mass by pairing with chiral AM Higgs fermions. One need
only substitute the vevs given above into the PS decomposition of the gaugino Yukawa
terms which have the form
1
1
F
F
F
g 2
+ H.c.
+
ij kl im mj
im
im
kl
ij kln
mj kln
ij kln
mj kln
3!
2 4!
(27)
One finds the following gaugino masses.
2
G6
mG ei G4 + ei G5 + H.c.,
2
2 (R0)
3 (15)
G6
,
5
5
44
G4 R ,
2
R+
G5 = 44 .
2
(28)
The naming conventions for the chiral states are given in Section 4 and Appendix A. Here
, are the phases of , . Since the representation is real, the mass matrix G in this
sector is symmetric. The complete G[1, 1, 0] sector mass matrix (including gauginos) G is
6 6 while its pure chiral part G (which arises only from the superpotential) is 5 5 and
symmetric and the 5tuple (0, 0, 0, , ) is both a left and right null eigenvector of Gas
will be obvious when it is presented further
on (Section 2 and Appendix A).
1, 4 ] J [3, 1, 4 ]: mJ = g 8a2 + 162 + 2 2 .
(ii) J[3,
3
3
In this case (J4 ) = 4 , (J4 ) = 4 pair up with thecombinations corresponding
to the
T
=
N
(,
2a,
2
2) of
left and right null eigenvectors v0J L = NJ ( , 2a, 2 2), v0J
J
R
the complex, nonsymmetric, upper left 3 3 submatrix J of the 4 4 mass matrix J in
the J sector. The gaugino mass terms are
ig J4 2 2a J2 + 4 J3 2 J1
ig 2 2a J2 + 4 J3 + 2 J1 J4 + H.c.
(29)
(iii) F [1, 1, 2] F [1, 1, 2]: mF = g 242 + 2 2 .
g F3 i 24F2 + 2 F1 + g i 24F2 2 F1 F3 .
(30)
284
2, 1 ] E[3, 2, 1 ]: mE = g (4a 2 + 2w p2 + 2 2 ).
3,
(iv) E[
3
3
5 4
The chiral partners of the gauginos E5 4
,E
1
correspond to the null eigenvec2
tors v0ER = (i, 2(a ), p)T ; v0EL = (i , 2(a ), p) of the upper left
3 3 corner E of the E sector 4 4 chiral fermion mass matrix E. E1 , E 1 do not mix with
other Esector multiplets. The mass terms are
g E 5 i 2 E2 + 2 a E3 + 2 p E4
+ g i 2 E 2 + 2 a E 3 + 2 p E 4 E5 .
(31)
2, 5 ] X[3, 2, 5 ]: mX = g 4a + 2 + 2p + 2 .
(v) X[3,
3
3
3 4
The chiral partners of the gauginos X3 4
,X
2
correspond to the null eigenvec1
tors v0XR = ( 2(a + ), + p)T = v0XL of the upper left 2 2 corner X of the 3 3
Xsector chiral fermion mass matrix X . The Xgaugino mass terms are
g X 3 2 a + X1 + 2 p + X2
+ g 2 a + X 1 + 2 p + X 2 X3 .
(32)
(s)
ij kl = m + (s)
+ (R)
(R) + (L)
(L)
4!
1 (a)
+ (a) 2 ,
2
(a)
M
+ 2
ij klm ij klm = M (a)
5!
(s)
+ (s)
+ (s)
(R)
(R)
(L)
1 3
2
(s)
= i 2i (s)
4!
3
+ (L)
2i (R)
(R)
(L)
(s)(L)
,
(33)
(34)
(s) (L)
(a) (s) (R)
(s)
+ (a)
(R) +
(L)
+ 2 (s)
(s) (R) + (s) (L)
(R)
(R)
(L)
(L)
1
(a)
(a) (R) + (a) (L)
2
2
285
(35)
1
= iH(a) (a) + H
H
4!
1
(s)
(s)
H (s) (s) (R)
+ (L)
2
1
(s)
H (a) (s) + H(a)
2
i
H
(R) (L)
2
(s)
+ H (a) (R)
(s)(R) + H(a)
(L)
(L)
1
(a) (a)
H
2
iH(a)
(a)
1 (a)
+ H
(a) ,
2
(36)
1
H = iH(a) (a) + H
4!
(s) (s)
1
H
+ (s)(L)
(R)
2
1
(s)
H (a) (s) + H(a)
2
i
+ H
(R) (L)
2
(s)(L)
(s)(R)
(a)
+ H (a) (L)
+ H
(R)
1 (a)
1 (a)
= 2i +
4!
+ 2i (s)(R)
+ (s)(L)
(s)(R)
(s)(L)
(37)
286
+ (s)(R) (s)(R) (s)(L) (s)(L)
(s)
+ i 2 (a) (s) + (a)
(s)
+ i 2 (a) (s) (a)
(s)
(s)
2i (s) (L) + (R)
(s)
(s)
(s)
(s)
2i (s) (R)
+ (L)
+ 2 (a) (L)
(L) + (a) (R)
(R)
2 2 (R) + (L)
(s)(R) (s)
(s)(L) (s)
(R)
(R)
(L)
(L)
(s)(L)
(s)
+ i 2 (a) + (a)
(R)
(s)(R)
(s)
+ i 2
(a)
(a)
(L)
(38)
The purely chiral superheavy supermultiplet masses can be determined from these expressions simply by substituting in the AM Higgs vevs and breaking up the contributions
according to MSSM labels.
It is again easiest to keep track of chiral fermion masses since all others follow using
supersymmetry and the organization provided by the gauge superHiggs effect.
There are three types of mass terms involving fermions from chiral supermultiplets in
such models: (A) unmixed chiral, (B) mixed pure chiral, (C) mixed chiral and gaugino.
3.1. Unmixed chiral
A pair of chiral fermions transforming as SU(3) SU(2)L U (1)Y conjugates pairs
up to form a massive Dirac fermion. For example for the properly normalized fields
44
(R+)
1, 4] = 44(R)
,
A[1, 1, 4] =
A[1,
2
2
one obtains the mass term
= mA AA.
2 M + (p + 3a + 6) AA
(39)
The physical Dirac fermion mass is then mA  since the phase can be absorbed by a field
redefinition. By supersymmetry this mass is shared by a pair of complex scalar fields with
the same quantum numbers. If the representation is real rather than complex one obtains
an extra factor of 2 in the masses. There are in fact 19 types of such multiplets and their
(roman letter) labels are given along with their masses and SO(10) origins in Table 1 in Appendix A. The case of the sectors C[8, 2, 1] and D[3, 2, 1] bears special mention. The
mass terms for these multiplets arise only between pairs drawn one each from (15, 2, 2),
(15,
2, 2) and there is no mixing between a C, C or D, D drawn from the same SO(10)
287
multiplet simply because the superpotential does not contain any term containing 2 or
2
. This was the reason for the discrepancy in this sector between the results of [11,13]
and [14]: there simply is no such mixing.
3.2. Mixed pure chiral
In this case there are no contributions from the gaugino Yukawas or the Dterms to
the supermultiplet masses, but there is a mixing among several multiplets of the same SM
quantum numbers. There are only three such multiplet types:
).
(a) [8, 1, 0](R1 , R2 ) ( , (R0)
(m a)
2
R=2
2 m + (p a)
(40)
with both rows and columns labeled by (R1 , R2 ). The masses are the magnitudes of the
eigenvalues of the matrix R:
2
p
p
2
R  = 2m 1 +
(41)
a
+ 2 = mR .
2
4
While the corresponding eigenvectors can be found by diagonalizing the matrix RR .
2, 1] and colour triplets
The mass matrices of the electroweak doublets h[1, 2, 1], h[1,
1, 2 ] which mix with the multiplets contained in the 10 plet FM Higgs
t[3, 1, 23 ], t[3,
3
are the most crucial ones for determining the phenomenology of the effective MSSM that
arises from this GUT. These matrices were first calculated in [11] (v2) and later, stimulated
by a contradiction with a recent paper [14], the d = 5 baryon violating operators induced
by the exchange of heavy higgsinos were computed and added to a revised version [11]
(v4) by using the Clebsches for the 16 16 126 and 16 16 10 invariants calculated earlier
by us. Thus one has.
(b) [1, 2, 1](h 1 , h 2 , h 3 , h 4 ) [1, 2, 1](h1 , h2 , h3 , h4 ) (H2 , 2
(15)
(15) 441
2
(15) 44
, )
2
, 2
(H 1 , 1 , 1 , ).
2
These multiplets label the 4 rows and columns of the 4 4 mass matrix H [11] which is
given in the collection of mixing matrices in Appendix A. We note that we have redefined
our mass parameters m, M by a factor of 2 relative to those we used in [11]. To achieve
the MSSM spectrum of one pair of light doublets, it is necessary to fine tune one of the
parameters of the superpotential (e.g., MH ) so that Det H = 0. By extracting the null eigenvectors of H H and HH one can compute the coefficients of the various bidoublets in
the light doublet pair, and, in particular, we can find those for the doublets coming from the
10, 126 multiplets which couple to the matter sector (see Section 6). In this way the SO(10)
can be brought into
constraints on the fitting of the Yukawa coupling matrices hAB , fAB
focus and the invalid assumption that the squares of these coefficients [5,7] add up to 1 can
be dispensed with.
(15)
288
,
1, 2 ](t1 , t2 , t3 , t4 , t5 ) [3, 1, 2 ](t1 , t2 , t3 , t4 , t5 ) (H 4
4
(c) [3,
(a) , (a) , R0 ,
3
3
(R)
For generic values of the couplings all these particles are superheavy. These triplets and
antitriplets participate in baryon violating process since the exchange of (t1 , t2 , t4 )
The strength of
(t1 , t2 ) higgsinos generates d = 5 operators of type QQQL and lu u d.
the operator is controlled by the inverse of the t t mass matrix T which we computed in
[11] and is given in Appendix A. We shall examine how d = 5 baryon and lepton number
violating operators are generated in Section 5.
(15)
44 / 2,
44(R+) / 2,
(a) [1, 1, 0](G1 , G2 , G3 , G4 , G5 , G6 ) (, (15) , (R0) , (R)
( 2(R0) 315 )/ 5)
which mix via a 6 6 mass matrix G given in Appendix A. The complex conjugates of
the 6th row and column form left and right null eigenvectors v0GL , v0GR of the upper left
5 5 block G of G. The determinant of G is generically nonzero although the determinant
of the submatrix G vanishes. It will clearly not affect the evolution of the MSSM gauge
couplings at one loop due to the singlet quantum numbers.
(a)4
(s)
(a)
4 ,
, 4
, 1 ).
2 ) (
2
4
1
1
289
via a righthanded vev could only come from the following two terms in Eq. (38):
(s)
(s)
= 2i (R)
.
(s) 2i (s) (R)
4!
(15,
2, 2) with E3 coming from (10, 2, 2) but not E 1 coming from (15, 2, 2) with
E3 coming from (10, 2, 2). Similar considerations account for the other decouplings
between E1 , E 1 and the rest of the E sector. Ultimately this correlation is accounted for by
the correlation between the duality properties of SO(6) decuplets and SO(4) triplets within
the SO(10) selfdual and antiselfdual multiplets , .
(15)
44 ,
(c) [1, 1, 2](F1 , F2 , F3 ) [1, 1, 2](F1 , F2 , F3 ) ( 44(R0) , (R) , (R) ) ((R0)
(15)
(R+) , (R+) ).
The mixing matrix F has the usual structure. The residual massive eigenstates after separating off the two Dirac fermions of mass
1/2
mF = g 242 + 2 2
(42)
is a Dirac fermion of mass
mF =  2 + 122
F = NF (i 12F1 + F2 )ei( ) ,
F = NF (i 12F1 + F2 ),
NF1 = 122 +  2 .
(43)
(44)
(R0)
The 4 4 mass matrix J has the usual superHiggs structure: complex conjugates of the
4th row and column are left and right null eigenvectors of the upper left 3 3 submatrix J.
J has nonzero determinant although the determinant of J vanishes.
4(s)
4(a)
(s)
(e) [3, 2, 53 ](X 1 , X 2 , X 3 ) [3, 2, 53 ](X1 , X2 , X3 ) ( 1 , 1 , 1 ) (4
,
2
(a)
, 2 ).
4
2 4
Mix via a 3 3 symmetric matrix X so the left and right null eigenvectors of the upper left
2 2 submatrix X, formed by the complex conjugates of the third row and column of X ,
are the same. Separating off the two Dirac [3, 2, 53 ] gauge fermions of mass
mX = g 4a + 2 + 2p + 2
one is left with a Dirac fermion of mass
2m(2x2 + 1 x2 )
mX = 2 2m + (a + ) + m +  =
1 x
290
+ 1 x2
i( )
e m 1x
2xX1 + (1 x)X2 , 2x X 1 + (1 x)X 2 .
2x2
(45)
This concludes our description of the superheavy mass spectrum of the minimal SUSY
GUT. As mentioned earlier our results were calculated in collaboration with the authors
of [13] and are in agreement with the chiral spectra calculated in [13]: whose results also
confirm our earlier results [11] on the phenomenologically important matrices H, T . Moreover we have evaluated the mixing of the gauginos with the chiral fermions explicitly and
calculated the gauge spectra and eigenstates besides furnishing all the couplings in the
superpotential sector explicitly. The gauge couplings and the gauge Yukawa couplings to
matter will be given in Section 5. The gaugino mixing with the chiral fields will be useful
to us when we examine B + L violation mediated by gaugino exchange as well by higgsino triplet exchange and when one wishes to examine the flow of gauge couplings past
the gauge thresholds.
4. RG analysis
The first phenomenological success of GUTs was the 1loop calculation of the numerical value of Weinberg angle [23]. This was followed by the prediction [24] and then the
verification [26] of an amazingly exact compatibility between UV gauge coupling convergence in the MSSM and the precision LEP data. The large mass at which the top quark was
eventually discovered and the associated large value Sin2 W 0.23 verified the originally
somewhat far fetched conjecture of [24]: a historical fact that is still not always appreciated. The proposal of Weinberg [21] for calculating threshold effects within an effective
field theory picture using mass independent renormalization schemes such as the standard
MS renormalization scheme was taken up and developed in detail in [22]. Thereafter, using
these results, it was argued [18] that highprecision calculations in SO(10), and particularly in supersymmetric SO(10) models which used large representations such as 210, 54,
126, etc., were futile. This was due to the huge corrections to the one loop predictions that
they expected in view of the large number of superheavy fields and the expected span in
their masses. It should be remarked however that without an explicit calculation cancellations that might naturally occur would be overlooked. Such calculations were never done.
These negative expectations were a motivation for the development [29] of a whole genre
of SO(10) models that eschewed large representations (and thus parameter counting minimality) in favor of models with a plethora of small representations and nonrenormalizable
interactions.
The other approach [14,19,20] approach has all along been to retain renormalizability
of the fundamental theory. We regard retention of Higgs multiplets just adequate to account
for the gauge and fermion spectrum via renormalizable couplings as a sine qua non for
even being clear as to what is testable about a given model. The inverse approach where
representations (hypotheses) are multiplied without necessity seems regressive to us.
291
Thus the proposal of the SUSY SO(10) GUT based on the 21012612610 [1,2] Higgs
system as being the Minimal SUSY GUT [3,4] must live or die by the criterion: Are
the one loop values of Sin2 W and MX generically stable against superheavy threshold
calculations?. By generically we mean: for a nonsingular subset of the parameter space.
So far this question could not be answered definitively since no complete mass spectrum
was available in any SUSY SO(10) model to settle the issue. Partly this was due to the lack
of accessible techniques to calculate mass spectra and couplings in these models due to the
difficulty in obtaining the relevant SO(10) Clebsches. Over the last few years we have
developed [11] a complete technology for translating SO(10) tensor and spinor labels into
those of the unitary labels of the PatiSalam maximal subgroup SU(4) SU(2)L SU(2)R
of SO(10). This allowed us to compute first the mass matrices of SM type doublets and
proton decay mediating triplets and then the complete spectrum and couplings reported
in this paper. The partial technology of [3,10] has also been used to compute [13,14] this
spectrum (but not the couplings). With the correct spectrum in hand we can apply the
standard formulae of Hall [22] to compute the changes in the 1loop GUT predictions as
functions of the few MSGUT parameters ( = M/m, , , , , m, g, MH ) which are
relevant at the GUT scale. Thereafter we can scan the parameter space to see how the
corrections vary with these parameters.
A few remarks on the role of the parameters are in order. The parameter = M/m
is the only numerical parameter that enters into the cubic equation (25) that determines
the parameter x in terms of which all the superheavy vevs are given. It is thus the most
crucial determinant of the mass spectrum. The dependence of the threshold corrections on
the parameters , , , seems quite mild (logarithmic) (this is especially obvious for the
unmixed chiral multiplets) and is also suggested by our preliminary scans of the parameter
space. Thus changing , by a factor of 100 each yields plots vs that seem indistinguishable from the ones presented below. From Eq. (26) we see that m/ can be extracted as the
overall scale parameter of the vevs. Since the threshold corrections we calculate are dependent only on (logarithms of) ratios of masses the parameter m does not play any crucial
role in our scan of the parameter space: it is simply fixed in terms of the (threshold and
two loop corrected) mass MV = MX of the lightest superheavy vector particles mediating
proton decay: which mass is chosen, in the approach of Hall, as the common physical
matching point in the equations relating the running MSSM couplings to the SO(10) coupling [22]. Inasmuch as we take the parameters , as given, and the parameter m is set
by the overall mass scale, the freedom in the parameter is essentially that of choosing the
126126 mass parameter M, i.e., the freedom in choosing the dimensionless parameter ,
is essentially that of the ratio M/m: which ratio is already known to be a crucial control
parameter of symmetry breaking in renormalizable models that utilize the 126, 126 to complete and enforce the symmetry breaking down to the SM symmetry [20]. As for MH it is
fine tuned to keep a pair of doublets light. The relation between the MSSM couplings at
the SUSY breaking scale MS 1 TeV and the GUT coupling at the scale MX is given by
[21,22]
bij
1
1
MX
=
+ 8bi ln
+ 4
ln Xj 4i (MX )
i (MS ) G (MX )
MS
bj
j
(46)
292
here
M0
Xj = 1 + 8bj G MX0 ln X
MS
(47)
is understood to be evaluated at the values of MX0 , G (MX0 ) determined from the one loop
calculations. In this equation the contribution of the Yukawa couplings has not been taken
into account and this should also be done in a full investigation [27]. Here we will confine
ourselves to estimating the corrections using the equations as given above, since these were
already conjectured [18] to lead to a breakdown of the unification scenario. The coefficients
33
1
{b1 , b2 , b3 } =
, 1, 3 ,
16 2 5
1
[bij ] =
(16 2 )2
(48)
(49)
are the standard one loop and two loop gauge evolution coefficients for the MSSM [28].
The term containing i represents the leading contribution of the superheavy thresholds:
i () =
2
MV
MV
MF
(biV + biGB ) + 2(biV + biGB ) ln
+ 2biS ln
+ 2biF ln
,
21
(50)
where V , GB, S, F refer to vectors, Goldstone bosons, scalars and fermions respectively
and a sum over heavy mass eigenstates is implicit. The formulae for the threshold corrections are
(th) (ln MX ) =
(51)
M
(th) (Log10 MX ) = 0.0217 + 0.0167(5b1 + 3b2 8b3 ) Log10 0 ,
MX
10(MS )
ij k (bi bj )k MX0
(th) sin2 W (MS ) =
(5b1 + 3b2 8b3 )
(52)
ij k
M
MX0
(53)
where bi = 16 2 bi are the 1loop beta function coefficients for multiplets with mass M .
To evaluate these formulae it is convenient to group the gaugino contributions along with
the chiral fermions they mix with. The values of the indices S1 , S2 , S3 combined as in
Eqs. (52), (53), i.e., SW = 4S1 9.6S2 + 5.6S3 ; SX = 5S1 + 3S2 8S3 are given in Table 2
in Appendix A.
293
5b1j + 3b2j 8b3j
1
(ln MX ) =
ln Xj ,
10b1 + 6b2 16b3
bj
j
(2loop) sin2 w (MS ) =
10(MS )
bkl
ij k (bi bj )
ln Xl .
(5b1 + 3b2 8b3 )
bl
(54)
ij kl
11 (MS ) = 57.45,
MS = 1 TeV,
1
3 (MS ) = 11.04
extrapolated from the global averages of current data, the two loops effects give
MX
2loop
log10
= 0.08,
2loop sin2 W (MS ) = 0.0026.
MS
(55)
(56)
= 0.21,
= 0.23,
= 0.35.
(57)
The results for different values of these parameters (but with the same ) are very similar. We will therefore keep them fixed at these values throughout since here we only wish
to illustrate the feasibility of precision RG calculations in the SO(10) MSGUT.
For real values of the superpotential parameters the cubic equation (25) that determines
the vevs has one real and two complex (conjugate) solutions. The latter give essentially
identical corrections. So for real we need to present plots for two solutions only. These
are given as Figs. 16.
From Figs. 1, 3 see that for most real values of the threshold effects on sin2 W (MS )
are less than 10 % of the 1loop values. There are three exceptional values of very close
to which this limit is breached but even then the change is only about 25%. For large magnitudes of an asymptotic regime of around 10% change seems to supervene. Similarly
Figs. 2, 4 show that the change in MX is also not drastic (though possibly phenomenologically interesting since the gauge contribution to the nucleon lifetime goes as MX4 )
except at certain special points among which one recognizes certain known points of enhanced symmetry [4,13] such as = 5, 10 (SU(5)), = 3 (GLR ), = 2/3 (flipped
SU(5) U (1)). It is natural to expect that something similar accounts for the other sharp
peaks and dips in these plots. Moreover their narrowness emphasizes that for generic values of the parameters one may expect the threshold corrections to be small for the real
real x cases. There are also regions in which the threshold corrections to Log10 MX are
as large as 5 and these need special examination with regard to their phenomenological
294
Fig. 1. Plot of the threshold corrections to Sin2 w vs for real : real solution for x.
viability and consistency with the one scale breaking picture. It is interesting that in this
way one can scan the parameter space of the MSGUT and obtain a global tomograph of
the variation in its character with the ratio M/m.
Figs. 3, 4 give a magnified view of the region   < 2. A comparison of the graphs shows
clearly that the peaks in the threshold corrections coincide by either measure, obviously
because some particles are becoming very light and enhancing the mass ratios that enter
the formulae. It will be amusing to use these plots to identify and unravel the special regions
of the MSGUT parameter space.
295
Fig. 3. Magnified plot of the threshold corrections to Sin2 w vs for real : real solution for x.
Let us turn next to the complex solutions of the cubic equation for x but still with real
values of . We obtain the typical plots Fig. 5, 6. The corrections to sin2 W (MS ) are very
small for small   < 2, with a minimum close to = 1. From Fig. 6 we see that apart
from the two peaks near = 5 the corrections to the unification scale are quite small for
small .
When we consider complex values of as shown in Figs. 712 we see that the behavior
is quite regular (like the case of the complex solutions for real ) and once again there are
large regions of parameter space where the corrections are less than 10% for Sin2 w while
296
Fig. 5. Plot of the threshold corrections to Sin2 w vs for real : complex solution for x.
MX changes by a factor of 10 or less. Thus even this cursory scan of the MSGUT parameter space shows that, quite contrary to expectations in the literature [18], precision RG
analysis of the SO(10) MSGUT is far from being futile, since the hierarchy of magnitudes
between MSSM one loop gauge coupling convergence values (O( 1 ) effects) and the one
loop threshold and two loop gauge coupling corrections (O(1) effects [22]) is generically
maintained at the level of 10% or less. Furthermore the RG analysis and parameter scan
in terms of the single parameter can teach us much about the structure of the parameter
297
Fig. 7. Plot of the threshold corrections to Sin2 w vs Re( ) for complex : Im = 1.2, first solution for x.
space since it shows a sharp sensitivity to points of enhanced symmetry. We will return to
these questions at length in the sequel [27].
5. d = 5 operators for B, L violation
5.1. Higgsino exchange mediated violation
In SO(10) B L is a gauge symmetry. Thus the vertices preserve this symmetry and the
vevs are just
leading effects of the spontaneous violation of B L by the superheavy ,
298
Fig. 9. Plot of the threshold corrections to Sin2 w vs Re( ) for complex : Im = 1.2, second solution for x.
299
Fig. 11. Plot of the threshold corrections to Sin2 w vs Re( ) for complex : Im = 1.2, third solution for x.
H
WFM
= hAB AT C2 i B Hi
= 2hAB H A B + H A
B H A B + A B
= 2 2hAB t1 ( u A dB + QA LB ) + t1 QA QB + u A eB dA B
2
2 2hAB h 1 [dA QB + eA LB ] + 2 2hAB h1 [u A QB + A LB ],
(58)
(5) (5)
300
1 T (5)
C2 i1 i5 i1 ,...,i5
5!
(a)
+ 4 2 +
= 2 2 (a)
+
+ 4
WFM
=
(59)
whence
t2 QA QB u A eB + A dB + t2 (QA LB u A dB )
WFM
= 4 2fAB
2
i
+ 4 2fAB
h 2 (dA QB 3eA LB ) h2 (u A QB 3A LB )
3
+ 2(C1 dA QB C2 u A QB ) + 2(E1 dA LB D2 u A LB )
+ 2(D2 eA QB E2 A QB )
A Q
+ 2i(A eA eB G5 A B ) 2 2i F1 eA B
+ 4fAB
Q
B
A LB )
+ (W QA QB + 2P QA LB + 2OL
2it4 (dA B + u A eB ) + 2i 2(K dA eB J1 u A B ) .
(60)
We have suppressed G321 indices and used a sub multiplet naming convention specified
in Section 2 and Table 1 in Appendix A.
In order that the exchange of a higgsino that couples to matter with a given B + L
lead to a B + L violating d = 5 operator in the effective theory at sub GUT energies it
is necessary that it have a nonzero contraction with a conjugate (MSSM) representation
higgsino that couples to a matter chiral bilinear with a B + L different from the conjugate
of the first B + L value. Inspecting the above superpotentials one finds that only {t(1) , t(2) }
and {t(1) , t(2) , t(4) } satisfy this requirement. Terms containing the right handed neutrinos A
must be further processed to integrate out the heavy field A in favor of the light neutrinos
(62)
(63)
and
301
fAB = 4 2fAB
.
hAB = 2 2hAB ,
(64)
This expression and the Clebsches contained in it, as well as the new baryon decay
126 (10, 1, 3) (the same PS multiplet
channel mediated by the triplets (t(4) ) contained in
that contains the Higgs field responsible for the right handed neutrino Majorana mass),
126 mediated decay focused on the multiplets
were given in [11]. Previous work [12] on
(2)
(2)
t , t and found that there was no contribution of t (4) , t(4) in their models. This new
channel nominally strengthens the emergent link between neutrino mass and baryon decay. Note however that t(4) couples only to the RR combinations d + u e and as such its
exchange will contribute only to the RRRR channel which, at least in SO(10), seems [12]
generically suppressed except at very large tan . However the mixing in the triplet mass
matrix could also strengthen the effects of this channel.
5.2. Novel d = 5, (B + L) = 0 operators via superheavy gaugino exchange?
A novel situation apparently arises in this GUT due to exchange of superheavy gaugino
Dirac multiplets that couple to matter both via the gauge Yukawa couplings of their gaugino part and the superpotential couplings of their (126) chiral components to the matter
sector. Such gauginos are not present in the case of SU(5) As is evident from Eq. (60)
2, 1/3], F1 [1, 1, 2], J1 [3, 1, 4/3] (which mix with the
the 126 submultiplet fields E 2 [3,
superheavy SO(10)/SU(5) coset gauginos: see Sections 2 and 3) couple only to terms
containing at least one superheavy neutrino field A . Thus, to leading order in MU1 , the
exchange of such gaugino Dirac multiplets will not lead to any d = 5 operator with 4 light
external fields. However a puzzle remains.
The superheavy neutrinos mix with the usual light neutrinos via Dirac masses. So in the
effective theory one trades them for the light neutrinos by using their equations of motion
Majorana
to leading order in their (inverse) masses (effectively = 2(mDirac
/M
) + ).
The chiral parts E2 , F1 , J1 of the gauge Dirac E, F , J multiplets therefore couple to light
neutrinos and another light matter field with a small coupling O(mDirac /MMajorana ). Exchange of the gaugino Dirac fermion between a gauge Yukawa vertex and a 126 16 16
vertex can lead to effective operators involving 4 light matter fields of which at least one
is a light neutrino and one is antichiral. This appears to violate the usual argument that in
the effective MSSM arising from a SUSY GUT, supersymmetric D terms involving 4 light
(mixed chiral and antichiral) fields must be d 6 or equivalently that the d = 5, B, L vi F . Exchange of SO(10)/SU(5) coset
olating terms are either of form [QQQL]F or [eu u d]
gauginos peculiar to SO(10) however appears to lead to (admittedly suppressed) d = 5
chiralantichiral operators with 4 light fields. These operators arise once the electroweak
scale vev that gives rise to neutrino Dirac masses is turned on. This vev is smaller than MS
and arises after soft SUSY breaking terms are included. In this theory B L is spontaneously broken giving rise to the Majorana mass for conjugate neutrinos (which was used
to eliminate them in favor of the SM neutrinos). Thus perhaps the contradiction is not as
violent as it seems at first. We emphasize that there are no analogous processes in SU(5)
302
SUSY GUTs since there the 12 coset gauginos acquire partners from the purely AM type
24 plets which do not couple to the matter sector.
The couplings of the gauginos of SO(10) to the matter fields are easily computed by
adapting the PS reduction of the SO(10) covariant derivative for the spinor 16 [11]:
A
tA
A t
A
+
LgY = 2ig
2
2
L
+
+
2
2
+
(65)
+ + H.c.
2
The terms carrying B, L are
L(B+L)=0
= 2g L J4 Q + Q J4 L 2g d J4 e + u J4 + e J4 d + J4 u
+ (g/ 2) d X 3 L u E (5) L + e X 3 Q + E 5 Q + d E5 Q + u X3 Q
(X3 e E5 ) + Q (E 5 d X 3 u)
+ (g/ 2) L (X3 d E5 u)
Q
+ .
(66)
There are no X[3, 1, 5/3] sector submultiplets in the 126. Thus we can focus on just
the E[3, 2, 1/3] and the J [3, 1, 4/3] sectors here. As discussed in Section 2, the super 2, 1/3].
heavy gauginos J4 , E5 mix with 126 derived fermions J1 [3, 1, 4/3] and E 2 [3,
Examining Eq. (60) we see that J1 , E2 couple only to operators involving at least one
superheavy field (E1 126 does not mix with Egauginos):
WFM
= 8 2fAB
[E 2 QB + iJ1 u B ]A
= [E 2 QA + iJ1 u A ] f M 1 mD
B + .
(67)
AB
(68)
d E5 Q + Q E5 L E5 u .
2
Only the first terms in (67), (68) are relevant and thus we find the following effective
Lagrangian due to superheavy gaugino exchange:
1 (D)
A B ) ,
m
E 15 2 d C QC (QA B + Q
L(B+L)=2 = 4g f M
(69)
AB
where E 15 2 is essentially the mass of the exchanged gaugino times mixing factors written
compactly in terms of the inverse of the relevant fermion mass matrix (in the E[3, 2, 13 ]
sector). By dressing this with MSSM gauginos we obtain B = L = 1 violating 4Fermi
vertices responsible for processes like
uL dL dL + L .
(70)
303
This is a vertex quite distinct from the higgsino mediated vertices since it involves exchange of massive gauginos between a chiral and an antichiral vertex. It requires non
zero external momenta for the fermions and vanishes in the limit of zero external momenta. Thus the coefficient of the corresponding 4Fermi operators for B violation in the
effective Lagrangian is M D mNucl /MX2 MS2 where MS is the SUSY breaking scale. This
magnitude seems hopelessly suppressed (relative even to gauge boson exchange) to be observable. Nevertheless the contrast of its structure with that of the standard QQQL and
u u d e operators perhaps warrants a more thorough investigation of the conditions for the
possibility of its appearance in the effective theory.
MAB
(71)
44
= 4 2fAB
= 4i 2fAB
.
Similarly the Majorana mass matrix for the left neutrinos A is (Eq. (60)).
11
MAB
O = 8ifAB
= 4 2fAB
O
,
(72)
where O
is the small vev of the SU(2)L triplet in the (10, 3, 1) induced by a tadpole
that arises as a consequence of SU(2)L breaking (see below).
In addition to this there is the Dirac mass which mixes the left and right neutrinos:
(2)
D
mAB
(73)
= 2 2hAB h(1)
2 + 4i 6fAB h2 .
We must make the fine tuning Det H = 0 necessary to keep a pair of Higgs doublets H(1) ,
H (1) (which is to develop the EW scale vev) light. Then these doublets will be the left and
right null eigenstates of the mass matrix H. If the biunitary transformations responsible
for diagonalizing H H and HH are U , U , i.e.,
(1) (2)
Diag mH , mH , . . . = U T HU
(74)
then writing
h(i) = Uij H (j ) ,
h (i) = U ij H (j ) ,
(75)
where H (j ) , H(j ) are the mass eigenstate doublets, the contributions of any coupling in
which h(i) , h (i) enter can be accounted for in the effective MSSM below the heavy thresholds of the GUT just by replacing h(i) i H (1) , h (i) i H (1) where i = Ui1 , i = U i1
and we have numbered the massless doublet pair 1. These components are easily obtained from the normalized null eigenvectors V , V of H H and HH to be Ui1 = Vi , U i1 =
Vi . Thus the neutrino Dirac mass matrix becomes
D
MAB
(76)
= (2 2hAB 1 + 4i 6fAB
2 )vu .
304
To obtain the final formula for the neutrino masses and mixings we must eliminate the
fields which are superheavy and evaluate the tadpole that gives rise to the type II seesaw.
44
= MO O O H 44 O H O
WFM
2
2
44
2 2i 4 4 44 O + 4 4 O
(77)
(4)
44
(4)
since 4 4 1 = 2 3i h (3) , 4 4 2 = 23i h(2) , 441
= 2h
, 2
= 2h , one gets for
the relevant terms
i
i
2
MO O O+ + O 1 + 2i 3 3 4 vd O+ 1 + i2 32 4 vu2 .
2
2
(78)
vu
i
O
= 1 + 2i 32 4
(79)
MO
2
and MO can be read off from Table 1 to be MO = 2(M + (3a p)).
The quark and charged lepton mass matrices are
2
d
if 2 vd ,
M = 2 2h 1 4
3
2
u
M = 2 2h 1 4
if 2 vu ,
3
(80)
(81)
(82)
These formulae are now in a form ready to use for fitting the fermion mass and mixing data
after lifting it via the RG equations of the MSSM to the GUT scale.
7. Discussion and outlook
In this paper we have calculated the complete superheavy spectrum of the minimal supersymmetric GUT along with the gauge and chiral couplings of all MSSM multiplets in
a readily accessible form. Partial calculations of these spectra and couplings [3,11,13,14]
have been published earlier but our method is different from the computer based method
of [3,13,14] and is more complete, especially regarding couplings. Being analytic and explicit it also allows us to trace and resolve discrepancies arising within the computer based
approach. We used the calculated spectra to perform a preliminary scan of the parameter
space of the MSGUT as regards the magnitude of the threshold corrections to two crucial
phenomenological parameters of the MSGUT: the Weinberg angle at low energy and the
mass of the X leptoquark gauge supermultiplet. We obtained a result that is in sharp contrast with expectations in the literature [18] that precision RG calculations in SO(10) are
305
futile. On the contrary we find that the 1loop GUT threshold and gauge two loop contributions are modest but significant. Thus, on the one hand, the basic GUT picture suggested
by the convergence of gauge couplings in the MSSM is in fact not destroyed by the contributions of the large number of superheavy fields. On the other hand extant precision
calculations that ignore threshold effects in SO(10) GUTs seem to be of dubious validity. In
particular the proper RG analysis of the MSGUT taking into account EWRSB, all fermion
masses and GUT threshold effects still remains to be done. This calculation is now being
performed [27]. In view of the other phenomenological successes of renormalizable SUSY
SO(10) [57] and the unforeseen correlations between disparate phenomena like neutrino
oscillations and nucleon decay that have emerged [11,12] the mildness and calculability
of threshold effects in the MSGUT is a most welcome and promising development. Our
preliminary scans of the MSGUT parameter space (whose very feasibilitybased on there
being just one sensitive control parameter ( )is a matter of some astonishment) show
that the threshold effects can potentially narrow down the allowed regions of the MSGUT
parameter space and indicate correlations between the GUT scale and the B L violating scale which can be of crucial significance when cross checking the particle physics
phenomenology against cosmology. We have also argued that above the perturbative unification scale realistic renormalizable SO(10) GUTs are necessarily strongly coupled [8,9].
We have recently reported the results of 2loop calculations of MSGUT RG equations
above the SO(10) restoration scale which we used to show that the strong growth of the
SO(10) coupling above MX cannot be evaded by taking shelter in a weakly coupled fixed
point [30]. On the other hand our work [8,9] has shown that a scenario of a calculable
dynamical symmetry breaking of the GUT symmetry which utilizes the nearly exact supersymmetry at the GUT scale offers rich possibilities for the significance of the new length
scale associated with the condensation of SO(10)/G321 coset gauginos implied by both
holomorphic analysis and by the Konishi anomaly. The present calculations show the way
for crossing the threshold and entering into the SO(10) regime in a controlled way. The
emerging coherence of the low energy phenomenology, B and L violation, perturbative
GUT structures (such as the natural Rparity preservation [20,31], successful seesaw scenarios, leptogenesis, etc.) and exciting hints of deeper mysteries, perhaps unavailable [8,9],
carry to our hopeful nostrils the spoor of a grail perhaps within reach.
Note added
After this paper was posted on the arXive as hepph/0405074 the authors of hepph/0405300 claimed that the mass spectra listed in Appendix A were not internally consistent with the requirements of SU(5) or SU(5) U (1) symmetry (at the special vevs where
p = a = ). However this is incorrect. The mass spectra we derived via a PS decomposition of SO(10) organize straightforwardly and termwise into appropriate SU(5) invariants
for SU(5) invariant vevs as given in Appendix B (added in hepph/0205074 v2). This termwise reorganization of several hundred G123 invariant mass terms into SU(5) invariant
mass terms is a more stringent consistency test than the tests of hepph/0405300 which are
based on traces and determinants and valid only for their conventions. The phase conventions and field normalizations of hepph/0405300 are quite different from our work. Thus
the blind application of their trace and determinant consistency tests to our results can
306
not but fail. We maintain unit field normalizations throughout by using only unitary field
redefinitions of fields with canonical kinetic terms. Finally our results for chiral spectra
also coincide, upto minor convention related adjustments, with those obtained in a parallel
computation reported in [13].
Finally we stress that our method [11] yields all coupling coefficients between both
spinor and tensor irreps and not just the tensor irrep ones relevant for masses and symmetry
breaking which were obtained using the method of [10]. Moreover we note that the most
complex of the mass matrices given here namely those of the Higgs doublets and triplets
relevant for proton decay were already derived by us in hepph/0204097 v2 (2003) [11].
Further note added
After version 2 of this paper (including the SU(5) reorganization given in Appendix B)
was accepted for publication the authors of hepph/0405300 issued yet another preprint
(hepph/0412348 v1) this time claiming that although our results pass the SU(5) reorganization test for = = 0 they failed to do so for = = 0 and that the counting
of Goldstone modes and distinct mass eigenvalues was, in their opinion incorrect. Further
they claimed that our results were inconsistent since they failed to pass certain trace and
hermiticity tests that they had applied successfully to their own results. All these claims
are incorrect. Our results in fact pass all three tests. We have issued a preprint showing this
explcitly [33]. Here we only remark that once the superHiggs effect for SO(10) MSSM
has been verified it is scarcely feasible that the GoldstoneHiggs counting could fail for
SO(10) SU(5) since the latter is a special case of the same spectra! However the reader
can easily check that the SU(5) singlet and 10plet mass matrices have zero determinant confirming that the required Goldstone supermultiplets 1 + 10 + 10 are present. The
demonstration that the trace constraints and hermiticity tests of hepph/0205300 are also
satisfied is also straightforward once proper account is taken of the difference in the phase
conventions of the two calculations. Details may be found in [33]. Finally as this paper goes
to press the authors of hepph/0412348 v1 have reissued the preprint hepph/0412348 v2
in which all the claims of the inconsistency of our results are totally retracted.
Acknowledgements
It is a pleasure for C.S.A. to acknowledge much correspondence and collaboration with
B. Bajc, A. Melfo, G. Senjanovic and F. Vissani while calculating the chiral spectra reported here. C.S.A. is also grateful to B. Bajc and Prof. D.R.T Jones for correspondence
related to RG analysis, to Sumit Kumar for technical help with Appendix B and to the
High Energy Group of the Abdus Salam ICTP, Trieste for hospitality while completing
Appendix B. A.G. thanks H.R.I., Allahabad for its facilities and hospitality. This work was
done under Project SP/S2/K07/99 of the Department of Science and Technology of the
Government of India.
Appendix A. Tables of masses and mixings
In this appendix we collect our results for the chiral fermion/gaugino states, masses
and mixing matrices for the readers convenience. Apart from the discussion of gauge
307
multiplet masses our results have been obtained in parallel with and are compatible with
those of [13], which, however, are computed with a different normalization for the ,
fields resulting in a difference between the mass and Yukawa coupling parameters M, of
these multiplets in the two starting actions. Moreover certain minor phase differences also
exist between the definitions of representative states used by them and our definitions for
the same states (which follow directly from our consistent definitions of PS tensors from
SO(10) submultiplets). Mixing matrix rows are labeled by barred irreps and columns by
unbarred. Unmixed cases (i) are given as Table 1.
Table 1
(i) Masses of the unmixed states in terms of the superheavy vevs. The SU(2)L contraction order is always F F .
The primed fields defined for SU(3)c sextets maintain unit norm. The absolute value of the expressions in the
column Mass is understood
Field [SU(3), SU(2), Y ]
1, 4]
A[1, 1, 4], A[1,
C1 [8, 2, 1], C 1 [8, 2, 1]
C2 [8, 2, 1], C 2 [8, 2, 1]
2, 7 ]
D1 [3, 2, 73 ], D 1 [3,
3
2, 7 ]
D2 [3, 2, 73 ], D 2 [3,
3
2, 1 ]
E1 [3, 2, 13 ], E 1 [3,
3
1, 8 ]
3,
K[3, 1, 83 ], K[
3
1, 2 ]
6,
L[6, 1, 23 ], L[
3
1, 8 ]
6,
M[6, 1, 83 ], M[
3
1, 4 ]
N [6, 1, 43 ], N [6,
3
PS fields
Mass
44
(R+)
, 44(R)
2
2
,
1
2
,
1
2
4 ,
4 2
1
4 ,
4 2
1
4 ,
2
4 1
4
4(R) , (R+)
(R0)
( , (R0) )
= , =
=
2
(R+)
( (R+) , (R) )
(R)
( , (R+) )
44
44(L) (L)
,
2
2
4
4(L)
, (L)
(L) ,
(L)
4
(R+) , 4(R)
(15)
(L)
(L)
4
(L)
, 4(L)
2(M + (p + 3a + 6))
2(M + (a + ))
2(M + (a ))
2(M + (a + ))
2(M + (a + 3))
2(M + (a ))
2(M + (a + p + 2))
2(M + (p a))
2(M + (p a + 2))
2(M + (p a 2))
2(M + (3a p))
2(M + (a p))
2(M (a + p))
2(m + (p + a + 4))
2(m + (2a p))
2(m (a + p))
2(m (p a))
44
44 2 1
,
2
2
( , )
1 2
( , )
2 1
(R+)
(R)
2(m + 3(a + ))
2(m + ( a))
2(m (a + ))
2(m + (p a))
308
Table 2
Index values for the 26 different chiral multiplet types (used in the threshold corrections). Except for Q, R, S all
other reps come in complex pairs. SW = 4S1 9.6S2 + 5.6S3 , SX = 5S1 + 3S2 8S3 are the combinations that
enter the threshold corrections to Sin2 W and to Log10 MX
Field [SU(3), SU(2), Y ]
{S3 , S2 , S1 }
SW
SX
A[1, 1, 4]
{0, 0, 12/5}
9.6
12
B[6, 2, 5/3]
C[8, 2, 1]
D[3, 2, 7/3]
E[3, 2, 1/3]
F [1, 1, 2]
G[1, 1, 0]
{5, 3, 5}
{6, 4, 12/5}
{1, 3/2, 49/10}
{1, 3/2, 1/10}
{0, 0, 3/5}
{0, 0, 0}
19.2
4.8
10.8
8.4
2.4
0
6
24
21
3
3
0
h[1, 2, 1]
I [3, 1, 10/3]
J [3, 1, 4/3]
K[3, 1, 8/3]
L[6, 1, 2/3]
M[6, 1, 8/3]
3.6
22.8
6
15.6
15.6
39.6
3
21
0
12
18
12
N [6, 1, 4/3]
O[1, 3, 2]
P [3, 3, 2/3]
Q[8, 3, 0]
R[8, 1, 0]
S[1, 3, 0]
{5/2, 0, 8/5}
{0, 2, 9/5}
{3/2, 6, 3/5}
{9, 16, 0}
{3, 0, 0}
{0, 2, 0}
20.4
12
46.8
103.2
16.8
19.2
12
15
9
24
24
6
t[3, 1, 2/3]
U [3, 3, 4/3]
V [1, 2, 3]
W [6, 3, 2/3]
X[3, 2, 5/3]
Y [6, 2, 1/3]
{1/2, 0, 1/5}
{3/2, 6, 12/5}
{0, 1/2, 27/10}
{15/2, 12, 6/5}
{1, 3/2, 5/2}
{5, 3, 1/5}
3.6
39.6
6
68.4
1.2
0
3
18
15
18
9
30
Z[8, 1, 2]
{3, 0, 24/5}
36
(m a)
2
.
R=2
2 m + (p a)
mR
2
p
p
a
= R  = 2m 1 +
+ 2 2 .
2
2
(A.1)
(A.2)
309
(15)
(15) 44
(b) [1, 2, 1](h 1 , h 2 , h 3 , h 4 ) [1, 2, 1](h1 , h2 , h3 , h4 ) (H2 , 2
, 2
, )
2
441
, (15)
, )
(H 1 , (15)
1
1
2
MH
+ 3( a)
3( + a)
3( + a)
0
(2M + 4(a + ))
0
.
H=
3( a) (2M + 4(a ))
0
2 3
0
2m + 6( a)
2 3
The above matrix is to be diagonalized after imposing the fine tuning condition Det H = 0
to keep one pair of doublets light.
,
1, 2 ](t1 , t2 , t3 , t4 , t5 ) [3, 1, 2 ](t1 , t2 , t3 , t4 , t5 ) (H 4
4
(c) [3,
(a) , (a) , R0 ,
3
3
(a)4
) (H4
, 4(R0)
, 4
)
,
, 4(a)
(R)
4(R+)
(a + p)
(p a)
2M
2M
4 2i
2i
MH
(p a)
T = (p + a)
2 2i
i
2 2i
0
4 2i
2M + 2p + 2a
2 2
0
2i
2 2
2m 2(a + p 4)
(15)
44 / 2,
44((R+) / 2,
(a) [1, 1, 0](G1 , G2 , G3 , G4 , G5 , G6 ) (, (15) , (R0) , (R)
(R0) (15)
( 2
3 )/ 5)
i
i
m
0
6
0
2
0
m + 2a
2 2
i 32
i 32
6 22 m + (p + 2a)
i 3
i 3
G = 2
i
i 32
i 3
0
M + (p + 3a 6)
2
i
i 3
M + (p + 3a 6)
0
2 i 32
5g
2
5g
2
0
5g
2
5g
2
0
4
(a)4
4 , (s) , (a) ,
2 ) (
1 )
2
4
1
4
1
2(M + (a 3))
2 2i
2i
2i 2
2(m + (a ))
2 2
E =
2i
2 2
2(m )
ig 2
2g(a )
g 2( p )
ig 2
2g(a )
.
2g( p )
0
(A.3)
310
(15)
(R+) , (R+) )
2(M + (p + 3a))
F =
2i 3
g 2
2i 3
2(m + (p + 2a))
24ig
g 2
24ig ) .
0
(A.4)
(R0)
ig 2
2(M + (a + p 2))
2
2 2
2
2(m + a)
2 2
2ig 2a
J =
.
2 2
2 2
2(m + (a + p))
4ig
ig 2
2 2iga
4ig
0
(A.5)
(s)4
, 2 )
4
2 4
2(m + (a + ))
X=
2 2
2g(a + )
2 2
2(m + )
2g( + p )
2g(a + )
2g( + p ) .
0
(a)4
, 1
(s)
, 1 ) (4
,
2
(A.6)
(B.1)
where the unbroken symmetry includes SU(5) the MSSM labeled mass spectra and couplings given in Appendix A do indeed reassemble into SU(5) invariant form. For the mass
spectra this is fairly straightforward to check and is reported explicitly below. A similar
calculation [32] for the superpotential couplings is much more tedious but furnishes an
SO(10) SU(5) U (1) analog of the SO(10)PS Clebsches reported here.
The decomposition of the chiral multiplets of the MSGUT into SU(5) U (1) multiplets and of those into MSSM multiplets (named as per the alphabetic convention of
Appendix A) is given below. The only complication is that certain MSSM multiplet types
occur in several copies and (orthogonal) mixtures of these are present in the different SU(5)
mutiplets. Thus, for instance, the 210 contains a 24 and a 75 of SU(5) both of which contain mixtures of the G123 multiplets R1 (8, 1, 0) and R2 (8, 1, 0). These mixtures must be
orthogonal and must be precisely the eigenstates of the mass matrices in this G123 sector
which have the same masses as the rest of the G123 submultiplet sets within the 24plet and
311
75plets as two wholes. The fact that this follows in every case from our results appears to
confirm their reliability. The decompositions we need are
H = 10 = 51 + 5 1 ,
= 126 = 15 (G4 ) + 5 1 + 103 + 153 + 451 + 501 ,
1, 2 ,
51 = h 3 (1, 2, 1) + t3,4 3,
3
4
1
103 = F1 (1, 1, 2) + J1 3, 1,
+ E2 3, 2,
,
3
3
2, 1 + N 6,
1, 4 ,
153 = O(1, 3, 2) + E 1 3,
3
3
2
2
2, 7
1, 8 + D 1 3,
451 = h3 (1, 2, 1) + t3 3, 1,
+ P 3, 3,
+ K 3,
3
3
3
3
2
1,
+ C1 (8, 2, 1),
+ L 6,
3
1, 2 + D2 3, 2, 7 + W 6, 3, 2
501 = A(1, 1, 4) + t3,4 3,
3
3
3
8
1,
+ M 6,
+ C 2 (8, 2, 1),
3
= 126 = 15 (G5 ) + 51 + 103 + 153 + 451 + 501 ,
2
51 = h2 (1, 2, 1) + t2,4 3, 1, ,
3
4
2, 1 ,
103 = F1 (1, 1, 2) + J1 3, 1,
+ E 2 3,
3
3
1
4
+ N 6, 1, ,
153 = O(1,
3, 2) + E1 3, 2,
3
3
2
3, 2 + K 3, 1, 8 + D1 3, 2, 7
451 = h 2 (1, 2, 1) + t2 3, 1,
+ P 3,
3
3
3
3
2
+ C 1 (8, 2, 1),
+ L 6, 1,
3
2, 7 + W 6,
3, 2
1, 4) + t2,4 3, 1, 2 + D 2 3,
501 = A(1,
3
3
3
8
+ C2 (8, 2, 1),
+ M 6, 1,
3
= 210 = 10 + 54 + 5 4 + 102 + 102 + 240 + 402 + 402 + 750 ,
10 = G1,2,3 ,
2
54 = h4 (1, 2, 1) + t5 3, 1, ,
3
2
1,
5 4 = h 4 (1, 2, 1) + t5 3,
,
3
312
1, 4 + E3,4 3, 2, 1 ,
102 = F2 (1, 1, 2) + J2,3 3,
3
3
4
2, 1 ,
+ E 3,4 3,
102 = F2 (1, 1, 2) + J2,3 3, 1,
3
3
5
2, 5
240 = (1, 1, 0)G1,2,3 + S(1, 3, 0) + X1,2 3, 2,
+ X 1,2 3,
3
3
+ R1,2 (8, 1, 0),
1
1, 4 + U 3,
3, 4
+ J2,3 3,
402 = V (1, 2, 3) + E3,4 3, 2,
3
3
3
1
2,
+ Z(8, 1, 2) + Y 6,
,
3
1
4
4
+ J2,3 3, 1,
+ U 3, 3,
402 = V (1, 2, 3) + E 3,4 3, 2,
3
3
3
1
1, 2) + Y 6, 2, ,
+ Z(8,
3
10
1, 10 + X1,2 3, 2, 5
75 = (1, 1, 0)G1,2,3 + I 3, 1,
+ I 3,
3
3
3
5
5
5
2,
2,
+ B 6, 2,
+ B 6,
+ R1,2 (8, 1, 0)
+ X 1,2 3,
3
3
3
+ Q(8, 3, 0).
(B.2)
If we insert a = = p in the mass matrices of Appendix A we find that, after diagonalizing the mass matrices of the submultiplets that mix, the resultant spectra group
precisely as indicated by the decompositions above with all the subreps of a given SU(5)
irrep obtaining the same mass. One obtains the SU(5) invariant mass terms
2(M + 10p)1 1 + 2(M + 4p)5 5 + 2(M 2p)50 50
+ 2(M + 4p)10 10 + 2(M + 2p)15 15 + 2M45 45
+ MH 5 H 5H + (m + 6p)(1 )2 + 2(m + 6p)5 5 + 2(m + 3p)10 10
+ (m + p)(24 )2 + 2m40 40 + (m 2p)(75 )2
+ 2 3 (5 5 + 10 10 ) + (5 5 + 10 10 )
+ 2 3p( 5 5H + 5 H 5 ) + 2i 5( 1 1 1 1 )
+ 5 5H + 5 H 5 ,
(B.3)
where every SU(5) invariant has been normalized so that the individual G123 subrep
masses can be read off directly from the coefficient of the invariant for complex SU(5)
representations which pair into Dirac supermultiplets and is 2 times the coefficient for
the real representations which remain unpaired Majorana/chiral supermultiplets. For a =
= p, = = 0 the 20 Goldstone supermultiplets G, J , J, F , F , E, E of the coset
SO(10)/SU(5) U (1) remain heavy as they should since they are eaten in the spontaneous breaking SO(10) SU(5) U (1) while the 12 fields in the {X3 , X 3 } multiplets
313
lose their mass terms with {X1,2 , X 1,2 } since they form part of the unbroken SU(5) gauge
supermultiplet. When a = = p, i.e., for flipped SU(5), the roles of the {X3 , X 3 } and
{E5 , E 5 } gauge multiplets are interchanged, with the Es remaining massless and the Xs
becoming heavy, so that one obtains the SU(5) invariant groupings corresponding to the
flipped SU(5) U (1) SO(10) embedding. Note that this successful SU(5) reassembly
is a much more finegrained consistency test than any overall trace or determinant test.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
[33]
Abstract
Generic heterotic Mtheory compactifications contain fivebranes wrapping nonisolated genus
zero or higher genus curves in a CalabiYau threefold. Nonperturbative superpotentials do not depend on moduli of such fivebranes. We show that fluxes and nonperturbative effects can stabilize
them in a nonsupersymmetric AdS vacuum. We also show that these fivebranes can be stabilized in
a dS vacuum, if we modify the supergravity potential energy by FayetIliopoulos terms. This allows
us to stabilize all heterotic Mtheory moduli in a dS vacuum in the most general compactification
scenarios. In addition, we demonstrate that, by this modification, one can create an inflationary potential. The inflationary phase is represented by a fivebrane approaching the visible brane. We give
a qualitative argument how extra states becoming light, when the fivebrane comes too close, can
terminate inflation. Eventually, the fivebrane hits the visible brane and disappears through a small
instanton transition. The postinflationary system of moduli has simpler stability properties. It can be
stabilized in a dS vacuum with a small cosmological constant.
2005 Elsevier B.V. All rights reserved.
PACS: 11.25.Mj
1. Introduction
Recently, there has been a considerable interest in cosmological aspects of string theory
(see [1] for a recent review). In [26], several methods of producing de Sitter (dS) vacua
Email address: evgeny@sns.ias.edu (E.I. Buchbinder).
05503213/$ see front matter 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.nuclphysb.2005.01.015
315
in string compactifications were presented. In [2,3,5], it was suggested that various corrections to the supergravity potential energy can raise a supersymmetric antide Sitter (AdS)
vacuum to a metastable dS vacuum. In [4], based on the earlier work [7], a dS vacuum was
created by balancing various exponential superpotentials and in [6], it was argued that a
dS vacuum can be created by taking into account higher order corrections to the moduli
Kahler potential. In addition, in [8,9], it was studied how effects of gravity and quantum
particle production could trap moduli at enhanced symmetry points. Furthermore, a substantial progress has been achieved towards inflation in string theory [1030]. In these
models, inflation was studied within the context of Dbranes. Under certain conditions the
Dbrane modulus can be treated as an inflaton. However, all these models usually have two
common problem. Inflation is often realized in compactification or brane world scenarios
which do not correspond to realistic fourdimensional physics. The other problem is that,
in addition to the inflaton, there are, usually, other moduli whose stabilization could be a
problem and whose presence can violate the slow roll conditions.
In this paper, we would like to explore the possibility of creating an inflationary potential within the framework of strongly coupled heterotic string theory, or heterotic Mtheory
[3133]. Such compactifications have a lot of attractive phenomenological features (see
[34] for a review on phenomenological aspects of Mtheory). Various GUT and Standard
Modellike theories were obtained from heterotic compactifications on CalabiYau threefold [3538]. For example, in [38], vector bundles on CalabiYau manifolds with Z3 Z3
homotopy group were constructed. A compactification on such a manifold can lead to the
Standard Model with suppressed nucleon decay. The actual particle spectrum in such theories was studied in [3941]. One more attractive feature of such compactifications is that
it is possible to stabilize moduli in a phenomenologically acceptable range [5,42]. The
set of moduli considered in [5,42] was very general. Nevertheless, it was not complete.
In [5,42], it was assumed that the CalabiYau threefold had enough isolated genus zero
curves to stabilize all h1,1 moduli. It was also assumed that the fivebranes in the bulk
wrapped only isolated genus zero curves. Even though such compactifications can certainly exist, a generic compactification with h1,1 greater than one contains various, not
necessarily isolated genus zero, cycles as well fivebranes wrapped on them. In this case, it
is quite possible that not all h1,1 moduli can be stabilized by methods presented in [5,42].
The moduli of a fivebrane wrapped on a nonisolated genus zero curve or a higher genus
curve cannot be stabilized by methods of [5,42] either. In this paper, we add these new
moduli. We show that this new additional h1,1 moduli can be stabilized in a supersymmetric AdS minimum by the slight modification of ideas of [5,42]. The fivebrane moduli
cannot be stabilized this way. Surprisingly, we find that they can be stabilized in a nonsupersymmetric AdS minimum. Of course, what we really mean by this is that the system
of moduli containing the moduli of this new fivebrane admits a nonsupersymmetric AdS
vacuum. However, the potential energy has one more minimum when the fivebrane coincides with the visible brane. A heterotic Mtheory vacuum can contain several fivebranes
wrapped on nonisolated genus zero or higher genus curves. Those which are located relatively far away from the visible brane will be stabilized. On the other hand, those which
are located close enough to the visible brane will roll towards it and, eventually, collide
with it. We show that these fivebranes can be stabilized as well by balancing the supergravity potential energy against the FayetIliopoulos terms [43] induced by an anomalous
316
U (1) gauge group in the hidden sector. This shows that the most general set of heterotic
Mtheory moduli can be stabilized. Furthermore, the cosmological constant can be positive and fine tuned to be very small. By balancing the supergravity potential energy against
FayetIliopoulos terms, it is also possible to create a positive potential satisfying the slow
roll conditions and treat the fivebrane translational modulus as an inflaton. Inflation takes
place when the fivebrane approaches the visible brane. However, this potential has one
negative feature. It has a vanishing first derivative when the fivebrane coincides with the
visible brane. This means that it takes infinite time for the branes to collide. This also means
that the primordial fluctuations will become infinite. On the other hand, at very short distances, one cannot trust the lowenergy field theory because new states are expected to
become massless. At the present time, physics at short distances in heterotic Mtheory is
not known. Nevertheless, we present an argument how these new state can terminate inflation before the fluctuations became too big. Once the fivebrane hits the visible brane, it
gets dissolved into it and turns into new moduli of the vector bundle, socalled transition
moduli, studied in [44,45]. This process is called small instanton transition [4649]. Thus,
the postinflationary phase does not have the inflaton but has extra moduli of the vector
bundle. These moduli are easier to stabilize [42]. Therefore, the new system of moduli can
be stabilized, whereas during inflation this was not the case. In addition, we argue that, after
a small instanton transition, generically, the cosmological constant changes. It is possible
to decrease the cosmological constant and, by fine tuning, make it consistent with observations. Let us point out that, though, besides the inflaton, there are various other moduli
during inflation, they are all taken into account. The potential energy has a minimum in
all the other directions. Therefore, dynamically, one expects that all these moduli will roll
very fast in their minimum leaving the fivebrane to roll slowly towards the visible brane.
This paper is organized as follows. In Section 2, we discuss the system of moduli in
compactifications with h1,1 greater than one. The reason is twofold. First, we would like to
obtain more general results on moduli stabilization. In particular, we would like to stabilize
the h1,1 moduli that do not have a nonperturbative superpotentials and, hence, cannot
be stabilized by methods of [5,42]. Second, before we begin to study potentials for the
fivebrane, whose stability properties are more complicated, it is important to understand
how the remaining moduli of the system are stabilized. The system of (complex) moduli
considered in this section includes the complex structure moduli, the volume modulus, two
h1,1 moduli and the moduli of the fivebrane wrapped on an isolated genus zero cycle.
One of the two h1,1 moduli is assumed to be associated with an isolated genus zero curve
and, hence, has a nonperturbative superpotential. The other one is associated with a nonisolated genus zero curve or a higher genus curve. The nonperturbative superpotential does
not depend on this modulus. By the slight modification of ideas of [5,42], we show that this
system can be stabilized in a supersymmetric AdS vacuum. The crucial moment is that, if
h1,1 is greater that one, it is possible to choose one of the contributions to the tension of
the hidden brane to be positive without having the gauge coupling constant stronger in the
visible sector. The system of moduli can be supplemented by vector bundle moduli [45].
They can be stabilized as well [42]. For simplicity, we will ignore them. In Section 3, we
add one more fivebrane to the system. This fivebrane is wrapped on a nonisolated genus
zero or higher genus curve. Approximately, we can treat the rest of the moduli fixed and
consider an effective potential for the remaining fivebrane modulus. This potential is very
317
difficult to analyze analytically. A graphical analysis shows that, generically, it has a nonsupersymmetric AdS minimum. Nevertheless, if a fivebrane was originally located close
to the visible brane, it will roll towards it. In the rest of the paper, we concentrate only on
dynamics of rolling fivebranes. We modify this effective potential with FayetIliopoulos
terms and show that addition of a FayetIliopoulos term in the hidden sector can stabilize
a rolling fivebrane. We also show that the cosmological constant in such a vacuum can
be positive and small. The results from Sections 2 and 3 provide stabilization of heterotic
moduli in the most general setup in a vacuum with a positive cosmological constant. They
also indicate that it is conceivable to obtain two distinct dS vacua. One of them is the
lift of the nonsupersymmetric AdS minimum. The other one is created by addition of a
FayetIliopoulos term in the hidden sector. In Section 4, we argue that, by balancing the
supergravity potential energy and FayetIliopoulos terms, it is also possible to construct
a potential with inflationary properties. One of the slow roll parameters turns out to be
naturally much less than one. The other one can be (not necessarily fine) tuned to be much
less than one. We also discuss the amount of inflation and primordial fluctuations. In the
last subsection of Section 4, we discuss how the system can escape from inflation at very
short distances. We give a qualitative argument how the appearance of new light states in
the field theory can provide such an escape. In Section 5, we discuss the postinflationary
phase. After inflation, the fivebrane hits the visible brane and disappears through a small
instanton transition. The new system of moduli does not contain the fivebrane but has
extra vector bundle moduli. Unlike the fivebrane modulus, these moduli can be stabilized.
We show that the new system of moduli can be stabilized in a vacuum with a positive
cosmological constant which can be fine tuned to be very small.
In this paper, we work in the context of strongly coupled heterotic string theory [31,32]
compactified on a CalabiYau threefold [33,50]. To one of the orbifold fixed planes we will
refer as to the visible brane (or the visible sector). To the other one we will refer as to the
hidden brane (or the hidden sector). Such compactifications also allow fivebranes wrapped
on holomorphic cycles in the CalabiYau manifold and parallel to the orbifold fixed planes.
Moduli stabilization in this theory was performed in a relatively general setting in [5,42].
Nevertheless, it was assumed in [5,42] that the CalabiYau manifold has enough isolated
genus zero curves to stabilize all the h1,1 and fivebrane moduli. In this paper, we would
like to consider a more complicated setup when the CalabiYau threefold has twocycles,
one represented by isolated genus zero curve and the other one by curves of a different type.
They could be either nonisolated genus zero curves or curves of a higher genus. In both of
these two cases, no nonperturbative superpotential for the corresponding h1,1 modulus can
318
be generated by string or, more precisely, open membrane instantons [51,52].1 This also
means that one cannot generate a nonperturbative superpotential for moduli of a fivebrane
wrapped on such a cycle. However, compactifications on a CalabiYau manifold with h1,1
represented only by isolated genus zero cycles are, of course, very restrictive. A generic
compactification scenario involves a CalabiYau manifold with nonisolated genus zero or
higher genus cycles and fivebranes wrapped on such cycles. As an example, consider a
CalabiYau manifold elliptically fibered over the Hirzebruch surface Fr , r = 0, 1, . . . . The
Hirzebruch surface Fr is a P1 bundle over P1 . We denote the class of the base of this bundle
by S and the class of the fiber by E. These CalabiYau threefolds are simply connected
and admit a (generically unique) global holomorphic section which we denote by . For
such a manifold, generically, we have
h1,1 = 3
(2.1)
E,
F.
(2.2)
Here is the projection map from the threefold onto the base Fr and F is the class of the
elliptic fiber. The curves S and E have genus zero. The curve F has genus
one. The curve S has a selfintersection r and, thus, is an isolated genus zero
curve for r > 0. It is a nonisolated genus zero curve for r = 0. The curve S has a
selfintersection zero for any r and, thus, is nonisolated genus zero curve. Therefore, it is
important to stabilize the h1,1 moduli corresponding to nonisolated genus zero or higher
genus curves. It is also important to understand whether or not it is possible to stabilize fivebranes wrapped on such cycles. In this paper, for simplicity, we consider the case h1,1 = 2.
We will assume that there is one isolated genus zero curve and one curve of a different
type. The generalization to the case involving many curves isolated curves of various types
is conceptually straightforward but technically more difficult.
At this point, we would like to make a remark. It may happen that the pullback of
more than one harmonic form I onto a given isolated curve is nonzero. As a result,
the nonperturbative superpotential associated with this isolated curve may depend on the
linear combination of more than one h1,1 modulus. In particular, it may depend on all
h1,1 moduli. In this case, all h1,1 moduli can be stabilized by methods presented in [42].
However, one might expect that, generically, there can be h1,1 moduli of two sorts, those
that appear in the nonperturbative superpotential and those which do not. In this paper, we
simply assume that our compactification has one modulus of each sort. In this section, we
will not consider fivebranes wrapped on nonisolated genus zero or higher genus curves.
We will add such a fivebrane in the next section.
The system of moduli that we would like to consider in this section includes the following complex moduli
S,
T 1,
T 2,
Y,
Z .
(2.3)
1 The statement that strings on nonisolated genus zero curves do not contribute to the nonperturbative su
perpotential was conjectured by Witten [51]. The author is very grateful to Edward Witten for discussions on this
issue.
319
(2.4)
where is the axion. The moduli T 1 and T 2 are the h1,1 moduli. They are defined as
follows [50,53]
T I = RbI + ip I ,
I = 1, 2,
(2.5)
bI
where R is the size of the eleventh dimension, are the Kahler moduli of the CalabiYau
threefold and p I s come from the components of the Mtheory threeform C along the
interval and the CalabiYau manifold. The moduli bI are not all independent. They satisfy
the constraint
2
dI J K bI bJ bK = 6,
(2.6)
I,J,K=1
(2.7)
CY
The constraint (2.6) reduces the number of independent bmoduli by one. We will take
T 1 to correspond to the area of the isolated genus zero curve and T 2 to the area of the
remaining curve. Y is the modulus of the fivebrane wrapped on the isolated genus zero
curve. In this case, there is only one fivebrane modulus [54], whose real is the position of
the fivebrane in the bulk
p1
,
Y=y +i a+
(2.8)
Rb1
where a is the axion arising from dualizing the threeform field strength propagating on
the fivebrane worldvolume. At last, by Z we denote the complex structure moduli. The
actual number of them is not relevant for us. A generic heterotic compactification contains
also instanton moduli [45]. Their stabilization was considered in [42]. In this section, for
simplicity, we will ignore them. They can be added and treated as in [42]. However, we
will come back to them in the last section. The moduli V , T 1 , T 2 and y are assumed to be
dimensionless normalized with respect to the following reference scales
1/6
(2.9)
R 1.
(2.10)
(2.11)
320
where
KZ = ln i ,
(2.12)
and
ln dI J K T I + T I T J + T J T K + T K
KS,T ,Y = ln(S + S)
2
(Y + Y)
+ 25
.
1 + T 1 )
(S + S)(T
(2.13)
T5 v5 ()2
,
2
MPl
(2.14)
where v5 is the area of the cycle on which the fivebrane is wrapped and T5 is
1 2/3
T5 = (2)1/3
,
2
211
(2.15)
(2.17)
(2.18)
(2.19)
CY
where G is the Mtheory fourform flux. The order of magnitude of Wf was estimated in
3 . In fact, this is flexible. The super[42] and was found to be, generically, of order 108 MPl
potential Wf may receive certain higher order corrections from ChernSimons invariants.
In [61] it was argued that these ChernSimons invariants can reduce the order of magnitude
of Wf .
By Wg we denote the superpotential induced by a gaugino condensate in the hidden
sector [6265]. A nonvanishing gaugino condensate has important phenomenological consequences. Among other things, it is responsible for supersymmetry breaking in the hidden
sector. When that symmetry breaking is transported to the observable brane, it leads to
321
soft supersymmetry breaking terms for the gravitino, gaugino and matter fields [6669].
See [70] for a good review on gaugino condensation in string theory. This superpotential
has the following structure
Y2
(2) 1
(2) 2
S + 1 T + 2 T 1 .
T
3
Wg = hMPl
exp
(2.20)
The order of magnitude of h is approximately 106 [64]. The coefficient is related to the
coefficient b0 of the oneloop betafunction and is given by
=
6
.
b0 GUT
(2.21)
For example, for the E8 gauge group 5. The coefficients I(2) represents the tension
(up to the minus sign) of the hidden brane measured with respect to the Kahler form I
(2)
I =
11 2/3
1
I tr F (2) F (2) tr R R ,
16vCY 4
2
(2.22)
CY
where F (2) is the curvature of the gauge bundle on the hidden brane. Similarly, the coefficient is the tension of the fivebrane. It is given by [71]
2 2 11 2/3
= 2/3
1 W,
4
v
CY
(2.23)
CY
where W is the fourform Poincar dual to the holomorphic curve on which the fivebrane
(2)
is wrapped. Generically both I and are of order one. In fact, from Eqs. (2.14), (2.15)
and (2.23) it follows that
5 .
(2.24)
Let us note the following fact which will be crucial for stabilization of T 2 . If h1,1 = 1,
apparently, it is important to have (2) positive (and, correspondingly, the tension negative)
in the hidden sector. This, in particular, happens when the bundle on the hidden brane is
trivial. The reason is that the quantity
(2)
Y2
Re S
I T I + 1
T
(2.25)
represents the inverse square of the gauge coupling constant in the hidden sector
Furthermore, the quantity
Re S
I
(1)
I T I
+T
Y2
1 1
T
1
.
2
ghidden
(2.26)
322
represents the inverse square of the gauge coupling constant in the visible sector
Here (1) is the tension (up to the minus sign) of the visible brane
11 2/3
1
(1)
(1)
(1)
=
tr F F tr R R ,
16vCY 4
2
1
.
2
gvisible
(2.27)
CY
where F (1) is the curvature of the gauge bundle on the visible brane. If, for example,
h1,1 = 1 and there are no fivebranes, we have
1
= Re S (2) T
(2.28)
2
ghidden
and
1
2
gvisible
= Re S (1) T .
(2.29)
(2.30)
(2) = (1) .
(2.31)
sets
Now it is clear that if (2) < 0, the gauge coupling constant in the hidden sector is weaker
that the gauge coupling constant in the visible sector and the whole assumption about the
gaugino condensation in the hidden sector breaks down. It is unlikely that this statement
changes when fivebranes are included. However, when h(1,1) is greater than zero, there
(2)
is nothing wrong with having some I s negative. It is still possible to keep the gauge
coupling constant stronger in the hidden sector. We will assume that
(2)
(2.32)
(2)
(2.33)
1 > 0
and
2 < 0.
It is important to note that the quantity given by Eq. (2.25) must be positive. This means
that the superpotential (2.20) cannot be trusted for large values of the interval size R. One
should expect that higher order corrections to the combination (2.25) will make the gauge
coupling constant 2 1 well defined for large values of R. Partial support for this comes
ghidden
from [72,73].
The last contribution to the superpotential that we have to discuss is the nonperturbative
superpotential Wnp [52,7482]. Such a superpotential is induced by open membranes
wrapped on an isolated genus zero curve. Therefore, it depends on the h1,1 modulus T 1
and on the fivebrane modulus Y. However, it does not depend on the h1,1 modulus T 2 .
The nonperturbative superpotential has three parts
Wnp = Wvh + Wv5 + W5h .
(2.34)
323
Wvh is induced by a membrane stretched between the visible and the hidden branes. It
behaves as
Wvh e T .
1
(2.35)
Wv5 is induced by a membrane stretched between the visible brane and the fivebrane. It
behaves as
Wv5 e Y .
(2.36)
At last, W5h is induced by a membrane stretched between the fivebrane and the hidden
brane. It behaves as
W5h e (T
1 Y)
(2.37)
(2.38)
where vi is the reference area of the isolated curve. Generically, is much bigger than
one. As in [5,42], we will assume that the fivebrane is close to the hidden sector. It was
argued in [5,42] that only in this case it is possible to stabilize the size of the interval in
a phenomenologically acceptable range. Therefore, the contributions Wvh and Wv5 decay
very fast and we have
3
ae (T
Wnp = W5h = MPl
1 Y)
(2.39)
1
0,
Im Y 0.
(2.40)
(2)
The imaginary part of the linear combination S 2 T 2 is stabilized in such a way that the
superpotentials Wf and Wg are out of phase. Similarly, Wf and Wnp are also out phase.
We already took this into account in Eq. (2.18) by putting the minus sign in appropriate
places. Unfortunately, the superpotential of the form (2.18)(2.20) and (2.39) does not
allow us to stabilize the remaining linear combination of S and T 2 . It can be shown to be
a flat direction. This problem cannot be resolved even by considering a multiple gaugino
condensation in the hidden sector. Nevertheless, it is easy to realize that this problematic
linear combination can be stabilized by taking into account the higher order T corrections
to the gauge coupling in the hidden sector. We will make a more detailed comment on it
later in this subsection. Therefore, stabilization of the remaining imaginary part does not
324
represent a conceptual problem. In the rest of the paper, we will concentrate only on the
real parts of the moduli ignoring their imaginary parts. Now let us consider the real parts of
the moduli and show that the system under study indeed has an AdS minimum satisfying
Dall fields W = 0,
(2.41)
where D is the Kahler covariant derivative. We will not distinguish between the superpotentials and their absolute values. First, we consider equations
DZ W = 0.
(2.42)
Assuming that
Wf Wg , Wnp
(2.43)
DS W = 0.
(2.45)
(2.46)
where
y2
1
.
1 + 5
F1 =
2V
(Rb1 )2
(2.47)
(2.48)
y2
F
+
F
1
2 Wf ,
(Rb1 )2
(2.49)
where
3 I J d1I J bI bJ
5 y 2
F2 =
+
.
R I J K dI J K b I b J b K
V (Rb1 )2
(2.50)
(2.51)
T 2,
T2
325
obtain
Wg = F3 Wf ,
(2.52)
where
3 I J d2I J bI bJ
.
F3 = (2)
2 R I J K dI J K bI bJ bK
(2.53)
Eqs. (2.46) and (2.52) are consistent only if F1 and F3 are equal to each other. In particular,
(2)
they must have the same sign. This is possible only if 2 is negative. As we argued before,
this does not lead to any contradictions. Note, that F1 and F3 are both real. This is the
reason why only one linear combination of the imaginary parts of S and T 2 moduli can
be stabilized. On the other hand, if higher order T corrections to the quantity (2.25) are
present, F3 is really complex and, hence, the imaginary parts of both S and T 2 moduli can
be stabilized. The last equation to consider is
DY W = 0.
(2.54)
(2.55)
Eqs. (2.46), (2.49), (2.52) and (2.55) are the four equations with four independent variables
V , R, y and one of two bI s. Equations of this type were analyzed in detail in [5,42] in the
case of only one h1,1 modulus. It was shown that they admit a solution with the following
properties
V is of order one;
R is of order one;
2
does not become imaginary;
The gauge coupling constant ghidden
The fivebrane is close to the hidden brane (R y 0.1).
In this paper, we will not perform a detailed analysis. Let us just point out that in Eqs. (2.46)
and (2.49), the runaway moduli are stabilized by fluxes. Eqs. (2.46) and (2.52) lead to
equation
F1 = F 3 ,
(2.56)
(2)
2
326
(2)
2
(X + X)
.
2 + T 2 )
(S + S)(T
(3.1)
Wg Wg e
X2
T
(3.2)
The coefficients 5 and are given by expressions similar to Eqs. (2.14) and (2.23). Unfortunately, if a fivebrane wraps a nonisolated cycle, one should expect other fivebrane
moduli in addition to X. Such moduli have never been considered in the literature in detail.
Nevertheless, one should expect that the gaugino condensate superpotential (3.2) depends
on them. This might provide their stabilization. Thus, we will assume that these additional
moduli are fixed and not consider them in this paper. In principle, one can avoid this issue
by taking a fivebrane wrapping an isolated higher genus curve. Let us first see if we can
stabilize X in an AdS vacuum. By using Eqs. (3.1) and (3.2), the equation
DX W = 0
can be written as
X
x
2 Wg + 5
Wf = 0.
T
V Rb2
It is easy to realize that the only solution for X is
X = 0.
(3.3)
(3.4)
(3.5)
The point x = 0 corresponds to the fivebrane coinciding with the visible brane. Such a
vacuum is unstable in the sense that the fivebrane will disappear through a small instanton
transition [44,4749] and turn into new vector bundle moduli.
As an approximation, we will assume that the presence of this extra fivebrane will not
modify much the vacuum constructed in the previous section. As a result, we can talk about
327
the effective potential U (x) describing dynamics of the fivebrane. In fact, it is possible
to show that the vacuum value of the moduli S, T I , Y receive corrections of order x 2 .
Therefore, for very small values of x, x 1, their vacuum values will not shift much. This
suggest that the effective potential U (x) is a decent approximation. Of course, in order to
describe the system exactly, one has to solve all the equations for moduli including the
equations for the imaginary parts. This is not possible to do analytically. However, it is
natural to argue that the qualitative behavior of this system will be captured assuming that
there is the effective potential U (x) with the rest of the moduli fixed along the lines of
the discussion in the previous section. Thus, we consider dynamics of one field X with the
Kahler potential
1
K(X)
2
= K0 + K1 (X + X)
2
4
MPl
(3.6)
(3.7)
(3.8)
W0 Wf ,
(3.9)
.
T2
(3.10)
(3.11)
K(X)
2
3W (X)W (X)
,
(X)
U (X) = e MPl G1
W
DX W (X)DX
XX
(3.12)
(3.13)
1
K(X)W (X).
2 X
MPl
(3.14)
As was argued before, the imaginary part of X can be stabilized by this potential. Therefore,
the potential U (X) can be treated as an effective potential for one real field x. We will
328
(3.15)
2 F1
.
K1
(3.16)
Eq. (3.15) gives an effective potential U (x). Unfortunately, it is very difficult to analyze
this potential analytically. A graphical analysis shows that, generically, this potential has a
nonsupersymmetric AdS vacuum for a nonzero value of x. The form of U (x) for various
choices of parameters is shown on Figs. 1 and 2. It is possible to adjust parameters in
such a way that the vacuum becomes dS. However, in this case, the parameter has to
Fig. 1. The graph of UU(x) for K1 = 3, = 3, = 1, = 10. There exists a nonsupersymmetric AdS minimum.
0
Fig. 2. The graph of UU(x) for K1 = 5, = 2.5, = 2, = 10. There exists a nonsupersymmetric AdS minimum.
0
329
be taken to be sufficiently greater than one, whereas Eqs. (3.16), (3.8) and (3.12) require
that be of order one. Therefore, for reasonable values of the parameters, the minimum is
always AdS. It is possible to adjust parameters so that x is less than the size of the interval,
which, as discussed in the previous section, can be stabilized at a value of order one. This
AdS vacuum can be raised to a metastable dS vacuum by methods discussed in [5]. This
demonstrates that the most general system of heterotic Mtheory moduli can be stabilized
in a dS vacuum.
In the rest of the paper, we will be interested in dynamics of a fivebrane in the regime
x 1. Heterotic Mtheory vacua can contain several fivebranes wrapped on nonisolated
genus zero or higher genus curves. We have just argued that those fivebranes which are
located sufficiently far away from the visible brane can be stabilized. Now we would like to
understand the fate of the fivebranes which are close to the visible sector. Such fivebrane
will roll towards x = 0. The potential U (x) in this regime does not lead to any interesting
physics. It does not provide stabilization of x. It is also hard to imagine how to use it in
any cosmological framework. On the one hand, it is negative and, hence, cannot be used
for inflation. On the other hand, it does not satisfy conditions necessary for Ekpyrotic
cosmology [8486]. To make use of this potential, we will modify it by FayetIliopoulos
terms. Depending on relations among various coefficients, FayetIliopoulos terms can lead
to either stabilization of x or a potential with certain inflationary properties.
3.2. FayetIliopoulos terms and dS vacua
In both weakly and strongly coupled heterotic string models, there can be anomalous
U (1) gauge groups. They can arise in both the visible and the hidden sectors. The anomaly
is canceled by a fourdimensional version of the GreenSchwarz mechanism. This anomalous U (1) gives rise to the FayetIliopoulos term [43], which, in turn, gives rise to the
moduli effective potential of the form
b
,
(3.17)
V2
where b is a constant and g is the gauge coupling constant. In the context of the strongly
coupled heterotic string theory, the coefficient b was estimated in [5] and was found to be,
generically, of order
4 2
g
UD = MPl
b 1018 .
(3.18)
The potential UD depends on in what sector there appears an anomalous U (1). The reason
is that the coupling constants in the visible and the hidden sectors are different. They are
given by [65]
2
=
gvisible
g02
Re(S + 1(1) T 1 + 2(1) T 2 + (T 1
Y2
) + (T 2
T1
X2
))
T2
(3.19)
and
2
ghidden
=
g02
(2)
(2)
(3.20)
330
where g0 is a moduli independent constant of order GUT . In [3,5], FayetIliopoulos potentials UD were used to raise AdS vacua to dS vacua. In this paper, we will be interested in
the x dependence of UD . If the anomalous U (1) appears in the hidden sector, the potential
UD (x) takes the form
UDvisible (x) =
B1
,
B2 x 2
(3.21)
whereas, if the anomalous U (1) appears in the hidden sector, the potential UD is
UDhidden (x) =
C1
,
C2 + x 2
(3.22)
where, B1 , B2 , C1 and C2 can be read off from Eqs. (3.17), (3.19) and (3.20) and is given
by Eq. (3.12).
We would like to modify our potential U (x) by UD (x). In this section, we take UD (x)
to be UDhidden (x). We will now show that the potential energy
U (x) = U (x) + UDhidden (x)
(3.23)
(3.25)
U (x)
(3.26)
=0
x
under mild assumptions. For x 1, the potential U (x) is governed by the quadratic function
U (x) = 3U0 + a2 U0 x 2 ,
where a2 is given by
a2 = K1 2(1 + )2 3 .
(3.27)
(3.28)
Using Eqs. (3.8), (3.12), (3.16) and (2.47), one can show that a2 is greater than zero for
any choice of the parameters. It is straightforward to solve Eq. (3.26) in this regime. The
approximate solution is
1/2
C1
1
C2
.
xmin
a 2 U0
331
(3.29)
It is possible to adjust the parameters so that xmin is real and much less than one. It is also
straightforward to show that, if the solution for xmin exists, it is always a minimum. The
simplest way to do it is prove that, if the solution (3.29) exists, then x = 0 is always a
maximum. Since xmin 1, the value of the cosmological constant is approximately given
by
3U0 +
C1
.
C2
(3.30)
(3.31)
which is consistent with observations. The form of the potential U (x) in the regime x 1
is shown on Fig. 3.
In Sections 2 and 3, we showed that the most general system of heterotic Mtheory
moduli can be stabilized in a dS vacuum. In addition to moduli considered in [5], we
also provided stabilization for extra h1,1 moduli and an extra fivebrane associated with
a nonisolated genus zero curve or with a higher genus curve. In the presence of such a
fivebrane, the system of moduli can be stabilized by fluxes and nonperturbative effects in
a nonsupersymmetric AdS vacuum which then can be lifted to a dS vacuum as in [5]. This
fivebrane can also be stabilized by balancing the supergravity potential energy against a
FayetIliopoulos term induced by an anomalous U (1) gauge group in the hidden sector.
Thus, the potential energy U (x) might admit two dS vacua. One of the them is the lift
of the nonsupersymmetric AdS vacuum. The other one can additionally arise for x 1,
though it did not existed in the absence of the FayetIliopoulos term.
Fig. 3. The graph of UU(x) in the regime x 1 for a2 = 2.95, = 1, U1 = 3.01, C2 = 1. There exists a dS
0
0
minimum.
332
(4.1)
does not have a nontrivial minimum for x 1 and x rolls towards x = 0. We will assume
that the potential U (x) is positive. The potential (4.1) has the following form
2
2 2
+
U (x) = U0 eK1 x 3 + 2K1 x 2 1 + e x
B1
.
B2 x 2
(4.2)
Let us recall that the coefficients , K1 , and are given by Eqs. (2.21), (3.8), (3.12) and
2 and B and B can be read off from Eqs. (3.17)
(3.16), U0 is a constant of order Wf2 /MPl
1
2
and (3.19). We assume that this potential is positive, that is,
B1
> 3U0 .
B2
(4.3)
Our goal is to examine whether this potential can satisfy the slow roll conditions required
by inflation. As in the previous section, we are interested in the regime x 1. Then we
can expand U (x) in powers of x. For our purposes, it is enough to keep only two leading
terms. We obtain
U (x) A0 + A2 x 2 ,
(4.4)
where
A0 = 3U0 +
B1
B2
(4.5)
and
B1
B1
A2 = K1 2(1 + )2 3 U0 + 2 = a2 + 2 .
B2
B2
(4.6)
In order to study the standard slow roll parameters (x) and (x), we have to canonically
normalize the kinetic energy. From the Kahler potential (3.6), it follows that we have to
redefine x as
2 x
x
(4.7)
.
2
K1 MPl
This new x is canonically normalized and has dimension one. The potential energy now
looks as follows
2A2 2
x .
U (x) = A0 +
2
K1 MPl
(4.8)
333
(4.9)
and
2
(x) = MPl
U (x)
U (x)
(4.10)
2A22
2
A20 K12 MPl
x2.
(4.11)
Clearly, for x MPl , (x) is naturally much less than one. For the parameter (x) we have
(x) =
4A2
.
K1 A0
(4.12)
(4.13)
4 B1
K1 B22
B1
B2
1.
(4.14)
(4.15)
4
1.
K1 B2
(4.16)
and
Condition (4.15) is a relatively mild constraint. Using Eqs. (3.8), (3.11), (3.17) and (3.19),
condition (4.16) can be rewritten as
2V
Re(S
(1)
+ 1 T 1
(1)
+ 2 T 2
+ (T 1
Y2
) + T 2)
T1
1.
(4.17)
Unfortunately, it does not seem to be possible to fulfill this condition, at least in the context
of lowenergy field theory. Inequality (4.17) requires that some of the tensions I(1) , or
be much greater than one. In this case, one cannot trust, even approximately, expressions
(3.19) and (3.20) for the coupling constants because they can be substantially modified by
higher order corrections [72,73]. On the other hand, Eq. (4.17) may make perfect sense
in the context of Mtheory. However, we would like to stay within the context of lowenergy field theory. All we have to do to make the parameter (x) small is to decrease
334
the parameter A2 in Eq. (4.4). This, in fact, can easily be done. We just have to replace a
FayetIliopoulos term in the visible sector by a FayetIliopoulos term in the hidden sector.
Equivalently, we could just add the FayetIliopoulos term in the hidden sector to Eq. (4.1).
In both cases, addition of such a FayetIliopoulos term increases A0 and decreases A2 .
It is possible to (not necessarily fine) tune the parameter to be much less than one. Let
us consider it in slightly more detail. Assuming, for simplicity, that only a hidden sector
FayetIliopoulos term is present and using (3.22), we have
(x) =
4(2(1 + )2 3)U0
3U0 +
4 C1
K1 C22
C1
C2
1.
(4.18)
4(2(1 + )2 3
3
K1 C2 )U0
A0
4
K1 C2 A0
1.
(4.19)
Since the quantities U0 and A0