www.elsevier.com/locate/npe
Abstract
We investigate supersymmetric SU(5) grand unified theories (GUTs) realized in 5 space time
dimensions and broken down to the MSSM by SU(5)violating boundary conditions on a S 1 /(Z2
Z2 ) orbifold with two 3branes. The doublettriplet splitting problem is entirely avoided by locating
the MSSM Higgs doublets on the brane on which SU(5) is not a good symmetry. An extremely
simple model is then described in which the MSSM matter is also located on this SU(5)violating
brane. Although this model does not unify the MSSM matter within SU(5) multiplets, it explains
gauge coupling unification. A second model with MSSM matter in the SU(5)symmetric bulk
preserves both the SU(5) explanation of fermion quantum numbers as well as gaugecoupling
unification. Both models naturally avoid problematic SU(5) predictions for the Yukawa couplings
of the first two generations and are consistent with proton decay constraints. We analyse the running
of gauge couplings above the compactification scale in terms of a 5d effective action and derive the
implications for the values of compactification scale, unification scale and of the scale at which the
bulk gauge theory becomes strongly coupled. 2001 Elsevier Science B.V. All rights reserved.
PACS: 12.10.Dm; 11.25.M; 12.60.Jv
1. Introduction
Since the pioneering work of Georgi and Glashow [1] (also [2]) the compelling concept
of a grand unified theory (GUT) of the standard model (SM) gauge interactions has
dominated our thinking about physics at high energies. The success of gauge coupling
unification [3] in minimal supersymmetric standard model (MSSM) extensions of these
theories [4,5] has further supported this idea [6]. However, despite these great successes
(and others, such as the generation of small neutrino masses) this GUT concept is not
without its faults. In particular we recall the problems of the Higgs structure of the highscale theory, especially the doublettriplet splitting problem, the issue of too fast proton
decay, and the mismatch of the GUT scale with the naive scale of unification with gravity.
Email address: arthur.hebecker@cern.ch (A. Hebecker).
05503213/01/$ see front matter 2001 Elsevier Science B.V. All rights reserved.
PII: S 0 5 5 0  3 2 1 3 ( 0 1 ) 0 0 3 7 4  1
Recently, a new possibility for the embedding of the SM into a form of GUT has been
suggested by Kawamura [79] and further extended by Altarelli and Feruglio [10] and
Hall and Nomura [11]. The basic idea is that the GUT gauge symmetry is realized in
5 or more spacetime dimensions and only broken down to the SM by utilizing GUTsymmetry violating boundary conditions on a singular orbifold compactification. Given
the success of the traditional supersymmetric gaugecoupling unification predictions,
the most attractive class of models are ones with both supersymmetry and (at least)
SU(5) gauge symmetry in 5 dimensions. In this case both the GUT group and 5d
supersymmetry are broken down to a N = 1 supersymmetric model with SM gauge group
by compactification on S 1 /(Z2 Z2 ) (a related idea was employed for EWSB in Ref. [12]).
This construction allows one to avoid some unsatisfactory features of conventional GUTs
with Higgs breaking, such as doublettriplet splitting, while maintaining, at least at leading
order, the desired MSSM gauge coupling unification. Moreover, given the fact that string
theory requires additional dimensions, as well as branes located at orbifold fixed points,
the necessary (inverseGUTscalesized) 5th dimension is not unreasonable. 1
In this paper we present two GUT models of this type, which we believe to be the
simplest yet discovered, our particular focus being the realization of the MSSM Higgs and
the MSSM matter. Specifically in Section 2 we review the basic physics of the construction
of Refs. [711], applying this in Section 2.1 to the gauge sector of the model. Section 2.2
contains the novel observation that the electroweak Higgs can be a localized field on
the SU(5)violating brane, with no triplet partners whatsoever, thus solving the doublettriplet problem of traditional GUTs in a particularly simple way. Section 2.3 contains
a detailed discussion of two novel constructions of the MSSM matter in such theories.
The first of these two models locates the MSSM matter on the SU(5)violating brane and
has both an exceptionally simple structure and is evidently consistent with lowenergy
phenomenology. The model explains gauge coupling unification but, intriguingly, has no
SU(5) multiplet structure for the MSSM matter, so in itself does not explain the quantum
numbers of the MSSM matter. In our second model we present a novel way of realising
the MSSM spectrum from SU(5)representations in the 5d bulk (thus explaining the matter
quantum numbers), which nevertheless leads to zeromodes which couple to the branelocalized Higgs in a way that does not imply the incorrect SU(5) relations for the Yukawa
matrices for the first two generations. Thus both models possess a number of appealing
features. In Section 3, we address the nature of gauge coupling unification in these models
(developing the treatment of [11]), with special emphasis on the corrections to exact
unification arising from SU(5)violating brane kinetic terms. We analyse the running of
gauge couplings above the compactification scale in terms of a 5d effective action and
derive the implications for the values of the compactification scale, the unification scale
and of the scale at which the bulk gauge theory becomes strongly coupled. Most of the
1 We will take an effectivefieldtheory viewpoint of orbifold branes, namely we will impose all constraints
necessary for a consistent lowenergy theory, but not require an explicit string theory (or other quantum gravity)
realisation.
discussion of Section 3 is generic and applies to other models with orbifold GUT breaking
as well. Section 4 contains our conclusions.
P : y y .
(1)
It is important to the whole construction that under these equivalences there are two
inequivalent fixed 3branes (or orbifold planes) located at y = 0, and y = R/2 ,
which we denote O and O , respectively. It is consistent to work with the theory obtained
by truncating to the physically irreducible interval y [0, ] with the 3branes at y = 0,
acting as endoftheworld branes, and henceforth we so do.
The action of the equivalences P , P on the fields of a quantum field theory living on
M4 S 1 /(Z2 Z2 ) is not fully specified by the action Eq. (1) on the coordinates. One
must also define the action within the space of fields. To this end, let (x, y) be a vector
comprising all bulk fields, then the action of P and P is given by
P : (x, y) P (x, y),
P : (x, y
) P (x, y ).
(2)
(3)
Here, on the rhs, P and P are matrix representations of the two Z2 operator actions,
which necessarily have eigenvalues 1 when diagonalized. Let us from now on work in
this diagonal basis of fields, and classify the fields by their eigenvalues (1, 1). Then
these fields have the KK expansions:
++ (x, y) =
+ (x, y) =
+ (x, y) =
(x, y) =
4
1
2ny
(2n)
,
++ (x) cos
R
R
( 2) n,0
(4)
n=0
(2n+1)
(x) cos
(2n + 1)y
,
R
(5)
(2n+1)
(x) sin
(2n + 1)y
,
R
(6)
(2n + 2)y
4 (2n+2)
.
(x) sin
R
R
(7)
4
R
4
R
n=0
+
+
n=0
n=0
From the 4d perspective the KK fields (k) (x) acquire a mass k/R, so only the ++
possess a massless zero mode. Moreover, only ++ and + have nonzero values at
y = 0, while only ++ and + are nonvanishing at y = .
It is important to realize that the action of the identifications P , P on the fields (namely
the matrices P and P ) can utilize all of the symmetries of the bulk theory. Thus P and P
can involve gauge transformations, discrete parity transformations, and most importantly
in the supersymmetric case, Rsymmetry transformations. This last feature allows one
to break the higherdimensional supersymmetry to a phenomenologically acceptable, and
desirable N = 1 SUSY theory in 4d.
2.1. The gauge structure of the minimal model
To reproduce the good predictions of a minimal supersymmetric GUT, we start from a
5d SU(5) gauge theory with minimal SUSY in 5d (with 8 real supercharges, corresponding
to N = 2 SUSY in 4d). Thus, at minimum, the bulk must have the 5d vector superfield,
which in terms of 4d N = 1 SUSY language contains a vector supermultiplet V with
physical components A , , and a chiral multiplet with components , . Both V and
transform in the adjoint representation of SU(5).
Now choose the matrix representation of the parity assignments, expressed in the
fundamental representation of SU(5), to be
P = diag(+1, +1, +1, +1, +1),
(8)
so that
P : V a (x, y)T a V a (x, y)P T a P 1 ,
P : V a (x, y )T a V a (x, y )P T a P 1 ,
(9)
where the action on the rhs is as matrices. For the same assignments are taken apart
from an overall sign for both P and P , so that
P : a (x, y)T a a (x, y)P T a P 1 ,
P : a (x, y)T a a (x, y)P T a P 1 .
(10)
O
brane at
These boundary conditions then break SU(5) to the SM gauge group on the
y = , and 4d N = 2 SUSY to 4d N = 1 SUSY on both the O and O branes. This can
Table 1
Parity assignment and KK masses of fields in the 4d vector and chiral adjoint supermultiplet. The
index a labels the unbroken SU(3) SU(2) U(1) generators of SU(5), while a labels the broken
generators
(P , P )
4d superfield
4d mass
(+, +)
(+, )
(, +)
(, )
Va
V a
a
a
2n/R
(2n + 1)/R
(2n + 1)/R
(2n + 2)/R
be seen explicitly by examining the masses of the KK towers of the fields as displayed in
Table 1. Only the (+, +) fields possess massless zero modes, and at low energies the gauge
and gaugino content of the 4d N = 1 MSSM is apparent.
2.2. The electroweak Higgs
We now come to the first significant difference with respect to the treatments of Refs. [7
11]: our model does not include bulk Higgs multiplets. If we were to include such bulk
Higgs, then the unbroken SU(5) gauge invariance of the bulk would force them to transform
as 5s (and 5s) of SU(5). As such representations contain more than the SU(2) doublets
necessary for electroweak symmetry breaking, we are faced with a version of the infamous
doublettriplet splitting problem. Of course a success of the models of Refs. [711] is that
by extending the action of P , P to this bulk Higgs sector, the triplet components can
be made heavy. Nevertheless it is unarguably simpler (and phenomenologically different)
never to have had the Higgs triplet components in the first place!
How then are the Higgs doublets to be realized? To understand the possibilities it is
important to realize that after the orbifolding, Eqs. (8)(10), the amount of unbroken gauge
symmetry is position dependent. On the O brane at y = 0 and in the bulk the entire
SU(5) symmetry is a good symmetry. In particular at position y = the allowed gauge
transformations are of the form U = exp(i a a (x, y)T a + i a a (x, y)T a ), with both
the gauge transformation parameters a (x, y) and a (x, y) nonvanishing. Only at the O
brane are the a (x, ) = 0, and purely the SM gauge symmetry is defined. The implication
of this for the gauge sector of the theory is that the form of the 5d effective Lagrangian (at
scale ) is:
1
(y)
2
+
Leff () = d
tr W W
2g5 ()2 2g0 ()2
3
(y )
tr Wi W ,
(11)
2gi ()2
i=1
where W,i (i = 1, 2, 3) are the supersymmetric field strengths of the unbroken U(1)
SU(2) SU(3) SM gauge group on the O brane, and W is the SU(5) field strength. In
other words, the effective action of the O brane need only respect the SM gauge group,
while the bulk theory must respect the full SU(5) gauge symmetry.
The symmetries of the theory allow one to add interactions, and additional matter, on the
O brane which only respect the SM gauge symmetry. Thus, unlike the models of Refs. [7
11], we choose to add a pair of weakSU(2) doublet N = 1 chiral superfields Hu , Hd to
the O 3brane theory at y = . This is a particularly natural and attractive definition of
the theory precisely because the Higgs is the one matter field in the MSSM that does not
have partners that effectively fill out a full SU(5) multiplet as far as quantum numbers are
concerned. As we do not have any Higgses which transform as a triplet of SU(3), there is
automatically no doublettripletsplitting problem either.
2.3. The MSSM matter
The next question is the location of the SM matter (and their 4d N = 1 superpartners).
In our framework it cannot be on the SU(5)invariant O brane at y = 0 as it would not then
be able to interact with the electroweak Higgs. This leaves us with two possibilities, both
of which lead to attractive models.
I. SM matter on the SM brane
If, together with the Higgs, the matter is also located on the O brane, then the Yukawa
couplings only need be invariant under the SM gauge group. Thus, the Yukawa couplings
are no more restricted than in the traditional 4d MSSM, and automatically there is enough
flexibility to accommodate fermion masses and mixings. Therefore the Yukawa terms in
the superpotential of the 5d effective Lagrangian for the light SMcharged fields of our
minimal theory are localized entirely on the O brane at y = :
J + hd QI Hd D
J + hl LI Hd E
J ,
WYukawa = d 2 (y )
huIJ QI Hu U
IJ
IJ
IJ
(12)
where I, J = 1, 2, 3 are generation indices, and all fields depend only upon
This
theory has considerable advantages as compared to the setups of both Altarelli and
Feruglio [10] (AF) and Hall and Nomura [11] (HN). Specifically, the construction of AF
considers the Higgs to be a bulk field (which therefore must be extended to a full 5), and
locates the SM on the SU(5) invariant brane (the SM matter therefore transform as full
5 + 10 multiplets of SU(5)). On this brane AF then assume SU(5)noninvariant Yukawa
interactions between the bulk Higgs and the SM matter. But such a setup is not consistent
with unbroken gauge invariance on the O brane. In particular it does not allow the freedom
to move between unitary and covariant gauges to show the simultaneous unitarity and
Lorentzinvariance of the lowenergy theory. (Lack of renormalizability, at least at high
scales, is not really the important issue after all the basic Kawamura setup considers a
5d nonrenormalizable theory. This is perfectly OK if the scale of the loss of calculability
is sufficiently high.) Indeed unless one includes further SU(5)breaking mass terms for
the SU(5) gauge bosons, m2 Aa A,a (localized to the formerly SU(5)invariant brane), the
theory is not unitary at low energy. The whole theory then has a rather different, and much
x.
more complicated character than that discussed by Altarelli and Feruglio. We therefore
consider this theory to be more complicated than necessary.
On the other hand the theory of HN, where again the Higgs is taken to be a bulk
field, and the SM matter is located on the SU(5)invariant brane, assumes SU(5) invariant
Yukawa interactions localized to the SU(5)invariant brane. This is a consistent setup.
Unlike in our model where we place the SM matter on the O brane, the HN model can
explain the b mass ratio [13] via SU(5) relations for the Yukawas. Unfortunately it also
leads to SU(5) predictions for the first and second generation fermion mass ratios, md /me
and ms /m , which are in strong disagreement with experiment. Thus HN are forced to
introduce additional fields in the bulk which mix with the first and second generations of
MSSM fields so as to correct the wrong predictions for the first two generations of quarks
and leptons. Although possible, this procedure is ad hoc.
Our model does not suffer from either of the above shortcomings. It is also physically
different in advantageous ways. First, since there are no Higgs triplets at all, it immediately
follows that there are no problems with triplethiggsinomediated proton decay. (The
models of AF and HN, where triplet higgsinos exist as KK excitations, also avoid
dangerous higgsinomediated proton decay, but for different reasons.) There are no
dimension5 proton decay mediating operators from higgsinos in our theory. Of course,
there are still protondecay processes mediated by the heavy (+, ) KK mode X and Y
gauge bosons. These are present due to terms in the O brane action involving the (nonvanishing) 5 derivatives of Aa . (Note that such KK excitations in the X and Y gauge boson
directions carry hypercharge, which is correctly normalized because it sits inside the bulk
SU(5), so their interactions with the branelocalized SM states also correctly normalize the
hypercharge of these SM states.) Being dimension 6 operators these lead to much less of
a constraint on the effective GUT scale 1/R. In addition, however, there can in principle
be branelocalized dimension 5 baryonnumberviolating operators from globalquantumnumberviolating quantum gravity effects [14]. Because of the existence of a relatively
large 5th dimension the quantum gravity scale is lower than the usual Planck mass, so these
operators, if they are there, are more problematic than usual. We leave a detailed discussion
of this issue to a future publication. Second, this model has several very nice features when
one considers the breaking of the remaining N = 1 supersymmetry down to the SM at
the TeV scale. An attractive solution of the supersymmetric FCNC problem follows if
the MSSM matter is sequestered from the SUSYbreaking sector. The localization of the
MSSM matter to the O brane naturally allows us to achieve this separation by having the
SUSYbreaking sector on the SU(5)symmetric Obrane, and communicating the SUSY
breaking to the MSSM matter by the MSSM gauginos which live in the bulk. This is
similar but not identical to the models of [15] considered by HN [11] (see also the earlier
works [16]). For example, one difference arises from the fact that in our model the gauginos
necessarily pick up their soft SUSYbreaking mass from an SU(5)symmetric interaction
(e.g., (S tr W W )F cpt if there is a singlet chiral superfield S on O with FS = 0).
Thus the gaugino masses will satisfy the traditional SUSY SU(5) relations to the same
high degree of accuracy as the gaugecouplings themselves unify. This and other details of
SUSY breaking in this (and the following) model will be discussed in a future publication.
10
P : c P c ,
(13)
and similarly for P . This difference leads to a chiral spectrum for the zero modes. Indeed
the KK spectrum of 5d bulk hypermultiplets in the representation 10 + 5 (whose 4d chiral
+ Tc + F
c ) resulting from the P , P actions is given in
components we denote T + F
+
Table 2. Note that since P does not commute with SU(5), components of the T + F
c fields in different SU(3) SU(2) U(1) representations have different KK mode
Tc+F
, D,
L, E,
etc., to indicate
structures. Thus in this table we label the expansions by Q, U
their SM transformation properties.
We immediately note from Table 2 that the zero mode structure (the n = 0 components
=
of the (+, +) fields) do not fill out a full generation of SM matter. We just get U
(3, 1)2/3 , E = (1, 1)1 , and L = (1, 2)1/2 leftchiral N = 1 superfields. This is rectified
by taking another copy of 10 + 5 in the bulk (with N = 1 chiral components denoted
c
T + F + T c + F ), and using the freedom to flip the overall action of the P parity on
11
Table 2
Parity assignments and KK masses of fields in the 4d chiral supermultiplets resulting from the
) representation. The subscript labels the SM
decomposition of 5d hypermultiplets in the (T + F
= (3, 2)1/6 , U
= (3, 1)2/3 , etc.
transformation properties, e.g., Q = (3, 2)1/6 , Q
(P , P )
4d superfield
4d mass
(+, +)
L
TU
, TE
, F
2n/R
(+, )
TQ , F
D
(2n + 1)/R
(, +)
c ,F
c
T
Q D
c
TUc , TEc , F
L
(2n + 1)/R
(, )
(2n + 2)/R
Table 3
Parity assignments and KK masses of fields in 4d chiral supermultiplets resulting from the
decomposition of 5d hypermultiplets in the (T + F ) representation. The alteration of the action
of P causes an effective interchange of the 1st and 2nd rows, and of the 3rd and 4th rows, relative
to Table 2
(P , P )
(+, +)
(+, )
(, +)
(, )
4d superfield
,F
TQ
D
T , T , F L
E
U
c
TUc , TEc , F L
c , F
c
T
D
Q
4d mass
2n/R
(2n + 1)/R
(2n + 1)/R
(2n + 2)/R
+ T c + F
c . 2 This difference leads to a different
these multiplets by a sign relative to T + F
selection of zero mode components, the KK spectrum being given in Table 3.
Now a marvellous thing has happened! Combining the results of Tables 2 and 3 we have
zero modes which fill out the full matter content of a SM generation. Remarkably, this
occurs in such a way as to explain charge quantization and the hypercharge assignments
despite the fact that what we think of as a single generation filling out a 10 + 5, has
components that arise from different (10 + 5)s in the higherdimension. This means that
when we couple three copies (for the three generations) of this combination of fields to
the Higgs on the O brane at y = , the Yukawa couplings do not have to satisfy the usual
SU(5) relations. 3 Explicitly, the Yukawa couplings for the light zeromode MSSM matter
fields now result from 3 different combinations of the 5d fields,
J
l
I J
huIJ TQI TUJ Hu + hdIJ TQI F D
WYukawa = d 2 (y )
Hd + hI J TE
FL Hd ,
IJ
(14)
2 Hall and Nomura make this observation in the 2nd footnote of the final version of [11].
3 These results are reminiscent of those from studies of string theory carried out in the midtolate 1980s [17].
12
where we have indicated the zero mode components contained in the 5d T , T , F and F
fields (using the same notation for the components as in Tables 2 and 3). As before, Hu
and Hd are 4d chiral fields localized to the O brane at y = . Thus we have no problems
with the standard incorrect 1st and 2nd generation massratio predictions of SU(5), and do
not require the contortions of other models to avoid these difficulties.
Having discussed how a successful understanding of quantum numbers and interactions
may be achieved, we now turn to the question of the nature of gauge coupling unification
in these models.
12 1
1
R 103 R 1 .
2
U
g5
(15)
13
We see that the range of validity of weakly coupled 5d gauge theory can span up to three
orders of magnitude and may lead us directly to, e.g., string physics at the Planck scale.
We now turn to loop corrections to the values of gi at the scale Mc 1/R. One approach
is to integrate out the fifth dimension at the maximal scale M at which our model is valid
and to run the 4d effective theory down to Mc = 1/R in a conventional manner. (Note
that M can be significantly smaller than its obvious upper bound if, for example, the
internal structure of the brane is resolved at a lower energy.) The 4d gauge couplings are
obtained by applying gi2 = g52 / (for i = 1, 2, 3) at scale M and integrating out the KK
modes between M and Mc :
N
N
1
M
M
M
1
1
+ bi
ln
+ ci
ln
. (16)
i (Mc ) = U +
ai ln
2
Mc
(2n 1)Mc
2nMc
n=1
n=1
Here ai is the zeromode contribution (corresponding to the usual MSSM running) while
bi and ci are the contributions of the higher KK modes. The masses of these modes are
even or odd multiples of Mc , depending on the boundary conditions at O and O . The
upper limit of the sum is determined by the UV cutoff N M/2Mc . Rewriting Eq. (16) as
N
N
1
M
M
1
1
1
i (Mc ) = U +
ai ln
,
+ (bi + ci )
ln
bi
ln 1
2
Mc
2nMc
2n
n=1
n=1
(17)
it seems impossible to extract any useful information about lowenergy gauge couplings
since the sum multiplying (bi + ci ) produces a powerlike dependence on the cutoff M.
The reason is that a powerlike cutoff dependence has an error which is not parametrically
smaller than the effect itself. Note however that this powerlike behaviour is a manifestation
of the 5d SU(5)symmetric bulk. Thus, even without a detailed calculation, we know the
sum (bi + ci ) to be independent of i. As a result, the physically important differences
ij (Mc ) i1 (Mc ) j1 (Mc ) are only logarithmically sensitive to the cutoff scale and
are therefore calculable:
1
1
M
.
ij (Mc ) =
(18)
aij + bij ln
2
2
Mc
Here aij = ai aj , bij = bi bj , and the sum multiplying bi in Eq. (17) has been
evaluated at leading logarithmic accuracy. The effect described by Eq. (18) has been called
differential running in [18], where it was discussed in the framework of GUT breaking
by an adjoint Higgs on the brane.
Let us now compare this to the running of the 5d effective action from M to Mc . Due to
unbroken gauge invariance in the bulk, the most general form of the effective Lagrangian
(disregarding SM matter) at scale (with M Mc ) is given by Eq. (11). This
Lagrangian, which is obtained from the original Lagrangian Leff (M) by integrating out
all modes with momenta above , has to be evaluated on fields that are smooth on a
scale . Thus, the functions in Eq. (11) have to be understood as approximate functions,
localized on a scale .
14
.
aij + bij ln
ij (Mc ) = ij () +
(19)
2
2
Mc
At the same time, one can start from the original Lagrangian LM and derive
1
1
M
ij (Mc ) = ij (M) +
.
aij + bij ln
2
2
Mc
(20)
This implies
1
1
M
ij () = ij (M) +
aij + bij ln ,
2
2
(21)
thus determining the evolution of the SU(5)breaking braneterms in Eq. (11) with .
The differences ij (M) in Eq. (21) are parameters of the (for our purposes most
fundamental) action at scale M. Since Eq. (21) holds only in the leading logarithmic
approximation, we have to ascribe an uncertainty of approximately 1 to the expression
ln(M/). Equivalently, this uncertainty can be attributed to the value of ij (M). This
implies that it is unnatural for the quantity ij (M) (which corresponds to a coefficient of a
term in the effective Lagrangian at scale M) to be smaller than this renormalization scale
uncertainty. In the most optimistic scenario, where no new physics at scale M contributes
to ij (M), we have therefore to assume that
1
1
ij (M)
(22)
aij + bij ,
2
2
where means that an O(1) coefficient remains undetermined. 5
To develop some intuition for the numerical values of the relevant scales, focus first
on model I of Section 2.3. The relevant function coefficients are ai = (33/5, 1, 3)
(MSSM) and bi = (10, 6, 4), ci = (0, 4, 6) (KK modes) (cf. [11,18]). With
i1 (mZ ) = (59.0, 29.6, 8.4) and an effective SUSY breaking scale mZ , conventional
unification occurs at MGUT 2 1016 GeV. An obvious possibility is to have Mc only
slightly below MGUT , in which case unification (within the uncertainty specified by
Eq. (22)) occurs slightly above MGUT (since above Mc the ij run somewhat slower than
in the MSSM). In this scenario, even though is near the Planck scale, the scale M at
which couplings unify and our orbifold model breaks down is near Mc . In the opposite
4 A careful treatment using effective field theory with dimensional regularization is possible [20]. We thank
R. Rattazzi for discussions of this issue.
5 In Ref. [11] the size of these brane localized gaugekinetic terms is estimated to be 1 1/(4 ) using NDA
i
and the assumption that the 4d gauge theory on the boundary becomes strongly interacting at the same scale as
the 5d bulk theory. Our result, which is similar numerically, does not rely on this assumption and is also valid in
the case of a weakly coupled 4d theory.
15
extreme, one can have Mc 2 1014 GeV and unification at M 2 1017 GeV. In this
scenario, the 5d orbifold model is valid over 3 orders of magnitude and the 5d Planck
scale is reached. In model II of Section 2.3, the zeromode ai coefficients are identical to
model I, while the nonzero KK mode function coefficients are bi = (2, 6, 8), and ci =
(12, 8, 6). Note that not only is the difference aij = ai aj the same for the two models,
but so is bij = bi bj , namely (b12 , b23, b31 ) = (4, 2, 6). The reason for this is that the
alteration between the two models is that model II has, effectively, at every nonzero KK
level a number of complete SU(5) multiplets (as can be seen by combining the quantum
numbers at each nonzeromode mass level from Tables 2 and 3). Thus the differences bij
are determined solely by the gauge structure of the bulk which is unchanged. Therefore,
apart from a somewhat reduced cutoff scale due to the faster SU(5)invariant running of
the bulk theory, the two models are very similar in this regard.
Summarizing, the primary phenomenological result of this section is that in both
models I and II the uncertainty in the gauge coupling unification (quantified by Eq. (22))
is comparable to the GUT scale threshold corrections of a conventional GUT.
4. Conclusions
In this paper we have investigated supersymmetric SU(5) grandunified theories (GUTs)
realized in 5 spacetime dimensions and broken down to the MSSM by SU(5)violating
boundary conditions on a S 1 /(Z2 Z2 ) orbifold with two endoftheworld 3branes.
Because of the position dependence of the amount of gauge symmetry, the MSSM Higgs
doublets can be located on the brane on which SU(5) is not a good symmetry. Thus SU(3)
triplet partners are never required, and the doublettriplet splitting problem is entirely
avoided. In Section 2.3 we constructed an extremely simple model in which the MSSM
matter is also located on this SU(5)violating brane. In this case, despite the loss of the
SU(5) understanding of the multiplet structure of the MSSM matter, one maintains gauge
coupling unification up to small corrections. As we discuss in Section 3, this is due to the
yindependence of the SM gauge boson zero modes, which allow the contribution of the
SU(5)symmetric bulk gauge kinetic term to dominate that of the SU(5)violating branelocalized kinetic terms. We find it intriguing that such a simple model is consistent. In
Section 2.3 we also constructed a second model, this time with MSSM matter located
in the SU(5)symmetric bulk. The model is slightly more involved, but preserves the
SU(5) predictions for fermion quantum numbers as well as gaugecoupling unification.
Both models naturally avoid problematic SU(5) predictions for the Yukawa couplings
of the first two generations. Finally, in Section 3, we analysed the running of the gauge
couplings above the compactification scale in terms of a 5d effective action and derived
the implications for the values of the compactification scale, the unification scale and of
the scale at which the bulk gauge theory becomes strongly coupled, finding values that are
phenomenologically attractive.
16
Acknowledgements
We are very grateful to Guido Altarelli, Wilfried Buchmller, Lawrence Hall, and
Riccardo Rattazzi for helpful conversations. J.M.R. thanks the members of the Theory
Group, LBNL, for their kind hospitality during the completion of this work.
References
[1] H. Georgi, S. Glashow, Phys. Rev. Lett. 32 (1974) 438;
H. Georgi, Unified gauge theories, in: Proc. Coral Gables 1975, Theories and Experiments In
High Energy Physics, New York, 1975.
[2] J. Pati, A. Salam, Phys. Rev. D 8 (1973) 1240;
J. Pati, A. Salam, Phys. Rev. D 10 (1974) 275.
[3] H. Georgi, H. Quinn, S. Weinberg, Phys. Rev. Lett. 33 (1974) 451.
[4] S. Dimopoulos, S. Raby, F. Wilczek, Phys. Rev. D 24 (1981) 1681.
[5] S. Dimopoulos, H. Georgi, Nucl. Phys. B 193 (1981) 150.
[6] N. Sakai, Z. Phys. C 11 (1981) 153;
L.E. Ibanez, G.G. Ross, Phys. Lett. B 105 (1981) 439;
L.E. Ibanez, G.G. Ross, Phys. Lett. B 110 (1982) 215;
M.B. Einhorn, D.R.T. Jones, Nucl. Phys. B 196 (1982) 475;
W.J. Marciano, G. Senjanovic, Phys. Rev. D 25 (1982) 3092;
P. Langacker, M.X. Luo, Phys. Rev. D 44 (1991) 817;
J. Ellis, S. Kelley, D.V. Nanopoulos, Phys. Lett. B 260 (1991) 131;
U. Amaldi, W. de Boer, H. Furstenau, Phys. Lett. B 260 (1991) 447.
[7] Y. Kawamura, Prog. Theor. Phys. 103 (2000) 613, hepph/9902423.
[8] Y. Kawamura, hepph/0012125.
[9] Y. Kawamura, hepph/0012352.
[10] G. Altarelli, F. Feruglio, hepph/0102301.
[11] L. Hall, Y. Nomura, hepph/0103125.
[12] R. Barbieri, L. Hall, Y. Nomura, Phys. Rev. D 63 (2001) 105007, hepph/0011311.
[13] M.S. Chanowitz, J. Ellis, M.K. Gaillard, Nucl. Phys. B 128 (1977) 506;
A.J. Buras, J. Ellis, M.K. Gaillard, D.V. Nanopoulos, Nucl. Phys. B 135 (1978) 66.
[14] M. Kamionkowski, J. MarchRussell, Phys. Lett. B 282 (1992) 137, hepth/9202003;
M. Kamionkowski, J. MarchRussell, Phys. Rev. Lett. 69 (1992) 1485, hepth/9201063;
R. Holman et al., Phys. Lett. B 282 (1992) 132, hepph/9203206;
R. Holman et al., Phys. Rev. Lett. 69 (1992) 1489.
[15] D.E. Kaplan, G.D. Kribs, M. Schmaltz, Phys. Rev. D 62 (2000) 035010, hepph/9911293;
Z. Chacko et al., JHEP 0001 (2000) 003, hepph/9911323.
[16] L. Randall, R. Sundrum, Nucl. Phys. B 557 (1999) 79, hepth/9810155;
Z. Chacko et al., JHEP 0001 (2000) 003, hepph/9911323.
[17] For early works see, for example, P. Candelas et al., Nucl. Phys. B 258 (1985) 46;
E. Witten, Nucl. Phys. B 258 (1985) 75.
[18] Y. Nomura, D. Smith, N. Weiner, hepph/0104041.
[19] Z. Chacko, M.A. Luty, E. Ponton, JHEP 0007 (2000) 036, hepph/9909248.
[20] R. Contino, L. Pilo, R. Rattazzi, E. Trincherini, hepph/0108102.
Abstract
For the socalled sourceprobe configuration in Matrix theory, we prove the following theorem
concerning the power of supersymmetry (SUSY): let be a quantumcorrected effective
SUSY
transformation operator expandable in powers of the coupling constant g as = n0 g 2n (n) ,
where (0) is of the treelevel form. Then, apart from an overall constant, the SUSY Ward identity
= 0 determines the offshell effective action uniquely to arbitrary order of perturbation theory,
provided that the SO(9) symmetry is preserved. Our proof depends only on the properties of the treelevel SUSY transformation laws and does not require the detailed knowledge of quantum corrections.
2001 Elsevier Science B.V. All rights reserved.
PACS: 11.25.w; 11.30.Pb
1. Introduction
Ordinarily, the role of a global symmetry with a finite number of parameters is to
a large extent kinematical. It provides conservation laws for the currents and charges,
relates the Greens functions, restricts the possible form of counter terms and so on, but
does not control the dynamics completely. The associated Ward identity is of an integrated
form unlike the more powerful SchwingerDyson equation and from grouptheoretic point
of view it only organizes a set of terms into invariants for which independent coupling
constants can be assigned. Especially, when one deals with the effective action, which can
contain arbitrary number of derivatives of fields, there may be many such invariants and
global symmetries are not expected to be able to fix their relative strengths.
Despite these wellknown facts, certain theories with a large number of global
supersymmetries (SUSY) might prove to be exceptions under some favorable conditions.
Notably, for the socalled Matrix theory for Mtheory, proposed by Banks, Fischler,
Email addresses: kazama@hep3.c.utokyo.ac.jp (Y. Kazama), tetsu@hep1.c.utokyo.ac.jp (T. Muramatsu).
05503213/01/$ see front matter 2001 Elsevier Science B.V. All rights reserved.
PII: S 0 5 5 0  3 2 1 3 ( 0 1 ) 0 0 3 9 6  0
18
Shenker and Susskind [1,2], evidence has been accumulating [310] that its high degree of
SUSY, namely, with the maximally allowed 16 supercharges, may be powerful enough to
determine the effective action of a Dparticle in interaction with an aggregate of coincident
source Dparticles, i.e., for the socalled sourceprobe configuration. Judged from our
common wisdom this may sound rather surprising but on the other hand impressive
agreement [1114] with 11dimensional supergravity lends support to such a conjecture,
as the latter is considered to be unique.
A clearcut settlement of this issue is hampered by the lack of unconstrained superfield
formalism for this system. One is forced to deal with the component formalism, where
SUSY is intertwined with gauge (BRST) symmetry and its algebra does not close without
the use of the full equations of motion. A related difficulty is that the SUSY transformation
laws for the effective action receive complicated quantum corrections and one cannot easily
identify possible SUSY invariants.
In order to gain insight into the issue of the power of SUSY in Matrix theory, in a
series of papers [9,10,15] we have (i) derived the SUSY Ward identity for the background
gauge in the form where the quantumcorrected SUSY transformation laws can be read
off in closed forms, (ii) computed the effective action and the SUSY transformation
operator explicitly for arbitrary offshell trajectory of the probe Dparticle including all
the spin effects at 1loop at order 4 in the derivative expansion, 1 (iii) checked that they
indeed satisfy the SUSY Ward identity = 0, and (iv) finally demonstrated that, given
such , the solution to the Ward identity regarded as a functional differential equation
for is unique to the order specified above. All of these calculations were extremely
cumbersome and elaborate codes had to be developed for the algebraic manipulation
program M ATHEMATICA, including a new fast algorithm for generating SO(9) Fierz
identities of considerable complexity. Although restricted to the sourceprobe situation, the
fact that the fully offshell effective action was uniquely determined from the knowledge
of the SUSY transformation laws was remarkable and strongly suggested that this feature
would persist to higher orders.
Indeed our result was not an accident that occurred at a particular low order. In
this article, we shall be able to prove the following theorem valid for the sourceprobe
configuration:
Theorem. Let be a quantumcorrected effective SUSY transformation operator expand
able in powers of the coupling constant g as = n0 g 2n (n) , where (0) is of the treelevel form. Then, apart from an overall constant, the SUSY Ward identity = 0 determines the offshell effective action uniquely to arbitrary order of perturbation theory,
provided that the SO(9) symmetry is preserved.
Actually the main effort will be devoted to proving the following proposition, from
which our theorem follows straightforwardly:
1 As usual, the concept of order in the derivative expansion is defined as the number of derivatives plus half
the number of fermions.
19
Proposition. Let (0) be the tree level SUSY transformation operator. Then assuming that
SO(9) is a good symmetry, the solution to the functional differential equation (0) = 0 is
unique up to an overall constant, and is given by the tree level action.
We wish to emphasize that our proof of the Proposition and hence the Theorem will
depend only on the properties of the treelevel SUSY transformation laws and will not
require the detailed knowledge of quantum corrections.
To avoid any possible misconception, however, let us state clearly that our theorem does
not yet prove that supersymmetry (together with SO(9) symmetry) completely determines
the sourceprobe dynamics in Matrix theory. What it states is that given an appropriate ,
which, for example, can be computed independently of by the method developed in [9]
in some gauge, the Ward identity fixes completely. In order to claim that SUSY fully
determines the dynamics, one must show that it is capable of determining both and
simultaneously in a selfconsistent manner, up to field redefinitions, without assuming the
knowledge of the underlying action of the Matrix theory. How this more ambitious program
should be formulated and attempted will be discussed in the final section.
Nevertheless, we believe that our theorem discloses another aspect of remarkably
powerful features of maximal supersymmetry: it clears a highly nontrivial necessary
requirement for SUSY to be able to determine everything, under certain conditions.
The rest of the article is organized as follows: in Section 2, after setting up our problem,
we derive what will be called the basic equation on which the proof of our Proposition
will be based. It is a reformulation
of the consistency
or integrability condition for the
(0)
(0)
(0) (0)
treetype Ward identity, namely,
= 0. Although it provides only a
necessary condition, it will prove to be sufficiently restrictive. Proofs of the Proposition
and the Theorem will be given in Section 3. As the proofs are somewhat involved, we will
first illustrate our basic ideas using a simple example. These ideas are then sharpened and
extended into a number of Lemmas which apply for general situation. These Lemmas step
by step reduce the possible solutions to be of more and more restricted form and lead to the
proof of the Proposition. This in turn immediately yields our main Theorem, by a recursive
argument with respect to the powers of the coupling constant. Finally, in Section 4 we will
discuss and indicate the line of attack for the important remaining problems, namely, the
investigation of the multibody case and of the possibility of complete determination of the
dynamics by SUSY.
20
1
g2
[D , Xm ]2 [Xm , Xn ]2
2
4
1 T
1
T m
+ [D , ] g [Xm , ] ,
2
2
D igA.
S0 = Tr
(2.1)
(2.2)
In this expression, Xijm ( ), Aij ( ) and ,ij ( ) are the (N + 1) (N + 1) Hermitian matrix
fields, representing the bosonic part of the Dparticles, the gauge fields, and the fermionic
part of the Dparticles, respectively. m are the real symmetric 16 16 SO(9) matrices,
and the vector index m runs from 1 to 9.
This action is known to possess a number of important symmetries. Among them, the
main focus in this article will be the supersymmetry, which carries 16 spinorial parameters
A =
T ,
Xm = i
T m ,
g
m
mn
= i [D , Xm ] + [Xm , Xn ]
.
2
(2.3)
(2.4)
1
Bm ( ) + Ym ( ),
g
N
( ) =
Bm ( ) = diag rm ( ), 0, 0, . . . , 0 ,
1
( ) + ( ),
g
(2.5)
N
( ) = diag ( ), 0, 0, . . . , 0 ,
(2.6)
where Bm ( ) and ( ) are the bosonic and the fermionic backgrounds expressing the
positions and the spin degrees of freedom of the Dparticles respectively and Ym ( ) and
( ) denote the quantum fluctuations around them. We will be interested in the general
case where rm ( ) and ( ) are arbitrary functions of not satisfying equations of motion.
At the tree level, the action and the SUSY transformations for the probe Dparticle are
of the form
2
r
(0)
= d
(2.7)
,
+
2g 2 2g 2
1
(0) rm =
m ,
(2.8)
i
(0) = i(/r
) ,
(2.9)
21
where the dot means differentiation with respect to and stands for , etc. After a
gauge is fixed and the functional integral over the fluctuations is performed, one obtains
the quantum effective action
= d L[rm , ].
(2.10)
Although a general argument does not exist, explicit calculations performed so far in the
socalled background gauge strongly indicate that L[rm , ] is a sum of local products
of derivatives of rm ( ) and ( ). As it will be emphasized again at the beginning of
Section 3, our proof of the Theorem and the Proposition will operate at each order in g 2
and in the derivative expansion, so that effectively we will be dealing with a finite number
of local expressions at each step regardless of whether the entire L is fully local or not.
satisfies the SUSY Ward identity
= 0,
(2.11)
where
is the quantumcorrected effective SUSY transformation operator. In a previous
article [9], we have derived the explicit form of this Ward identity in the background gauge,
where
is given in closed form in terms of expectation values of certain products of fields.
Further in [15] a complete calculation of and
at 1loop at order 4 in the derivative
expansion was performed and the validity of (2.11) was explicitly verified at that order.
As was stated in the introduction, an interesting question is, for a given
how unique
is, without the use of the equations of motion. For the simplest case, namely at 1loop
at order 2, it was found that is completely determined by
[9]. Moreover, a highly
nontrivial calculation at order 4 again showed that is unique [10]. In fact, these results
were not accidental: we shall prove that this feature persists to an arbitrary order of
perturbation theory.
2.2. Consistency equation
Since our main Theorem can be shown to follow rather easily from the Proposition, we
will first concentrate on the proof of the latter.
(0)
What we wish to show is that the solution to the equation
= 0 is unique up
(0)
to an overall constant, where the action of
is as defined in (2.8) and (2.9). This
equation, despite its simple appearance, is a highly complicated partial differential equation
involving Grassmann variables for the integrand L of . In fact, written out in full, it reads
+
(0) ( )
(0) = d
(0) rm ( )
(2.12)
= 0,
rm ( )
( )
where, in terms of L, the functional derivative /rm means (suppressing dependence)
= D[L, rm ],
rm
L
L
L
D[L, rm ]
+ 2
+ ,
rm
rm
rm
(2.13)
(2.14)
22
and similarly for / . Later we will often use the symbol DN [L, rm ] to denote
(N)
(N)
D[L, rm ] truncated at (and including) (1)N N (L/rm
), where rm
is the N th
derivative of rm with respect to .
Our basic idea for analyzing (2.12) is to look at its consistency or integrability
equation, namely,
(0) (0)
(0) (0) = 0.
(2.15)
Although this only gives a necessary condition, it will prove to be powerful enough to fix
the fermionic part of completely. The original equation
(0) = 0 then immediately
determines the bosonic part as well.
Now computing the left hand side of (2.15) by using the functional derivative as in
(2.12), one finds that the parts containing the second functional derivatives of cancel
and the result is
(0) (0)
(0) (0)
(0)
(0)
= d ,
(0) rm ( )
(2.16)
+ ,
(0) ( )
.
rm ( )
( )
The closure of the algebra (0) ,
(0) is easily calculated. On rm it simply gives the usual
timetranslation
(0) (0)
drm
,
,
rm = 2(
)
(2.17)
d
while on one finds
(0) (0)
,
= (m )(m
) (
m )(m ) .
(2.18)
By using a wellknown SO(9) matrix identity
m m
+ (cyclic in , , ) = + (cyclic in , , ),
it can be rewritten as
(0) (0)
d
,
= 2(
) + (,
),
d
where
(,
) m
(m ) +
( ) (
) + (
) .
(2.19)
(2.20)
(2.21)
Let us put these results back into (2.16). Then, by using the definition of the functional
derivative (2.14) and integration by parts, one finds that the contributions from the
canonical timetranslation part of the closure relation add up to an integral of a total
derivative as
d ( )
dL
drm ( )
+
= 2(
) d
2(
) d
(2.22)
d rm ( )
d ( )
d
and vanish. Thus, the consistency equation is reduced to
d (,
)F = 0,
(2.23)
23
where for simplicity we defined F / . This notation will be used throughout the
rest of the article. Further, since this equation must hold for arbitrary and
, we may
remove these global spinor parameters and obtain
d F = 0,
(2.24)
m
( m F ) + F + F ( F ).
F
(2.25)
There are two ways that this equation may be satisfied; either (a) F vanishes identically
or (b) F is a total derivative.
Let us begin with case (a). First, take the trace over the index set (, ). Since m
is traceless, we immediately get F = 0. Putting this back into the equation, we have
m ( F ) + F + F = 0. Now contract this with n . Then, using Tr n m =
(2.26)
n F = 0, as well.
Note that this guarantees the previous two relations, F =
There are two types of possible solutions to this algebraic equation. One is of the form
F = G,
(2.27)
where G is some SO(9) scalar. Due to the Grassmann nature of , this obviously satisfies
(2.26). The other is the case where F contains the product of 16 s, namely,
16
(2.28)
=1
which is easily checked to be SO(9) invariant. In this case, F vanishes identically for
any and and hence (2.26) is trivially satisfied.
Now let us turn to case (b), for which F = dG /d for some nonconstant G .
Going through exactly the same procedure as in case (a), we find
dG
,
(2.29)
d
m
= G 1 Tr G 1
G
(2.30)
Tr(m G).
14
14
d
( F + F ) ( F + F ), we must demand that the
Rewriting F + F = d
2 Remember that implicitly we are dealing with expressions order by order in the double sense and hence
locality argument is valid.
24
Now consider case (b2). For this to be possible at all, F must contain the derivative
of , namely, it must be of the form F = f for some SO(9) invariant f . Then, we
have F + F = ( + )f . As this does not contain f, needed to form a
total derivative, f can only be a constant. This reduces the possibility to the expression
+ . But this is clearly not a total derivative and hence case (b2) is also
excluded.
Summarizing, we have shown that only case (a) can be possible and the consistency
equation is reduced to
= D[L, ] = G + H ,
(2.31)
25
(3.2)
2 (L/ ).
(3.3)
On the righthand side (RHS), the first term is a total derivative and can be dropped, while
due to antisymmetry the second term is actually the negative of the lefthand side (LHS).
Thus we get
1
E = E .
(3.4)
2
This does not contain any more and can be absorbed in the C term.
3. Having disposed of the E term, look at the remaining term with , namely, the
D term. It can be rewritten as
D = ( D ) D D .
(3.5)
Thus, arises only from the total derivative term, which we may discard, and the D term
can be absorbed by the other terms with less number of derivatives on the spinors. At this
stage all the terms containing SPLD have disappeared.
4. Next consider the C term, with two powers of , and look for terms with in
D2 [L; ]. It is produced only from (L/ ) and is given by C . This must vanish
and hence C = 0.
5. We are now left with L = (1/2) A + B . The B term can be rewritten as
B = ( B ) + B T B.
(3.6)
26
(A ) + 2(BS ) + (B S ) = G.
(3.8)
(3.9)
.
E1 2 16
16! 1 2 16
Then, if L contains , it must be of the form
(N) (N)
(3.10)
(3.11)
(N)
L = 1 2 n Z1 2 n
(N) (N)
(N)
= 1 2 16 1 2 n E1 2 16 Z1 2 n ,
(3.12)
where Z1 2 n may contain derivatives of the spinor up to (N1) except for s, which
are explicitly exhibited. Now look at the terms of the form ( )15 (N+1) ( (N) )n1 in
DN [L, ]. There are only two sources for such a structure. One is produced as
L
16n2 16 (N+1)
(N)
(N)
E2 16 Z1 2 n .
n
1
2
(3.13)
27
(3.14)
Since the product E1 2 16 Z1 2 n cannot be totally antisymmetric in its entire index
set (1 16 , 1 n ), (3.13) and (3.14) are independent and cannot cancel each other.
Moreover, neither term can be proportional to due to the total antisymmetry of E1 16 .
(For (3.13) it is obvious. For (3.14), if there exists 1 , then there must also be 16 ,
(N+1)
instead of .) Therefore, they must vanish separately and we find
which produces
Z1 n = 0, which proves the lemma.
The second lemma will severely restrict the way a product of more than two SPLDs can
appear in the effective Lagrangian L.
Lemma 2. Let (N) , with N 2, be the SPLD in L and assume that there are at least two
of them. Then, they can only be contained through the special combination (N) .
Proof. In general such structures with different number of (N) s may appear in L.
However, as they lead to algebraically independent expressions in the basic equation, we
may treat each term separately. Thus, we consider a generic structure of that type and write
it as
(N)
(N)
Z1 2 m ;1 2 n ,
L = 1 2 m (N)
n
1
2
(3.15)
where Z1 2 m ;1 2 n does not contain nor (N) and is separately antisymmetric in the
indices {1 2 m } and in {1 2 n }. Now look at the structure containing (2N) in
DN [L, ]. Such a structure can only be produced from N (L/ (N) ) and is proportional
to
(N)
(N)
(2N) Z1 2 m ;1 2 n1 .
1 2 m (N)
1
2
n2 n1
(3.16)
28
L=
m
(N) Zm ,
(3.17)
m2
where Zm contains derivatives of only up to (N1) . We now look at the part involving
(2N1) and in DN [L, ]. We will distinguish two cases.
First suppose Zm does not contain a term of the form ( (N1) )n . Then, the only way
(N)
that such a structure is produced is through N L/ and is given by
N
L
(N)
(N) m2
Nm(m 1) (2N1)
Zm .
(3.18)
Since this is not proportional to nor to (whose existence has already been excluded
by Lemma 1) and since m 2, this must vanish and hence Zm = 0.
On the other hand, if Zm contains ( (N1) )n , there will be another source. In such a
case, we may write L in the form
n
m
(N) (N1) Zmn .
L=
(3.19)
m2,n
Suppressing the summation symbol, the contribution of the previous type takes the form
N
L
(N)
(N) m2 (N1) n
Zmn ,
Nm(m 1) (2N1)
(3.20)
L
(N1)
m n
Zm n .
(3.21)
n = n + 1,
(3.22)
(3.23)
(3.24)
Now let the largest m and n be mmax and nmax , respectively. From the above equation,
we find that Zmmax ,nmax must be proportional to Zmmax 1,nmax +1 , which, however, does not
exist. Thus Zmmax ,nmax must vanish. This contradicts the assumption that mmax , nmax are
the maximum values for m, n. Applying this logic recursively, we conclude that Zmn must
vanish for all m, n and L may depend on (N) only linearly.
The last lemma will completely dispose of SPLD with more than two time derivatives.
Lemma 4. Let the SPLD in L be (N) with N 2 and assume that L depends on (N)
linearly. Then, the dependence of L on (N) can only be through a total derivative.
29
(3.26)
(N1) (N)
Z; ,
L =
(3.28)
and a priori there is no symmetry property required of Z; . Now look at the terms
(N)
and
containing (2N1) in DN [L, ]. Such terms can only be produced by N L/
(N1)
N1
L/
and never arise by acting on Z; . Explicitly, these two contributions
add up to give
(1)N1 (2N1) (Z; + Z; ).
(3.29)
30
L is then
L=
1 2 n 1 2 m Z1 2 n ;1 2 m .
(3.31)
First consider the terms with more than two s. Then from L/ , we will get a term
proportional to
1 2 n 2 3 m Z1 2 n ;2 m ,
(3.32)
which cannot be produced in any other way. Such a term is allowed if and only if it is
proportional to . For this to happen, Z1 2 n ;2 m must contain 3 and similar
Kronecker s. However, due to the total antisymmetry of the index set (, 2 , . . . , m ),
this is not possible. Hence such terms cannot be present in L.
We are left with the possibility with zero or one and the most general form of L is
1 2 n Z1 2 n + 1 2 n 1 Z1 2 n ;1 .
L=
(3.33)
The terms with a can be rewritten as
L1 1 2 n 1 Z1 2 n ;1
= (1 2 n 1 Z1 2 n ;1 )
n1 2 n 1 Z1 2 n ;1 1 2 n 1 Z 1 2 n ;1 .
(3.34)
Consider first the case where n 2. Then if Z1 2 n ;1 is not totally antisymmetric, the
second and the third term vanish and hence L1 is a total derivative, which can be dropped.
On the other hand if Z1 2 n ;1 is totally antisymmetric, then the second term is equal to
nL1 . Thus we may solve for L1 and get
1
(1 2 n 1 Z1 2 n ;1 ) 1 2 n 1 Z 1 2 n ;1 .
n+1
(3.35)
Z; = ( Z; ) + Z; .
(3.36)
(3.38)
31
ZS is a constant and, since the only symmetric constant bispinor is , it must be of the
form (ZS ) = c , with c a constant. Then the basic equation is satisfied with G = 2c.
Thus, adding the independent contribution [r] which has hitherto been suppressed,
we have found that the solution to the basic consistency equation can only be of the form
=c
d ( ) + [r].
(3.39)
g 2n (n) ,
 =
n0
g 2m2  (m) .
(3.40)
m1
k1
(n)  kn .
(3.41)
n=0
(3.42)
Our Proposition dictates that the only possible solution to the first equation is  (1) =
with c a constant independent of g. Such an expression does not have the
c d (r 2 + ),
right dimension and hence  (1) must vanish. This in turn makes the second equation to
be of the form (0)  (2) = 0 and for the same reason we get  (2) = 0. Repetition of this
reasoning clearly shows that  = 0 altogether. This completes the proof.
32
4. Discussions
As the results of our present investigation have already been wellsummarized, we shall
not repeat them. Below we shall discuss two important future problems and indicate how
they should be attacked.
One immediate question is whether similar results hold for the multibody case.
Replacing rm and by rm,i and ,i , where the extra index i refers to different Dparticles,
it is easy to see that our consistency equation (2.31) is generalized to
( ,i F,i + ,i F ,i ) = 0.
(4.1)
i
Apart from the special case where F,i contains a product of 16 s for each particle, this
equation can be algebraically satisfied if and only if
F,i = Sij ,j ,
(4.2)
where Sij is a symmetric matrix. There are many such matrices, examples of which are
rim rjm , i mn j rm rn , etc. Therefore, the analysis of its solution becomes more involved
compared to the single Dparticle case treated in this work. Nevertheless, it would be quite
interesting and important to see, first, if one can construct a nontrivial solution, thereby
proving that SUSY alone cannot determine the structure of the effective action uniquely
for the multibody case. If serious efforts in this direction should all fail, that would then
indicate a possibility to be able to construct a uniqueness proof.
The second and more ambitious enterprise is, as was mentioned in the introduction,
to study if SUSY alone truly determines the dynamics of the theory. In other words,
one would like to know whether the requirement of SUSY determines both the effective
transformation law and the effective action at the same time, up to some field redefinitions.
To formulate this problem more precisely, we must define what we mean by the effective
SUSY transformation to begin with. In the absence of offshell superfield formalism, this
is already nontrivial since the closure relations, only by which one can guarantee that the
transformation is SUSY, themselves must depend on the form of the full effective action.
Consider, for example, the simplest case of the sourceprobe situation. Assuming that the
noncanonical parts of the closure relations are linear in the equations of motion, 3 the set
of equations characterizing SUSY would be of the form
= T
,
(4.3)
rm = m
,
(4.4)
+ B n
rn
[
, ]rm = 2(
)rm + Cm
+ Dm n
rn
= 0,
[
, ] = 2(
) + A
(4.5)
(4.6)
(4.7)
3 Although we have no proof, judging from the results at tree and 1loop levels, nonlinear dependence is highly
unlikely.
33
where
and are the global SUSY parameters and A D are some, in general field
dependent, coefficients. To show that SUSY determines the dynamics completely, one must
solve this set of nonlinear functional differential equations, with the treelevel information
as the only input, and demonstrate that all the quantities are fixed uniquely, up to field
redefinitions. A preliminary investigation indicates that at least at low orders this might be
feasible.
We hope to report on these and related matters in future communications.
Acknowledgement
Y.K. is grateful to T. Yoneya for a useful discussion. The research of Y.K. is supported in
part by GrantinAid for Scientific Research on Priority Area #707 Supersymmetry and
Unified Theory of Elementary Particles No. 10209204 and GrantinAid for Scientific
Research (B) No. 12440060, while that of T.M. is supported in part by the Japan Society
for Promotion of Science under the Predoctoral Research Program No. 129617, all from
the Japan Ministry of Education, Culture, Sports, Science and Technology.
References
[1] T. Banks, W. Fischler, S.H. Shenker, L. Susskind, Phys. Rev. D 55 (1997) 5112, hepth/9610043.
[2] L. Susskind, hepth/9704080.
[3] S. Paban, S. Sethi, M. Stern, Nucl. Phys. B 534 (1998) 137, hepth/9805018.
[4] S. Paban, S. Sethi, M. Stern, J. High Energy Phys. 06 (1998) 012, hepth/9806028.
[5] D.A. Lowe, J. High Energy Phys. 11 (1998) 009, hepth/9810075.
[6] S. Sethi, M. Stern, J. High Energy Phys. 06 (1999) 004, hepth/9903049.
[7] S. Hyun, Y. Kiem, H. Shin, Nucl. Phys. B 558 (1999) 349, hepth/9903022.
[8] H. Nicolai, J. Plefka, Phys. Lett. B 477 (2000) 309, hepth/0001106.
[9] Y. Kazama, T. Muramatsu, Nucl. Phys. 584 (2000) 171, hepth/0003161.
[10] Y. Kazama, T. Muramatsu, hepth/0106218.
[11] K. Becker, M. Becker, J. Polchinski, A. Tseytlin, Phys. Rev. D 56 (1997) 3174, hepth/9706072.
[12] Y. Okawa, T. Yoneya, Nucl. Phys. B 538 (1999) 67, hepth/9806108.
[13] Y. Okawa, T. Yoneya, Nucl. Phys. B 541 (1999) 163, hepth/9808188.
[14] W. Taylor, M.V. Raamsdonk, J. High Energy Phys. 04 (1999) 013, hepth/9812239.
[15] Y. Kazama, T. Muramatsu, Class. Quantum Grav. 18 (2001) 2277, hepth/0103116.
Protected operators in N = 2, 4
supersymmetric theories
Nicola Maggiore a , Alessandro Tanzini b
a Dipartimento di Fisica, Universit di Genova, via Dodecaneso 33 and INFN, Sezione di Genova,
Abstract
The anomalous dimension of single and multitrace composite operators of scalar fields is shown to
vanish at all orders of the perturbative series. The proof hold for theories with N = 2 supersymmetry
with any number of hypermultiplets in a generic representation of the gauge group. It then applies
to the finite N = 4 theory as well as to nonconformal N = 2 models. 2001 Published by Elsevier
Science B.V.
PACS: 03.70.+k; 11.10.Gh; 11.30.Pb
Keywords: Extended supersymmetry; BRS quantization; AdSCFT correspondence
1. Introduction
It has been known for a long time that quantum field theories with extended
supersymmetry display a set of remarkable nonrenormalization properties. The function
of the gauge coupling was argued to vanish to all orders of perturbation theory for the
N = 4 Super YangMills theory (SYM) [1] and to receive only oneloop corrections
for the N = 2 case [2]. The conjectured AdS/CFT correspondence [3] has renewed the
interest on the finiteness properties of these theories; in fact, many of the tests of the
correspondence have relied on nonrenormalization properties, which are crucial in order
to ensure a meaningful comparison between the strong coupling regime, accessible by
type IIB supergravity computations, and the weak coupling one, where the usual field
theory techniques are reliable.
In particular, the prototype example of the correspondence [3] establish a duality
between type IIB superstring on AdS5 S 5 and N = 4 SYM in the superconformal
Email addresses: maggiore@ge.infn.it (N. Maggiore), tanzini@lpthe.jussieu.fr (A. Tanzini).
05503213/01/$ see front matter 2001 Published by Elsevier Science B.V.
PII: S 0 5 5 0  3 2 1 3 ( 0 1 ) 0 0 3 9 8  4
35
phase. In this context, it has been shown that chiral gauge invariant operators belonging
to short multiplets of the superconformal group SU(2, 24) have vanishing anomalous
dimensions [4]. The analysis of correlation functions of Chiral Primary Operators (CPO),
which are the lowest scalar components of such short superconformal multiplets, provided
a highly nontrivial check for the AdS/CFT correspondence [5,6]. Moreover, it recently led
to discover new doubletrace operators which are protected despite they do not obey any
of the known shortening conditions [7,8].
We remark that the nonrenormalization properties of some CPOs does not necessarily
rely on superconformal algebra, and it can actually be shown also for theories with less
number of supersymmetries. This is indeed the case for the gaugeinvariant operator Tr 2 ,
where in the N = 2 formalism corresponds to the complex scalar field of the vector
multiplet. The anomalous dimension of this operator has been shown to vanish to all orders
of the perturbative series in pure N = 2 SYM [9]; this allowed to provide a rigorous proof
of the celebrated nonrenormalization theorems for the beta functions of N = 2 and N = 4
theories [9,10].
In this work we extend the analysis of [9] to higher rank gaugeinvariant polynomials
including both single and multitrace operators. It is worth
of the scalar field and ,
to underline that our proof holds for N = 2 theories with any number of hypermultiplets
in a generic representation of the gauge group. It then applies to the finite N = 4 theory
as well as to nonconformal models, which are at present object of intensive research in
extensions of the AdS/CFT duality to phenomenologically more interesting gauge theories
with a nontrivial renormalization group flow [11,12].
A geometrical interpretation of such nonrenormalization properties can be given by
resorting to the topologically twisted formulation of the N = 2 theory. In fact, in this
context, the operators we consider can be written as components of the Chern classes of
the universal bundle defined in [22].
This work is organized as follows. In Section 2 we sketch the basic features of the
twisting procedure. In Section 3 we give the proof of finiteness of the operators Tr k ,
k 2 (singletraced) and i Tr ki , i ki = k (multitraced). Our conclusions are drawn in
Section 4, while we confined into appendices the unavoidable technicalities of our proofs
and the explicit examples k = 2, 3, 4.
2. The twist
It is well known that N = 2 SYM theory can be given a twisted formulation [13],
which on flat manifolds is completely equivalent to the conventional one. Nonetheless,
the use of the twisted variables makes more evident and easily understandable some
important features, like the topological nature of a subset of observables [13] and
the nonrenormalization theorem concerning the betafunction of the gauge coupling
constant [9]. In view of these advantages, we choose to work in the twisted version of
N = 2 SYM theory. We address the interested reader to the existing literature [13,15] for
36
what concerns the many and deep aspects of the twisting procedure. In this section, we
prefer rather to set up the field theoretical background in which we shall work.
j ) are characterized by an index i = 1, 2, which
The usual N = 2 supercharges (Qi , Q
counts the number of supersymmetries, and a Weyl spinor index , which runs on the same
values as i : = 1, 2. At the origin of the twist, is Wittens idea of identifying the indices
i .
(2.1)
1
Q ,
(2.2)
2
1
Q ,
(2.3)
2
1
( ) Q .
(2.4)
2
The N = 2 supersymmetry algebra, written for the twisted supercharges, contains a
subalgebra, formed by , and by the translations :
2 = 0,
{, } = ,
{ , } = 0.
(2.5)
We call the subalgebra (2.5) topological, since it appears to be common to all known
topological field theories [17], both of the Witten and Schwartz type [18]. The three twisted
supercharges turn out to be redundant, since they do not play any role either in the
classical definition or in the quantum extension of the theory [14]. For this reason, we shall
discard them throughout the rest of our reasoning.
( ) ,
( ) () ,
(2.7)
are the anticommuting fields resulting from the twist of the spinor fields (i , i ) into
and the gauge boson A are not altered by
( , ). Notice that the scalar fields (, )
the twisting operation.
The N = 2 SYM pure gauge action twists to the TYM action
N =2
A , i , i , , STYM A , , , , , ,
SSYM
(2.8)
which, in the 4D flat Euclidean spacetime reads [14]:
1
+ +
STYM = 2 Tr d 4 x 12 F
F
(D D )+ + D
g
D + 1 , 1 ,
12 D
2
2
1
18 [, ] 32
, , ,
(2.9)
37
1
+ =F
, F
+ = F+ = 1
+ and
where F
+ 2 F
2 F
1
(D D )+ = (D D ) + D D .
2
The symmetries of the action (2.9) are the usual gauge invariance
gaugeSTYM = 0,
(2.10)
(2.11)
It is apparent from (2.7), that the 4D flat Euclidean twist simply corresponds to a linear
change of variables, therefore, it is not felt by the partition function
Z = DeS[] .
(2.12)
Moreover, the stressenergy tensor of the theory is modified by the twist only by a
total derivative term which do not affect the translation generators [15]. This observation
underlies the equivalence of the two, twistrelated, theories N = 2 SYM and TYM when
formulated on flat manifolds. The theory has been extended to the quantum level both in
its untwisted [19] and twisted [14] formulation. Moreover, in [9] has been demonstrated
that a renormalization scheme does exist, in which the function of the gauge coupling
constant g receives one loop contributions only.
(3.1)
where s is the nilpotent BRS operator, is the twisted scalar supercharge (2.2), and is a
global commuting ghost.
The operator Q describes an invariance of the action STYM
QSTYM = 0,
(3.2)
and it is nilpotent
Q2 = 0,
since s 2 = 2 = {s, } = 0.
(3.3)
38
For our purposes, the Qvariations of the scalar field and of the FaddeevPopov ghost
c are sufficient
Q a = f abc cb c ,
1
Qca = f abc cb cc 2 a ,
2
(3.4)
where the index a = 1, . . . , dim G counts the adjoint representation matrices of the gauge
group G, and f abc are the structure constants of the corresponding Lie algebra.
The operators Tr k and i Tr ki , i ki = k are the observables of TYM (hence N = 2
SYM) theory [13,14]; we can write them as
D a1 ak a1 ak ,
(3.5)
D a1 a2 = a1 a2 ,
D a1 a2 a3 = a1 a2 a3 ,
..
.
,
p=0
(3.8)
39
(k 1)!k!
(1)kp
,
2(kp) p!(2k p 1)!
0 p k 1,
(3.9)
(3.10)
Notice that the coefficient of the last term on the r.h.s. of (3.7) is universal
(k)
fk1 =
1
.
2
(3.11)
It is worth to remark that the operators (3.5) we are considering have an interesting
geometrical interpretation [22]. Let us consider for the moment k = 2; by redefining the
scalar field as 2 , (3.8) can be written as
Tr 2 = Q Tr cQc 23 ccc .
(3.12)
If we define the universal connection A = A + c, (3.12) can be seen as the (p = 0, q = 4)
component of the equation
F
= dA
CS = d Tr A dA 2 A A A ,
Tr F
(3.13)
3
where by (p, q) we indicate the grading of the forms with respect to the extended exterior
derivative d = d + Q, p being the spacetime form degree and q the ghost number. Thus
P (2)(c, ) can be understood as the (0, 3) component of the ChernSimons form ACS
while as the (0, 2) component of the corresponding
of the universal connection A,
. Analogously, higher rank single trace operators (3.5) can be expressed in
curvature F
k .
terms of (0, k) components of higher Chern classes Tr F
We now come to the proof of finiteness; what is actually important in this sense is that
operators (3.5) can be expressed as the Qvariations of polynomials in the fields and
c, and, in particular, that they are related to D a1 ak ca1 sca2 scak (the first term on the
r.h.s. of (3.7)), which belongs to the cohomology of the ordinary BRS operator, and is
known to be finite to all orders of perturbation theory [20]. It is then natural to expect
that the nonrenormalization property of the BRS invariant odd polynomials in the ghost c
translates to (3.5) as well. As we shall see, this is indeed the case.
Let us consider now the classical action
= STYM + Sgf + Sext + Sk .
(3.14)
STYM is the twisted N = 2 SYM action (2.9), Sgf is the (Landau) gauge fixed term
A ,
Sgf = Q Tr d 4 x c
(3.15)
c being the FaddeevPopov antighost. According to the usual BRS procedure to implement
at the quantum level classical field theories, in Sext the nonlinear Qvariations of the
40
(3.16)
(3.17)
p=2
In the above expression for Sk , the s are external sources coupled to (k) (), P (k)(c, )
(defined by (3.8)), and its derivatives with respect to the ghost c (a /ca ). They are
antisymmetric in their color indices as, and are commuting or anticommuting objects
depending from the number of ghosts c contained in the composite operators they are
coupled to. Such a number is represented by the lower indices, hence their statistics is given
by (1)2kp . Notice that, thanks to the overall multiplying factor 2k , the functional Sk
is, as it should for being part of the action, an analytic expression in .
The whole action satisfies the SlavnovTaylor (ST) identity
S() = S old () + S (k) = 0,
(3.18)
S old () =
Tr d 4 x
+
c
(3.19)
(k)
4
mnap1 mna1 ap2
S = d x 2k1
f
+
2kp1
a1 ap1 .
0
2
2kp
p=2
(3.20)
Notice that the sources, transforming one into the other, enter the ST identity (3.18) as
BRS doublets, which have the nice property of keeping unaltered the cohomology of the
(linearized) ST operator.
For the proof of finiteness, it is crucial that an additional constraint on the theory is
satisfied: the ghost equation, peculiar of the Landau gauge [21]. Its main consequence is
a nonrenormalization theorem, which states that, in the Landau gauge, the ghost field c
enters the quantum theory only if differentiated ( c). It is precisely for the purpose of
being able to take advantage of this additional symmetry of the theory, that in Sk external
sources have been coupled not only to (k) () and P (k) (c, ), but also to the operators
a1 ap1 P (k)(c, ). The ghost equation reads (see Appendix C)
F a = a ,
where
Fa =
(3.21)
d 4x
2k3
a
a
1
p1
+ f abc cb c
(1)2kp 2kp
aa1 ap1 ,
ca
b
2kp1
p=1
(3.22)
41
(3.23)
indeed, the operator a a1 a2k3 P (k)(c, ) is independent from the field , and it is
linear in the ghost field c.
Once observed that both the ST identity (3.18) and the ghost equation (3.21), being
anomalyfree, can safely be extended to the quantum level [14]
S( ) = 0,
F a = a ,
(3.24)
where is the 1PI generating functional = + O(h ), we are able to prove our main
result, i.e., that the quantum insertions Tr k = Tr k + O(h ) have vanishing anomalous
dimensions, namely, satisfy the CallanSymanzik (CS) equation
C Tr k (x) = 0,
(3.25)
where C is the CS operator
C
+ h g
h
N
(3.26)
satisfying
C = 0.
(3.27)
(3.28)
S( )
= 0,
2k1
=0
where we put to zero the set of external sources appearing in Sk (3.17). From the
expression of the ST operator, Eq. (3.28) writes
b
(3.29)
=
.
2k1 =0
0 =0
The operator b appearing in the above equation is the (offshell nilpotent) linearized
quantum ST operator
4
+
+ c
b = d x Tr
+ 2k1
b
0
2k3
p(1 p)
and it can be recognized as the quantum functional translation of the operator Q. The
relation (3.29) is the quantum extension of the classical one (3.8) between the composite
42
D a1 ak a1 ak = b P (k) (c, ) .
(3.31)
Owing to the ghost equation (3.21), the quantum insertion P (k)(c, ) has vanishing
anomalous dimension, since the undifferentiated ghost field c does not get quantum
corrections
C P (k) (c, ) = 0.
(3.32)
Now, since the CS operator commutes with the linearized quantum ST operator b
[C, b ] = 0,
(3.33)
from Eqs. (3.32) and (3.33) we derive our result (3.25): the composite operators Tr k are
finite to all orders of perturbation theory. This result can be trivially extended to the class
of operators Tr k .
We would like to stress once again that the presence of matter does not alter in any way
our result (3.25). Indeed, we needed only to consider the symmetry transformations (3.4)
of the scalar field of the N = 2 gauge supermultiplet and of the FaddeevPopov ghost
field c. These transformations are not altered by the presence of matter fields, hence the
matter sector is completely decoupled in our proof of existence of protected operators.
Under this respect, matter is only an avoidable graphical charge, which we preferred to
omit. The class of protected operators discussed in this paper refers, therefore, to N = 2
SYM theories coupled to matter in an arbitrary representation of the gauge group, hence to
N = 4 theory as well, since for the particular, choice of matter in the adjoint representation,
N = 4 SYM is recovered from the N = 2 case.
4. Conclusions
We have shown the vanishing to all orders of the perturbative expansion of the
anomalous dimension of local single and multitrace composite operators of scalar fields
in theories with N = 2 supersymmetry. Although the proof has been given explicitly for
the pure gauge case, it holds unaltered for N = 2 SYM coupled to matter belonging to
a generic representation of the gauge group G, hence in particular it remains valid for
the N = 4 case, obtained by putting matter multiplets in the adjoint representation of G.
Nonrenormalization properties of these operators can be better displayed using the twisted
formulation of the theory, where they have a nice geometrical interpretation as components
of the Chern classes of a suitable universal bundle [22]. The study of correlation functions
of such operators play a relevant role in tests of the AdS/CFT correspondence [5,6]. In
addition, it recently led to conjecture the existence of a new class of gauge invariant
operators which are protected despite they do not obey any of the known shortening
conditions [7,8].
We stress that our analysis does not rely on the superconformal algebra. It would then
be interesting to extend it to the above mentioned operators. Moreover, the approach we
43
followed can be applied also to nonconformal N = 2 theories which have been recently
considered in extensions of the AdS/CFT duality [11,12]. Even if a precise map between
bulk supergravity fields and boundary gaugeinvariant field theory operators, as that of
the N = 4 superconformal theory, is not presently known for these cases, the appearance
of protected operators in theories with a nontrivial renormalization group flow is a very
interesting feature by itself, and could motivate further analysis on the correlation functions
of these operators.
Acknowledgements
We acknowledge helpful discussions with L. Baulieu, M. Bianchi, S. Kovacs and
C. Schweigert. A.T. is supported by a fellowship of the MURST project. This work
is partially supported by the European Commission, RTN programme HPRNCT200000131 and by MURST.
dim .
Rcharge
charge
R +
1
2
0
2
0
0
1
1
1/2
1
1
0
(A.1)
Therefore, the most general expression for (k) () = D a1 ak a1 ak having homogeneous R + quantum number is the following
k1
a1 ak a1
ak
a1 ak
(k) a1 a2
akp akp+1
ak
= Q D
fp c sc sc
.
D
(A.2)
p=0
(k)
In (A.2), the coefficients fp depend on the global ghost appearing in the definition of
the symmetry operator Q = s + (3.1). The first term of the sum (A.2) is proportional to
D a1 ak ca1 sca2 scak , which is the c(2k1) cocycle, belonging to the cohomology of the
BRS operator [23], and known to be finite to all orders of perturbation theory [20]. The
properties of the invariant polynomials (k) () are mostly determined by the relationship
existing between them and the BRS cocycles c(2k1) .
Our first goal is to find the explicit expression for the coefficients fp(k) in (A.2). In order
to obtain it, let us observe that
Qsca = 2 s a ,
(A.3)
44
and that
D a1 an ca1 sca2 scam am+1 an1 s an
2
D a1 an sca1 scam am+1 an .
=
(A.4)
nm
Eq. (A.4) is a consequence of the nilpotency of the BRS operator s and of the fact that the
completely symmetric rankk tensor D a1 ak is invariant, hence it satisfies the Jacobylike
identity
D p1 a2 ak f p1 qa1 + D a1 p2 a3 ak f p2 qa2 + + D a1 ak1 pk f pk qak = 0.
(A.5)
Performing the Qvariation in (A.2) and exploiting the identities (A.3), (A.4), we get
D a1 ak a1 ak
=
k1
p=1
k2
a1 ak
fp(k) 2
p=0
2k
sca1 scakp1 akp ak
1
p+1
fk1 2 D a1 ak a1 ak .
(k)
k1
p=1
2k
(k)
fp1
D a1 ak sca1 scakp akp+1 ak 2 1
fp(k) . (A.6)
p
2k
(k)
(k)
2
fp = 1
fp1 ,
p
(k)
fk1
=
(A.7)
1 p k 1.
(A.8)
(k 1)!k!
(1)kp
,
2(kp) p!(2k p 1)!
0 p k 1,
(A.9)
as can be verified by induction. This proves the relations (3.7) and (3.9).
For future reference, it may be useful to write (3.7) for the lowest values k = 2, 3, 4.
For k = 2, the invariant tensor D ab is just the Kronecker delta ab , and from (3.7)
and (3.9) we immediately get
(2)
1 a a
1 a a
(2)
c
sc
a a = Q f0 ca sca + f1 ca a = Q
(A.10)
.
34
2
45
For k = 3, there is only one completely symmetric tensor D abc = d abc , and
d abc a b c = Qd abc f0(3) ca scb scc + f1(3) ca scb c + f2(3) ca b c
1 a b c
1 a b c
1 a b c
abc
= Qd
c sc sc + 4 c sc 2 c .
106
2
(A.11)
+ f2(4) ca scb c d + f3(4) ca b c d
1 a b c d
1
= QD abcd
c sc sc sc 6 ca scb scc d
8
35
5
3 a b c d
1 a b c d
+ 4 c sc 2 c .
5
(A.12)
by demanding
QSk = 0.
(B.2)
We have
QSk =
2k
d x (Q0 ) (k) +
4
2k2
a1 ap1
Q2kp
a1 ap1 P (k)
p=1
a1 ap1
Qa1
+ (1)2kp 2kp
ap1 P
(k)
,
(B.3)
where we taked into account the Qinvariance of the polynomials (k) () and the statistics
of the sources, given by (1)2kp .
In order to evaluate (B.3), we observe that the anticommutator between the derivative
with respect to the ghost field c (a ) and the symmetry operator Q, gives the generators of
the rigid gauge transformations Ha
{a , Q}Xb = Ha Xb = f abc Xc ,
(B.4)
where Xa is a generic function of the fields and c and belongs to the adjoint
representation of the gauge group. From the obvious relations
Ha (k) () = Ha P (k)(c, ) = a QP (k) (c, ) = 0,
(B.5)
46
we derive
Qa P (k) = 0
(B.6)
and, in general,
a1 am Qa1 am P (k) = (1)m+1
m(m 1) a1 a2 p
a3 am p P (k) a1 am .
f
2
(B.7)
QSk =
2k
2k2
a1 ap1
Q2kp
a1 ap1 P (k)
p=1
(p 1)(p 2) a1 ap1 a1 a2 q
(k)
2kp f
+
a3 ap1 aq P
2
2k
4
=
d x (Q0 ) (k) 2k1 (k) + (Q2k1 )P (k)
a a
+ Q2 1 2k3 a1 a2k3 P (k)
2k3
a1 ap1 p(p 1) rsa1 ap2
2kp1 f rsap1 a1 ap1 P (k) .
Q2kp
+
+
2
p=2
(B.8)
In order that the whole expression QSk vanish, by identifying terms containing equal
number of ghost fields c, we have
Q0 = 2k1 ,
a a
Q2 1 2k3
(B.9)
= 0,
(B.10)
Q2k1 = 0,
(B.11)
and
a a
1
p1
Q2kp
=
2 p 2k 3.
(B.12)
(k)
4
mnap1 mna1 ap2
f
+
2kp1
S = d x 2k1
a1 ap1 ,
0
2
2kp
p=2
(B.13)
which coincides with (3.20).
47
2k2
a a
1
p1
(1)2kp 2kp
a a1 ap1 P (k)
p=1
2k3
a a
1
p1
(1)2kp 2kp
p=1
a a2k3
d 4 x 2 1
aa1 ap1
2kp1
a a1 a2k3 P (k).
(C.1)
Notice that, according to the expression (3.7), the maximum number of ghosts c present
in P (k)(c, ) is 2k 1, in the term c(2k1), whereas the maximum number of its ghost
derivatives in the last term at the r.h.s. of (C.1) is 2k 2, hence this term is linear in c and
independent from , i.e., it represents a classical breaking.
Therefore, the complete ghost equation reads
F a = a ,
where
F =
a
(C.2)
d x
2k3
abc b
2kp a1 ap1
+ f c
(1)
2kp
aa1 ap1 ,
ca
bc
2kp1
(C.3)
p=1
(C.4)
References
[1] M.F. Sohnius, P.C. West, Phys. Lett. B 100 (1981) 245;
S. Mandelstam, Nucl. Phys. B 213 (1983) 149;
P.S. Howe, K.S. Stelle, P.K. Townsend, Nucl. Phys. B 236 (1984) 125.
[2] P.S. Howe, K.S. Stelle, P.C. West, Phys. Lett. B 124 (1983) 55.
[3] J. Maldacena, Supergravity, Adv. Theor. Math. Phys. 2 (1998) 231;
J. Maldacena, Int. J. Theor. Phys. 38 (1998) 231, hepth/9711200.
[4] S. Ferrara, C. Fronsdal, A. Zaffaroni, Nucl. Phys. B 532 (1998) 153, hepth/9802203;
L. Andrianopoli, S. Ferrara, Phys. Lett. B 430 (1998) 248, hepth/9803171;
L. Andrianopoli, S. Ferrara, Lett. Math. Phys. 46 (1998) 265, hepth/9807150;
L. Andrianopoli, S. Ferrara, Lett. Math. Phys. 48 (1999) 145, hepth/9812067.
[5] M. Bianchi, S. Kovacs, Phys. Lett. B 468 (1999) 102, hepth/9910016;
M. Bianchi, S. Kovacs, G. Rossi, Y.S. Stanev, Nucl. Phys. B 584 (2000) 216, hepth/0003203;
M. Bianchi, S. Kovacs, G. Rossi, Y.S. Stanev, JHEP 9908 (1999) 020, hepth/9906188.
48
[6] B. Eden, P.S. Howe, P.C. West, Phys. Lett. B 463 (1999) 19, hepth/9905085;
B. Eden, P.S. Howe, C. Schubert, E. Sokatchev, P.C. West, Phys. Lett. B 472 (2000) 323, hepth/9910150;
P.S. Howe, C. Schubert, E. Sokatchev, P.C. West, Nucl. Phys. B 571 (2000) 71, hepth/9910011;
B.U. Eden, P.S. Howe, E. Sokatchev, P.C. West, Phys. Lett. B 494 (2000) 141, hepth/0004102.
[7] M. Bianchi, S. Kovacs, G. Rossi, Y.S. Stanev, hepth/0104016.
[8] G. Arutyunov, B. Eden, A.C. Petkou, E. Sokatchev, hepth/0103230.
[9] A. Blasi, V.E. Lemes, N. Maggiore, S.P. Sorella, A. Tanzini, O.S. Ventura, L.C. Vilar,
JHEP 0005 (2000) 039, hepth/0004048.
[10] V.E. Lemes, M.S. Sarandy, S.P. Sorella, A. Tanzini, O.S. Ventura, Operator, JHEP 0101 (2001)
016, hepth/0011001;
V.E.R. Lemes, N. Maggiore, M.S. Sarandy, S.P. Sorella, A. Tanzini, O.S. Ventura, Nonrenormalization theorems for N = 2 super YangMills, in: J. Bagger, S. Duplij, W. Siegel (Eds.),
Concise Encyclopedia of Supersymmetry, Kluwer, Dordrecht, hepth/0012197;
V.E. Lemes, M.S. Sarandy, S.P. Sorella, O.S. Ventura, L.C. Vilar, hepth/0103110.
[11] J. Polchinski, Int. J. Mod. Phys. A 16 (2001) 707, hepth/0011193;
M. Bertolini, P. Di Vecchia, M. Frau, A. Lerda, R. Marotta, I. Pesando, JHEP 0102 (2001) 014,
hepth/0011077;
M. Petrini, R. Russo, A. Zaffaroni, hepth/0104026.
[12] N. Evans, C.V. Johnson, M. Petrini, JHEP 0010 (2000) 022, hepth/0008081;
K. Pilch, N.P. Warner, Nucl. Phys. B 594 (2001) 209, hepth/0004063;
A. Brandhuber, K. Sfetsos, Phys. Lett. B 488 (2000) 373, hepth/0004148.
[13] E. Witten, Commun. Math. Phys. 117 (1988) 353.
[14] F. Fucito, A. Tanzini, L.C. Vilar, O.S. Ventura, C.A. Sasaki, S.P. Sorella, Algebraic renormalization: perturbative twisted considerations on topological YangMills theory and on N = 2
supersymmetric gauge theories, lectures given at the First School on Field Theory and Gravitation, Vitria, Esprito Santo, Brazil, hepth/9707209;
A. Tanzini, O.S. Ventura, L.C. Vilar, S.P. Sorella, J. Phys. G 26 (2000) 1117.
[15] E. Witten, Supersymmetric YangMills theory on a fourmanifold, J. Math. Phys. 35 (1994)
5101, hepth/9403195;
M. Marino, The geometry of supersymmetric gauge theories in four dimensions, hepth/9701128.
[16] J. Bagger, J. Wess, Supersymmetry and Supergravity, Princeton Univ. Press.
[17] D. Birmingham, M. Rakowski, Phys. Lett. B 269 (1991) 103.
[18] D. Birmingham, M. Blau, M. Rakowski, G. Thompson, Phys. Rep. 209 (1991) 129.
[19] N. Maggiore, Int. J. Mod. Phys. A 10 (1995) 3781, hepth/9501057;
N. Maggiore, Int. J. Mod. Phys. A 10 (1995) 3937, hepth/9412092.
[20] O. Piguet, S.P. Sorella, Nucl. Phys. B 381 (1992) 373.
[21] A. Blasi, O. Piguet, S.P. Sorella, Nucl. Phys. B 356 (1991) 154.
[22] L. Baulieu, I.M. Singer, PARLPTHE88/18 Presented at Annecy Mtg. on Conformal Field
Theory and Related Topics, Annecy, France, March 1416, 1988, Nucl. Phys. B (Proc. Suppl.) 5
(1988) 12.
[23] J. Manes, R. Stora, B. Zumino, Commun. Math. Phys. 102 (1985) 157;
B. Zumino, Nucl. Phys. B 253 (1985) 477.
Abstract
We perform an explicit twoloop computation of the Higgs mass in a fivedimensional supersymmetric theory compactified on the orbifold S 1 /Z2 , with superpotential localized on a fixedpoint
brane and supersymmetry broken in the bulk of the extra dimension by ScherkSchwarz boundary
conditions. At oneloop, the Higgs mass is finite and proportional to 1/R 2 , R being the radius of the
extra dimension. The twoloop corrections contain finite ( 1/R 2 ) and linearly divergent ( /R)
terms, where is the cutoff of the fivedimensional theory. The divergent terms are cancelled by
diagrams containing oneloop wavefunction renormalization counterterms, i.e., by linear threshold
effects of Yukawa (and gauge) couplings, at the scale 1/R. As a result there is no counterterm for the
Higgs mass, which is finite and calculable to the considered order. 2001 Elsevier Science B.V. All
rights reserved.
PACS: 12.60.Jv; 04.50.+h; 11.30.Pb; 11.10.Kk
1. Introduction
There has been recently a lot of effort in order to provide an explanation for electroweak
symmetry breaking (EWSB) by means of supersymmetric theories with one compact
(radius R) extra dimension [18]. Such theories arise naturally as low energy limits of
string theories [9] and are to be understood as effective theories, when all heavy string
modes are integrated out, with an ultraviolet (UV) cutoff of the order of the string
scale Ms .
A common claim of these models is that electroweak symmetry breaking (EWSB) is
triggered by radiative corrections which are insensitive to the scale . Some one loop
Work supported in part by CICYT, Spain, under contract AEN980816, and by EU under contracts HPRNCT200000152 and HPRNCT200000148.
Email address: mariano@makoki.iem.csic.es (M. Quirs).
05503213/01/$ see front matter 2001 Elsevier Science B.V. All rights reserved.
PII: S 0 5 5 0  3 2 1 3 ( 0 1 ) 0 0 3 8 0  7
50
calculations of the effective potential, or the Higgs mass, have been performed in order to
establish this result in field [18], and string [10] theory, and also some general arguments
have been given [5,1113].
One uses typically the socalled KaluzaKlein (KK) regularization which exploits the
fact that sums over KK modes (integration over the fifth momentum) turn out to be finite
(this is already known from finite temperature field theories). Since the summation goes
over modes with masses far beyond the cutoff of the effective theory, there have been some
doubts on the validity of such a regularization [14]. However, it has been shown that several
cutoffs which respect all symmetries of the theory lead to a insensitive result at oneloop,
which equals the one obtained in the KK regularization [15,16]. This is a very interesting
conclusion since it means that EWSB does not depend on the microscopic theory at all and
we have an effective theory with an unexpected amount of predictability. However, since
no rigorous proof has yet been given which guarantees the independence of EWSB on the
high energy theory, it seems worthwhile to explicitly check this in a two loop calculation.
In this article we will perform a twoloop calculation of the Higgs mass from the
top/stop sector of the model presented in Ref. [8]. It displays the way how supersymmetric
cancellations survive the soft breaking. There remains a linear sensitivity to the cutoff
which is, however, entirely absorbed by the supersymmetric running of the Yukawa
coupling. 1 As a consequence there is no mass counterterm to the considered order and
the Higgs mass is calculable and finite. These features would remain in the twoloop
calculation of the Higgs mass when the gauge sector is included.
The outline of this paper is as follows. In Section 2 we will summarize the main features
of the fivedimensional (5D) supersymmetric model that will be used in the twoloop
calculation. In particular the fourdimensional Lagrangian, after integration over the fifth
dimension, with all relevant interactions will be explicitly provided. The corresponding
Feynman rules can be easily deduced from the Lagrangian. In Section 3 we review, for
completeness, the diagrammatic calculation of the oneloop Higgs mass, and the twoloop
contribution is provided in Section 4, where the relevant oneloop counterterms due to the
wavefunction renormalization of the superfields, and their contribution to the twoloop
Higgs mass, is computed. Our conclusions are drawn in Section 5.
2. The model
The model we will use for the calculation is the one of Ref. [8]. The 5D spacetime is
compactified on the orbifold M S 1 /Z2 which has two fixed points at y = 0, ( = R
is the segment length) where two 3branes are located. There are two types of fields: those
living in the bulk of the extra dimension (U states), organized in N = 2 supermultiplets,
and those living on the branes and localized at the fixed points (T states), organized in
N = 1 supermultiplets. U states are: gauge fields, which appear in N = 2 vector multiplets
V = (V , 1 ; + iV5 , 2 ) and transform under the adjoint representation of the SU(3)
1 In the supersymmetric limit there is only a wavefunction renormalization for the fields which is absorbed by
the running of the Yukawa coupling. This running is of course linear for D = 5.
51
W = ht QH2 U (y)
(2.1)
which triggers EWSB at oneloop as was shown in Ref. [8]. In the rest of this paper we
will consider the leading contribution from the topstop sector and neglect, in explicit
calculations, the contributions from gauge couplings. The latter can be easily included
and will not modify qualitatively the result of this paper.
Let us express the top/stop sector of the theory in terms of fourdimensional N = 1
superfields, using the formalism of Ref. [17]
L = U U + U U + Q Q(y) + H2 H2 (y)
+ U y U + h(5)
(2.2)
t U QH2 (y) + h.c. .
The superpotential term is given by the second expression and depends on the chiral
superfields Q, H2 , U and y U . The inclusion of the derivative term, on top of the
superpotential given in (2.1), is pure convention and it is motivated by the fact that,
from a fourdimensional viewpoint, it constitutes a mass term for the modes. Notice
that we have not included a H2 H1 (y) term in the superpotential. The parameter
is a supersymmetric mass and thus its presence is completely decoupled from the
actual calculation of the possible soft supersymmetry Higgs mass counterterms, while
it makes the calculation more involved. For that reason we have decided to work in the
(phenomenologically unrealistic) case = 0. In Section 5 we will make some comments
about the impact, in the present calculation, of the term. We can anticipate, not
unexpectedly, that the final result (in fact the absence of a mass counterterm) will be
unchanged by the presence of the term.
In terms of component fields we have
L = U U + U U + Q Q(y) + H2 H2 (y)
+ i u u + i u u + i q q(y) + i h 2 h2 (y)
FU 2 FU 2 FQ 2 (y) FH2 2 (y)
W
W
W
W
FU +
FQ +
+
FU +
FH
U
U
Q
H2 2
2W
2W
2W
2W
uu
qh
qu
h2 u + h.c.
2
U U
QU
QH2
H2 U
The auxiliary fields are computed to be:
FU =
W
(5)
= ht (y)QH2 y U ,
U
(2.3)
(2.4)
2 We are using the same notation, upper case symbols, for the N = 1 chiral superfields and their scalar
components, while lower case symbols are reserved for the fermionic components, e.g., U = (U, u).
52
W
= y U,
U
W
(5)
= ht (y)U H2,
(y)FQ =
Q
W
(5)
= ht (y)U Q.
(y)FH 2 =
H2
FU =
(2.5)
(2.6)
(2.7)
(5)
(5)
(2.8)
cos(ky/R)u(k),
(2.9)
1
u =
sin(ky/R)u(k),
R k=
(2.10)
1
cos (k + )y/R U (k) ,
U=
R k=
(2.11)
1
sin (k + )y/R U (k) ,
U =
R k=
(2.12)
R k=
1 F (k) (k) 1 F (k) (k)
(k)
(k)
i u u mk u u mk u u
+
2
2
k
+ i q q + i h 2 h2
U (k) U (l) H2 H2 + U (k) U (l)Q Q
h2t
Q QH2 H2 h2t
k
k,l
ht
53
(k)
QH2 + h.c. ,
H2 qu(k) + U (k) qh2 + Qh2 u(k) + mB
kU
(2.13)
n=
p2
R
1
n+ 2 = 2p cot (pR + ) + cot (pR ) ,
( R )
(3.1)
n+
R
(p) =
D
2 ( n+ )2
p
R
n=
R
cot (pR + ) cot (pR ) .
=
2
(3.2)
(p) =
coth (pR + i) coth (pR i) ,
D
E
2i
DE
(p) =
(3.3)
DE
(p) =
R
1 + 2e2Rp cos(2) ,
p
(3.4)
E
D
(p) = 2Re2Rp sin(2),
(3.5)
+
mH = iNc ht
+ D (q)
2D (q) +
2
(2)4
q2
q 2 q 2 (mB
k)
k
Nc h2t
4 2
0
dq DE
(q) DE
(q) .
(3.6)
54
Nc h2t 2i
Li3 e
+ Li3 e2i 2(3) .
(3.7)
4
2
16 R
There is a cancellation of tree level infinities between the second and third diagrams. This
cancellation always occurs according to the diagrammatical rule
m2H =
= ih2t q 2 D (q).
(3.8)
The corresponding rule will be to drop all diagrams containing (0) and replace in the
remaining ones
m2k
q 2 m2k
q 2 D (q).
(3.9)
55
occurs due to symmetric summation over an odd function of k. For this reason diagrams
0a and 0b are zero when summed over.
Contractions are done using Weyl fermions. The relevant propagators are:
pm m
T (p) = 2 2 ,
p m
pm
T (p) = 2 m 2
p m
(4.1)
(4.2)
Tr m n r s = 2 mn rs mr ns + ms nr
(4.3)
are used.
There will be two kinds of cancellation between bosons and fermions:
If a loop involves only particles which are localized on the brane, there will be
the usual supersymmetric cancellations of quadratic divergences. Supersymmetry
breaking does not affect these particles (both fermions and bosons remain massless).
If a loop involves particles moving in the bulk, one expects cubic divergences by
power counting. In the supersymmetric theory ( = 0) these terms are cancelled exactly by the sfermion. For = 0 fermions and bosons get different masses. However, after summation over the modes (integration over the loop fifth momentum) the
dependence lies in the propagators D and D 0 whose difference is exponentially
suppressed for large Euclidian fourmomentum (see Eq. (3.4)). So the cancellation is
no longer exact but leaves over a UVfinite piece.
We use in the following the definitions k = p + q, l = p q and the identities
2p q = k 2 p2 q 2 = l 2 + p2 + q 2 ,
2p k = q 2 + p2 + k 2 ,
2q k = p2 + q 2 + k 2 ,
(4.4)
to eliminate the scalar products and express all integrands in terms of p2 , q 2 and k 2 . We
will suppress the integrations and the common factor of h4t . The diagrams of the first group
are shown in Fig. 3.
The calculation of the diagrams in Fig. 3 leads to expressions d1 = d1a + d1b + d1c + d1d :
2 p k
2
q2
1
1
= iNc D (q) 2 2 + 2 + 2
d1a = 2iNc D (q)
p2 k 2
p k
p
k
2
2
2
q
= iNc D (q) 2 2 + 2 .
(4.5)
p k
p
56
The possibilityof the renaming k p becomes obvious if one writes the integration over
k explicitly as d 4 k (p + q k) and observes that the sharp cutoff has to be introduced
symmetrically for all three momenta:
2 1
(q)
,
d1b = iNc D
k 2 p2
(4.6)
2 1
.
d1c = d1d = iNc D (q)
p2
(4.7)
(q)
d1 = iNc D (q)
+ iNc D
.
2
2
p k
k 2 p2
(4.8)
(q)
=
iN
(q)
+
D
d2a = 2iNc D
(4.9)
c
q 2 p2 k 2
q 2 p2 p2 k 2 q 2 k 2
2
1
2
(q)
= iNc D
(4.10)
,
q 2 p2 p2 k 2
Fig. 3. Twoloop diagrams computed in Eqs. (4.5)(4.7). Diagrams 1c and 1d have the same value.
Fig. 4. Twoloop diagrams computed in Eqs. (4.9)(4.12). Diagrams 2b and 2c are equal. The last
three diagrams combine according to the rule (3.8). Note there is indeed a combinatorial factor of 2
for diagram 2e, as needed.
2 1
(q)
d2b = d2c = iNc D
,
q 2 p2
2 q 2
.
d2d + d2e + d2f = iNc D (q)
p2 k 2
57
(4.11)
(4.12)
(q)
d2 = iNc D (q)
+ iNc D
.
2
2
p k
k 2 p2
(4.13)
(4.14)
Adding d1 to d3 we get
2
2 q 2
2 1
(q)
+ 2iNc D
. (4.15)
d1 + d2 + d3 = 2iNc D (q) D 0 (q)
2
2
p k
k 2 p2
Now the qintegration is completely finite. The remaining logarithmic divergence of the
pintegration will be associated to the wavefunction renormalization (WFR) of the bulkfield.
Fig. 6. Twoloop diagrams computed in Eqs. (4.16)(4.18). They have to be combined according to
rule (3.8), i.e., 4a + 4b, 4c + 4d, 4e + 4f + 4g + 4h.
58
Next we consider the diagrams from Fig. 6. All of them are infinite by themselves
(having (0) vertices or divergent sums) but combining them as below these divergences
cancel:
pk
d4a + d4b = 2iNc D (q)D 0 (p) 2 2
q k
1
p2
1
= iNc D (q)D 0 (p) 2 + 2 2 + 2 ,
(4.16)
k
q k
q
d4c + d4d = iNc D (q)D (p)
1
,
q2
(4.17)
p2
.
q 2k2
(4.18)
0
+ 2 + iNc D (q)D 0 (p) 2 .
d4 = iNc D (q) D (p) D (p)
(4.19)
2
2
q k
q
k
We also have the corrections to the fermionic (q) propagator (see Fig. 7):
q k
p2
1
1
d5a = 2iNc D (p)D 0 (q) 2 2 = iNc D (p)D 0 (q) 2 2 + 2 + 2 ,
q k
q k
k
q
(4.20)
d5b = 2iNc D 0 (q)D 0 (p)
pq
1
1
p2
0
0
=
iN
D
(q)D
(p)
+
+
.
c
q 2l2
q 2 k2 q 2k2
(4.21)
They combine to give a supersymmetric cancellation of the cubic divergence in the
integration over p:
p2
1
0
0
d5 = iNc D (q) D (p) D (p) 2 2 + 2
q k
q
1
iNc D 0 (q) D (p) + D 0 (p) 2 .
(4.22)
k
Summing up the two contributions we get:
1
d4 + d5 = iNc D (q) D 0 (q) D (p) D 0 (p) 2
q
p2
+ iNc D (q) + D 0 (q) D (p) D 0 (p) 2 2
q k
1
+ iNc D (q)D 0 (p) D 0 (q) D (p) + D 0 (p) 2 .
k
59
(4.23)
There is a linear divergence in the pintegration coming from the last term. It is
associated with the WFR of the matter fields living on the brane and will be cancelled
by the usual supersymmetric counterterm.
To make the qintegration finite we look at the three remaining diagrams (see Fig. 8)
1
,
k2
(q) 1 ,
(p)D
d7 = iNc2 D
q 2 p2
(q) 1 ,
(p)D
d8 = 2iNc D
k 2 p2
d6 = iNc D (q)D (p)
(4.24)
(4.25)
(4.26)
d7 and d8 are UV finite according to the expansion (3.5), while d6 is exactly the missing
piece in Eq. (4.23).
We end up with:
Nc h4t
d 4p d 4q
(2)4 (2)4
2 q 2
2 1
2 0
E (q)
(q) DE (q)
2
D
2 DE
p2 k 2
k 2 p2
1
0
0
+ DE (q) DE
(q) DE
(p) DE
(p) 2
q
p2
0
0
+ DE (q) + DE
(q) DE
(p) DE
(p) 2 2
q k
1
0
0
+ DE (q) DE
(q) DE
(p) + DE
(p) 2
k
1
1
(4.27)
One sees that the only remaining divergences are in the pintegration, more specifically
there is the logarithmic divergence in the first line and the linear divergence in the fourth.
The expression vanishes for = 0.
Fig. 8. Last three twoloop contributions, computed in Eqs. (4.24)(4.26). Diagrams 7 and 8 are
UVfinite.
60
R (0) hbare
ht = 1
(4.32)
t .
2
2
2
By inspection of Eq. (2.2) it is obvious that this leaves the Lagrangian unchanged if
expressed in rescaled quantities. The different sign for U and y U guarantees that the first
term of the superpotential remains unaltered.
The following counterterms now appear in the 4D Lagrangian:
LCT = Q Q Q H2 H2 H2 U R U U + y U y U y=0
Q q q H2 h2 h 2 U R u u y=0
Q h2t R U U y=0 H2 H2 H2 h2t R U U y=0 Q Q
2
U h2t R(0) Q QH2 H2 + U ht R(0) QH2 y U y=0 + h.c.
= Q Q Q H2 H2 H2 U
U (k) U (l)
U
k,l
mk U
(k)
ml U
(l)
k,l
Q q q H2 h2 h 2 U
Q h2t
k,l
U h2t
u(k) u (l)
k,l
(k)
(l)
H2 H2
H2 h2t
U (k) U (l) Q Q
k,l
2
(k)
mk U QH2 + h.c. .
R(0) Q QH2 H2 + U ht R(0)
k
(4.33)
61
The values of U and Q are determined by the infinite parts of the oneloop two point
functions at nonvanishing momentum q. One finds:
B
2
d 4 p mB
k ml + q
Uk Ul = h2t
(2)4
p2 k 2
2
B B
2
2 1
+ finite,
= iht mk ml + q
log
(4.34)
8 2
2
p2
q2 0
d 4p
0
1
+
+
(p)
D
(p)
D (p)
D
(2)4
k2
k2
R
+ finite,
= ih2t q 2
(4.35)
8
where is an infrared cutoff. The infinities can be absorbed if one chooses
2
2 1
U = ht
(4.36)
,
log
8 2
2
R
.
Q = h2t
(4.37)
8
There are now the additional diagrams shown in Figs. 9 and 10 (U counterterms are
denoted by and Q counterterms by ).
The diagrams in Fig. 9 contribute:
2Q D (q) D 0 (q) .
(4.38)
Q Q = h2t
62
The diagrams of Fig. 10 give (observe the factor of two for the third diagram):
1
m2k
mk
ml 2
1
U
+
m
m
q
+
2
k l
q 2 q 2 m2k q 2 m2k
q 2 q 2 m2k
k,l
2
2
1
1
1
q + mk ml + 2 2U q 2 D 0 (q)
+
2
2
2
2
q
q mk q mk
1
1
1
q 2 + mk ml mk ml + 2 q 2 m2l m2k
= U
2
2
2
2
2
q q mk q mk
k,l
2
+ q 2 q 2 + mk ml + q 2 m2k q 2 m2l 2U q 2 D 0 (q)
2
2 2
(q) .
= 2U q 2 D (q) q 2 D 0 (q) + D
(4.39)
E (q) .
2U q 2 DE
(q) q 2 DE
(q) D
(4.40)
d 4p 1
U
(2)4 k 2 p2
1
d 4p
0
0
(4.41)
(q) DE
(q) h2t
(p) 2 + 2Q
DE (p) + DE
+ DE
4
(2)
k
5. Conclusions
In this paper we have made an explicit twoloop calculation of the Higgs mass in the
fivedimensional model of Ref. [8], where the fifth dimension is compactified on the
orbifold S 1 /Z2 and supersymmetry is broken in the bulk by means of ScherkSchwarz
boundary conditions. We have considered only (top) Yukawa couplings in the calculation
(ht ) and disregarded gauge couplings (g). We have also considered O(0 ) results, which
amounts to disregard the parameter in the superpotential. Our result is that no Higgs
mass counterterm is generated at twoloop and the only divergent terms in the twoloop
correction can be absorbed by the renormalization group running of ht between the cutoff
and the scale where the theory becomes fourdimensional, 1/R. The latter can be
interpreted as threshold effects at 1/R when matching the 4D theory below it with the
complete KK theory above 1/R. As a result our explicit calculation shows that the Higgs
mass is finite and calculable to the considered order in perturbation theory.
63
Inclusion of gauge couplings has been done at oneloop, O(g 2 ) terms, for an
arbitrary gauge group [1,2]. At twoloop we would have O(g 2 h2t ) and O(g 4 ) terms. We
expect divergent terms which are cancelled by supersymmetry or by counterterms from
supersymmetric wavefunction renormalization, which can be absorbed by the running of
gauge and Yukawa couplings, yielding again a finite result.
Inclusion of the term (a supersymmetric mass) in the superpotential would introduce
divergences, even at the oneloop level, that can be interpreted as the supersymmetric
running of the tree level Higgs mass terms, while the analysis becomes much more involved
since the Higgs mass M2 becomes a twobytwo matrix. All infinities, in this case, are
expected to be absorbed by supersymmetry as well as by the running of gauge and Yukawa
couplings, and of the parameter. The resulting absence of Higgs mass counterterms
should hold in this case. At oneloop, we have checked that the field rescaling H2 = Nc Q
is indeed sufficient to render M2 finite.
Acknowledgements
The work of A.D. was supported by the Spanish Education Office (MEC) under an FPI
scholarship. The work of G.G. was supported by the DAAD.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
Abstract
We consider the N = 4 supersymmetric YangMills theory in four dimensions. We compute the
oneloop contributions to the effective action with five external vector fields and compare them with
corresponding results in open superstring theory. Our calculation determines the structure of the F 5
terms that appear in the nonAbelian generalization of the BornInfeld action. The trace operation on
the gauge group indices receives contributions from the symmetric as well as the antisymmetric part.
We find that in order to study corrections to the symmetrized trace prescription one has to consistently
take into account derivative contributions not only with antisymmetrized products [ ] but also
with symmetrized ones ( ) . 2001 Elsevier Science B.V. All rights reserved.
PACS: 11.15.q; 11.30.Pb
1. Introduction
Scattering amplitudes of massless modes in string theories can be described in terms
of effective Lagrangians. For the open string case the Abelian BornInfeld action [1]
represents a remarkable example of an effective field theory which contains string
corrections to all orders in [24]. The contributions can be summed up to all orders in the
case of a constant, Abelian field strength. As soon as these two conditions are relaxed the
problem becomes complicated and a complete action for such fields has not been obtained.
A nonAbelian generalization of the BornInfeld action can be defined as suggested in [5]
using a symmetrized trace operation over the gauge group matrices. However there are
Email addresses: andrea.refolli@mi.infn.it (A. Refolli), alberto.santambrogio@mi.infn.it
(A. Santambrogio), niccolo.terzi@mi.infn.it (N. Terzi), daniela.zanon@mi.infn.it (D. Zanon).
05503213/01/$ see front matter 2001 Elsevier Science B.V. All rights reserved.
PII: S 0 5 5 0  3 2 1 3 ( 0 1 ) 0 0 3 6 8  6
65
indications that this prescription might not be sufficient to include all the contributions that
one would obtain from an open string approach [68].
The best available way to construct the effective action for the nonAbelian, non constant
curvature case, is to proceed order by order. One has to compute corrections with increasing
number of derivatives and of field strengths. A well established result is up to the order
2 [9,10]. As soon as one focuses on higher orders the calculations become difficult.
The inclusion of supersymmetry seems to be quite useful to gain insights and it might
set enough constraints to fix uniquely the form of the action. Several attempts in different
directions are under consideration [7,8,11].
In this paper we attack the problem as follows: since we want to make contact
with the tendimensional open superstring we consider its fourdimensional field theory
limit, i.e., the N = 4 supersymmetric YangMills theory. Then we compute at oneloop
perturbatively: the idea is that in this way we construct an effective action which is
supersymmetric and generalizes the YangMills theory. If supersymmetry determines the
form of the allowed deformations [12] it should correspond to the nonAbelian Born
Infeld theory. We set the external background onshell and compute oneloop npoint
vector amplitudes. These functions are in general nonlocal expressions that contain a loopmomentum integral. For a given npoint amplitude we perform a lowenergy expansion
in the external momenta: this is achieved through the introduction of an infrared mass
regulator that in the expansion plays the same role as , i.e., it keeps track of the order of
the derivative term under consideration. The net result is that each npoint amplitude gives
rise to an infinite series of local terms with higher and higher number of derivatives. The
leading term is automatically gauge invariant. The subleading contributions which contain
derivatives are not gaugeinvariant by themselves. The correct covariantization is obtained
including corrections from amplitudes with a higher number of external background fields
through a mechanism similar to the one in [13] .
We present the calculation of the four and fivepoint functions, and of the part of
the sixpoint function which is needed for the above mentioned covariantization of the
lower order results. These computations will allow us to evaluate the complete gauge
invariant structures for the F 4 and the F 4 , F 5 contributions. Since the symmetric trace
prescription would rule out F 5 terms, the nonvanishing result that we find confirms that
the actual form of the nonAbelian BornInfeld action is definitely richer. In fact the trace
operation on the gauge group indices receives contributions from the symmetric as well as
the antisymmetric part.
In the next section we briefly recall the N = 1 superspace formulation of the N = 4
YangMills action and describe the main ingredients that enter the quantization via the
background field approach. Moreover we outline the general procedure that, starting from
the oneloop YangMills amplitudes, allows us to determine corrections to the nonAbelian
BornInfeld action. In Section 3 we compute the oneloop amplitude with four external
vector fields and study its lowenergy expansion. We extract from the superfield result its
component content and in particular we study the bosonic contributions which contain the
field strengths F . In Section 4 we compute in the same manner the fivepoint function.
The calculation is quite complicated, on one hand because of the gauge group structure,
66
on the other hand because of the existence of several onshell identities which makes it
difficult to express the result in a canonical form. Finally in Section 5 we compute the part
of the sixpoint function that we use to complete the covariantization of the terms with
four field strengths and two derivatives. In Section 6 we collect all our results and compare
them with corresponding results from open superstring theory [14].
W
d x d e i e +
d x d W W +
d4 x d2 W
S = 2 Tr
g
2
2
j k 1
1
4
2
i
4
2 ij k
+
(2.1)
d x d iij k , +
d x d i i [j , k ] ,
3!
3!
2 (eV D eV )
where the i with i = 1, 2, 3 are three chiral superfields, and the W = i D
are the gauge superfield strengths. All the fields are Liealgebra valued, e.g., i = ai T a ,
in the adjoint representation of the gauge group, with [Ta , Tb ] = ifabc Tc .
At the component level the three chiral superfields i contain six spin0 particles and
three spin 21 Weyl spinors, while W describes one spin 21 spinor and one gauge vector
particle. Since we are interested in computing field strength corrections to the BornInfeld
action, we will consider amplitudes with vector fields as external background. Moreover
we will extract from them only the gauge field bosonic components using the relations
1
D( W) =0 = f = ( ) F ,
2
( W
=0 = f = 1 ( ) F
D
(2.2)
)
2
with and defined as in (A.3). In this way we will be able to isolate the relevant
contributions.
In order to evaluate oneloop amplitudes we need to quantize the theory and this is
most efficiently done using the background field method [16,17] which guarantees a gauge
invariant result for the effective action. For the gauge multiplet one performs a non linear
splitting between the quantum prepotential V and the background superfield, via covariant
derivatives (in quantum chiral representation [15])
= eV D eV eV B eV ,
B .
= D
(2.3)
In this way the external background enters the quantum action implicitly in the background
covariant derivatives through the connections
B = D i ,
B = D
i
,
aB = a i a
(2.4)
67
B , {
B , B }]. Background
and explicitly in the background field strength W = 2i [
SNK = Tr d4 x d4 bb.
(2.6)
(2.7)
c =
The three ghosts c, c and b are background covariantly chiral superfields, i.e.,
B
B c = B b = 0. In the following we drop the suffix B from the covariant derivatives.
After gaugefixing the quadratic quantum V action becomes
1
1
i
4
4
S 2 Tr d x d V i
2
2
2g
(D
i ) V .
iW (D i ) i W
(2.8)
As mentioned above we want to compute oneloop contributions to the effective action
with external vector fields. The N = 4 theory is particularly simple since this type of terms
are produced only by quantum vector loops. The loops with the three chiral matter fields
are cancelled by the three ghosts: in fact each ghost contributes to the oneloop effective
action exactly as a standard chiral superfield would do, the only difference being an overall
opposite sign because of the statistics.
Thus we focus on the action in (2.8). The interactions with the background are at most
are needed in the loop
linear in the Dspinor derivatives and at least two Ds and two Ds
in order to complete the Dalgebra. Thus the first nonvanishing result is at the level of
the fourpoint function [18]. Here we describe the general procedure we have adopted in
order to extract from the super YangMills effective action the wanted contributions to the
nonAbelian BornInfeld action.
A oneloop npoint amplitude will contain n external background fields from the
in order to complete the Dalgebra.
interactions in (2.8), with at least two W and two W
We label the external momenta p1 , p2 , . . . , pn and introduce a general notation for the
dependence on the gauge group (with structure constants fabc )
g(a1 , a2 , . . . , an ) = fx1 a1 x2 fx2 a2 x3 fxn an x1 .
(2.9)
In the following we treat the gauge fields as matrices in the adjoint representation, i.e.,
Wac fabc Wb . We define
TrAd (AB ) Aa Bb g(a, b, . . .).
(2.10)
68
Moreover we set the external fields onshell, i.e., we freely use the equations of motion
W = 0,
= 0.
W
(2.11)
In this way a npoint function gives rise to a background field dependence which contains
of
a total of n fields, with a given number of connections , and of field strengths W, W,
the general form
(pi ) W (pj ) (pk ) W
(pl )
W
d4 TrAd W (p1 ) W
(pn ) .
(2.12)
From (2.12) we want to extract the relevant structures that might produce F s at the
component level. To this end we need convert
1
D
d4 d2 d2 D D D
(2.13)
4
using the relations in (2.2).
and act with these four spinor derivatives on the W and W
One can reconstruct gauge covariant expressions also from the connections , but this
can happen only at the level of the sixpoint function and beyond. Since in this paper we
consider structures up to the fivepoint functions we leave this point for a future discussion.
In conclusion we obtain the structures we are looking for if our oneloop diagram
.
( W
has produced products of fields D( W) and their hermitian conjugate ones D
)
In addition we have to deal with a loopmomentum integral which contains n scalar
propagators and momentum factors directly from the vertices in (2.8) and/or from
commutators of spinor derivatives produced while performing the Dalgebra
hn (k, pi )
.
In = d4 k 2
(2.14)
k (k + p2 )2 (k + p2 + p3 )2 (k + p2 + + pn )2
We can rewrite (2.14) as an infinite series of local terms in a lowenergy expansion with
higher derivatives. We introduce an IR mass M and expand the propagators keeping the
external momenta small as compared to M. First we Feynman combine the propagators in
(2.14)
1
= (n)
A1 A2 An
1
x1
dx2
dx1
0
xn2
dxn1
0
k 2 + M 2 + (1 x1 )(p2 + + pn )2 + + (xn2 xn1 )(p2 )2
2 n
(1 xn1 )p2 + + (1 x1 )pn
(2.15)
1
n!
2
(k + M 2 )n
d4 k
dxi
Hn (xi , pj )
+ ,
(k 2 + M 2 )n+1
69
(2.16)
where
Hn (xi , pj ) = (1 x1 )(p2 + + pn )2 + + (xn2 xn1 )(p2 )2
[(1 xn1 )p2 + + (1 x1 )pn ]2 .
All the integrals over the kmomentum are finite and give
1
2
1
d4 k 2
=
,
(k + M 2 )n (n 2)(n 1) (M 2 )n2
k k
1
2
=
C C ,
d4 k 2
(k + M 2 )n (n 3)(n 2)(n 1) (M 2 )n3
..
.
(2.17)
(2.18)
The nonlocal expression In has been traded by an infinite sum of local terms. In the next
two sections we apply this procedure to the computation of the four and the fivepoint
functions and show explicitly how we can determine order by order structures that we
interpret as corrections to the nonAbelian BornInfeld action.
(3.2)
70
The complete answer is obtained by summing to the above expression all permutations in
the 2, 3, 4 indices. We are interested in rewriting the superspace result in components and
in particular we want to study the bosonic gauge field contributions. To this end we use the
definitions in (2.2) and the relations in the appendices. First we obtain
(p3 , a3 )W
(p4 , a4 )
d4 W (p1 , a1 )W (p2 , a2 )W
f4 ) ,
(f1 ) (f2 ) (f3 ) (
(3.3)
b ) .
F ab i(Fa ) (F
(fa ) (fb ) =
4
In general we use the notation
n
F a an (Fa1 )1 2 (Fa2 )2 3 (Fan )n 1
(3.4)
(3.5)
(3.6)
(3.7)
which we keep with indices contracted in standard manner. Making use of (A.10) and
the final result becomes
(A.11) to eliminate F
1
g(a1 , a2 , a3 , a4 )
d4 k 2
k (k + p2 )2 (k + p2 + p3 )2 (k + p2 + p3 + p4 )2
1
4 F4 a a a a + 4 F4 a a a a + 4 F4 a a a a
1 2 3 4
1 2 4 3
1 3 2 4
8
2
2
2
2
2
F a a F a a F a a F a a F a a F2 a a .
(3.8)
1 2
3 4
1 3
2 4
1 4
2 3
Now following the general procedure outlined in the previous section we rewrite (3.8) as an
infinite series of local terms in a lowenergy expansion with higher derivatives. Feynmann
combining the propagators in (3.1) we obtain
1
I4 =
d4 k 2
2
k (k + p2 ) (k + p2 + p3 )2 (k + p2 + p3 + p4 )2
1
x1
x2
4
3! d k dx1 dx2 dx3
4!
d
k
dx
+
O
,
i
2
4
2
2
5
(k + M )
(k + M )
M8
2
(3.9)
71
where
H4 (xi , pj ) = (1 x1 )(p2 + p3 + p4 )2 + (x1 x2 )(p2 + p3 )2 + (x2 x3 )(p2 )2
2
(1 x3 )p2 + (1 x2 )p3 + (1 x1 )p4 .
(3.10)
First we look at the leading term in (3.9). It does not depend on the external momenta
and therefore in (3.8) we can freely symmetrize on them. We obtain the F 4 nonAbelian
contribution to the BornInfeld action
4 g(a1 , a2 , a3 , a4 ) F 4 a a a a + 2 F 4 a a a a .
1 2 3 4
1 2 4 3
1 2
F a a F2 a a + 2 F2 a a F2 a a
1
3
2
4
1
2
3
4
4
= TrAd F F F F + 2F F F F
1
F F F F + 2F F F F .
4
(3.11)
We notice that if the gauge group is SU(N) we can express the gauge group matrices which
appear in (3.11) using the standard fundamental representation. If we do so we find that
the planar sector (large N limit) reproduces the symmetrized trace structure [5]
4 Tr T a1 T a2 T a3 T a4 + permutations F 4 a
1 2 2
4
STr F F
.
4
1 a2 a3 a4
1 2
F a a F2 a a
1
2
3 4
4
(3.12)
Next we study the subleading contribution which corresponds to terms with two derivatives. Including the overall factor (1/2)4 (1/2!)223 from the vertices and combinatorics,
the integral containing H4 (xi , pj ) in (3.10) gives
1 2 1
10(p2 )2 + 8p2 p3 + 2p2 p4 g(a1 , a2 , a3 , a4 )
6
40 12 M
1 4
4 F a a a a + 4 F4 a a a a + 4 F4 a a a a F2 a a F2 a a
1
2
3
4
1
2
4
3
1
3
2
4
1
2
3 4
8
2
2
2
2
F a a F a a F a a F a a + permutations a2 , a3 , a4 .
4(der) =
1 3
2 4
1 4
2 3
(3.13)
These terms give rise to contributions that are not gauge invariant since from p i
ordinary derivatives are produced. The correct covariantization of the result is obtained
adding terms with one and two background connections from the five and sixpoint
functions, respectively. In the next sections we will compute these terms explicitly.
72
Including the combinatorial factors and the loopmomentum integration we obtain the
corresponding bosonic contributions in terms of the field strengths
5(1)
1 51 1 4
.
(4.2)
(k + p2 + p3 + p4 + p5 )2
(a)
(b)
Fig. 2. Fivepoint amplitudes.
73
(4.3)
It is easy to show that (f n )a1 a2 an is symmetric under cyclic permutation of the indices
a1 a2 an for n even, while it is antisymmetric for n odd. Using the definitions in (4.3)
(1)
we denote the contributions corresponding to 5 simply as (f 2 )a1 a2 (f3 )a3 a4 a5 . In the
same way we obtain another graph interchanging the external background W(p2 , a2 ) with
3 , a3 ), e.g.,
W(p
1 5 1 1 4 2
(2)
2 f a a f3 a a a g(a1 , a3 , a2 , a4 , a5 )
5 =
1 2
3 4 5
2 2! 3!
1
d4 k 2
k (k + p3 )2 (k + p2 + p3 )2 (k + p2 + p3 + p4 )2
1
,
(4.4)
(k + p2 + p3 + p4 + p5 )2
In addition to (4.2) and (4.4) we have graphs obtained through a permutation of couples
of indices (pi , ai ), i = 3, 4, 5 and (p1 , a1 )(p2 , a2 ). In this way we end up with a total
of 3! 2! 2 = 4! inequivalent contributions. The final result can be rewritten in terms of
the field strengths F . Indeed using the formulas (B.5) and (B.6) in Appendix B we can
express the product of three fs (or equivalently three f s) as
1 3
c )
2 F abc + i(Fa ) (Fb ) (F
8
a )
+ i(F
(Fb ) (Fc ) .
(4.5)
In the same way using (B.3) and (B.4) two fs (or two f s) can be rewritten as in (3.4) and
(3.5). The product of (3.4) and (4.5) gives
1 2
a2 )
F a a + i(Fa1 ) (F
1
2
32
a5 ) + i(F
a3 ) (Fa4 ) (Fa5 ) .
2 F 3 a a a + i(Fa3 ) (Fa4 ) (F
3 4 5
(4.6)
1 5
2Fa1 a4 a3 a2 a5 2Fa51 a4 a5 a2 a3 + 2Fa51 a5 a4 a2 a3
f 2 a a f3 a a a =
1 2
3 4 5
32
2Fa51 a3 a4 a2 a5 + 2Fa21 a2 Fa33 a4 a5 Fa21 a5 Fa32 a3 a4
Fa22 a5 Fa31 a3 a4 Fa21 a3 Fa32 a4 a5 Fa22 a3 Fa31 a4 a5 .
(4.7)
Now, as we have done in the previous section, we study the loop momentum integral in the
lowenergy approximation
1
d4 k 2
I5 =
2
k (k + p2 ) (k + p2 + p3 )2 (k + p2 + p3 + p4 )2
74
1
(k + p2 + p3 + p4 + p5 )2
1
x1
x2
x3
4
4! d k dx1 dx2 dx3 dx4
2
3
1
1
(tot)
4
d k 2
f
f
5 =
g(a1 {a2 , a3 }, a4 , a5 )
a1 a2
a3 a4 a5
12
(k + M 2 )5
+ permutations of a1 , a2 and a3 , a4 , a5 .
(4.10)
Taking into account the symmetry properties of (f n )a1 a2 an , we obtain
3
2
1
1
(tot)
4
5 =
f
d k 2
f
a1 a2
a3 a4 a5
12
(k + M 2 )5
2 g(a1 {a2 , a3 }[a4, a5 ]) + g(a1 {a2 , a4 }[a5, a3 ]) + g(a1 {a2 , a5 }[a3 , a4 ]) .
(4.11)
Finally we can use the result in (4.7) and perform explicitly the commutator algebra in the
gauge group structures
1 2 1
(tot)
5 =
g(a1 , a2 , a3 , a4 , a5 )
32 12 M 6
2 F 5 a a a a a F 5 a a a a a + 3 F 5 a a a a a
1 2 3 4 5
1 4 2 5 3
1 4 3 2 5
3
3
2
2
+ 1/2 F a a F a a a 1/2 F a a F a a a .
(4.12)
2 4
1 3 5
3 4
1 2 5
The above expression can be manipulated further so that the F 2 F 3 terms are eliminated in
favor of the F 5 terms, using the identities (C.4) depicted graphically in Fig. 5.
1 2 1
(tot)
5 =
g(a1 , a2 , a3 , a4 , a5 )
32 12 M 6
8
4
F5 a a a a a F5 a a a a a + 4 F5 a a a a a .
1 2 3 4 5
1 4 2 5 3
1 4 3 2 5
5
5
75
(4.13)
In Section 6 we will come back to (4.13) and assemble all our results. Now we turn to the
computation of the second class of diagrams that contribute to the fivepoint function.
W
fivepoint terms
4.2. The W W W
In this subsection we analyze contributions from the fivepoint function with a
W
. These terms are not gauge invariant,
background dependence of the form WWW
but as we have previously stated, they do contribute to the covariantization of derivative
terms obtained from the fourpoint function. In order to prove this we start considering a
typical graph of this type as shown in Fig. 2(b). After completion of the Dalgebra it gives
rise to
(p3 , a3 )W
(p4 , a4 )
d4 (p5 , a5 )W (p1 , a1 )W (p2 , a2 )W
(4.14)
i 2 1
g(a1 , a2 , a3 , a4 , a5 )[4p1 + 3p2 + 2p3 + p4 ]
=
40 12 M 6
(4.17)
Now we want to show that (4.17) gives the correct contribution linear in to covariantize
(3.13). (The part quadratic in is produced at the level of the sixpoint function and it will
be discussed in the next section.) Thus we go back to (3.13) and there we substitute every
ordinary derivative with a covariant derivative whose expression in momentum space can
be written as
()c (p) = ()c (p) + d4 q ( a )(q)b (p q)f abc .
(4.18)
76
In so doing we find that the terms which are linear in exactly amount to the ones in
(4.17). This can be checked in a simple manner as follows: first we write the derivative
contributions to the fourpoint function as
1 2 1
(4 )der =
10(p2 )2 + 8p2 p3 + 2p2 p4
6
40 12 M
1
TrAd A(p1 )B(p2 )C(p3 )D(p4 ) + permutations of A, B, C, D ,
4
(4.19)
the momenta in the order p1 , p2 , p3 , p4 . Then we can use the ciclicity and inversion
property of the trace so that we always have the two derivatives acting both on the first
field or on the first and the second or on the first and the third. In this way the rules for the
covariantization are simply given by (here we consider the terms linear in only)
TrAd 2 ABCD 2 TrAd ([ , A]BCD)
2TrAd ( ABCD) 2 TrAd ( BCDA),
TrAd (ABCD) TrAd (A[ , B]CD) + TrAd ([ , A]BCD)
TrAd ( BCDA) TrAd ( CDAB)
+ TrAd ( ABCD) TrAd ( BCDA),
TrAd (ABCD) TrAd (AB[ , C]D) + TrAd ([ , A]BCD)
TrAd ( CDAB) TrAd ( DABC)
+ TrAd ( ABCD) TrAd ( BCDA).
(4.20)
In order to streamline the notation we have written the above expressions in the gauge
= 0, a choice actually not necessary to prove the result. Since we have to include
a sum on all permutations we can freely rename the fields and write the final result in the
form
i 2 1
(4 )cov1 =
[4p1 + 3p2 + 2p3 + p4 ]
40 12 M 6
TrAd (p5 )A(p1 )B(p2 )C(p3 )D(p4 )
+ permutations of A, B, C, D
(4.21)
which exactly reproduces the formula (4.17).
(a)
77
(b)
Fig. 3. Sixpoint amplitude.
(5.1)
24
6 =
2
4
2!
g(a1 , a2 , a3 , a4 , a5 , a6 )( 5 ) ( 6 ) (f1 ) (f2 ) (f3 ) (f4 )
1
d4 k 2
2
k (k + p1 ) (k + p1 + p2 )2 (k + p1 + p2 + p3 )2
1
.
(5.2)
(k + p1 + p2 + p3 + p4 )2
To the above contribution we have to sum all the ones obtained permuting the a1 , . . . , a4
indices. Once again we compute the momentum integral at the leading order in the lowenergy expansion. We have
1 2 1
g(a1 , a2 , a3 , a4 , a5 , a6 )
6
16 12 M 6
=
( 5 ) ( 6 ) (f1 ) (f2 ) (f3 ) (f4 )
16 12 M 6
78
g(a1 , a2 , a3 , a4 , a5 , a6 ) + permutations of a1 a4 .
(5.3)
Notice that the permutations are only on the indices of the Ws and the Ws.
W
5.2. Diagrams ( )( )W W W
Depending on the relative position of the two vertices one obtains three
different diagrams. The first one contains the two adjacent vertices and produces a
contribution of the form
1 6 1 3 5
=
2
d4 k( 5 ) ( 6 ) (f1 ) (f2 ) (f3 ) (f4 )
6
2
2!
k (k + p6 )
2
k (k + p6 )2 (k + p6 + p1 )2 (k + p6 + p1 + p2 )2
1
g(a1 , a2 , a3 , a4 , a5 , a6 ).
(5.4)
(k p5 p4 )2 (k p5 )2
The leading order in the lowenergy expansion gives
6
5!
16
1
x4
dx1
d k
0
1
2
g(a1 , a2 , a3 , a4 , a5 , a6 )
(k + M 2 )6
k k + p6 (1 x5 ) + p1 (1 x4 )
+ p2 (1 x3 ) + p3 (1 x2 ) + p4 (1 x1 )
p6 (1 x5 ) + p1 (1 x4 ) + p2 (1 x3 )
+ p3 (1 x2 ) + p4 (1 x1 ) p6 .
(5.5)
6 tot =
( 5 ) ( 6 ) (f1 ) (f2 ) (f3 ) (f4 )
32 30 M 6
2g(a1 , a2 , a3 , a4 , a5 , a6 ) + 2g(a1 , a6 , a2 , a3 , a4 , a5 )
+ g(a1 , a2 , a6 , a3 , a4 , a5 ) + permutations of a1 , a2 , a3 , a4 .
(5.6)
Notice that the first two graphs have a multiplicity which is double as compared to the third
one. All together they reconstruct the 5! permutations of the a2 , . . . , a6 indices.
5.3. Covariantization of 4 with two s
The starting point is the result in (4.19). We proceed in complete analogy with what
we have done for the part linear in , so that we obtain simple rules also for the terms
79
quadratic in :
TrAd 2 ABCD TrAd ([ , [ , A]]BCD)
TrAd ( ABCD) 2TrAd ( A BCD) + TrAd ( BCDA),
TrAd (ABCD) TrAd ([ , A][ , B]CD)
TrAd ( A BCD) TrAd ( AB CD)
TrAd ( BCDA) + TrAd ( B CDA),
TrAd (ABCD) TrAd ([ , A]B[ , C]D)
TrAd ( AB CD) TrAd ( ABC D)
TrAd ( B CDA) + TrAd ( BC DA).
(5.7)
Once again, using the properties of the trace operation it is possible to set one of the s
in first position, and the other one either in second or in third or in fourth position. Then
renaming the fields and using the formulas (2.18) we obtain
1 2 1
3TrAd (p5 ) (p6 )A(p1 )B(p2 )C(p3 )D(p4 )
(4 )cov2 =
80 12 M 6
+ 2TrAd (p5 )A(p1 ) (p6 )B(p2 )C(p3 )D(p4 )
+ (p5 )A(p1 )B(p2 ) (p6 )C(p3 )D(p4 )
+ permutations of A, B, C, D
(5.8)
which is just the sum of (5.3) and of (5.6). This completes the proof of the covariantization
of the twoderivative terms for the fourpoint function.
80
1
F F F F ,
20
(6.1)
where it is understood that the indices on the two are to be contracted. We integrate by
parts in such a way to have the derivatives acting either on the same or on two adjacent
field strengths. We obtain
1 2 1
Tr
F F F F
(4 )der =
Ad
20 12 M 6
+ F F F F + F F F F
1
F F F F + F F F F + F F F F
4
1 2 1
Tr
2 2 F F F F 4 2 F F F F
Ad
20 12 M 6
1
+ 2 F F F F + 2 F F F F .
(6.2)
2
Now we show that quite generally we can rewrite terms which contain 2 and four F s
as F 5 contributions. Indeed, using the equations of motion F = 0 and the Bianchi
identities [ F] = 0, we have the following identities
2 F = F
Bianchi
( F + F )
e.o.m
= [ , ]F [ , ]F
= +i[F , F ] + i[F , F ] = 2i(F F F F ),
(6.3)
(6.4)
(6.5)
Thus in order to study corrections to the symmetrized trace prescription one has to
consistently take into account derivative contributions. In particular the above relations
clearly show that for the case of two covariant derivatives one has to consider not only the
antisymmetrized products [ ] , as it was already noticed [7,8], but also the symmetrized
ones ( ) .
Using (6.5), we can rewrite the last two lines of the two derivative result in (6.2) as
5
5
1 2 1
5
4
F
+
4
F
+
8
F aaaaa
a
a
a
a
a
a
a
a
a
a
1 2 3 4 5
1 2 4 3 5
1 4 5 2 3
20 12 M 6
3
3
2
2
+ 4 F a a F a a a + 2 F a a F a a a g(a1 , a2 , a3 , a4 , a5 ).
(6.6)
3 4
1 2 5
2 5
1 3 4
81
These terms can be combined with the F 5 contributions we have obtained from the fivepoint function in (4.13). Making use of the identities (C.4) the sum of (6.6) and of (4.13)
can be written as
1 2 1
5
7
3
F a a a a a F5 a a a a a F5 a a a a a
6
1 2 3 4 5
1 2 4 3 5
1 4 2 5 3
20 12 M
2
2
5
+ 2 F a a a a a g(a1 , a2 , a3 , a4 , a5 )
1 3 2 5 4
1 2 1
7
Tr
=
Ad F F F F F F F F F F
6
20 12 M
2
3
(6.7)
F F F F F + 2F F F F F .
2
Thus our final result to the M 6 order is:
1 2 1
tot =
Tr
F F F F
Ad
6
20 12 M
+ F F F F + F F F F
1
F F F F + F F F F + F F F F
4
7
1 2 1
Tr
+
Ad F F F F F F F F F F
20 12 M 6
2
3
F F F F F + 2F F F F F .
(6.8)
2
Up to an overall numerical factor, the first two lines in (6.8) reproduce the corresponding
result in Ref. [14], formula (3.3), obtained from an open superstring scattering amplitude
(see also [19]). So, our first result is that, at the level of the fourpoint function, the
supersymmetric YangMills effective action exactly reproduces the structure of the nonAbelian BornInfeld theory, including the first derivative corrections. The terms F 5 , which
were also computed in [14], are more difficult to compare with ours: this is essentially
due to the fact that the result quoted in [14] is not written in a canonical form and
moreover it requires additional symmetrizations which we are not clear how to interpret
unambiguously.
We would like to conclude summarizing our results. We have considered the N = 4
supersymmetric YangMills theory and computed at oneloop the four and fivepoint
functions with external vector fields. From the superfield result we have extracted the part
of the bosonic components which contain the field strengths F . These nonlocal oneloop
contributions have been expanded in a low energy approximation and expressed as a sum
of an infinite series of local terms. We have argued that these local expressions reproduce
contributions to the nonAbelian BornInfeld action, if supersymmetry has to determine
its structure. We have explicitly computed the leading contributions from the four and the
fivepoint functions and the subleading terms of the fourpoint function. These latter ones
contain two derivatives acting on the four F . From the fourpoint function calculation
one simply obtains ordinary derivatives, thus the result is not in a gauge invariant form. We
82
have checked that the correct covariantization of the result is obtained via contributions
from the five and the sixpoint functions with one and two connections, respectively.
We have confirmed that beyond the F 4 order the nonAbelian BornInfeld action
contains terms which are not included in the symmetrized trace prescription. In order
to compute corrections one has to consider terms with derivatives, and in particular
we have shown that one cannot disregard terms with symmetrized derivatives. Both the
antisymmetrized products [ ] and the symmetrized ones ( ) are on equal footing.
The part of the sixpoint function needed in order to check the gauge invariance of
our result has been computationally rather difficult, but clearly not impossible. Taking
advantage of the superfield approach and of the background field method the determination
of the full sixpoint function seems at hand.
Acknowledgements
The authors have been partially supported by INFN, MURST, and the European
Commission RTN program HPRNCT200000113 in which they are associated to the
University of Torino.
(A.1)
C =
,
C =
i 0
i
i
,
0
C = C .
C = C ,
(A.2)
( )
C C ( ) = (1, ),
1
( ) ( ) ,
4
1
( ) ( ) .
4
Vector indices are transformed into spinor notation by
(A.3)
V = V
(A.4)
F = 2f C + 2f C ,
(A.5)
V = ( )
V .
2
For an antisymmetric tensor of rank two we have
83
where f and f are symmetric bispinors. The above relations can be inverted
1
f = ( ) F ,
2
1
f = ( ) F .
(A.6)
2
Therefore an antisymmetric tensor is always expressible in terms of two symmetric
bispinors.
F = ( ) f ( ) f .
(A.7)
(A.8)
(A.9)
(A.11)
(B.1)
= ( ) + i .
(B.2)
These allow to obtain the following results for the trace of two matrices
1
i
Tr( ) = + +
4
2
+ completely antisymmetrized in (, )(, )
(B.3)
84
and similarly
1
i
Tr( ) = +
4
2
+ completely antisymmetrized in (, )(, ).
(B.4)
Tr( ) = ( + ) +
8
16
32
+ completely antisymmetrized in (, )(, )(, ).
(B.6)
4 5
Clearly not all the possible structures are independent, since we can freely use the ciclicity
and inversion property of the trace. It is easy to realize that among the F 5 structures only
four are independent: we can list them as follows
A: TrAd (F F F F F ) = F 5 a a a a a g(a1 , a2 , a3 , a4 , a5 ),
1 2 3 4 5
5
B: TrAd (F F F F F ) = F a a a a a g(a1 , a2 , a3 , a4 , a5 ),
1 2 3 5 4
5
C: TrAd (F F F F F ) = F a a a a a g(a1 , a2 , a3 , a4 , a5 ),
1 3 2 5 4
5
D: TrAd (F F F F F ) = F a a a a a g(a1 , a2 , a3 , a4 , a5 ).
(C.2)
1 4 2 5 3
In the same way one finds that there exist only two independent structures F 2 F 3
E: TrAd (F F F F F ) = F 3 a a a F 2 a a g(a1 , a2 , a3 , a4 , a5 ),
1 2 3
4 5
2
3
F : TrAd (F F F F F ) = F a a a F a a g(a1 , a2 , a3 , a4 , a5 ).
1 3 4
2 5
(C.3)
The terms in (C.2) and (C.3) can be represented graphically as shown in Fig. 4 and
explained in [8]. Every vertex corresponds to a F field strength and the trace operation
85
Fig. 4. F 5 structures.
has to be taken following the vertices of the pentagon clockwise. The arrows denote how
the indices are contracted.
A quite remarkable result is that the F 2 F 3 structures can be reexpressed in terms of the
F 5 structures as follows:
3
1
F = A B + C + D,
5
5
3
1
E = A + B + C + D.
5
5
(C.4)
86
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
Abstract
When eightdimensional instantons, satisfying F F = 8 (F F ), shrink to zero size, we
find stringy objects in higher order tendimensional YangMills (viewed as a lowenergy limit of
open string theory). The associated F 4 action is a combination of two independent parts having a
singletrace and a doubletrace structure. As a result we get a Dstring from the singletrace term
and a fundamental string from the doubletrace. The latter has (8, 0) supersymmetry on the worldsheet and couplings to the background gauge fields of a heterotic string. A correlation between the
conformal factor of the instanton and the tachyon field is conjectured. 2001 Elsevier Science B.V.
All rights reserved.
PACS: 11.25.w; 11.15.q; 11.10.z
1. Introduction
Recent developments in string theory have given some hope for answering important
questions which require offshell formulation of a second quantized string theory. In the
case of open string field theory, several offshell formulations are currently known [13].
The situation with the closed string field theory is more complicated. One might consider
the possibility that in fact the complete and consistent offshell string theory can originate
from properly defined open string field theory alone, with closed string being understood
in terms of classical open string field theory (for motivations and references see, [4,5]).
The possibility that a closed string worldsheet action arises from classical solutions in
open string field theory will be the main theme of this paper. Although a complete picture
might require incorporating the whole field theory of open strings, one might hope to get
Email addresses: ruben.minasian@cpht.polytechnique.fr (R. Minasian), samson.shatashvili@yale.edu
(S.L. Shatashvili), vanhove@spht.saclay.cea.fr (P. Vanhove).
1 On leave of absence from St. Petersburg Steklov Mathematical Institute, St. Petersburg, Russia.
2 Unit mixte du CNRS et de lEP, UMR 7644.
05503213/01/$ see front matter 2001 Elsevier Science B.V. All rights reserved.
PII: S 0 5 5 0  3 2 1 3 ( 0 1 ) 0 0 3 6 9  8
88
important insights by looking at low energy sector only. This is what we will attempt here.
We start from a nonabelian superYangMills theory including the higherderivative F 4
terms in ten dimensions in the background of a SO(8) instanton and find stringy objects. 3
The presence of higherderivative terms in the effective action is crucial for our analysis
since not only the lowest order F 2 term cannot lead to stringy objects with a finite tension
but, as we will argue, does not play any role. In zero size limit of the instanton, a Dstring and a fundamental heterotic string are found. The heterotic string solution will be
supersymmetric only under a linear combination of the standard supersymmetry of the
YangMills theory together with an extra supersymmetry transformation. It is tempting
to interpret this extra supersymmetry transformation as originating from the nonlinear
supersymmetry of a nonBPS Dbrane sitting in our vacuum.
The consistency of the heterotic string background is related to the original symmetries
of the YangMills background; the chirality of the background will be reflected in the
chirality of the string worldsheet theory. In particular, starting from an anomaly free
open string theory (a correct choice of the gauge group for which the top term in the
anomaly polynomial vanishes is what is meant here) will surely result in a closed heterotic
string theory in a consistent gravitational and YangMills background.
The Dstring appears from the presence of tr F 4 in the effective action, and its tension
is quantized according the value of 7 (G) = Z for G allowing nontrivial embedding
of SO(8) groups [6,7]. For obvious reasons, we chose G = SO(32). The novelty of our
study resides in the origin of a fundamental heterotic string from the doubletrace structure
(tr F 2 )2 and its supersymmetric completion in the effective action.
In the next section, the precise structure of these terms will be shown from the
supersymmetry analysis but we would like to make some qualitative comments here from
a different point of view. Some relevant ideas come just from looking at the (bosonic part
of) open string action
S = Sopen (, B, T ) d 10 x det(G + F ) f (T )
(1.1)
written here for the constant tachyon field T (the precise analog of this action for nonabelian gauge groups is not known yet but can be conceptually derived in any given string
field theory). According to the wellknown conjectures of Sen [8], f (T ) has zero at the
minimum of the tachyon potential where the closed string vacuum is expected. It can
be calculated using the formalism of [2]; for bosonic string field theory it is given by
f (T ) = eT while the tachyon potential is VB = (1 + T )eT and for the superstring,
2
f (T ) coincides with the tachyon potential VF (T ) = eT [9]. Since our analysis relies on
supersymmetry, from now on we will replace f by V .
In a general string field theory the Lagrangian is defined up to field redefinitions and a
correct choice of fields is dictated by the principle that equations of motion are proportional
to worldsheet function with some Zamolodchikov metric G: i S = Gij j . Here G is
3 The structure and the supersymmetry of F 4 terms will be discussed in details in Section 2. For now we just
remark that due to different ways of contracting the gauge indices they can be of two types singletrace terms
tr F 4 and doubletrace (tr F 2 )2 terms.
89
the metric on the space of fields and has to be nondegenerate. One can note that the perfect
square part of doubletrace terms, (tr F F )2 , in a classical open string field theory
Lagrangian can be generated by a field redefinition T T + tr F 2 . The compatibility
of the supersymmetry and the above principle should fix these terms unambiguously as
well as complete the needed t(8)(tr F 2 )2 .
We conclude this introduction by a comment on the worldsheet topology. We already
mentioned that by expanding around the solution we obtain couplings for collective modes
which turn out to reproduce those of the heterotic string, and this should already indicate
that the string under the consideration is closed. Much of this structure is directly inherited
form the SO(8) instanton construction. It is yet another feature of the construction that the
solution has in fact SO(9) symmetry and the fact that we are using only the SO(8) part
leads to the periodicity of one of the string worldsheet coordinates.
A brief outline of the paper is the following. In Section 2, we present the Yang
Mills action and the supersymmetry transformations. In Section 3, we look for BPS
configurations associated with eightdimensional instanton configurations having SO(8)
symmetry. We consider the supersymmetric excitations in the Section 4 and show the
emergence of the twodimensional action with couplings of the heterotic string with
a particular embedding of the gauge group. Finally the Section 5 contains a general
discussion of our results. We have collected various definitions and identities needed in
the main text in two appendices.
32
+ ( )2 MABCD 4F A F B C (D )D
+ 2F A F B C (D )D
+ quartic fermions.
(2.1)
90
This action (2.1) is invariant under the linear supersymmetry transformations (by a
straightforward generalization of the results of [11])
C D
C D
( )2 A
A
& A A
M BCD 2 & B F
=
4
&
F 8 & B F
F
8
(2.2)
& 1 4 B FC1 2 FD3 4 ,
()
1
( )2 A
A
+ 5
7 8 & 1 6 & FB1 2 FC3 4 FD5 6 ,
M BCD t(8)
& A = &F
8
2 4!
(2.3)
are the SO(1, 9) matrices, as well as all the spinors in this expression; the tendimensional spacetime indices are labeled by Greek letters from the middle of the alphabet
, = 0, 1, . . . , 9.
All the fields are in the adjoint of the gauge group G = SO(32), labeled by a capital
letter from the beginning of Latin alphabet. The quadratic expressions in the field are
contracted with the metric AB = trfund (TA TB ) and quartic expressions with the completely
symmetric four rank tensor MABCD [10,12]. For any group, there are two independent fourrank symmetric tensors given by the symmetrized trace in the fundamental representation
trfund(T(A TB TC TD) ) = Str( ) and the doubletrace (AB CD) trfund ( ) trfund ( ) in
the fundamental representation
MABCD = m1 trfund (T(A TB TC TD) ) + m2 (AB CD) .
(2.4)
91
The presence of the two independent terms in (2.4) will be fundamental for our analysis.
The singletrace term will carry the charge of the instanton introduced in Section 3 and
the doubletrace term will be at the origin of the interactions of the fluctuations around the
instanton.
2.2. The nonlinear supersymmetry
Looking at the effective action as related to Dbrane configurations in a type II vacuum,
it is natural to look for the possibility of nonlinear supersymmetry resulting from the
breaking of some of the original supersymmetries of the vacuum. There are various ways
to introduce nonlinear supersymmetry depending on which fields are coupled to the
supersymmetry parameter [10,15] in = + O().
For the case of open string actions with tachyon field and abelian gauge field, the
quadratic effective action takes the form (abelian field have tilde)
1
V (T )
/ + 1 T T +
1 F
L=
(2.5)
F + D
gs
4
2
and is invariant under the nonlinear supersymmetry transformations
1
= 1 + T ,
2
T = 2 2 T
2 T T ,
A =
.
(2.6)
For later reference we notice that the supersymmetry transformation on abelian gaugino
take the form = g(T ) where g(T ) is function of the tachyon field only, taken its
values in the spinorial representation of SO(1, 9).
In a way, the nonlinear supersymmetry involving the tachyon field is forced on us
by the desire of working with pure YangMills without the presence of the supergravity
multiplet. The open string vacuum, where this second symmetry appears, will be important
for us since as we will see in Section 4 a linear combination of the linear and nonlinear
supersymmetries guarantees the supersymmetry of the fluctuations around the instanton.
92
(x 2
1
P = (1 (8)).
2
The SO(1, 9) fermionic indices in (2.2) and (2.3) are split according to SO(8) indices as
& = (&a , &a ) (note all the tendimensional spinors are MajoranaWeyl). Most importantly
for our analysis the antisymmetric product of two SO(1, 9) matrices decompose under
SO(8) matrices as
mn = mn P+ + mn P .
As shown in Appendix A, the eightdimensional solution (3.1) satisfies a very important
identity
1
r1 r8
r1 r8
MABCD t(8)
(3.2)
+ (8)
FrA1 r2 FrB3 r4 FrC5 r6 = 0,
2
with the relative positive sign for our choice of the SO(8) representation for the instanton.
(For the singletrace part this has been noticed in [6].)
For the solution (3.1), using (3.2) and the previous conventions, the ( )2 correction to
the gaugino supersymmetry transformation (2.3) takes the form
( )2 A
(3.3)
M BCD m1 m6 P & FmB1 m2 FmC3 m4 FmD5 m6 ,
4
2 4!
the appearance of the projector P is related to the choice of the SO(8) representation used
to define the instanton.
Since we have already restricted ourselves to eightdimensional configurations, it is easy
to see that there may be further relations between &(0) and &(2) by virtue of Hodge
duality. But in order to see these we need some facts about eightdimensional selfduality
equations and their solutions [16,17], discussed in more details in Appendix A. The gauge
field (3.1) is a solution of the following selfduality equation in flat R8 [16,17]
MABCD F C F D 8 F C F D = 0.
(3.4)
&(2) A =
This equation is conformal, and the selfduality is guaranteed by the properties of the SO(8)
matrices.
4 We refer to Appendix A for details about SO(8) instantons, in particular the chirality discussion and their
generalisations to multiinstantons solutions.
93
One can check than the solution (3.1) also satisfies a relation
2
1
m1 4f (, x) 8 F A = M A BCD F B F C F D .
(3.5)
2
Note that we are not using the curved (nonconformal) counterpart of (3.4), but stay in the
flat space and simply use another property of the solution. Eq. (3.5) allows to establish a
(0)
connection between the quadratic and quartic term in the gauge field in (2.1) and & and
(2)
& . Notice that in the lhs of (3.5) the dependence on m2 has dropped out by virtue of the
first Pontryagin class p1 (F ) vanishing on the solution. We finally rewrite the full resulting
supersymmetry as
2
1 rs A
4
A
2
& = Frs 1 + ( ) m1 f (, x) P &.
(3.6)
8
3
Using (B.7) to work out the contraction between the spacetime matrices and the SO(8)
matrices used to construct the instanton we see that the supersymmetry transformation
of the SO(8) part of the gaugino ab takes the form
2
4
2
&
(
& ab = 7f (, x)(P+ )ab
(3.7)
f
(,
x)
)
m
P
,
1
+
1
(3.8)
by remarking that the operator (P ) projects on the chirality space opposite to the one
used to construct the instanton
(P+ )ab
P = 0.
Therefore the ( )2 contribution to the supersymmetry transformation & vanishes
(2)
identically. Since there is no background for the fermions & A = 0 for the instanton (3.1)
and the ( )2 piece of the Lagrangian (2.1), in the SO(8) bosonic background (3.1), is
invariant under the lowestorder linear supersymmetry transformations
&(0) L(2)
for (3.1) = 0.
(3.9)
(2)
94
zero, and the limit corresponds to the usual open string vacuum
, for all x,
open string vacuum: lim f (, x) = 0 for
0,
and x = 0.
(3.10)
In this regime the linear supersymmetry (2.3) is not broken and the nonlinear supersymmetry (2.6) is present. There is no relation between these two supersymmetries.
When sitting on the instanton x = 0, the zerosize limit behavior is different as the form
factor blows up to infinity: this is the closed string regime
closed string vacuum: lim f (, 0) 2 = .
0
(3.11)
In this regime the instanton dissolves and the nonlinear supersymmetry get promoted to
linear supersymmetry and a linear combination of the two supersymmetries (3.8) and (2.6)
is expected to survive.
2
+
+
z
z
+
.
d
m
2
gs
1920 gs
60 gs
95
We would like to take this a bit further and in particular study the coupling to background
gauge fields.
Before we go on we comment on the coefficients in (4.3). The first constant term is the
energy of the instanton, and thanks to the identities of Section A.2, it receives contributions
from the singletrace term only. It is therefore proportional to m1 . The second term
from the fluctuations around the instanton in the SO(8) part of the gauge group receives
contributions from both the single and the doubletrace part, and depends on m1 and m2 .
We observe here that this leads to emergence of two different string tensions here the
singletrace part m1 /(2 gs ) 1/(2 gs ) has a tension of a Dstring while the doubletrace part m2 /(2 gs ) 1/(2 ) has a fundamental string tension.
A more interesting test comes from considering the next order perturbation on the space
(1)
(0)
of connections Am = A(0)
m + Am (m = 1, . . . , 8) around the instantonic solution (3.1) Am .
(1)
The fluctuation Am = am lives in the SO(24) part of the gauge group. Since we choose a
purely bosonic instanton background, the fermions are all of the first order type = (1) .
m and
Fermions, respectively in SO(8) and SO(24), are associated with the fluctuations F
(1)
A . At this point another crucial difference between the single and doubletrace parts
emerges. The former does not allow any mixing of the SO(8) and SO(24) parts and an
action (4.3) with half of the original supersymmetries preserved is all one gets. The doubletrace term however contains the supersymmetric extension
F
)
8L(2) 4m2 trSO(8) (
D ) trSO(8)(F
F
) ,
D ) trSO(8)(F
+ 2m2 trSO(8) (
(4.4)
whose supersymmetry variation under (3.8) is
3
& 8L(2) m2 f (, x) (& P+ D).
(4.5)
Contrary to the action evaluated for the classical background, the action for the fluctuations
is not invariant by itself under the supersymmetry generated by &. We have observed
already the similarity between the linear and nonlinear supersymmetries, and it looks
natural to try to use the latter in order to compensate for & (8L(2) ). As noticed below (2.6)
the geometry of the nonlinear supersymmetry transformation is controlled by the profile
of the tachyon field. Moreover, beyond the group theory identifications of the gauge SO(8)
and the Lorentz SO(8), a tachyon field dependence can occur in the map between the
two gaugini and . We confine this freedom to a function W (T ) whose relation to the
potential is yet unclear. Acting with the nonlinear supersymmetry (2.6) on (4.4) we get
D),
8L(2) m2 W (T ) (
(4.6)
The fluctuations around the instanton (3.1) can be made supersymmetric under a linear
combination of the two supersymmetry transformations & and provided the following
identification between the supersymmetry parameters
=
(f (, x))3
P+ &.
W (T )
(4.7)
96
(4.8)
is satisfied and the positivechirality component of + survives. Notice that this chirality
is the one of the SO(8) representation used to construct the instanton. The remaining
supersymmetry, generated by + , will give rise to lightcone GreenSchwarz fermions
for an heterotic string.
As we will see in the next subsection, these fermionic zero modes will give rise to
both the fermions of a chiral (8, 0) supersymmetric sigmamodel and the gauge degrees
of freedom for the wordsheet theory of an heterotic string obtained from the fluctuations
around the instantons (3.1).
4.1. Gauge coupling
(1)
(4.10)
Notice that (3.8) implies that (x) scales as f (, x) and the previous integral has a finite
nonzero limit when goes to zero. The other fermion bilinear in (2.1) involving
does not contribute to the kinetic term for the lightcone fermions S.
97
From the interactions between the SO(8) deformation (4.2) and the SO(24) fermions
m2 d 10 x trSO(24) (
) trSO(8) (Fn Fn )
we derive the kinetic term for the twentyfour fermions
)h .
m2 d 2 trSO(24)(
(4.11)
5. Discussion
We would like to underline some of the important points and list some open questions.
Starting from pure tendimensional YangMills action we are able to produce a Dstring and more surprisingly/controversially a heterotic string in the background of
eightdimensional SO(8) instanton. The former has a nonvanishing vacuum energy and
supersymmetric (8, 0) worldsheet. Its very plausible connections to noncommutative
6 Curiously enough, one of very few features of ordinary fourdimensional instantons that is replicated by the
SO(8) instanton is the possibility of extending the solution to a curved space by setting the gauge connection
equal to spin connection [19].
98
solitons and to D1D9 (D0D8) systems are not very clear to us at the moment and are
beyond the scope of this paper. There may be some interesting questions to be asked in
this context. As for the heterotic string to which we will turn now, the SO(9) symmetry
of the instanton solution gets reflected in the periodicity of one of the string worldsheet
coordinates.
Our main interest is in the appearance of the fundamental heterotic string. Technically
speaking, its existence is due to the presence of the doubletrace structures in the Yang
Mills F 4 terms responsible for mixing of SO(8) and SO(24). These terms were first seen
from the supersymmetry analysis on the tendimensional YangMills action, and as we
tried to argue here may also be related to the field redefinition freedom (fixed by the
requirement that the equations of motion are proportional to the worldsheet function
and by supersymmetry) in the classical open string field theory Lagrangian. As a result
by expanding around the soliton we obtain a string action in the background where the
SO(8) gauge connection is set to the spin connection. It might be interesting to notice that
this doubletrace structure also allows to generate kinetic terms for YangMills fields in
the bulk by giving classical values to tr F 2 . Indeed, in the SO(8) instanton background,
t(8) (tr F 2 )2 tr F F , and in a way the SO(8) part is used as the center of the gauge
group.
Finally, a crucial point in the analysis was the mutual cancellation of linear and nonlinear supersymmetry. The latter is known for the abelian case but applicable to the
SO(8) group by the magic of the SO(8) instantons, which realizes an identification of
the Lorentz and gauge SO(8) groups. The resulting connection between linear and nonlinear supersymmetry parameters contains a dependence of the tachyon and the size of the
instanton via (f (, x))3 /W (T ). Of course it is natural to expect that this ratio is a constant
and thus one gets a correlation between the tachyon and the size of the instanton (e.g.,
since f (, x) is singular when the instanton shrinks to the zero size, W (T ) should behave
accordingly). This correlation is not very explicit though, since it is not yet completely
clear how W (T ) is related to the tachyon potential. One could hope to understand better
what are the respective regimes where the Dstring and the heterotic string discussed above
dominate. Such a possibility of at least partially recovering the tachyon potential from the
lowenergy analysis in our opinion deserves further study.
Acknowledgements
It is a pleasure to thank Carlo Angelantonj, Costas Bachas, Emilian Dudas, Elias Kiritis
and Hong Liu for discussions, as well as Tigran Tchrakian for guidance into the SO(8)
instanton literature. The work of R.M. and P.V. is supported in part by EEC contract HPRNCT200000122; R.M. is also supported by the INTAS contract 551590. The work of S.S.
is supported in part by the OJI award from DOE.
99
rm = 2
2xm
(m = 1, . . . , 8),
2 + x2
r9 =
2 x2
.
2 + x2
Thanks to the SO(9) rotational invariance of the solution, we could have considered the
complement of r9 = +, without any change in the discussion of the main text. The
solution with no singularity at the origin is given by
(A )m = 4imn P
x n 2
( 2 + x 2 )2
(A.2)
2
,
( 2 + x 2 )2
(A.3)
x n 2
x 2 ( 2 + x 2 )2
(A.4)
6
.
x 4 ( 2 + x 2 )2
(A.5)
100
(A.6)
(A.7)
(A.5) has a form factor singular at the origin and therefore does not qualify for such nonconformal generalisation.
The solution (A.3) satisfies a relation
2
8 f (, x) 8 F = F F F ,
(A.8)
with the flat metric gmn = mn . Some of the discussion in the main text relies heavily on
this relation.
One can show that in 4d dimensions the following doubleduality relation holds for
any Einstein metric
m m
(2d)!2 d R 1 2d 1 2d
n n
= & m1 m2d n1 n2d d R 1 2d 1 2d & 1 2d 1 2d ,
where d R is the completely antisymmetrized product of 4ddimensional Riemann
curvatures. Similarly to fourdimensional case [21], for d = 2 this doubleduality implies
a single duality relation (A.1) for a gauge field A [19], provided
1 mn
1 mn
F
= R
P mn .
A
(A.9)
= P mn ,
2
2
Thus the procedure of setting the gauge connection to spin connection enables one to
extend the construction of the instanton to curved backgrounds.
This setup is generalized to multiple instanton solutions classified by maps of the
eightsphere S 8 onto itself [22]:
n wp m wq
pq P ,
(1 + wl wl )2
N
7
wl l = x8 I +
x m m .
(F )mn = 2
(A.10)
m=1
101
(A.12)
for any matrix N in the adjoint of the group, generalising the identity used in [6]. The
completely symmetric tensor MABCD of Eq. (2.4) (defined in [10]) has two parts the
symmetrized singletrace and the doubletrace. It is therefore sufficient to check that
tr t(8)r1 r8 (F )r1 r2 (F )r3 r4 (F )r5 r6 Nr7 r8
2
= 25 (7d 11)(d 4) f (, x) tr (F )r7 r8 Nr7 r8 ,
tr (8)r1 r8 (F )r1 r2 (F )r3 r4 (F )r5 r6 Nr7 r8
2
= 24 6! f (, x) tr (F )r7 r8 Mr7 r8
therefore for d = 8 (A.12) is satisfied.
t(8) r1 r8 tr[(F )r1 r2 (F )r3 r4 ] tr[(F )r5 r6 Nr7 r8 ]
2
= 28 (d 1)(d 8) f (, x) tr (F )r7 r8 Mr7 r8 ,
B
tr (8)r1 r8 (F )A
r1 r2 (F )r3 r4 (F )r5 r6 Nr7 r8 = 0
therefore for d = 8 (A.12) is satisfied.
A.3. An action for the instanton
This solution has been shown [6,18] to be of importance for constructing the string as
a soliton for the fivebrane in ten dimension. It was there shown that the solution satisfies
the equationofmotion [23]
E (0) (A) =
1
D f (, x)F = 0,
f (, x)
and
E (2) (A) =
( )2 e
M A BCD
go2 3! 8gs
(A.13)
102
C D 1 B C D
D F B F
F D F F F
4
(A.14)
(B.1)
(B.2)
The SO(1, 9) gamma matrices are decomposed into SO(8) gamma matrices as
0
.
=
(B.3)
0
acts from the space of spinor of positive SO(8) chirality into the space of negative SO(8)
chirality: : S+ S . does the opposite. Antisymmetric products of even numbers
of Gamma matrices respects the direct sum S+ S . In particular
0
=
(B.4)
= P+ + P .
0
The antisymmetric product of two SO(8) matrices can be expressed in terms of the
completely antisymmetric octonionic structure constants
ab
+ cmn ab .
(mn )ab = mn
(B.5)
We need the following matrices identities (all antisymmetrizations are with unit
weight)
{ , } = 2 ,
pq
rs
= pq
rs
(B.6)
[s
+ 4[p q]
r]
sr
+ 2pq
,
= (d 2) + (d 1) ,
sn
,
tr r1 rn sm s1 = 24 n! rs11r
n
(8)r1 r8 = r1 r8 (8) ,
(8)
r1 rn
s1 s8n
rn r1
(B.7)
(B.8)
(B.9)
(B.10)
= n! s1 s8n (8) .
(B.11)
103
References
[1] E. Witten, Interacting field theory of open superstrings, Nucl. Phys. B 276 (86) 291.
[2] E. Witten, On background independent open string field theory, Phys. Rev. D 46 (1992) 5467,
hepth/9208027;
E. Witten, Some computations in background independent offshell string theory, Phys. Rev.
D 47 (1993) 3405, hepth/9210065;
S.L. Shatashvili, Comment on the background independent open string theory, Phys. Lett. B 311
(1993) 83, hepth/9303143;
S.L. Shatashvili, On the problems with background independence in string theory, Algebra and
Anal. 6 (1994) 215, hepth/9311177.
[3] N. Berkovits, A new approach to superstring field theory, Fortsch. Phys. 48 (2000) 31, hepth/9912121.
[4] S.L. Shatashvili, On field theory of open strings, tachyon condensation and closed strings, Talk
at Strings 2001, Mumbai, India, hepth/0105076.
[5] D.J. Gross, W. Taylor, Split string field theory II, hepth/0106036.
[6] M.J. Duff, J.X. Lu, Strings from fivebranes, Phys. Rev. Lett. 66 (1991) 1402.
[7] E. Witten, Dbranes and Ktheory, JHEP 12 (1998) 019, hepth/9810188.
[8] A. Sen, Descent relations among bosonic Dbranes, Int. J. Mod. Phys. A 14 (1999) 4061, hepth/9902105;
A. Sen, NonBPS states and branes in string theory, hepth/9904207;
A. Sen, Universality of the tachyon potential, JHEP 9912 (1999) 027, hepth/9911116.
[9] A.A. Gerasimov, S.L. Shatashvili, On exact tachyon potential in open string field theory,
JHEP 0010 (2000) 034, hepth/0009103;
D. Kutasov, M. Mario, G. Moore, Some exact results on tachyon condensation in string field
theory, JHEP 0010 (2000) 045, hepth/0009148;
D. Kutasov, M. Mario, G. Moore, Remarks on tachyon condensation in superstring field
theory, hepth/0010108;
A.A. Gerasimov, S.L. Shatashvili, Stringy Higgs mechanism and the fate of open strings,
JHEP 0101 (2001) 019, hepth/0011009.
[10] M. Cederwall, B.E. Nilsson, D. Tsimpis, D = 10 super YangMills at O( 2 ), hepth/0104236.
[11] K. Peeters, P. Vanhove, A. Westerberg, Supersymmetric higherderivative actions in ten and
eleven dimensions, the associated superalgebras and their formulation in superspace, Class.
Quant. Grav. 18 (2001) 843, hepth/0010167.
[12] M. Cederwall, B.E. Nilsson, D. Tsimpis, The structure of maximally supersymmetric Yang
Mills theory: constraining higherorder corrections, hepth/0102009.
[13] N. Sakai, M. Abe, Coupling constant relations and effective Lagrangian in the type I superstring,
Prog. Theor. Phys. 80 (1988) 162.
[14] C. Bachas, C. Fabre, E. Kiritsis, N.A. Obers, P. Vanhove, Heterotic/typeI duality and Dbrane
instantons, Nucl. Phys. B 509 (1998) 33, hepth/9707126.
[15] J. Bagger, A. Galperin, A new Goldstone multiplet for partially broken supersymmetry, Phys.
Rev. D 55 (1997) 1091, hepth/9608177;
T. Hara, T. Yoneya, Nonlinear supersymmetry without the GSO projection and unstable D9brane, Nucl. Phys. B 602 (2001) 499, hepth/0010173;
E. Dudas, J. Mourad, Consistent gravitino couplings in nonsupersymmetric strings, hepth/0012071;
S. Terashima, T. Uesugi, On the supersymmetry of nonBPS Dbrane, JHEP 0105 (2001) 054,
hepth/0104176.
[16] D. Tchrakian, Spherically symmetric gauge field configurations with finite action in 4P dimensions (P = Integer), Phys. Lett. B 150 (1985) 360.
104
Abstract
We investigate SchwingerDyson equations for correlators of Wilson line operators in noncommutative gauge theories. We point out that, unlike what happens for closed Wilson loops, the
joining term survives in the planar equations. This fact may be used to relate the correlator of an
arbitrary number of Wilson lines eventually to a set of closed Wilson loops, obtained by joining
the individual Wilson lines together by a series of welldefined cutting and joining manipulations.
For closed loops, we find that the nonplanar contributions do not have a smooth limit in the
limit of vanishing noncommutativity and hence the equations do not reduce to their commutative
counterparts. We use the SchwingerDyson equations to derive loop equations for the correlators of
Wilson observables. In the planar limit, this gives us a new loop equation which relates the correlators
of Wilson lines to the expectation values of closed Wilson loops. We discuss perturbative verification
of the loop equation for the 2point function in some detail. We also suggest a possible connection
between Wilson line based on an arbitrary contour and the string field of closed string. 2001
Published by Elsevier Science B.V.
1. Introduction
Noncommutative gauge theories are realized on branes in the zero slope limit in the
presence of a large NSNS Bfield [18]. Recently these theories have attracted a lot of
attention. Various aspects of these theories have been studied in [936].
In ordinary gauge theories, a generic gaugeinvariant observable is provided by an
arbitrary closed Wilson loop. Noncommutative gauge theories have more general gaugeinvariant observables, defined on open contours. Such gaugeinvariant observables in noncommutative gauge theories were constructed in [16]. Different aspects of these were
* Corresponding author.
106
where
[x , x ] = i ,
(1.2)
and the trace in (1.1) is over both the gauge group, taken here to be U (N), as well as the
operator Hilbert space. These open Wilson lines are gaugeinvariant in noncommutative
gauge theories, unlike in ordinary gauge theories, provided the momentum k associated
with the Wilson line is fixed in terms of the straight line joining the end points of the
path C, given by y ( ) where 0 1, by the relation
y (1) y (0) = k .
(1.3)
The path C is otherwise completely arbitrary. When k vanishes, the two ends of the path
C must meet and we have a closed Wilson loop.
In an earlier work [41], based on a perturbative analysis of correlation functions of
straight Wilson lines with generic momenta, we had suggested that at large momenta the
Wilson lines are bound into the set of closed Wilson loops that can be formed by joining the
Wilson lines together in all possible different ways. In the present work we will establish
a more general connection between correlators of Wilson lines and expectation values of
Wilson loops in a nonperturbative setting, for arbitrary Wilson lines. In this generic case,
however, the closed Wilson loops to which the Wilson lines are related are not formed
by simply joining the Wilson lines together, but by more complicated cutting and joining
manipulations. Also, the statement is valid for arbitrary momenta, not necessarily large,
carried by the Wilson lines. We will use the framework of SchwingerDyson equations and
the closely related loop equations in the planar limit. In the context of noncommutative
gauge theories similar equations have been studied earlier in [4749].
This paper is organized as follows. In the next section we summarize some aspects of
operator formulation of noncommutative gauge theories, which is used throughout this
paper, and in particular list some useful identities. In Section 3 we derive Schwinger
Dyson equations for the correlators of open and closed Wilson observables and discuss
these at finite N as well as in the planar limit. As in commutative gauge theories, the
splitting term disappears from the planar equations. However, unlike in the case of closed
Wilson loops, the joining term survives in the planar equations for open Wilson lines. This
has the consequence of eventually relating them to closed Wilson loops. We also find that at
finite N , the splitting term does not reduce to the ordinary gauge theory result in the limit in
which the noncommutativity is removed. We trace this result to the UVIR mixing in noncommutative gauge theories. In Section 4 we use the results of Section 3 to write down loop
equations for the correlators of open and closed Wilson observables. We consider the loop
107
equation for the 2point function of the open Wilson lines in the planar limit and discuss
the verification of this equation in t Hooft perturbation theory in some detail. Section 5
contains a discussion of a possible connection between a Wilson line based on an arbitrary
contour and string field for closed string and the nonperturbative meaning of the new loop
equations derived here. In Appendix A we give details of the perturbative calculations.
23 = 32 = b ,
(2.2)
with all other components vanishing, and let us define the operators
x2 + x3
x0 + x1
,
b=
,
a=
2a
2b
which satisfy the standard harmonic oscillator algebra,
a, a = 1,
b, b = 1.
The operator Hilbert space trace is then defined by
x)n
x)
trH O(
na , nb O(
a , nb .
= (2)2 a b
(2.3)
(2.4)
(2.5)
na ,nb
Note that with this definition of the operator Hilbert space trace, the coupling constant g
appearing in the action (2.1) is dimensionless.
We use the standard Weyl operator ordering,
x)
O(
= d 4 y O(y) (4)(x y),
(2.6)
where the operator deltafunction is defined in terms of the Heisenberg group elements by
d 4 k ik.y ik.x
(4)
e
e .
(x y) =
(2.7)
(2)4
The use of this operator deltafunction simplifies many calculation because it shares some
properties of the usual deltafunction. For example, (2.6) and
trH (4)(x y) = 1.
(2.8)
108
There are, of course, differences as in the following identity which encodes the star product:
i
(2.9)
1 (x)t
2 (x)t
1 (x)
2 (x)
Tr O
a (4) (x z) Tr O
a (4) (x z) = Tr O
O
,
d 4z
a
(2.10)
and the second one splits two operators which are inside the same trace,
1 (x)t
2 (x)t
d 4z
Tr O
a (4) (x z)O
a (4) (x z)
a
1
(2)4 det
2 (x)
1 (x)
Tr O
.
Tr O
(2.11)
Here the t a s are the generators for the gauge group, which we have taken to be U (N),
with the normalization dictated by the completeness condition
a
(2.12)
tija tkl
= il j k .
a
Here the subscripts 0s indicate that the pathordered phase factor runs from = 0 to
C [y]s1 , which
= s along the curve C, and y(s) y(0) = ks . The Wilson operator W
runs from = s to = 1, is defined similarly:
i ks .x
where y(1) y(s) = ks . These operators are related to the Wilson observable as follows:
109
C [y]01 = e 2 ks k W
C [y]0s W
C [y]s1,
W
i
(2.16)
and 2 therefore,
i
C [y]0s W
C [y]s1
WC [y] = eik.y(0)e 2 ks k Tr W
Cs [ys ]
= eik.y(s) Tr W
WCs [ys ],
(2.17)
where in the second step we have used the cyclic property of the trace and recombined the
two operators in the opposite order. The contour Cs is given by
ys ( ) = y( + s),
0 (1 s)
= y( 1 + s) + y(1) y(0),
(1 s) 1.
(2.18)
It is obtained from the curve C by cutting it at a point = s and rejoining the two pieces
in the opposite order, as shown in Fig. 1.
If the original curve is a straight line, then the transformed curve is also a straight line
shifted by an amount s( k). It is easy to see more directly that such shifts are a symmetry
of the straight Wilson line. This symmetry was used very effectively in [41] to simplify
perturbative calculations. More generally, the cyclic symmetry relates Wilson observables
defined on contours that are nontrivially different. As we shall see later, the quantity
(k)
VC [y] = i dy ( ) eik.y( )
(2.19)
C
C [y]01 , which goes over the full parametric range, will be denoted by W
110
3. SchwingerDyson equation
In this section we will first derive the SchwingerDyson equation for multipoint
correlators of Wilson observables and then analyse it at finite N as well as in the
planar limit. The SchwingerDyson equation follows from the standard functional integral
identity
S
d 4z
0=
DAb (x)
e Tr WC1 [y1 ]t a (4) (x z)
a
A (z)
a
Cn [yn ] .
C2 [y2 ] Tr W
Tr W
(3.1)
Using
Aa (z)
C [y] = i
W
i
C [y]0s t a (4) (x z) W
C [y]s1,
dy (s) e 2 ks k W
(3.2)
and the joining and splitting identities, (2.10) and (2.11), we get
1
C2 [y2 ] Tr W
Cn [yn ]
(x)
Tr WC1 [y1 ]D
Tr W
2
g
n
i
dyl (s) e 2 kls kl
=i
l=2 C
Cl [yl ]s1 Tr W
C1 [y1 ]W
Cn [yn ]
C2 [y2 ] Tr W
Cl [yl ]0s W
Tr W
i
i
+
dy1 (s) e 2 k1s k1
4
(2) det
C1
C1 [y1 ]0s Tr W
Cn [yn ] .
C1 [y1 ]s1 Tr W
C2 [y2 ] Tr W
Tr W
(3.3)
In this equation yl (s) yl (0) = kls and each of the contours C1 , C2 , . . . , Cn may be open
or closed.
3.1. Closed Wilson loop
Let us first consider a single closed Wilson loop. In this case Eq. (3.3) reduces to
1
(x)
C [y]s1 .
C [y]0s Tr W
Tr
[y]
D
=
dy (s) Tr W
W
C
2
4
g
(2) det
C
(3.4)
Here C is a closed curve. The righthand side of (3.4) has a disconnected piece.
However, because of momentum conservation, the disconnected piece contributes only
C [y]0s ) and
Tr(W
111
1
(x)
F
Tr WC [y]D
2
g
i
C [y]0s Tr W
C [y]s1
dy (s) (4) y(s) y(0) Tr W
=
V
C
i
(2)4 det
C [y]s1
C [y]0s Tr W
dy (s) Tr W
.
conn.
(3.5)
The second term on the righthand side of (3.5) contains the connected part of the 2point
function of open Wilson lines. At finite N , it is easy to see that this term is down by a
factor of 1/N 2 relative to the other terms in the equation. In the planar limit, therefore, this
term drops out and the planar equation looks formally like the corresponding equation in
commutative gauge theory. This is consistent with the perturbative result that, except for in
an overall phase, the dependence on the noncommutative parameter drops out of planar
diagrams. However, there are new gaugeinvariant observables in noncommutative gauge
theory, the open Wilson lines, and so there are new equations. As we shall see shortly,
these new equations have a nontrivial planar limit. One might then say that it is these new
equations that reflect the new physics of noncommutative gauge theory.
At finite N the second term on the righthand side of (3.5) contributes. One might
wonder whether this term reduces to its commutative counterpart in the limit of small
noncommutative parameter. An argument has been presented in [49] suggesting that this
is the case. However, we find that, in fact, the small limit of this term is not smooth, at
least in perturbation theory, as we will now show.
At the lowest order in perturbation theory, the second term on the righthand side of
(3.5) evaluates to
ig 2 N
(2)4 det
C
1
dy (s) 2
ks
s
dy( ) e
0
iks .y( )
1
dy( ) e
iks .y( )
(3.6)
For simplicity, let us specialize to a rectangular contour of sides L1 and L2 . We will also
take to be of the form in (2.2) with a = b = 0 . In this case the diagrams that contribute
to (3.6) are shown in Fig. 2.
112
e
f
()
f
i(L
L
)
1
f
()
+
i
e
()
,
1
2
(2)4 02
(3.7)
where = L1 1 L2 is the magnetic flux passing through the rectangular contour and
ds
1 eis .
f () =
(3.8)
s
In the limit of small noncommutativity, is large, and then f () ln . In this case
the righthand side of (3.5) is divergent with the leading term going as ln /02 . So we
see that if we take the limit of small noncommutativity first, keeping N finite, we do
not recover the commutative result. It is easy to see in perturbation theory that the origin
of this problem lies in UVIR mixing. It has been argued in [22] that this phenomenon
renders loop diagrams finite in noncommutative field theory. Now, the diagrams in Fig. 2
that contribute to the righthand side of (3.5) at order 1/N 2 in the lowest order in t Hooft
perturbation theory actually come from nonplanar oneloop diagrams on the lefthand side
of this equation, as shown in Fig. 3.
In fact, the relevant amplitude is
d 4p
d 4 q eip(y1 y3 ) eiq(y2 y4 ) ipq
g 4 N dy1 dy3 dy2 dy4
e
.
(2)4
(2)4
p2
q2
(3.9)
We can estimate the above momentum integral as follows
d 4 q eip(y1y3 ) eiq(y2y4 ) ipq
d 4p
e
4
(2)
(2)4
p2
q2
1
1
1
1
4 2 (y1 y3 )2 4 2 (y2 y4 )2
=
(3.10)
1
1
ln
y
y
y

<
.
if
y
1
3 2
4
0
2(2)4 2
y1 y3 y2 y4 
0
In commutative gauge theory these diagrams have short distance singularities which
are linearly divergent. In the noncommutative theory they get regularized at the noncommutativity scale, as can be seen from the above expression. The singularities of
the commutative theory reappear in the limit of small noncommutativity. This is what
113
is reflected in the singular behaviour of the righthand side of (3.5) for small noncommutativity.
3.2. Open Wilson lines
The generic equation satisfied by the npoint function of Wilson lines is (3.3). The
second term on the righthand side of this equation has a disconnected part which is given
by
Cn [yn ]
C1 [y1 ]s1 Tr W
C2 [y2 ] Tr W
C1 [y1 ]0s Tr W
Tr W
C1 [y1 ]s1 Tr W
Cn [yn ] .
C1 [y1 ]0s Tr W
C2 [y2 ] Tr W
+ Tr W
(3.11)
Because of momentum conservation, the first term contributes only for y(s) = y(0), while
the second term contributes only for y(s) = y(1). In either term we get back the original npoint function. This is just like for the closed Wilson loop discussed above. The connected
part of the second term on the righthand side can be easily seen to be down by a factor of
1/N 2 compared to the other terms in the equation. In the planar limit, therefore, it drops
out, leaving only the joining term (the first term on the righthand side), apart from the
disconnected term mentioned above. We then have the result that the planar Schwinger
Dyson equation for Wilson lines expresses any npoint function entirely in terms of (n1)point functions. By iterating this procedure (n 1) times we may, in principle, express any
npoint function entirely in terms of closed Wilson loops.
The simplest example of the above phenomenon is provided by the 2point function. In
this case, the planar SchwingerDyson equation reads
1
(x)
C2 [y2 ]
Tr WC1 [y1 ]D
Tr W
2
g
i
C2 [y2 ]0s W
C2 [y2 ]s1
C1 [y1 ]W
= i dy2 (s) e 2 k2s k2 Tr W
C2
(2)4 det
C1
C1 [y1 ]s1 Tr W
C2 [y2 ]
C1 [y1 ]0s Tr W
Tr W
C1 [y1 ]0s Tr W
C2 [y2 ] .
C1 [y1 ]s1 Tr W
+ Tr W
(3.12)
We see that the righthand side involves closed Wilson loops, apart from the the 2point
function itself. The closed curves involved are obtained by first traversing the curve C1
given by y1 ( ), 0 1 and then the curve given by
y( ) = y2 ( + s) y2 (s) + y1 (1),
0 (1 s),
= y2 ( 1 + s) y2 (s) + y1 (0),
(1 s) 1,
(3.13)
for different values of s, 0 s 1. Note that the closed curves obtained in this way are
continuous because of momentum conservation.
114
C3 [y3 ] ,
C2 [y2]s1 Tr W
C2 [y2 ]0s W
C1 [y1 ]W
Tr W
C3 [y3 ]0s W
C2 [y2 ] ,
C3 [y3]s1 Tr W
C1 [y1 ]W
Tr W
(3.14)
depending on whether C1 and C2 or C1 and C3 combine into a single curve. These 2point
functions are themselves related to closed loops, as discussed above. Thus, the 3point
function can eventually be related to closed Wilson loops, there being two distinct sets
of structures for the closed contours involved. These closed contours can be obtained
explicitly, as we have done above for the case of the 2point function. For npoint function
one eventually gets (n 1)! distinct structures for the closed loops.
The above discussion establishes a general link between the correlators of Wilson lines
and the expectation value of closed Wilson loops. In a previous work [41] we have
presented a perturbative proof for long straight Wilson lines to be bound together into
closed loops. The connection we have found here is more general in that the Wilson lines
are based on arbitrary contours and the momenta need not be large. The closed loops we
now find are also more general. We believe that the planar SchwingerDyson equation
indeed supports our previous claim for the high energy behaviour of the Wilson lines.
This is because, firstly, it can be argued that the planar approximation is always valid in
the high energy limit (or large limit) since the splitting term is suppressed by 1/ det .
Secondly, we need to repeatedly insert the equation of the motion operator into the Wilson
line correlator in order to eventually relate it to Wilson loops. Such an operation picks
up contact terms in the sense that it makes two Wilson lines touch each other. In high
energy limit, we expect to rediscover such contact terms, although the real singularities are
expected to be regulated by the noncommutativity. Finally, the set of relevant closed loops
may collapse to that found in [41], which is the set of extreme configurations of Wilson
loops obtained by simply joining the Wilson lines endtoend.
C [y]
2 W
y ( )y ( )
i
(x)
C [y] i y ( )F
(x)
C [y] 1
= e 2 k k W
W
W
C [y]0 i y ( )F
i
C [y]0 i y ( )D
(x)
C [y] 1 ,
( )e 2 k k W
W
(4.1)
115
careful. However, if we vary both the ends keeping k fixed, and assume that the tangents to
the contour at the ends are identical, 3 then a very similar identity is valid at the ends also.
We need to separate out the equation of motion piece on the right hand side of (4.1).
This may formally be done by defining the loop laplacian [50]
2
lim
y 2 ( ) 00
0
dt
2
.
y ( + t/2)y ( t/2)
(4.2)
In principle, if the quantum theory is regularized then the first term in (4.1) does not have
any singularities as and so the loop laplacian picks up only the deltafunction term
on the righthand side of this equation. 4 We then get
C [y]
i
2W
C [y] 1 .
C [y]0 i y ( )D
(x)
= e 2 k k W
W
y 2 ( )
(4.3)
Using this in (3.3) we get the loop equation for Wilson line correlators
1 2
WC1 [y1 ]WC2 [y2 ] WCn [yn ]
2
2
g y1 ( )
n
ds i y1 ( ).is yl (s) eik1 .y1 ( )ikl .yl (s)
=
l=2 C
C1 [y1 ]W
1
s
1
(2)4 det
C1
C1 [y1 ]0s Tr W
Tr W
(4.4)
where the contours C1 and C1s are as defined in (2.18). Also, the operator WC [y] has
been defined in (2.13) and WC [y] is the gaugeinvariant observable defined in (1.1).
4.1. Twopoint function
We will now discuss the case of the 2point function in some detail. For the 2point
function, the loop equation reduces to
1 2
W
[y
]W
[y
]
C
1
C
2
1
2
g 2 y12 ( )
C1 [y1 ]W
C2s [y2s ]
= ds i y1 ( ).is y2 (s) eik1 .y1 ( )ik2 .y2 (s) Tr W
C2
3 Under these conditions the variation of contour at the end points is effectively like at an interior point.
4 In practice the separation of the two terms on the right hand side of (4.1) is a nontrivial issue. For a discussion
116
1
(2)4 det
C1
i
ds i y1 ( ).is y1 (s) eik1 .y1 ( )+ 2 k1s k1
C1 [y1 ]0s Tr W
(4.5)
We are interested in the planar limit of this equation. In this limit only the disconnected
part of the 3point function, appearing on the righthand side of (4.5), survives. This
disconnected part is
C1 [y1 ]0s Tr W
C1 [y1 ]s1 Tr W
C1 [y1 ]0s + Tr W
C1 [y1 ]s1
C1 [y1 ]0s + Tr W
C1 [y1 ]s1
eik1 .y1 ( ) Tr W
WC1 [y1 ]WC2 [y2 ] .
Using this in (4.5), together with the fact that y ( ) gives tangents at the two ends of the
contour C1 , which are equal by construction, we get
1 2
WC1 [y1 ]WC2 [y2 ]
2
2
g y1 ( )
( y1 ( ))2
C1 [y1 ]s1
C1 [y1 ]0s + Tr W
ds
Tr
W
=
(2)4 det
C1
WC1 [y1 ]WC2 [y2 ]
C2s [y2s ] .
C1 [y1 ]W
+ ds i y1 ( ).is y2 (s) eik1 .y1 ( )ik2 .y2 (s) Tr W
C2
(4.6)
This is the final form of the planar loop equation for the 2point function. Notice that the
first term on the righthand side of this equation is proportional to ( y1 ( ))2 and also
it involves the original 2point function. Taken together with the left hand side, the two
terms have the form of string hamiltonian acting on the 2point function. However, it is
not clear that this term is really physically meaningful. In fact, a corresponding term in the
loop equation for the commuting gauge theory is often ignored in the regularized theory.
In the present case also, an evaluation of the coefficient of ( y1 ( ))2 cannot be done
unambiguously. This is because such a calculation involves computation of amplitudes for
splitting off of tiny bits at the two ends of the Wilson line defined on the contour C1 . The
117
computation of this is delicate and needs a regulator. So the physical significance of this
term remains unclear.
We should mention here that Eqs. (4.4) and (4.6) are new type of loop equations since
there is no analogue of these in commutative gauge theory. Also, it is clear that the planar
equation (4.6) relates the 2point function of open Wilson lines to the expectation value of
a closed Wilson loop. The latter may be obtained by solving the planar loop equation for
a closed Wilson loop. Thus one needs equations for both types of Wilson observables to
form a closed system of equations.
4.2. Perturbative verification
Perhaps the most interesting aspect of the loop equation (4.6) is its stringy interpretation.
Investigating this aspect of the loop equation is bound to be inherently nonperturbative. In
fact, recently such an exercise has been successfully carried out in [51,52] for the loop
equation in commutative gauge theory, using the AdS/CFT correspondence. A similar
exercise for the present noncommutative case seems to require a better understanding
of the connection between noncommutative gauge theory and its conjectured gravity dual
[10,11], and is beyond the scope of the present work. Here we will restrict ourselves to a
perturbative verification of (4.6). We will do the computations upto the second order in the
t Hooft coupling.
Separating out the momentum conserving deltafunction, we may parametrize the
2point function as
WC1 [y1 ]WC2 [y2 ] = (2)4 (4) (k1 + k2 )GC1 C2 [y1 , y2 ],
(4.7)
where, in perturbation theory,
(1)
(2)
(4.8)
GC1 C2 [y1 , y2 ] =
1 (k1 )
(k )
V [y1 ].VC22 [y2 ],
k12 C1
(4.9)
(k)
where VC [y] has been defined in (2.19). Operating the loop laplacian on (4.9), we get the
lowest order expression for the lefthand side of (4.6) 5
(k )
i y1 ( )eik1 .y1 ( ) VC22 [y2].
(4.10)
5 Here and in the following we have omitted the momentum conserving deltafunction factor (2 )4 (4) (k +
1
k2 ) which is present on both sides of (4.6).
118
Fig. 5. Examples of diagrams contributing to the 2point function at second order in perturbation
theory.
(k)
In arriving at this expression we have used that k.VC [y] = 0, which is true because of the
identity k.y(1) = k.y(0) which follows from the definition of k, (1.3).
On the righthand side of (4.6), at the lowest order in , the first term does not contribute.
The relevant contribution comes from the second term by setting the gauge field to zero in
each of the two Wilson line operators involved in making the closed Wilson loop. Omitting
the momentum conserving deltafunction, we get precisely the expression in (4.10).
At the next order in , there are several different types of diagrams that contribute to the
2point function. Fig. 5 shows a representative example from each type.
(2)
We have done a calculation of the quantity GC1 C2 [y1 , y2 ], which gives the second order
contribution to the 2point function. Some details of this calculation and the result are given
in Appendix A.
At the second order in , both terms on the righthand side of (4.6) contribute. The
contribution of the first term comes from the lowest order calculation of the 2point
function, while that of the second term comes from a one gauge boson exchange. In
Appendix A we have discussed in detail how each of these contributions arises as a result
of operating the loop laplacian on G(2)
C1 C2 [y1 , y2 ]. Here we only mention that some of the
terms from different sets of diagrams that appear in the calculation of the lefthand side
of (4.6) have a structure that does not occur on the righthand side. However, there are
nontrivial cancellations between the contributions of different sets of diagrams. We have
checked that many such terms disappear from the overall result for the lefthand side as
well, but we have not attempted a complete verification of this.
It would be interesting to extend the present perturbative analysis to all orders in .
To verify the new loop equation (4.4) nonperturbatively, one needs to understand
119
what the multipoint correlators of Wilson lines based on arbitrary contours map on to
in the string/supergravity dual. A better understanding of the noncommutative gauge
theory/string theory duality than we have at present appears to be necessary for this.
5. Discussion
In this paper we have investigated SchwingerDyson and loop equations in noncommutative gauge theory. A major difference from the commutative case is the existence
of gaugeinvariant Wilson line observables based on open contours, in addition to those
on closed contours. The SchwingerDyson and loop equations in noncommutative gauge
theories, therefore, involve both types of gaugeinvariant Wilson observables. In the planar
limit, the equations for a closed Wilson loop simplify and, like in their commutative
counterparts, involve only closed loops. There are, however, new equations, those for
open Wilson lines. These involve closed Wilson loops as well, so both types of Wilson
observables are needed for a closed set of equations in the planar limit. In fact, as we have
seen, these latter equations determine correlators of open Wilson lines entirely in terms of
closed Wilson loops.
Recently in several works it has been argued [39,4246] that local operators in noncommutative gauge theory with straight Wilson lines attached to them are dual to bulk
supergravity modes. In this context it is relevant to ask what bulk observables are dual
to Wilson lines based on arbitrary open contours. This question is also important for a
nonperturbative study of the new loop equations derived here. Our proposal is to identify
a Wilson line based on an arbitrary open contour with the operator dual to bulk closed
string. This proposal is based on the following reasoning.
The momentum variable appearing in a Wilson line satisfies the condition (1.3). This is a
constraint on the contour enforced by gauge invariance. The contour is otherwise arbitrary.
This condition may be regarded as a boundary condition on the curves involved. A generic
curve with this boundary condition may be parametrized as y( ) = ( k) + y ( ), where
0 1 and y ( ) satisfies periodic boundary conditions. Thus the freedom contained
in a generic Wilson line is exactly the one needed to describe a closed string!
Actually, we can take this line of reasoning further. Let us confine our attention to smooth
curves, with the additional condition that the tangents to the curve at the two ends are equal.
In this case we may parametrize the curves as
y( ) = y0 ( ) + y( ),
y0 ( ) = y0 (0) + ( k),
y( ) =
n e2in + n e2in .
(5.1)
n=1
As the above parametrization suggests, what we are going to do is to assume that deviations
of the given curve from a straight line are small and expand the Wilson line, based on the
given curve, around the corresponding straight Wilson line. This gives
120
1
WC [y] = WC0 [y0 ] +
d y ( )
WC [y]
y ( )
1
+
2
1
1
d
y=y0
2 WC [y]
d y ( ) y ( )
y ( )y ( )
+ .
(5.2)
y=y0
1
d
d y ( ) y ( )
C0 (, )
C0 (0, ) il F
(x + y0 ( )) U
Tr U
C0 ( , 1)eik.x ( ) + ( )
(x + y0 ( )) U
il F
1
+
C0 (0, ) il D
C0 (, 1)eik.x
(x + y0 ( )) U
d y ( ) y ( ) Tr U
1
+
C0 (0, )F
C0 (, 1)eik.x ,
(x + y0 ( ))U
d y ( ) i y ( ) Tr U
(5.3)
C0 (1 , 2 ) is the pathordered phase factor, running along the straight line contour
where U
C (0, 1)eik.x ).
C0 , from the point 1 to 2 . Note that in this notation WC [y] = Tr(U
0
Substituting for y( ) from (5.1) in this expression and extracting the part of the n = 1
term symmetric in the indices , , we get precisely the operator that has been identified
in [44] as being dual to the bulk graviton (in the bosonic string), polarized along the brane
directions, modulo factors that connect the open string metric with the closed string metric
and terms involving the scalar fields. Note that the last term in (5.3) is purely antisymmetric
in the indices , and hence contributes only to the operator dual to the bulk antisymmetric
tensor field.
It seems quite likely that the above procedure gives us all the gauge theory operators
dual to bulk string modes. It is, therefore, tempting to identify the Wilson line based
on generic curves of the type described by (5.1) as dual to the bulk closed string. An
expansion of the Wilson line around the corresponding straight line contour would then be
like the expansion of the closed string field in terms of the various string modes carrying
a definite momentum. If this is true, then multipoint correlators of Wilson lines should be
121
identified with closed string scattering amplitudes. In particular, the 2point function would
then have the interpretation of closed string propagator and (4.6) would be the equation of
motion satisfied by the propagator. Such a nonperturbative interpretation of (4.6), or more
generally (4.4), should further enhance our understanding of gauge theory/string theory
duality. The above discussion applies to the bosonic string. It would be interesting to extend
these ideas to the case of the superstring.
Appendix A
(2)
In this appendix we will give some details of the calculation of GC1 C2 [y1 , y2 ]. We will
also describe how the action of the loop laplacian on it reproduces the righthand side of
the loop equation (4.6).
(2)
The diagrams that contribute to GC1 C2 [y1, y2 ] can be collected into four different types
of groups. A representative from each of these has been shown in Fig. 5. There are six
selfenergy type of diagrams, Fig. 5(a). Their total contribution to G(2)
C1 C2 [y1 , y2 ] is
1
(y1 (1 ).y2 (2 ))(y2 (2 ).y2 (2 )) ik1 .(y1 (1 )y2 (2 ))
e
4 2 y2 (2 ) y2 (2 )2
k12
1 2 >2 >
2
+ (1 2),
(A.1)
where a dot on y stands for a derivative with respect to the argument and the last
contribution is obtained by the 1 2 interchange of the subscripts on k, y and .
There are two diagrams in the second set represented by Fig. 5(b). Their total
contribution to G(2)
C1 C2 [y1 , y2 ] is
(y1 (1 ).y2 (2 ))(y1 (1 ).y2 (2 ))
4 ik1 .z
d ze
4 2 z + y1 (1 ) y2 (2 )2 4 2 z + y1 (1 ) y2 (2 ) + k1 2
1 >1 2 >2
(y1 (1 ).y2 (2 ))(y1 (1 ).y2 (2 ))
.
+
4 2 z + y1 (1 ) y2 (2 )2 4 2 z + y1 (1 ) y2 (2 )2
(A.2)
In the third set, represented by Fig. 5(c), also there are two diagrams. Their total
contribution to G(2)
C1 C2 [y1 , y2 ] is
1
eik1 .(z+y2 (2 )y1 (1 ))
2 d 4z
4 2 z2
k1
1 2 >2
ik1 y1 (1 ) y2 (2 )y2 (2 ) 2y2 (2 )y2 (2 )
+ y1 (1 ) y2 (2 )y2 (2 ) + y2 (2 )y2 (2 ) 2y2 (2 ).y2 (2 ) z
122
1
1
4 2 z + y2 (2 ) y2 (2 ) + k1 2 4 2 z + y2 (2 ) y2 (2 )2
+ (1 2).
(A.3)
Finally, we have the gauge boson selfenergy diagrams like Fig. 5(d), including those
with ghosts. Their total contribution to G(2)
C1 C2 [y1 , y2 ] is
1
y1 (1 )y2 (2 )
z + 2i k1 .z 5k1 + 8z z
.
4 2 z + k1 2 4 2 z2
(A.4)
Let us now evaluate the action of the loop laplacian, 2 /y12 ( ), on the expression for
the second order contribution to the 2point function given above. In the first term in (A.1),
(k )
the only dependence on y1 is in the form of VC11 [y1 ], which has been defined in (2.19).
Applying the loop laplacian on it gives the result
2
(k )
V 1 [y1 ] = k12 k1 k1 y1 ( )eik1 .y1 ( ) .
y12 ( ) C1
(A.5)
The k12 in the first term above cancels the factor of 1/k12 in front of the first term in (A.1).
The rest of this factor can be seen to precisely reproduce that contribution of the second
term on the righthand side of the loop equation (4.6) in which a selfenergy insertion
is present on the contour C2s . The three terms correspond to the three possibilities that
the marked point s on the contour C2s is entirely above, entirely below or inbetween the
points where the selfenergy insertion takes place. The second term in (A.5) gives rise to
the following contribution from the first term in (A.1):
4 2 y2 (2 ) y2 (2 )2
2 >2 >2
(A.6)
y2 (2 ).y2 (2 ) eik1 .y2 (2 ) eik1 .y2 (2 )
2 >2
1
1
.
4 2 y2 (2 ) y2 (2 ) + k1 2 4 2 y2 (2 ) y2 (2 )2
(A.7)
123
Now, it is easy to see that on the righthand side of (4.6) there are no terms at this order
having the above structure. Therefore, for consistency of the loop equation, (A.7) must get
cancelled by another term in the 2point function. In fact, this does happen and the required
(k )
term comes from the first term of (A.3). In this term also y1 is in the form of VC11 [y1], so
applying the loop laplacian results in two terms because of (A.5). Let us look at the second
term. It is
eik1 .(z+y2 (2 ))
k1 .y1 ( ) ik1 .y1 ( )
4
e
z
d
ik1.y2 (2 )k1 .y2 (2 )
4 2 z2
k12
2 >2
+ k1 y2 (2 )y2 (2 ) + y2 (2 )y2 (2 ) 2y2 (2 ).y2 (2 ) z
1
1
. (A.8)
4 2 z + y2 (2 ) y2 (2 ) + k1 2 4 2 z + y2 (2 ) y2 (2 )2
In the last term in the square brackets above, let us rewrite 2k1.z as i{(k1 + iz )2 +
z2 k12 } and then use (k1 + iz )2 (eik1 .z /z2 ) = eik1 .z z2 (1/z2 ) and z2 (1/z2 ) = 4 2 (4)(z).
As a result of this simplification, one of the terms we get is precisely (A.7), but with
opposite sign. So this unwanted term cancels in a rather nontrivial way, since the
cancellation involves terms which come from two entirely different diagrams.
Going back to (A.1), let us now look at the second term, which is obtained from the first
by 1 2 interchange:
(y2 (2 ).y1 (1 ))(y1 (1 ).y1 (1 )) ik1 .(y2 (2 )y1 (1 ))
1
e
4 2 y1 (1 ) y1 (1 )2
k12
2 1 >1 >
1
(A.9)
The y1 structure of this term is much more complicated than that of the first term in
(A.1). So the result of applying the loop laplacian on it is also more complicated. For
example, let us consider the first term in the above expression. If the loop laplacian acts
on y1 (1 )eik1 .y1 (1 ) , the result is simple and, in fact, just reproduces that contribution
of the second term on the righthand side of the loop equation (4.6) in which a selfenergy
insertion is present on the contour C1 entirely above the marked point . A similar
operation of the loop laplacian on the other two terms in (A.9) reproduces the other
two contributions in which the selfenergy insertion is either entirely below the marked
point or across it. On the other hand, if the loop laplacian acts on the propagator
1/4 2 y1 (1 ) y1(1 )2 , the result is a deltafunction type of contribution. Together with
a similar contribution from the last term in (A.9) (the middle term has no contribution
of this type since the deltafunction does not click), this precisely reproduces the entire
contribution of the first term on the righthand side of the loop equation in this order.
Let us now go to the next term, (A.2). Here, the loop laplacian may act on any of
the four propagators, resulting in a deltafunction. This gives four different terms and
124
these precisely reproduce that contribution of the second term on the righthand side of
the loop equation (4.6) in which a gauge boson is exchanged between the two contours
C1 and C2s . For this it is essential to remember that the loop on the righthand side,
C2s [y2s ]), involves the hatted operators. As defined in (2.17), these differ
C1 [y1 ]W
Tr(W
from the unhatted ones in that the argument of the gauge field is shifted by the starting
point of the contour. For the contours C1 and C2s , the starting points are, respectively,
y1 (0) = y1 ( ) and y2s (0) = y2 (s). The four terms mentioned above correspond to the
four different possibilities of the two ends of the gauge field propagator landing above or
below the marked points on the two contours.
Thus, we see that the action of the loop laplacian on the second order contribution
to the 2point function reproduces all the terms expected on the righthand side of the
loop equation (4.6). There are also other terms produced in the process of applying
the loop laplacian on the contribution of individual diagrams to the 2point function.
We have seen above an explicit example of one such term which, however, eventually
disappeared because of a nontrivial cancellation with another term. We expect that all other
similar terms (which are produced in the process of applying the loop laplacian on the
contributions of different diagrams to the 2point function, but are not present on the righthand side of the loop equation) will eventually disappear through similar cancellations, but
we have not attempted a complete verification of this.
References
[1] A. Connes, M. Douglas, A. Schwarz, Noncommutative geometry and Matrix theory:
compactification on tori, JHEP 9802 (1998) 002, hepth/9711162.
[2] M.R. Douglas, C. Hull, Dbranes and the noncommutative torus, JHEP 9802 (1998) 008, hepth/9711165.
[3] F. Ardalan, H. Arfaei, M.M. SheikhJabbari, Mixed branes and Matrix theory on noncommutative torus, hepth/9803067;
F. Ardalan, H. Arfaei, M.M. SheikhJabbari, Noncommutative geometry from strings and
branes, JHEP 9902 (1999) 016, hepth/9810072.
[4] Y. Cheung, M. Krogh, Noncommutative geometry from D0 branes in a background Bfield,
Nucl. Phys. B 528 (1999) 185, hepth/9803031.
[5] H. GarciaCompean, On the deformation quantization description of matrix compactifications,
Nucl. Phys. B 541 (1999) 651, hepth/9804188.
[6] C. Chu, P. Ho, Noncommutative open string and Dbrane, Nucl. Phys. B 550 (1999) 151, hepth/9812219.
[7] V. Schomerus, Dbranes and deformation quantization, JHEP 9906 (1999) 030, hepth/
9903205.
[8] N. Seiberg, E. Witten, String theory and noncommutative geometry, JHEP 9909 (1999) 032,
hepth/9908142.
[9] N. Nekrasov, A.S. Schwarz, Instantons on noncommutative R 4 and (2, 0) superconformal sixdimensional theory, Commun. Math. Phys. 198 (1998) 689, hepth/9802068.
[10] J. Maldacena, J. Russo, Large N limit of noncommutative gauge theories, hepth/9908134.
[11] A. Hashimoto, N. Itzhaki, Noncommutative YangMills and the AdS/CFT correspondence,
Phys. Lett. B 465 (1999) 142, hepth/9907166.
125
126
[37] S.J. Rey, R. von Unge, Sduality, noncritical open string and noncommutative gauge theory,
hepth/0007089.
[38] S.R. Das, S.J. Rey, Open Wilson lines in noncommutative gauge theory and tomography of
holographic dual supergravity, hepth/0008042.
[39] D.J. Gross, A. Hashimoto, N. Itzhaki, Observables of noncommutative gauge theories, hepth/
0008075.
[40] A. Dhar, S.R. Wadia, A note on gauge invariant operators in noncommutative gauge theories
and the matrix model, Phys. Lett. B 495 (2000) 413, hepth/0008144.
[41] A. Dhar, Y. Kitazawa, High energy behaviour of Wilson lines, JHEP 0102 (2001) 004, hepth/
0012170.
[42] H. Liu, Trek II, hepth/0011125.
[43] S.R. Das, S. Trivedi, Supergravity couplings to noncommutative branes, open Wilson line and
generalized star products, hepth/0011131.
[44] Y. Okawa, H. Ooguri, How noncommutative gauge theories couple to gravity, hepth/0012218;
Y. Okawa, H. Ooguri, Energy momentum tensors in Matrix theory and in noncommutative
gauge theories, hepth/0103124.
[45] H. Liu, J. Michelson, Supergravity couplings of noncommutative Dbranes, hepth/0101016.
[46] K. Okuyama, Comments on open Wilson lines and generalized star products, hepth/0101177.
[47] M. Fukuma, H. Kawai, Y. Kitazawa, A. Tsuchiya, String field theory from IIB matrix model,
Nucl. Phys. B 510 (1998) 158, hepth/9705128.
[48] A. Dhar, S.R. Wadia, unpublished.
[49] M. AbouZeid, H. Dorn, Dynamics of Wilson observables in noncommutative gauge theory,
hepth/0009231.
[50] A.M. Polyakov, Gauge Fields and Strings, Harwood Academic, 1997.
[51] A.M. Polyakov, V.S. Rychkov, Gauge fieldsstrings duality and the loop equation, hepth/
0002106;
A.M. Polyakov, V.S. Rychkov, Loop dynamics and AdS/CFT correspondence, hepth/0005173.
[52] N. Drukker, D.J. Gross, H. Ooguri, Wilson loops and minimal surfaces, Phys. Rev. D 60 (1999)
125006, hepth/9904191.
Abstract
We use a series of reductions, T dualities and liftings to construct connections between fractional
brane solutions in IIA, IIB and Mtheory. We find a number of phantom branes that are not supported
by the geometry, however materialize upon untwisting and/or Hopfreduction. 2001 Published by
Elsevier Science B.V.
PACS: 11.25.w; 04.65.+e
128
129
related to orientations of cycles. The T dual and the Mtheory lift of the KT solution and its
Sdual are discussed in Sections 3 and 4, respectively. Here we find a number of phantom
branes that are not supported by the geometry, however materialize upon Hopfreduction
and/or untwisting. Section 5 is more speculative and concerns fivebranes in the geometry of
AMV. Finally in Section 6 we explore the U (1) isometry of the deformed conifold and find
a new set of variables in which this isometry is explicit. Some useful formulae concerning
different reductions and T dualities are collected in Appendix A.
(2.1)
e5 = d + A1
(p,q)
A1
= p cos 1 d1 + q cos 2 d2 .
(2.2)
(p,q)
A1
is the connection 1form of the (p, q) U (1) bundle over S 2 S 2 . In terms of the
above base the metric (2.1) becomes
2
2 2
2 2
dsT2 = D e5 + A e1 + e2 + C e3 + e4 + 2B e1 e3 + e2 e4 .
(2.3)
Of particular interest to us is the space T 1,1 which is topologically S 3 S 2 . A basis for the
harmonic representatives of the (onedimensional) spaces H 2 (T 1,1 , Z), H 3 (T 1,1 , Z) was
constructed in [14]:
2 = (sin 1 d1 d1 sin 2 d2 d2 ) = e1 e2 e3 e4 ,
3 = e5 2 .
(2.4)
1 = cos 1 d1 cos 2 d2 .
(2.5)
130
We denote by S 2 , S 3 the basis of homology cycles for the spaces H2 (T 1,1 , Z), H3 (T 1,1 , Z)
so that
(2.6)
3 = 2 = 1.
S3
S2
Since b2 (T 1,1 ) = 1 (b3 (T 1,1 ) = 1), there is only one S 2 (S 3 ) homologically. Consider the
two different 3spheres corresponding to the orbits of each of the two SU(2) factors in
the isometries of T 1,1 . Let us denote them by 3(i) , i = 1, 2. They are parametrized by
(i)
(i , i , ). Similarly, let us denote by 2 the two 2spheres parametrized by (i , i ),
i = 1, 2. The homology basis can be written as [14,15]:
(1)
(2)
S 3 = 3 3 ,
(1)
(2)
S 2 = 2 2 .
(2.7)
Following [14,15] we will interpret a (fractional) brane wrapping (1) as being equivalent
to an antibrane wrapping (2) . Under the transformation 2 2 , 1 becomes the
U (1) connection over a T 1,1 with reversed orientation. We have,
A1 1 ,
2 dA1 ,
(2)
(2)
2,3
2,3
,
(2.8)
= 1.
(2.9)
(1)
(2)
(2.10)
(2)
Taking (2.8) into account, a similar set of relations can be easily derived for the integrals
of dA1 , e5 dA1 over 2,3 .
Finally we can define the T 1,1 space which differs from T 1,1 in the relative sign
between the two factors in the U (1) connection or, equivalently, in the relative orientation
of the two S 2 in the base. An equation similar to (2.8) relates the T 1,1 with its orientationreversal. Note that A1 can be either the U (1) connection of a T 1,1 or of a T 1,1 with
reversed orientation. Similarly, 1 can be either the U (1) connection of a T 1,1 or of
a T 1,1 with reversed orientation.
131
in Appendix A. Reducing along the U (1) to ninedimensional IIB, dualizing to IIA and
lifting back to ten dimensions, we find that the U (1) bundle gets untwisted so that the
transverse space is foliated by 2(1) 2(2) S 1 . We will see how in a further lifting
to eleven dimensions a new twisted direction develops so that the transverse space is
foliated by T 1,1 S 1 .
The KT solution reads
r
B2 = 3gs M ln 2 ,
F3 = M3 ,
e = gs ,
r0
r
F5 = F5 + F5 ,
(3.1)
2 3 ,
F5 = N + 3gs M 2 ln
r0
where 2 , 3 where defined in the previous section. The metric is
2
= h1/2 dx dx + h1/2 dr 2 + r 2 dsT2 ,
ds10
where
h(r) = b0 +
c gs
r
3
2
2
g
M
+
3g
M
ln
N
+
.
s
s
r4
4
r0
(3.2)
(3.3)
The metric dsT2 is the T 1,1 metric defined in the previous section for the special case D =
1/9, A = 1/6, B = 0. With this choice of constants T 1,1 becomes Einstein. The constant
b0 can be fixed by demanding that the asymptotically flat limit of the metric be the standard
Minkowski. Similarly the constant c can be fixed by requiring that for M = 0 the metric
reduce to the standard AdS5 S 5 in the nearhorizon limit. We also have, 3
r
F3 = M,
F5 = N + 3gs M 2 ln .
(3.4)
r0
S3
T 1,1
The identification of fractional D3branes with D5branes wrapping the S 2 of the conifold,
was justified in [16].
3.1. The IIA versions of KT solution
As already mentioned, reducing the KT solution along the U (1) fibre to ninedimensional IIB, dualizing to IIA and lifting to tendimensional IIA, we find that the
U (1) bundle gets untwisted so that the transverse space is foliated by 2(1) 2(2)
S 1 . The interpretation of the T dualized (IIA) version of KT, depends on our choice of
orientation for (2) . Geometrically speaking, there seems to be no reason for preferring
one orientation over the other, but the brane content is entirely different in each case.
It must be noted that while reversing the orientation of a solution is also a solution to
the supergravity equations, the amount of preserved supersymmetry may change. In fact,
if an Einstein space that is not a round sphere has Killing spinors, its orientation reversal
gives a space with no Killing spinors [17]. So trying to preserve the original amount of
supersymmetry will choose an orientation for us, however we will consider both cases
3 We are dropping numerical constants of proportionality.
132
here. Concerning the effect of T duality/Hopfreduction we note that T duality at the full
string theory level always preserves as much supersymmetry as has survived the U (1)
compactification. The U (1) compactification in the case where the compactifying direction
is naturally periodic (as is here) also leaves supersymmetry unbroken at the full string
theory level. At the level of supergravity however, it is known that these operations can
brake the supersymmetries of pbrane solutions (see [5] for a detailed discussion).
3.1.1. Tduality
Let us start with the orientation for 2(2) which is inherited from (2.7). Reducing to nine
dimensions, dualizing and lifting back to tendimensional type IIA, we obtain the following
solution
dr
2 + dA1 dz,
F3 = 3gs M
F2 = M2 ,
r
r
2 2 ,
F4 = N + 3gs M 2 ln
r0
(3.5)
1/4
2
1
1
2
1/8
(i)
2
d2
= gs
h1/2 dx dx + h1/2 dr 2 + r 2
ds10
h1/2 r 2
9
6
i=1
3/4
7/8 1 1/2 2
h r
+ gs
(3.6)
dz2 ,
9
where (d2(i) )2 := sin2 i di2 + di2 . We see that the dualization has untwisted the U (1)
bundle so that the transverse space is foliated by S 2 S 2 S 1 .
There is a charge Q(6) corresponding to D6branes wrapping an S 2 S 1 . This gives
fractional D3branes! Indeed, let us denote by 1,2 the two S 2 factors. We have
Q(6) =
(3.7)
F2
F2 = M,
(1)
(2)
2 S 1
(2)
2 S 1
(3.9)
(1)
(2)
2 2
The brane content is in agreement with the (naive) expectations from T duality: the D5
(fractional D3) become D6, the D3 become D4 wrapping the T duality circle.
133
(2)
2 S 1
(2)
2 S 1
(3.12)
(1)
(2)
2 2
The above brane content should be compared with the system of [15,1820] where there
are two NS5branes, one of them stretching along the directions x 05 , and the other along
x 03,7,8 . The x 6 direction is a circle. There are also two types of D4branes along x 03,6 .
N of them going around the circle and M of them stretching between the two NS5branes.
(1)
In order to compare to the situation at hand, we should identify x 4,5 with 2 , say, and
(2)
x 7,8 with 2 . The circle x 6 should be identified with the S 1 . The constant factor N on
the righthand side of (3.12) is coming from the D4branes going around the circle, while
the rdependent piece should be attributed to the D4branes stretching between the NS5branes. Having these two types of fourbranes will be important when the solution is lifted
to eleven dimensions.
3.2. The Mtheory versions of KT
Lifting to elevendimensional supergravity will produce two distinct solutions, corresponding to the two versions of KT in type IIA. In both cases a new twisted U (1) develops.
Let us start with the second case discussed in the last section (solution (3.10)(3.12))
and lift to eleven dimensions. The solution reads
2
1 1/2 2 1/3 1/2
1 2 (i) 2
2
1/2
2
ds11 =
h r
d2
dx dx + h
h
dr + r
9
6
i=1
1 1/2 2 2/3 1 2
h r
+
(3.13)
gs dz + gs (d M1 )2 .
9
Note that for M = 0 the bundle is untwisted and the transverse geometry is foliated by
S 2 S 2 S 1 S 1 . However in the presence of fractional branes M = 0 and a twist is
134
induced. The transverse space becomes foliated by T 1,1 S 1 . More accurately, this is a
T 1,1 space whose S 3 fibre (when T 1,1 is viewed as S 3 over S 2 ) is replaced by the lens space
S 3 /ZM . Indeed, under the redefinition M, d M1 becomes proportional to
the covariant displacement on the U (1)/ZM fibre of a T 1,1 with reversed orientation. This
is reminiscent of the situation in [5] (see also [21]).
We also have
r
2
2 2
F4 = N + 3gs M ln
r0
dr
2 + dA1 dz (d M1 ).
3gs M
(3.14)
r
Note that dF4 = 0 as it should. To prove that, use the fact that 2 dA1 = 0.
The solution contains M5brane charge wrapping the S 2 of T 1,1 ,
(5,1)
Q
=
F4 .
(3.15)
S 3 S 1
(2)
2 2
(1)
(2)
(5,2) charge can be taken
Since there is no nontrivial 2 2 cycle in T 1,1 such a Q
to zero, even though there is nonvanishing F4 flux. However upon reduction to IIA, these
phantom M5branes come to existence in the form of the D4branes wrapping the S 1 , as
in (3.12)! Also, as noted below (3.13), in the absence of fractional branes the U (1) bundle
is untwisted and again the phantom M5branes become ordinary M5branes wrapping
S 1 S 1 . In the following, all the phantom charges will carry tildes.
This is the first instance in which we encounter a phenomenon whereby there are
fluxes unsupported by the geometry, which however materialize into physical branes upon
untwisting and/or Hopfreduction. The U (1) (un)twisting is the key to this, and in a way
provides a discrete analogue of the large N duality between the fluxes and branes. We will
meet more examples of this in the sequel.
Let us compare to the situation in [12] where a string with nonzero winding number
in the background of a KaluzaKlein monopole was considered. The string can unwind
since the total space has 1 = 0, but the charge associated to the winding number of the
string is conserved. From this one concludes that there is a zero mode among the collective
coordinates of the KK monopole which couples to the charge associated with the winding
of the string. In our case 1 = ZM is nonzero, but there are still no nontrivial onecycles
in homology (b1 = 0).
The lifting of the solution (3.7)(3.9) is similar. The resulting geometry is now
T 1,1 S 1 (for M = 0) and the charge Q(5,1) = 0. As before the S 3 fibre of the T 1,1
135
is replaced by the lens space S 3 /ZM . Again, there is a flux through the contractible
(1)
(2)
2 2 . Upon reduction this becomes the D4brane flux.
4. The Sdual KT
We saw in the previous section that upon T dualizing the KT solution we get
(1)
(2)
a configuration with untwisted 2 2 S 1 geometry for the level surfaces. It
would be interesting to have a situation where by T dualizing a IIB solution we get level
surfaces with twisted geometry. The reason why this does not work with the KT solution
(3.1) is clear. From the T duality transformations (A.22) we see that in order to get a
twisted U (1) fibration upon going from IIB to IIA, the original IIB solution would have to
(3)
have ANS
1 = 0. Upon T dualizing this becomes the connection A1 of the U (1) fibration in
NS
IIA. Since H3 = dANS
2 dA1 dz, this means that the original IIB solution would have
to have H3 with nonzero flux through the Hopf fibre. We can indeed obtain a solution of
IIB with the aforementioned property by performing an Sduality transformation on the
KT solution:
a + b
H3
d c
H3
,
(4.1)
b a
F3
F3
c + d
with the rest of the fields inert and
a b
SL(2, Z).
:= + ie ,
c d
(4.2)
1/4
2
1 2 (i) 2
1/8 1 1/2 2
2
1/2
1/2
2
ds10 = gs
h
dr + r
h r
d2
dx dx + h
9
6
i=1
3/4
7/8 1 1/2 2
h r
+ gs
(4.4)
(dz + cM1 )2 .
9
e = gs d 2 +
Note that dz + cM1 is proportional to the covariant displacement on the U (1)/ZcM fibre
either of a T 1,1 with reversed orientation or of a T 1,1 . We have
136
F2 = (a c)M2 ,
dr
2 + dA1 (dz cM1 ),
F3 = 3dgs M
r
r
dr
F4 = N + 3gs M 2 ln
2 (dz cM1 ).
2 2 + 3(b d)Mgs
r0
r
(4.5)
The brane content is different in each case.
First we discuss the T 1,1 with reversed orientation. The solution has zero units of
D6branes wrapping S 3
(6)
Q = F2 = 0.
(4.6)
S2
Again, as in (3.16), note that this makes sense either upon Hopfreducing or in the absence
of fractional branes in which case the U (1) fibre gets untwisted.
Turning to the T 1,1 case we find Q(6) = (a c)M and Q(5) = 0. What was said
before about (4.8) is true here as well.
4.1. The Mtheory lift
Upon lifting to Mtheory, yet another twisted U (1) fibre develops. The resulting
transverse geometry is foliated by a nontrivial S 1 /ZaM bundle over T 1,1 ZcM such that
reducing the fibre along any one of the two circles produces a T 1,1 space (or a T (1,1) , by
a reasoning that should be familiar by now). In this sense there is a similarity with the AMV
solution in the next section. 4 The transverse geometry in that case is an S 3 over S 3 bundle.
Viewing the base as a Hopf fibration over S 2 and reducing along the U (1) fibre produces an
S 3 (up to moding out by a discrete group) over S 2 bundle, which is actually a T 1,1 space.
Analytically, the elevendimensional metric reads
2
1 1/2 2 1/3 1/2
1 2 (i) 2
2
1/2
2
d2
dx dx + h
h
dr + r
h r
ds11 =
9
6
i=1
4 In spite of the similarity it is easy to see that the resulting geometry cannot be that of AMV simply because
the original tendimensional one is not that of the resolved conifold.
1 1/2 2
+
h r
9
2/3
1
2
gs (d aM1 ) (dz cM1 )
+ gs (dz + cM1 )2 ,
137
(4.9)
(4.10)
(2)
2 2
Upon reduction to IIA these become the D4branes wrapping the Hopf fibre of T 1,1 , as
in (4.8). Depending on the orientation there can be phantom M5brane charge wrapping S 2
(5,2) =
Q
F4 ,
(4.12)
S 1 S 3
138
upon lifting becomes the base of a twisted U (1) fibration and thus is no longer able to
support an M5. This is in agreement with the fact that G2 holonomy manifolds do not have
calibrated twodimensional submanifolds.
We should therefore look for an elevendimensional G4 that has a flux through the
(elevendimensional) Hopf fibre. Reducing this will give an F3 flux, which upon integration
over the S 3 will give NS5 charge. Since there are two S 3 s in the original geometry, an
obvious possibility is to consider a flux
(5.1)
G4 ,
S 3 S1
which upon reduction will give an NS5 wrapping S 2 . Although the tendimensional
geometry is foliated by T 1,1 , it cannot be that of the resolved conifold, since the flux
of the NS threeform will stabilize the S 3 part as well. At the moment we are not able to
write down explicitly such a solution containing simultaneously D6branes wrapping S 3
and NS5branes wrapping S 2 and having four common noncompact directions.
Mtheory solutions with a flux as in (5.1) have recently appeared in the literature [26,
28]. It would clearly be interesting to obtain their IIA reductions and their IIB T duals.
The hope is to provide an explicit realization of the mirror symmetry between D5branes
on S 2 of the deformed conifold in type IIB and D6 on S 3 of the resolved in type IIA, as a
single T duality along the Hopf fibre of the T 1,1 . 5 This task is complicated, however, by
the form of the deformed conifold metric, as will be discussed in the next section.
6. The KS solution
Although we have already excluded the possibility that the KS solution [8] is the T dual
of the AMV solution since T dualizing the latter gives untwisted transverse geometry
whereas the former has a twisted one we can still try to perform the analysis of the
previous sections to this case. Technically this is complicated by the fact that the metric
of the deformed conifold is written in terms of variables in which the U (1) isometry is
implicit [13]. Here we make partial progress by identifying a set of new variables in which
the U (1) isometry of the metric will be manifest, but we are yet unable to write down
explicitly the metric in terms of these new variables.
The KS solution reads
2
= h1/2 ( ) dx dx + h1/2 ( ) ds62 ,
ds10
F3 = M e5 g 3 g 4 + d F ( ) g 1 g 3 + g 2 g 4 ,
B2 = gs M f ( )g 1 g 2 + k( )g 3 g 4 ,
F5 = gs M 2 l( )g 1 g 2 g 3 g 4 e5 ,
(6.1)
5 Note that as we have already remarked, the existence of NS5branes is required in addition to the Dbranes
for this duality to work.
where
ds62
3 2 4 2
1 4/3
1 2 5 2
2
d + e
g
+ cosh
= / K( )
+ g
2
3K( )
2
1 2 2 2
g
+ g
+ sinh2
2
139
(6.2)
and is a radial coordinate. The explicit form of the functions F, f, k, l, K will not be
important in the following. The oneforms g i are defined as
e1 e3
e2 e4
e1 + e3
g2 = ,
g3 = ,
,
2
2
2
2 + e4
e
g4 = ,
(6.3)
g 5 = e5 ,
2
so that the level surfaces of (6.2) are of the general form (2.3).
As we noted in Section 4, for this solution to T dualize to a twisted transverse geometry,
B2 would have to have a nonzero component along the direction corresponding to the U (1)
isometry.
Trying to Hopfreduce runs into trouble because the solution is written in terms of
coordinates in which the U (1) isometry is not simply a shift in . In fact from (2.2),
(2.3) we see that the isometry reads
cos c sin c
sin 2 d2
sin 2 d2
.
+ c,
(6.4)
sin c cos c
d2
d2
g1 =
We would like to find a coordinate transformation for 2 , 2 which has the effect of the
second line in (6.4). Clearly this would have to be a special case of the SU(2) group of
isometries of the sphere
az + b
,
+ a
bz
2
, a2 + b2 = 1.
(6.5)
2
Indeed for a = cos //2, b = sin //2, (6.5) implies (6.4) to order O(/), provided we identify
z
z := ei2 tan
sin 2
.
sin 2
Explicitly the coordinate transformation reads, to O(/) order,
c=/
(6.6)
sin 2
,
sin 2
/ sin 2 cot 2 ,
+/
2 2 + / cos 2 .
(6.7)
e5
sin 2
sin 2 cot 2
+ cos 2
.
sin 2
2
2
(6.8)
It remains to find a set of new variables , 2 , 2 in which k locally takes the form
k = / . Rewriting (6.2) in terms of these new variables would result in a metric
140
2 2
2 2 2
2 2
2 2
One can easily verify that
k2
0
k 2 = 0 .
k
1
sin 2
1
,
L (f1 + f2 ) =
2
sin 2
where
,
L := k k
1
L (f1 f2 ) = 1,
2
(6.9)
Lh(sin 2 sin 2 ) = 0,
sin 2 sin 2
f1,2 (2 , 2 ) := arctan
,
cos 2 cos 2
(6.10)
(6.11)
(6.12)
(6.13)
Finally one can derive the relation of the differentials of the old coordinates in terms of the
new,
cos 2
cot 2 sin 2 cot 2 sin 2
2
2
2
h (cos 2 + cos 2 sin 2 ) sin 2
d
d2
cos 2 sin 2
d
d2 =
cos 2
cos 2
2
2
sin 2
sin 2
cos 2 cot 2
1
sin 2
sin 2
h (cos2 2 + cos2 2 sin2 2 )
(6.14)
Using the above relation, one can examine whether or not B2 in (6.1) has a nontrivial
component along the U (1) fibre. The answer is that it does and therefore, as already
explained, the T dual version of KS has twisted transverse geometry.
6 This is not the most general form. One can add arbitrary functions of sin sin to the rhs. Also one could
2
2
take arbitrary linear combinations of 2 , 2 in the second and third lines.
141
Note added
After this paper was posted on the hepth archive, we were informed by I. Klebanov,
A. Tseytlin and L. Pando Zayas of their related independent work which overlaps with
Section 3 of this paper. See the talk by I. Klebanov at the Avatars of Mtheory conference
(UCSB) http://online.itp.ucsb.edu/online/mtheory_c01/klebanov/.
Acknowledgements
It is a pleasure to thank C. Bachas, M. Cvetic, M. Douglas and G. Ferretti for discussions.
The work of R.M. is supported in part by EU contract HPRNCT200000122 and by
INTAS contract 551590; D.T. is supported in part by EU contract HPRNCT200000122
and by the SNSRC. D.T. would like to thank Ecole Normale Suprieure and CERN for
hospitality.
(A.1)
(A.2)
Note that all eleven (ten)dimensional field strengths and potentials are denoted with
(without) a hat. Assuming the geometry has a U (1) isometry, we can put the metric in
the form:
1
2
2
= e 6 ds10
+ e 3 (A1 + dz)2 ,
ds11
(A.3)
(A.4)
(A.5)
142
where
F4 = dA3 + dA2 A1 ,
F3 = dA2 ,
F2 = dA1 ,
(A.6)
so that
4 = F4 + F3 (dz + A1 ).
F
(A.7)
3 = d A
2 ,
F
2 = d A
1 .
F
(A.8)
(A.9)
We assume that there is a U (1) isometry so that the tendimensional metric can be cast in
the form
1
2
7 (3)
2
= e 2 7 ds92 + e 2 A1 + dz ,
ds10
(A.10)
2
where , A(3)
1 and the ninedimensional metric ds9 do not depend on the U (1) coordinate z.
We reduce the potentials as follows
1 = A(1) + A0 dz,
A
1
7
1
1
1 3
1 3 (2) 2
7
e1 L = R ()2 ()2 e 2 2 (F1 )2 e
F2
2
2
2
4
1 3 + 1 (1)2 1 4 (3) 2
1 + 1 (1) 2
7
F3
e 2 2 7 F2
e 7 F2
e
4
4
12
1 1 5 (2) 2
1 1 + 3
e 2 2 7 F3
e 2 2 7 (F4 )2 + C.S.,
12
48
where
(A.12)
= ,
F1 = dA0,
(1)
(1)
(3)
F2 = dA1 + dA0 A1 ,
F2(2) = dA(2)
1 ,
F2(3) = dA(3)
1 ,
(2)
(3)
F3(1) = dA(1)
2 + dA1 A1 ,
(1)
(2)
(1)
F3(2) = dA(2)
2 + A1 dA1 + A0 dA2 ,
(1)
(2)
(3)
(2)
(1)
(3)
(1)
(3)
F4 = dA3 + dA(1)
2 A1 + dA2 A1 dA1 A1 A1 A0 dA2 A1 ,
(A.13)
143
so that
4 = F4 + F (2) dz + A(3) ,
F
3
1
3 = F (1) + F (2) dz + A(3) ,
F
3
2
1
2 = F (1) + F1 dz + A(3) .
F
2
1
(A.14)
(A.15)
where
1 NS R 1 R NS
R
5 = d A
F
4 d A2 A2 + d A2 A2 = F5 ,
2
2
R
NS
NS .
F3 = d A2 d A2 ,
H3 = d A
2
(A.16)
We assume that there is a U (1) isometry so that the tendimensional metric can be cast in
the form
2
=e
ds10
2 7
ds92 + e
7
2
(A1 + dz)2 ,
(A.17)
where , A1 and the ninedimensional metric ds92 do not depend on the U (1) coordinate z.
We reduce the potentials as follows
R
R
R
A
2 = A2 + A1 dz,
R
R
R
A
4 = A4 + A3 dz.
= ,
= ,
NS
F2R = dAR
F2NS = dANS
F2
1 + dA1 ,
1 ,
NS
NS
NS
F3 = dA2 + A1 dA1 ,
NS
R
NS
F3R = dAR
2 + dA1 A1 dA2 + dA1 A1 ,
= dA1 ,
144
F4R = dAR
3 +
1
R
NS
R
R
NS
R
NS
dANS
2 A1 A2 dA1 + dA2 A1 + A2 dA1 ,
2
(A.20)
so that
n = Fn + Fn1 (dz + A1 ).
F
(A.21)
A.4. 9D Tduality
The two ninedimensional Lagrangians described above, are in fact related to each other
by the following local field transformations:
3
7
B ,
A B
4
4
3
7
A
B B ,
4
4
R
NS
A(1)
1 A1 + A1 ,
A(2)
1 A1 ,
(1)
A0 ,
NS
A(3)
1 A1 ,
(2)
NS
A2 ANS
2 A1 A1 ,
R
A2 AR
2 + A1 A1 ,
1 NS
1 R
R
NS
NS
R
A3 AR
3 + A1 A2 A1 A2 A1 A1 A1 .
2
2
(A.22)
F2 F2 ,
(2)
(1)
F3 F3R ,
F2 F2R ,
(3)
F2 F2NS ,
(2)
F3 F3NS ,
F4 F4R .
(A.23)
str
.
g = e 2 g
(A.24)
str
ds10
= e2
( 1 )
7
ds92 + e 2 (+
7)
(dz + A1 )2 .
(A.25)
F3
F5 + C.S.
12
4 5!
(A.26)
145
References
[1] B.E.W. Nilsson, C.N. Pope, Hopf fibration of elevendimensional supergravity, Class. Quant.
Grav. 1 (1984) 499.
[2] M.J. Duff, P.S. Howe, T. Inami, K.S. Stelle, Superstrings in D = 10 from supermembranes in
D = 11, Phys. Lett. B 191 (1987) 70.
[3] M.J. Duff, H. Lu, C.N. Pope, Supersymmetry without supersymmetry, Phys. Lett. B 409 (1997)
136, hepth/9704186.
[4] M.J. Duff, H. Lu, C.N. Pope, AdS(5) S(5) untwisted, Nucl. Phys. B 532 (1998) 181, hepth/9803061.
[5] M.J. Duff, H. Lu, C.N. Pope, AdS(3) S 3 (un)twisted and squashed, and an O(2, 2, Z)
multiplet of dyonic strings, Nucl. Phys. B 544 (1999) 145, hepth/9807173.
[6] I.R. Klebanov, A.A. Tseytlin, Gravity duals of supersymmetric SU(N) SU(N + M) gauge
theories, Nucl. Phys. B 578 (2000) 123, hepth/0002159.
[7] P. Candelas, X.C. de la Ossa, Comments on conifolds, Nucl. Phys. B 342 (1990) 246.
[8] I.R. Klebanov, M.J. Strassler, Supergravity and a confining gauge theory: duality cascades and
chiSBresolution of naked singularities, JHEP 0008 (2000) 052, hepth/0007191.
[9] B.S. Acharya, On realising N = 1 superYangMills in Mtheory, hepth/0011089.
[10] M. Atiyah, J. Maldacena, C. Vafa, An Mtheory flop as a large n duality, hepth/0011256.
[11] C. Vafa, Superstrings and topological strings at large N , hepth/0008142.
[12] R. Gregory, J.A. Harvey, G. Moore, Unwinding strings and T duality of KaluzaKlein and
H monopoles, Adv. Theor. Math. Phys. 1 (1997) 283, hepth/9708086.
[13] R. Minasian, D. Tsimpis, On the geometry of nontrivially embedded branes, Nucl. Phys. B 572
(2000) 499, hepth/9911042.
[14] S.S. Gubser, I.R. Klebanov, Baryons and domain walls in an N = 1 superconformal gauge
theory, Phys. Rev. D 58 (1998) 125025, hepth/9808075.
[15] K. Dasgupta, S. Mukhi, Brane constructions, fractional branes and antide Sitter domain walls,
JHEP 9907 (1999) 008, hepth/9904131.
[16] I.R. Klebanov, N.A. Nekrasov, Gravity duals of fractional branes and logarithmic RG flow,
Nucl. Phys. B 574 (2000) 263, hepth/9911096.
[17] M.J. Duff, B.E.W. Nilsson, C.N. Pope, KaluzaKlein supergravity, Phys. Rep. 130 (1986) 1.
[18] E. Witten, Solutions of fourdimensional field theories via Mtheory, Nucl. Phys. B 500 (1997)
3, hepth/9703166.
[19] E. Witten, Branes and the dynamics of QCD, Nucl. Phys. B 507 (1997) 658, hepth/9706109.
[20] K. Dasgupta, S. Mukhi, Brane constructions, conifolds and Mtheory, Nucl. Phys. B 551 (1999)
204, hepth/9811139.
[21] M. Cvetic, H. Lu, C.N. Pope, Decoupling limit, lens spaces and TaubNUT: D = 4 black hole
microscopics from D = 5 black holes, Nucl. Phys. B 549 (1999) 194, hepth/9811107.
[22] L.A. Pando Zayas, A.A. Tseytlin, 3branes on resolved conifold, JHEP 0011 (2000) 028, hepth/0010088.
[23] J.M. Maldacena, C. Nunez, Towards the large N limit of pure N = 1 superYangMills, Phys.
Rev. Lett. 86 (2001) 588, hepth/0008001.
[24] J.D. Edelstein, C. Nunez, D6branes and Mtheory geometrical transitions from gauged
supergravity, JHEP 0104 (2001) 028, hepth/0103167.
[25] K. Dasgupta, K. Oh, R. Tatar, Geometric transition, large N dualities and MQCD dynamics,
hepth/0105066.
[26] M. Cvetic, H. Lu, C.N. Pope, Massless 3branes in Mtheory, hepth/0105096.
146
Abstract
We present a fivedimensional supersymmetric SU(5) theory in which the gauge symmetry is
broken maximally (i.e., at the 5D Planck scale M ) on the same 4D brane where chiral matter is
localized. Masses of the lightest KaluzaKlein modes for the colored Higgs and X and Y gauge fields
are determined by the compactification scale of the fifth dimension, MC 1015 GeV, rather than
by M . These fields wave functions are repelled from the GUTbreaking brane, so that proton decay
rates are suppressed below experimental limits. Above the compactification scale, the differences
between the standard model gauge couplings evolve logarithmically, so that ordinary logarithmic
gauge coupling unification is preserved. The maximal breaking of the grand unified group can also
lead to other effects, such as O(1) deviations from SU(5) predictions of Yukawa couplings, even in
models utilizing the FroggattNielsen mechanism. 2001 Published by Elsevier Science B.V.
PACS: 12.10.D; 04.50
1. Introduction
The standard model of particle physics, described by the gauge group SU(3) SU(2)
U(1) is one of the most successful physical theories ever. Nonetheless, the standard
model is theoretically unsatisfying for a number of reasons. For instance, the instability of
the weak scale against radiative corrections has motivated the study of numerous theories,
including technicolor, supersymmetry and theories with additional dimensions.
While the gauge hierarchy may be the most compelling motivation for new physics, there
are additional reasons to consider theories with further structure. It was long ago realized
that the gauge group of the standard model could be embedded into a simple group [1].
This work was supported in part by the US Department of Energy under Contracts DEAC0376SF00098, in
part by the National Science Foundation under grant PHY9514797.
1 Research fellow, Miller Institute for Basic Research in Science.
05503213/01/$ see front matter 2001 Published by Elsevier Science B.V.
PII: S 0 5 5 0  3 2 1 3 ( 0 1 ) 0 0 3 8 8  1
148
149
law running, the majority of it is SU(5) universal. Nonetheless, we can extract a nonuniversal piece, which is proportional to the MSSM running, and logarithmic, but at a rate
slowed compared to the MSSM, such that ordinary, logarithmic unification can occur at
the higherdimensional Planck scale. In Section 4 we will see how proton decay operators
are naturally suppressed by suppressing the KK wave functions on the brane. In Section 5
we will discuss deviations from SU(5) Yukawa relationships can arise from the maximal
breaking of the GUT on the brane. In Section 6 we discuss extensions to larger gauge
groups. Finally, we conclude in Section 7.
2. Framework
We consider a 5D supersymmetric SU(5) theory with the fifth dimension compactified
on an S1 /Z2 orbifold (in Section 6 we will consider the possibility of larger gauge groups).
The bulk exhibits N = 1 supersymmetry in five dimensions, which translates into an N = 2
supersymmetry in four dimensions. The fields that propagate in the bulk are contained
in a gauge supermultiplet and two Higgs hypermultiplets transforming as 5 and 5 under
SU(5). Under the 4D N = 1 supersymmetry preserved after the orbifold compactification,
the gauge supermultiplet decomposes into a vector superfield V and an associated chiral
adjoint . The two hypermultiplets yield four chiral multiplets H5 , H5c , H5 and H c . (H5
5
and H c transform as 5 under the SU(5), while H5c and H5 as 5.) Under the orbifold Z2 ,
5
which flips the sign of the fifth coordinate (y y), the various superfields transform as
H5,5 H5,5 ,
V V,
c
c
H5,5
H5,5
,
(1)
which leaves zero modes only for the MSSM fields and their GUT counterparts.
The S1 /Z2 orbifold has fixed points at y = 0 and y = R. We take chiral matter to be
localized to the y = R fixed point, along with an adjoint chiral superfield . The SU(5)
symmetry is broken by the vacuum expectation value (vev) of , which we assume to have
the form
2/5 0
0
0
0
0 2/5 0
0
0
ij =
(2)
0
0
2/5
0
0
.
0
0
0 3/5
0 2
0
3/5
We imagine that this breaking occurs at or near the fivedimensional Planck scale M ,
which we assume to be much larger than the conventional GUT scale of 2 1016 GeV. We
will refer to the y = R fixed point as the GUTbreaking brane.
Bulk fields even under the orbifold Z2 can couple directly to the fields on the GUTbreaking brane. For instance, the Higgs fields can couple to the chiral matter fields T10
and F5 as (u T10 T10 H5 + d T10 F5 H5 )(y R) in the superpotential. We also take there
to be a bare Higgs mass term, H5 H5 (y R), as well as Higgs coupling to the GUT
150
L (y R)g52 2 A,a Aa ,
(3)
where a indexes the broken generators of the group and g52 is the 5D gauge coupling, with
mass dimension 1. If we were to decompose each X and Y bulk vector field
in a naive KK
basis, we would estimate that the zero modes have mass MV = g5 / 2R = g4 .
Such an estimate incorrectly assumes that the gauge boson wave functions do not change
appreciably in the presence of the large localized mass term. To find the correct spectrum
we must solve the differential equation
y2 A + (y R)g52 2 A = m2 A ,
(4)
(5)
151
Fig. 1. Mass spectrum for the lowest KK modes of the gauge fields (a), and Higgs fields (b) in our
model. As explained in the text, the transition to degeneracy between the X, Y and 321 towers
occurs much more gradually than shown in (a). In (b), the limit of very large is taken, so that the
slight nondegeneracy between the colored hypermultiplet pairs at each level is not resolved. In both
(a) and (b), the triplet of numbers below each tower corresponds to the beta function contribution
(b1 , b2 , b3 ) that comes from each level in that tower (for the colored Higgs, the contributions from
the nearlydegenerate hypermultiplet pairs are combined).
spaced according to (n + 1/2)MC . The value of the wave function at the GUTbreaking
brane goes as cos(MnG R) = (2n + 1)MC /(g52 2 ), and there are corrections to the mass
which go like (2n + 1)MC2 /(g52 2 ). Modes with masses near or above the scale
g52 2 become essentially degenerate with the ordinary KK tower, mn = nMC .
We have not broken supersymmetry, and thus the gauginos corresponding to the broken
SU(5) generators have an identical spectrum. The KK towers for the gauge fields are
illustrated in Fig. 1(a). For each unbroken SU(3) SU(2) U(1) (321) generator there
is a massless vector multiplet and massive vector multiplets at MC , 2MC , 3MC , and
so on. For each broken (X, Y ) generator there is a tower of massive vector multiplets
whose low levels are offset from those of the 321 tower by MC /2, but whose higher
152
levels relax into alignment with the 321 tower. Although in Fig. 1(a) this transition to
degeneracy has already been achieved at 5MC , for the parameters of interest in our model,
the relaxation will actually occur much more gradually, over the span of g52 2 /MC
O(10100) KK levels. Note that for each X, Y massive vector multiplet but one there
is a corresponding 321 massive vector multiplet. The remaining one is paired with the
massless 321 vector multiplet, indicating a larger total number of states in the X, Y
tower relative to the 321 tower. These excess states are the eaten Goldstone degrees of
freedom from the multiplet. This means that above the scale g52 2 the KK towers of
gauge fields contributing to the renormalization group (RG) running are completely SU(5)
symmetric except that one chiral adjoint for each 321 vector multiplet is missing. These
missing degrees of freedom are provided by the physical fields which remain after the
Goldstone components of are eaten by the X, Y gauge multiplets. Therefore, above the
scale g52 2 and m (the mass of physical fields), the spectrum contributing to the
running is completely SU(5) symmetric.
2.2. Spectrum: Higgs fields
The spectrum analysis for the Higgs triplet fields HC and HC
is somewhat different,
because the brane coupling is dimensionless, and hence the naive mass that the zero
modes would pick up by not avoiding the brane is just MC which is comparable to MC
for order one . For our purposes, we will take to be a large parameter, roughly O(10),
so that the Higgs wave functions will in fact be strongly repelled from the brane. This size
for is quite consistent with the general features of the theory, which must be somewhat
strongly coupled in order to achieve order one top Yukawa and gauge couplings in spite of
the large radius.
The equations that determine the colored Higgsino spectrum are
c (y R)H
= 0,
i HC y H
C
C
c
i H
y H (y R)HC = 0,
C
C
c
C = 0,
i HC + y H
c
= 0.
i HC
+ y H
C
(6)
(7)
(8)
(9)
(10)
HC (c) =
(c)
(c)
C,n
(x)gC,n
(y),
(11)
c
c
where gC,n and gC,n
are even under y y, while gC,n and gC,n
are odd. If we integrate
the first two of these equations in a region (R /, R + /) where / 0, we obtain two
constraints,
c
(x) =
C,n
gC,n
(R)
c
2gC,n
(R /)
C,n
(x),
(12)
gC,n (R)
c (R /) C,n (x).
2gC,n
153
(13)
gI,n (R)
c (R /) ,
2gJ,n
(14)
(15)
c
MnH gC,n
y gC,n
(y) (C,C),n
(y) = 0,
(16)
c
MnH (C,C),n
gC,n (y) + y gC,n (y) = 0,
(17)
c
MnH (C,C),n
gC,n
(y) = 0,
(y) + y gC,n
(18)
H
c
gC,
(y) = Nn sin Mn y ,
C,n
(19)
1
arctan(/2)
MnH =
n+
,
R
(20)
(21)
where n runs from negative infinity to positive infinity (the physical masses are given by
the absolute value of MnH ). For large , arctan(/2) /2, so MnH (n + 1/2)MC , and
H
). Note that, in contrast to the
the masses fall into nearly degenerate pairs, (MnH , Mn1
case of the gauge fields, even for high values of n, the triplet Higgsinos do not become
degenerate with the doublet Higgsinos.
Supersymmetry fixes the masses and couplings of the Higgs scalars to be the same
as for the Higgsinos. The KK towers for the Higgs fields are shown in Fig. 1(b). The
doublet spectrum consists of a pair of massless N = 1 chiral multiplets, and pairs of exactly
degenerate hypermultiplets at MC , 2MC , 3MC , and so on. For large , the triplet spectrum
consists of nearly degenerate pairs of hypermultiplets at MC /2, 3MC /2, 5MC /2, and so
on. In Fig. 1(b), the nondegeneracy between these pairs is not resolved.
3. Differential running
A crucial question in our model is how the gauge couplings evolve above the
compactification scale. To address this question, it is useful to focus on the differential
running, i.e., the nonuniform evolution of the gauge couplings. Above MC /2 a whole
154
(22)
with 1 vanishing trivially. The evolution of i () above MC /2 (but below m ) takes the
form
i () = i (MC /2)
1 H
Ri () + RiG () ,
2
(23)
where RiH () represents contributions arising from Higgs loops and RiG () represents
contributions arising from gauge loops. Contributions from matter loops vanish because
they contribute universally to the running. Using the known spectra for the various KK
towers and their beta function contributions (shown in Fig. 1), we can calculate these
quantities. For the Higgs contributions we find
2
4
H
R2 () = log
log
log
+
,
5
MC /2
5
nMC
5 H
MnH 
0<nMC <
Mn <
(24)
3
6
log
log
+
R3H () = log
.
5
MC /2
5
nMC
5 H
MnH 
0<nMC <
Mn <
(25)
Here the MnH are the colored Higgs masses given in Eq. (21). Recall that there are
actually two slightly nondegenerate MnH  near each explicitly shown colored Higgs level
in Fig. 1(b).
3 The threshold correction from KK towers are also discussed in the context of powerlaw unification [15] and
orbifold breaking of the unified gauge symmetry [16].
G
log
log
R2 () = 6 log
+4
,
4
MC /2
nMC
MnG
0<nMC <
MnG <
G
log
log
R3 () = 9 log
+6
,
6
MC /2
nMC
MnG
G
0<nMC <
155
(26)
(27)
Mn <
where MnG are the X, Y vector multiplet masses, given by the solutions to Eq. (5).
3.1. Higgs contributions
The running of i () is given by
di
1 dRiH () dRiG ()
.
=
+
d
2
d
d
(28)
Let us consider the Higgs and gauge contributions to this running separately, starting with
the Higgs loops first. We have
dR2H
1
=
(# of doublets with mass ) (# of triplets with mass) ,
d
5
(29)
and
dR3H
3
=
(# of doublets with mass < ) (# of triplets with mass ) . (30)
d
10
Here we mean the number of doublet and triplet chiral superfields: there are two massless
doublets, four triplets with mass MC /2, four doublets with mass MC , and so on.
Both of the above expressions have the expected 1/ characteristic of logarithmic
running, but have coefficients which average out to zero. Below MC /2 only the massless
Hu and Hd doublets contribute, so R2H (R3H ) runs in the positive (negative) direction. At
MC /2 we encounter four triplet chiral superfields and R2H (R3H ) now runs in the negative
(positive) direction. The directions reverse again when we gain four doublets at MC , and
again at 3MC /2 when we gain another four triplets. These threshold effects continue, but
become increasingly negligible. Taking the triplet masses to be (n + 1/2 )MC , the
threshold effects from the states between nMc and (n + 1)MC are proportional to
log
+ log
log
nMC
(n + 1)MC
(n + 1/2 + )MC
156
We conclude that the differential running due to the Higgs fields dies off quickly above
the compactification scale. Thus the total Higgs contribution to the running between MC /2
and a high scale such as m or g52 2 is essentially independent of the high scale. Taking
the pairs of triplet hypermultiplets to be exactly degenerate, we find that for high scales,
2
2MC
MC
3MC /2
H
+ log
R2 =
log
log
5
MC /2
MC
3MC /2
5MC /2
log
+
2MC
2
,
= log
(32)
5
2
and similarly,
3
H
R3 = log
(33)
.
5
2
Note that the ratio R2H /R3H = 2/3 is the same as one would have calculated by including
the contributions from the massless doublets alone. This fact will be important when we
compare with the running in 4D minimal SU(5) in Section 3.3.
3.2. Gauge contributions
In contrast to the Higgs threshold effects, those arising from gauge field loops do not
quickly die off. In fact, we will see that these effects add up to an effective logarithmic
running all the way up to m or g52 2 .
The definitions we have used for RiG are very convenient for making connection
to ordinary, fourdimensional running. In particular, 1 receives no gauge contribution
below the GUT scale in fourdimensional theories, and likewise here R1G will not
receive contributions to differential running. R2G and R3G , on the other hand, will receive
corrections above the compactification scale.
Referring to Eqs. (26) and (27), we see that the quantities RiG are proportional to the
quadratic Casimir coefficients C2 (G) of the gauge groups. Thus, we can write the gauge
contributions as
RiG () = C2 (Gi )6(),
(34)
2 # of(V , )X,Y levels below .
(35)
The notation here is as in Fig. 1(a). Note that in contrast to the case for the Higgs multiplets,
here the coefficient multiplying 1/ does not average to zero. Rather, for low modes (such
that MnG (n + 1/2)MC ), the coefficient is 3 up to MC /2, 1 between MC /2 and MC , then
3 again until 3MC /2, and so on. For the higher mass modes, such that the X, Y and 321
multiplets are nearly degenerate, the coefficient becomes fixed at 1.
157
Fig. 2. The qualitative picture for gauge coupling unification in our 5D model. We define
i () i1 () 11 (). The conventional unification scale MGUT 2 1016 GeV is a derived
scale rather than a physical one. Here we assume m < g52 2 , so that unification is achieved near
g52 2 .
If the running were coming entirely from the massless 321 vector multiplet, the
coefficient multiplying 1/ would be 3, so we see that above the compactification scale,
the differential running due to the gauge loops is slowed somewhat relative to the ordinary
logarithmic evolution in 4D. Note that even once the X, Y and 321 multiplets become
degenerate, the differential running continues. In this regime, the running is due entirely to
the eaten Goldstone states contained in the X, Y KK tower. The differential running stops
completely only once we reach the scale m , when the rest of the degrees of freedom
begin to propagate in the loops. 5
The qualitative picture for gauge coupling unification in our model is depicted in Fig. 2.
Above the compactification scale, the nonuniform evolution of the gauge couplings slows,
so that unification occurs at a larger scale than usual. The unification scale will essentially
be the larger of m and g52 2 . Next we will verify that this picture is in fact correct, and
we will make more precise the connection to unification in ordinary minimal SU(5) in 4D.
3.3. Unification and minimal SU(5)
Up to this point we have discussed the differential evolution of the couplings above the
compactification scale, but have not explicitly demonstrated that the couplings in our model
unify in a manner consistent with electroweak scale measurements of the 321 couplings.
Here we will demonstrate that, at one loop, the successful prediction of sin2 W of minimal
SU(5) with MHC < MGUT [17] is equivalent to the successful prediction of sin2 W in this
5 Or, if m happens to be below the the scale at which the X, Y and 321 towers become degenerate
(determined by g52 2 ), the differential running due to the Goldstone degrees of freedom ceases above m ,
but the total differential running does not die off until a scale g52 2 .
158
model with MC MHC . That is, the fact that minimal SU(5) with a triplet Higgs below
the GUT scale can give sin2 W correctly guarantees that this model can do as well.
To see this we must reexamine the running of gauge couplings in fourdimensional,
minimal SU(5). Above the Higgs triplet mass, the gauge couplings evolve as
(4d)
1
(4d)
3
MGUT
4d
4d
C2 (Gi ) log
i (MGUT ) i MHC =
(37)
,
(4d)
2
MH
C
(4d)
, the
where MGUT is the conventional GUT scale 2 1016 GeV. Below the scale MH
C
running of the gauge couplings in the 5D theory is the same as for the 4D theory. Thus, if
(4d)
has the same form as Eq. (37), we will have
we can demonstrate that i5d () i5d MH
C
shown that successful unification in ordinary minimal SU(5) implies successful unification
(4d)
in our model. Here the superscript in MH
C is to emphasize that it is the colored Higgs mass
that gives successful unification in the 4D theory that is meant, rather than the lightest KK
mode of the colored Higgs in the 5D theory.
(4d)
are
The Higgs contributions to i5d () i5d MH
C
1
MC /2
1
,
c
+
log
ci log
(38)
i
(4d)
2
2
2
MHC
where (c2 , c3 ) = (2/5, 3/5). The first term comes from the ordinary running due to Hu
and Hd below MC /2, and the second term is the finite running above MC /2 found in
Eqs. (32) and (33). Since the two terms are both proportional to ci , we see that the Higgs
(4d)
contribution to the differential running in 4D (which takes place entirely below MHC )
is emulated in 5D with the proper choice of compactification scale. Had the differential
running due to the Higgs multiplets turned off right at MC /2, we would simply identify
(4d)
. Because we find a small amount of differential running above MC /2, we
MC /2 = MH
C
(4d)
instead identify MC / = MH
. We expect small additional corrections to this quantity
C
from further finite oneloop effects, but the precise value MC / is not important: what
(4d)
that leads to a successful prediction of sin2 W in
matters is that, given any value of MH
C
(4d)
the 4D theory, there is a value for MC MH
such that the total differential running
C
due to Higgs loops in the 5D theory will be the same as in the 4D theory.
(4d)
arises entirely from loops of
Having chosen this value for MC , i5d () i5d MH
C
gauge and physical fields:
(4d)
3
MC /2
5d
5d
C2 (Gi ) log
i () i MHC =
(4d)
2
MHC
1
159
where the first term comes from the ordinary 4D running below MC /2, and two terms in
the curly bracket represent the contributions from gauge KK towers and physical fields,
respectively (for the parameterization of the gauge contribution, see Eq. (34)). We find that
this has the same form as Eq. (37), with the required identification
MGUT
.
= 3 log
6() log
m
MC /2
(40)
Note that 6() log(/M) under , where M is some mass scale which is a
function of MC and g52 2 , so that the quantity in the lefthand side has a welldefined
value which is a function of MC , g52 2 and m . (g52 2 is the scale at which the
differential running almost ceases in the 5D theory.)
We see that the form of differential running above the Higgs triplet mass in minimal
SU(5) is precisely what arises from threshold effects in the case of the fivedimensional
theory. This holds irrespective of whether the stepfunction approximation of fields in the
RG running is very good or not. For instance, had we used the full oneloop calculation
as in Ref. [13], all gauge contributions would still be proportional to C2 (G). At one loop,
the structure of unification is equivalent to that of minimal SU(5). That this model can
successfully yield sin2 W is no more nor less remarkable than in fourdimensional unified
theories. The remarkable feature is that the grand unified group is broken maximally, at
a scale well above the masses of the lightest X and Y gauge bosons. This feature changes
the predictions of the theory, as we will see in Sections 4 and 5.
Let us make one final comment regarding the differential running of the couplings. In
a sense, its logarithmic (as opposed to power law) behavior is natural. The power law
evolution is characteristic of the bulk (i.e., fivedimensional) nature of the gauge coupling.
In contrast, the SU(5) breaking occurs strictly on a brane. Simply by Lorentz invariance,
the brane breaking cannot contribute to the bulk coupling, but instead should just generate
a brane contribution to the gauge kinetic term. Consequently, we would not a priori have
expected the effects of this SU(5) breaking to exhibit a power law behavior.
3.4. Scales
Now that we have shown the direct correspondence between the differential running
in our model to that of ordinary minimal SU(5), we can determine the scale at which
successful unification occurs. In other words, given a compactification scale MC , we can
find the values of m and that give sin2 W correctly.
The allowed compactification scales are already known: these are related to the allowed
values of MHC in minimal SU(5) by MC MHC . Then we need only to find the
values of m and for which Eq. (40) holds. These values are indicated in Fig. 3
for compactification scales between 6 1014 GeV and 6 1015 GeV (corresponding to
MHC 2 1014 GeV and 2 1015 GeV). For these compactification scales, the 5D Planck
scale M is (1 3) 1017 GeV, and Fig. 3 shows that the GUTbreaking scale can be
at or near this scale and the unification of three gauge couplings can be attained below
160
Fig. 3. Values of M and g52 2 which generate the same threshold effect above MC as is
generated in minimal supersymmetric SU(5) between MHC and MGUT . The dashed line corresponds
to MC = 6 1014 GeV, while the solid line corresponds to MC = 6 1015 GeV. To calculate these
lines we have used firstorder solutions in (g52 2 R)1 for the X and Y masses.
M depending on the parameters of the model. 6 In the parameter region where the field
theoretic unification works, the ratio of the cutoff to the compactification scales is O(10
100), and no gauge or Yukawa couplings become nonperturbative below the cutoff scale.
Thus, our oneloop treatment of the running is well justified. We also find that must be
somewhat smaller than M in this parameter region. This is consistent with the observation
that the theory is more or less strongly coupled at the scale M to have O(1) Yukawa and
gauge couplings in 4D; if the theory is strongly coupled at M , is naturally
M /(4), since the superpotential giving the vev of scales like M 2 + 4 3 . Even
then, however, the couples to the other fields in the combination of 4/M , so that
various operators feel order one GUT breaking, since 4 /M 1. Therefore, we treat
as if has a vev of order M in the following discussions, although all the arguments also
apply in the strongly coupled case, M /(4).
3.5. Brane operators
We finally comment on effects from brane operators like d 2 (/M )m W W (y
R) where m = 1, 2, . . . , and W is the field strength superfield. Although it is possible
that these operators are somehow suppressed at the cutoff scale, from effective field theory
point of view it is generically expected that they are present with order one coefficients.
They give treelevel splitting of three gauge couplings for SU(3), SU(2) and U(1), and
could affect the previous analysis of the gauge coupling unification. In particular, since
6 If m or g 2 2 is above the 5D Planck scale, the gauge unification is not completed in a field theoretic
5
regime. In this case, our field theoretic treatment is not fully trustable above M and would have some
161
we are considering M , one might think that they give O(1) correction to sin2 W .
However, we can expect that the effect of these operators are actually smaller by making
the following observations.
As an example, let us first consider the extreme case where all the interactions are
strongly coupled at the scale M . Then, the operators involving the field strength superfield
scale as
M
1 4
W
W
+
(y
R)
W
W
+
,
L5 = d 2
(41)
24 3
16 2 M
where 4 /M is an O(1) quantity. (Strictly speaking, 4 /M must be somewhat
smaller than 1 so that all higherdimensional operators involving (4 /M )m do not
equally contribute and make the theory unpredictable.) On integrating over y, the zeromode 4D gauge couplings g0 are given as
1
M R
1
+
.
2
2
2
12
16
g0
(42)
Here, the first and second terms are SU(5)preserving and SU(5)violating contributions,
respectively. Since we know that g0 1, we have to take M R 12 2 in this strongly
coupled case. This shows that the SU(5)violating contribution coming from brane
operators is suppressed by a factor of 1/(16 2 ) in this case.
In fact, the theory is not truly strongly coupled at the scale M in the realistic case
discussed in previous sections so that the oneloop treatment of gauge coupling evolutions
is reliable. Nevertheless, the above argument applies more generically; the SU(5)violating
brane contribution is small relative to the SU(5)preserving bulk contribution due to the
large volume factor 2RM . Thus, the correction to sin2 W is expected to be small. We
will not discuss possible effects of the brane operators further, and assume that they are
negligible in the subsequent discussions.
4. Proton decay
One of the key signals of grand unification is proton decay. X and Y gauge boson
exchange generates dimensionsix proton decay operators in the low energy theory, and
Higgsino triplet exchange generates dimensionfive operators. One might expect that since
the particles appear at a scale MC , this will be the suppression scale of the proton decay
operators. One interesting possibility offered by this framework is that the rate of proton
decay is not constrained by gauge invariance to be related to the strength of gauge and
Yukawa interactions in the usual way.
We have already noted that the dangerous particles have wave functions at the y = R
fixed point that are small compared with those of their 321 or doublet counterparts. On the
other hand, there is a tower of states that can mediate proton decay, and we must sum each
contribution. In this section, we investigate proton decay operators generated by exchanges
of X, Y gauge bosons and colored Higgsinos, and show that the present model can satisfy
the constraints coming from experimental lower bounds on the proton lifetime.
162
cos2 (M G R)
n=0
n
2
G
Mn
(43)
where MnG are the masses of the X and Y gauge boson KK modes. These masses satisfy
Eq. (5), which allows us to approximate the sum as
tanh(g52 2 /2MC )
4
=
.
g 4 4 + MC2 (2n + 1)2
g52 2 MC
n=0 5
(44)
1/2
cos2 (MnH R)
1
,
MHC
MnH
n=
(45)
where MnH are the masses of the colored Higgs KK modes, given in Eq. (21). Using this
spectrum, we obtain the sum
8
4
1
=
.
(46)
2
2
+ 4 nMC + MC arctan(/2)/
( + 4)MC
n=
Thus, proton decay from dimensionfive operators is suppressed by a factor 8/ 3
compared to the compactification scale. In ordinary minimal SU(5), proton decay limits
require MHC > 9.4 1016 GeV, so for a compactification scale of MC = 2 1015 GeV,
we would need a parameter of ten to adequately suppress proton decay operators. This
leaves open the possibility that proton decay could be observed in future experiments.
However, for larger detection becomes increasingly unlikely.
4.3. Derivative operators
In addition to the ordinary Yukawa couplings between the Higgs and matter fields on
the brane, there can be derivative couplings of the conjugate Higgs fields to the matter
163
fields as well. 7 These operators such as (u T10 T10 y H c + d T10 F5 y H5c )(y R) can
5
lead to proton decay. Of course, coefficients u and d of these operators are not related
to the usual Yukawa couplings, u and d , by SU(5), so that their actual significance is
unknown. Furthermore, there is an ambiguity in what we mean by the derivative of the
conjugate field, which is not differentiable at the point y = R. However, using Eqs. (6)
and (7), we can rewrite the coupling as
c
H
Mn gC,n (R)
g (R) C,n (x),
(y R)y HC = (y R)
2R m C,n
n
(47)
where the summation over m arises from rewriting the function as a sum of the KK mode.
While we can compute proton decay diagrams involving these vertices, it is now
apparent that such diagrams have a strong dependence on how we cut off the sum of the
KK mode. If the cutoff is done near the fundamental scale, these diagrams can, at least
in principle, give comparable contribution to those involving the usual Yukawa couplings.
However, it is not entirely clear what the cutoff for these diagrams should be. In the case
where the brane is dynamical, the summation in the above operators are cut off at the scale
of the brane tension [18]. For couplings at the orbifold fixed point, which is not dynamical
and cannot fluctuate, it is unclear whether the cutoff is the fundamental scale or a lower
scale, such as the radion mass.
For our purposes here, we will not address these issues further. Since we cannot a priori
know the size of the couplings of these operators and their flavor structure, estimating
the resulting proton decay rate is already very uncertain. However, it is possible that such
operators may provide an opportunity for detectable dimensionfive proton decay in the
near future.
5. Yukawa couplings
So far, we have a framework that looks like ordinary SU(5) except with suppressed
proton decay. While the breaking occurs at a scale near the 5D Planck scale M , the
X, Y gauge bosons are much lighter. Still, in a very real sense, the breaking of SU(5)
is maximal, which can manifest itself in deviations from SU(5) expectations. We here
consider the effects of such maximal GUT breaking on the fermion Yukawa couplings.
In the minimal SU(5), one important prediction is the unification of the Yukawa
couplings. A successful prediction of the theory is the unification of the bottom and
Yukawa couplings at the GUT scale [19]. However, it is well known that the SU(5) relations
fail in the lighter first two generations. For instance, an SU(5) relation me /m = md /ms
fails by a factor of ten. 8
7 We thank M. Graesser for bringing this to our attention.
8 One may be able to correct this prediction by means of contributions from supersymmetrybreaking A terms
164
In ordinary 4D SU(5) GUT, it has been suggested that the operators involving the field
can correct this discrepancy [21]. However, this mechanism does not work in most theories
where the fermion mass hierarchy is explained by the FroggattNielsen mechanism [22].
In this mechanism, the matter fields carry generation dependent flavor U(1) charges and
the various Yukawa couplings are generated through the U(1) breaking spurion. This is
an attractive mechanism in that it not only suppresses the firsttwo generation Yukawa
couplings but also suppresses dangerous treelevel dimensionfive proton decay operators.
However, having employed this mechanism, the GUTbreaking operators involving
can modify SU(5) mass relations only by an amount suppressed by a factors of MGUT /MPl ,
which is too small to accommodate order one deviations suggested by the observed quark
and lepton masses. 9
In the present model, on the other hand, there is no suppression of higherdimensional
operators involving GUTbreaking effects, since is near the cutoff scale M . In this
sense, fields living on the GUTbreaking brane see the GUT broken maximally, and their
Yukawa couplings need not respect the SU(5) symmetry. Thus, the failure of SU(5) to
describe (me /m )/(md /ms ) seems quite natural. In this framework, however, the success
of the b unification must be viewed as an accident, unless there is a reason why the
coupling of to the third generation is somewhat suppressed.
165
7. Conclusions
The possibility that the standard model gauge group is merely a subgroup of a larger,
simple group is an attractive one. Unfortunately, the simplest version of supersymmetric
SU(5) predicts proton decay at rates incompatible with experiment.
We have demonstrated that the incorporation of just one new ingredientan additional
dimension in which gauge and Higgs fields propagatebrings about crucial changes
relative to ordinary GUTs. The couplings of the lightest X and Y bosons are no longer
directly linked through gauge symmetry to those of the lightest 321 bosons, and the
couplings of the lightest Higgs triplets are no longer directly linked to those of the Higgs
doublets. The strong breaking of the GUT symmetry pushes away the wave functions
of these states, suppressing the generated proton decay below experimental limits, all
while retaining ordinary, logarithmic unification. Such features seem special to a fivedimensional theory, incapable of reproduction in simple fourdimensional theories.
The presence of extra dimensions of this size has been motivated previously as a means
to resolve the supersymmetric flavor problem [28]. Our GUTbreaking picture can nicely fit
into this framework of supersymmetrybreaking mediation. For instance, if supersymmetry
is broken at y = 0 fixed point by F term vev of singlet or nonsinglet field, it naturally
realizes the scenarios of Refs. [29] and [30], respectively. The incorporation of our GUT
picture within these scenarios may give interesting signatures, since we now have one
more piece of information about parameters of the model; the compactification radius
is determined by the gauge coupling unification. We leave an investigation of detailed
phenomenology for future work.
To summarize, the framework we have described is very simple, but can describe
various observed features such as gauge coupling unification, lack of unification in Yukawa
couplings, the absence of proton decay, and so on. It will be interesting to add nonminimal
structure to the model as a means of deriving experimental signatures.
Acknowledgement
We thank M.L. Graesser, L.J. Hall, D.B. Kaplan, D.E. Kaplan, A.E. Nelson, M.E. Peskin
and L. Randall for useful discussions.
References
[1] H. Georgi, S.L. Glashow, Phys. Rev. Lett. 32 (1974) 438.
[2] H. Georgi, H.R. Quinn, S. Weinberg, Phys. Rev. Lett. 33 (1974) 451.
[3] S. Dimopoulos, H. Georgi, Nucl. Phys. B 193 (1981) 150;
N. Sakai, Z. Phys. C 11 (1981) 153.
[4] S. Dimopoulos, S. Raby, F. Wilczek, Phys. Rev. D 24 (1981) 1681;
L.E. Ibanez, G.G. Ross, Phys. Lett. B 105 (1981) 439.
[5] N. Sakai, T. Yanagida, Nucl. Phys. B 197 (1982) 533;
S. Weinberg, Phys. Rev. D 26 (1982) 287.
166
[6] J. Hisano, H. Murayama, T. Yanagida, Nucl. Phys. B 402 (1993) 46, hepph/9207279;
For a recent analysis, see: T. Goto, T. Nihei, Phys. Rev. D 59 (1999) 115009, hepph/9808255.
[7] H. Murayama, A. Pierce, in press.
[8] P. Candelas, G.T. Horowitz, A. Strominger, E. Witten, Nucl. Phys. B 258 (1985) 46;
E. Witten, Nucl. Phys. B 258 (1985) 75.
[9] J.D. Breit, B.A. Ovrut, G.C. Segre, Phys. Lett. B 158 (1985) 33.
[10] A. Sen, Phys. Rev. Lett. 55 (1985) 33.
[11] Z. Chacko, M.A. Luty, E. Ponton, JHEP 0007 (2000) 036, hepph/9909248.
[12] N. ArkaniHamed, L. Hall, Y. Nomura, D. Smith, N. Weiner, hepph/0102090.
[13] K.R. Dienes, E. Dudas, T. Gherghetta, Phys. Lett. B 436 (1998) 55, hepph/9803466;
K.R. Dienes, E. Dudas, T. Gherghetta, Nucl. Phys. B 537 (1999) 47, hepph/9806292.
[14] I. Antoniadis, Phys. Lett. B 246 (1990) 377.
[15] D. Ghilencea, G.G. Ross, Phys. Lett. B 442 (1998) 165, hepph/9809217;
D. Ghilencea, G.G. Ross, Nucl. Phys. B 569 (2000) 391, hepph/9908369.
[16] L. Hall, Y. Nomura, hepph/0103125.
[17] J. Hisano, H. Murayama, T. Yanagida, Phys. Rev. Lett. 69 (1992) 1014.
[18] M. Bando, T. Kugo, T. Noguchi, K. Yoshioka, Phys. Rev. Lett. 83 (1999) 3601, hepph/9906549.
[19] M.S. Chanowitz, J. Ellis, M.K. Gaillard, Nucl. Phys. B 128 (1977) 506.
[20] J.L. DiazCruz, H. Murayama, A. Pierce, hepph/0012275.
[21] J. Ellis, M.K. Gaillard, Phys. Lett. B 88 (1979) 315.
[22] C.D. Froggatt, H.B. Nielsen, Nucl. Phys. B 147 (1979) 277.
[23] K.I. Izawa, K. Kurosawa, Y. Nomura, T. Yanagida, Phys. Rev. D 60 (1999) 115016, hepph/9904303.
[24] Y. Fukuda et al., SuperKamiokande Collaboration, Phys. Rev. Lett. 81 (1998) 1562, hepex/9807003.
[25] T. Yanagida, in: O. Sawada, A. Sugamoto (Eds.), Proc. of the Workshop on the Unified Theory
and Baryon Number in the Universe, KEK report 7918, 1979, p. 95;
M. GellMann, P. Ramond, R. Slansky, in: P. van Nieuwenhuizen, D.Z. Freedman (Eds.),
Supergravity, NorthHolland, Amsterdam, 1979, p. 315.
[26] L.J. Hall, M. Suzuki, Nucl. Phys. B 231 (1984) 419;
I. Lee, Nucl. Phys. B 246 (1984) 120.
[27] N. ArkaniHamed, L. Hall, H. Murayama, D. Smith, N. Weiner, hepph/0006312;
N. ArkaniHamed, L. Hall, H. Murayama, D. Smith, N. Weiner, hepph/0007001;
F. Borzumati, Y. Nomura, hepph/0007018.
[28] L. Randall, R. Sundrum, Nucl. Phys. B 557 (1999) 79, hepth/9810155.
[29] D.E. Kaplan, G.D. Kribs, M. Schmaltz, Phys. Rev. D 62 (2000) 035010, hepph/9911293;
Z. Chacko, M.A. Luty, A.E. Nelson, E. Ponton, JHEP 0001 (2000) 003, hepph/9911323.
[30] D.E. Kaplan, G.D. Kribs, JHEP 0009 (2000) 048, hepph/0009195.
Abstract
We construct supersymmetric M3brane solutions in D = 11 supergravity. They can be viewed as
deformations of backgrounds taking the form of a direct product of fourdimensional Minkowski
spacetime and a noncompact Ricciflat manifold of G2 holonomy. Although the 4form field
strength is turned on it carries no charge, and the 3branes are correspondingly massless. We also
obtain 3branes of a different type, arising as M5branes wrapped over S 2 . 2001 Published by
Elsevier Science B.V.
1. Introduction
The standard D3brane provides a natural supergravity dual of fourdimensional N = 4
superconformal YangMills theory, via the AdS/CFT correspondence [13]. Branes with
less supersymmetry can in general be constructed by replacing the spheres that form the
level surfaces in the flat transverse space by some other Einstein space that admits a lesser
number of Killing spinors [4]. It was proposed that D3branes on the sixdimensional
conifold, in which the level surfaces are the T 1,1 space, is dual to an N = 1 superconformal
theory in D = 4 with gauge group SU(N) SU(N) [5]. The conformal symmetry can
then itself be broken, by introducing fractional branes corresponding to the wrapping of
D5branes on 2cycles. The corresponding supergravity solutions were obtained in [6,7].
It has been proposed that Mtheory compactified on a certain singular sevendimensional
space with G2 holonomy might be related to a N = 1, D = 4 gauge theory [810], which
has no conformal symmetry to begin with. (See also the recent papers [1115].) This leads
to the question of whether there might exist a 3brane configuration in Mtheory, whose
Email address: cvetic@cvetic.hep.upenn.edu (M. Cvetic).
1 Research supported in part by DOE grant DEFG0295ER40893 and NATO grant 976951.
2 Research supported in full by DOE grant DEFG0295ER40899.
3 Research supported in part by DOE grant DEFG0395ER40917.
168
169
(1)
j
where i i = 1 and Di = di + ij k A(1) k , and Ai(1) is the SU(2) YangMills instanton
on S 4 , whose unit metric is d42 . The functions H , a and b will be taken to depend only
on the radial coordinate in the transverse space. This describes the case of the R3 bundle
over S 4 . The second possibility is obtained by replacing the S 4 by CP2 . This does not
affect the form of the equations for H , a and b. 4 The constrained i coordinates can be
expressed in terms of two angular coordinates on S 2 in a standard way,
1 = sin sin ,
2 = sin cos ,
3 = cos .
(2)
(3)
There is clearly a vacuum solution of the form (1) that is simply the direct product
of fourdimensional Minkowski spacetime and the associated Ricciflat sevendimensional
manifold with G2 holonomy. In the vacuum we shall have H = 1, with a and b being given
by [16,17]. We should now like to turn on the 4form field strength of elevendimensional
supergravity. The 4form ansatz that respects the symmetry of the metric is given by [17,
19]
F(4) = f1 (4) + f2 X(2) Y(2) + f3 d Y(3) ,
F (4) = H 4 a 2 b4 f1 (4) d X(2) + H 4 a 2 f2 (4) d Y(2)
+ H 4 f3 (4) X(3) ,
(4)
(5)
(Note that F(4) could in principle have had a term of the form d X(3) as well, but this
is ruled out by the field equation dF(4) = 0.) The Bianchi identity dF(4) = 0 implies that
4 In everything that follows, results obtained for the case of the S 2 bundle over S 4 apply equally, mutatis
mutandis, to the case of the S 2 bundle over CP2 .
170
(6)
In fact, F(4) = dA(3) with A(3) = 12 f Y(3) . The field equation dF(4) = 0 implies
1
2a 4 + b4 H 4 f a 2b4 H 4 f = 0.
(7)
2
(The F(4) F(4) term vanishes here.)
In order to impose the D = 11 Einstein equation it is convenient to perform a Kaluza
Klein reduction on the 4dimensional worldvolume of the 3brane, so that the problem
can be reformulated from a sevendimensional point of view. This allows us to make use of
curvature calculations for 7metrics of this type that were performed in [17]. The relevant
sevendimensional Lagrangian is given by
1
1
e1 L7 = R ()2 e
2
48
8
5
2
F(4)
,
(8)
=e
32
2
5
ds72
+e
1
3
5
2
dx dx .
(9)
(10)
Thus we have
dt = H 4/5d,
a = H 4/5a,
b = H 4/5b.
(11)
The original elevendimensional Einstein equation is now recast as the sevendimensional dilaton equation and Einstein equation. The dilaton equation gives
2 4 4 2 4 2 2f 2 4f 2 a 2
a b H
(12)
= H f + 2 +
,
3
a
b4
which can be rewritten as
f 2
6H 18H 2 12a H 24b H
2f 2
4f 2
+
+
=
+
+
+
.
H
H2
aH
bH
a 2 b4 a 4 b4
b8
From the Einstein equation we get three separate equations, namely,
5
5a 2
5a 28a H 5a 2 20a b 16b H 4H 12H 2
+
+ 2 +
+
+
+
2 4
2
a
aH
a
ab
bH
H
H
a
b
2
2
2
f
2f
6f
= 2 4 4 4+ 8 ,
4a b
a b
b
(13)
171
(14)
40bH
15H 2
5f2
5a 2 20a b 15b 2 20a H
+
,
ab
aH
bH
2a 2
b2
H2
8a 2b4
5
V = H 8 2a 2b4 a 4 + b4 6a 2 b2 2a 4 + b4 f 2 ,
(15)
4
together with the constraint T + V = 0. Note that here a dot is a derivative with respect
to the coordinate , defined by d = a 2 (bH )4 d. The kinetic term T can be expressed as
T = 12 gij i j , where a = e1 , b = e2 , H = e3 and f = 4 , with
5 20 20
0
20 30 40
0
.
gij =
(16)
20 40 30
0
T =
5
4a 2 b4
(18)
(19)
Solutions of these equations will necessarily satisfy the original secondorder equations,
implying that we shall have solutions of the D = 11 supergravity equations. One may note
172
(20)
b = H 1/2b,
a = H 1/2a,
f = H 3/2f.
(21)
H 1/2 d.
The
(22)
1
d b
=
,
dr
2b 3
(a 2 + b 2 )f2
1 dH
=
.
H dr
3a 2 b 4 (4a 2 b 4 + f2 )
(24)
In particular, the equations for a and b have now decoupled from the rest, and they are in
fact nothing but the firstorder equations for the G2 metrics on the R3 bundle over S 4 (see,
for example, [20]).
The system of firstorder equations is now completely solvable. After a final change of
radial variable r 12 r 4 , we find that the general solution for the metric in the ansatz (1) is
1
2
= H 2 dx dx + 2H 7 U 1 dr 2 + r 2 H 1 U Di Di + r 2 H 1 d42 ,
ds11
2
where
1/6
c2
/4
H = 1 + 12 2
,
U =1 4,
r
2r U
(25)
(26)
and c is a constant. The function f appearing in the 4form ansatz (6) is given by
f=
c
r 3 H 3/2U 1/2
(27)
173
1/2 1
i
i
2
D D + d4 + d 2 ,
dx dx +
/ = 0: ds11 =
2
1
2
/
= 0: ds11
(28)
= 2/5 dx dx + 1/5 Di Di + 4/5 d42 + d 2
2
as the proper distance tends to zero. We can also consider the the possibility of sending
/4 /4 . In this case, the metric behaviour at small proper distance (i.e., near r = 0)
becomes
1
2
1/4
i
i
ds11 =
(29)
dx dx + D D + 1/2 d4 + d 2 .
2
The 4form flux is given by
Q = 2f (4),
(30)
S4
and from (27) this can be seen to vanish when r is sent to infinity. Thus our M3brane
configurations do not carry any conserved charge, and might be described as 3branes
without 3branes. The solution should really be thought of as a gravitional monopole
involving the supergravity multiplet. The two constants / and c are both continuous
parameters.
What is perhaps the most interesting feature of the solution is that it is massless. The
leadingorder rdependence in the function H at large r is c2 /r 12 , whilst a mass term for
a 7dimensional transverse space would have a leadingorder rdependence of the form
m/r 5 . The existence of the naked singularity may be related to the fact that the solution
is massless. In fact, all the previously known massless pbrane solutions in supergravity
contain naked singularities [2125]. A repulsion mechanism was proposed in string theory
[26] to resolve such a naked singularity in the massless dyonic string [25]. A further
observation is there appears not to exist a natural and nontrivial decoupling limit. This
174
may also be a consequence of the masslessness. For example, a massive dyonic string has
a decoupling limit, but this ceases to exist when the charges are tuned for masslessness.
Unlike the dyonic string, where the masslessness is achieved by making an adjustment of
integration constants, in our new M3brane solution there is no mass integration constant.
From the point of view of supersymmetry, the masslessness is consistent with the absence
of a nonvanishing conserved 4form charge. However, we shall defer a more detailed
investigation of the supersymmetry of this solution until Section 4.
3. M3brane in background of R4 bundle over S 3
In this section we construct an analogous 3brane solution in the background of the third
manifold of G2 holonomy, whose topology is R4 S 3 . As we shall see, the configuration
is again a nobraner, which carries no conserved brane charge.
3.1. The ansatz
In this case, we consider the elevendimensional metric ansatz
2
ds11
= H 2 dx dx + d 2 + a 2 i2 + b2 i2 ,
(31)
where i i
and i and i are two sets of leftinvariant on two independent
SU(2) group manifolds. The level surfaces r = constant are therefore an S 3 bundle over
S 3 . Since the bundle is a trivial one, the level surfaces are topologically S 3 S 3 . There is a
vacuum solution which is a complete Ricciflat manifold, namely the direct product of fourdimensional Minkowski spacetime and the known sevendimensional manifold R4 S 3
with G2 holonomy [16,17].
Again we should now like to turn on the 4form field strength, in order to introduce
a 3brane configuration in this background. The ansatz for the 4form, respecting the
symmetries of the vacuum, can be written as [18]
1
2 i ,
F(4) = f1 i j i j + f2 d 1 2 3
1
+ f3 ij k d i j k .
2
The Bianchi identity dF(4) = 0 gives
1
1
f1 f2 + f3 = 0,
8
2
and the field equation d F(4) = 0 gives
(32)
(33)
2f1 H 4
3f1 H 4
+ f3 ab1 H 4 = 0,
+ 2f2 b3 a 3 H 4 = 0.
(34)
ab
ab
It is again convenient to derive the conditions implied by the elevendimensional Einstein
equation in terms of a dimensional reduction to D = 7. The dilaton equation gives
f32
2f12
f22
H 3H 2 3a H 3b H
+
+
= 0,
H
aH
bH
H2
a 4 b4 6a 6 2a 2 b4
(35)
175
15f 2
15f 2
60f 2
3
60 b43
90 + b43
120 b4 3 15f
b4
2
15f 2
60f 2
15f3
90 + 15f43
60 b43
120 + b4 3
b
b4 .
gij =
(38)
60f32
60f32
240f32
60f3
120 4
120
+
120
4
4
4
b
b
b
b
3
15f
b4
15f3
b4
3
60f
b4
15
b4
1
1
1
1
90
90
45
90 f3
1
7
1
1
90
90
45
90 f3
g ij =
(40)
.
1
1
1
1
90
90
f
3
72
18
V=
1
90
f3
7
90 f3
1
18
f3
3 4
45
b +
13 2
45 f3
176
If the integration constant is taken to be zero, 5 we find that the potential can be
expressed in terms of a superpotential W , i.e., V = 12 g ij i W j W , with
15
W = H 4 a 2 b1 a 2 + 4b2 b4 f32 .
(41)
4
From this we can obtain firstorder equations, given by
a 2 b4 4b6 2a 2 f32
2a 2b4 + (a 2 4b2)f32
,
b
,
=
8b4 K
8ab3K
H (a 2 + 4b2 )f32
f3 ((12b2 a 2 )b4 f32 (a 2 + 20b2))
H =
(42)
,
f
,
=
3
8ab4K
8ab4K
where K b4 f32 and a prime denotes a derivative with respect to the original
coordinate appearing in (31). Note that these firstorder equations again imply an algebraic
relation among the functions, analogous to (20). This time, we have
a =
a 3 bH 6f3 = .
(43)
4
2
= H 2 dx dx + 12H 4U 1 dr 2 + r 2 H 2 U i2 + r 2 H 2 i2 ,
ds11
3
where
1/6
c2
/3
H = 1 12 3
.
U 1 3,
r
r U
(44)
(45)
c
r 4 H 2 U 3/2
(46)
177
manifold with R4 S 3 topology, since the contribution to the metric is dominated at large
r by the instanton contribution /4 /r 4 , in comparison to the F(4) contribution which is of
order c2 /r 12 . At small distance, the solution has a naked singularity.
288
36
We make a standard 4 + 7 decomposition of the Dirac matrices, as follows:
A = DA
= 1,
a = 5 a .
(47)
(48)
Substituting into (47), and examining first the worldvolume directions , we find that a
Killing spinor of the form = must satisfy 5 = . For the case of 5 = + we
find
1/2
1/2
= g b2 + f3
(49)
1 + g b2 f3
2 ,
where 1 and 2 are constant spinors in the transverse 7space, satisfying the projection
conditions
(1 + i4 )2 = 0,
23 1 = 042 ,
(26 35 )1 = 2i04 2 .
(50)
Note that these conditions uniquely determine 1 and 2 , up to an overall scale. For the
case where 5 = , the associated spinor is given again by (49), but with f3 replaced
by f3 . The dependence of the overall function g on the coordinates of the transverse
space is undetermined by the components of A = 0.
The components of A = 0 lying in the directions A = a of the transverse space
will now determine the dependence of g on the transverse coordinates. It is easiest
first to examine the radial direction (i.e., the 0 direction), which determines the radial
178
dependence of the Killing spinor. Then, by looking at the remaining transverse directions,
we find that the Killing spinor has no dependence on the angular coordinates of the two
3spheres. The conclusion is that the function g in (49) is given by
g = b1 H 1/2.
(51)
Thus the firstorder equations (37) (where f1 and f2 are given by (42) with = 0) are
precisely the integrability conditions for the existence of a spinor satisfying A = 0 in
(47). Since we have 2 solutions (corresponding to two spinors in the M3brane worldvolume) for each of the cases 5 = , the general Killing spinor has four real solutions,
corresponding to N = 1 supersymmetry on the worldvolume of the M3brane. Of course,
if the constant c is set to zero, so that the 4form is turned off, these Killing spinors reduce
to the usual ones in the product of fourdimensional Minkowski spacetime and the Ricciflat metric of G2 holonomy.
4.2. Solutions in the R3 bundle over S 4 background
Here, we repeat the analysis of the supersymmetry transformations in the case of the R3
bundle over S 4 background, for which the M3brane solution was constructed in Section 2.
Again we begin by considering the components of A = 0 lying in the worldvolume
of the 3brane. From = 0 we deduce that a Killing spinor of the form = will
be given by 5 = , and for the case 5 = + we have
= gP (K f )1/2 1 + (K + f )1/2 2 ,
(52)
where K is given by (19), P is given by
sin 12
1
1
1
1
K
cos
if
sin
+
K
cos
+
if
sin
02
01
2
2
2
2
2ab2
+ cos 12 cos 12 + sin 12 12 ,
(53)
and the constant 8component spinors 1 and 2 are uniquely specified (up to scale) by the
projections
(0 3456)2 = 0,
(34 + 56 2012)2 = 0,
(54)
Here the explicit indices 1 and 2 on the Dirac matrices refer to the two directions on the S 2
fibres, as in (3), whilst the indices 3, 4, 5 and 6 refer to the directions in the S 4 base. The
index 0 refers to the radial direction. For the case when 5 = , the associated spinor
in the transverse space will be given again by (52), but now with f sent to f in (52) and
in (53).
The dependence of the overall prefactor g on the coordinates of the transverse space
is not determined by = 0 in the worldvolume directions. (We have, however, made
a convenient choice of and dependent overall factors, in anticipation of subsequent
results.)
179
From the radial component 0 = 0 in the transverse space, we can again determine the
radial dependence of the function g, finding
g=
g
,
H ba 1/2
(55)
where g depends only on the angular coordinates of the transverse space. Examination of
A = 0 in the S 2 directions then implies that g is independent of these two coordinates.
Finally, the components in the S 4 directions determine that g has the dependence associated
with the singlet fermion zeromode in the YangMills instanton background (as in [28,29]).
We have seen that again, the firstorder system of equations for this 3brane in the
background of the R3 bundle over S 4 have turned out to be precisely the integrability
conditions for the existence of a Killing spinor. There are in total four real solutions,
implying N = 1 supersymmetry on the worldvolume of the M3brane. As in the previous
example, if the field strength in the solution is taken to zero, by setting the constant
c = 0, the Killing spinor reduces to the standard one in the vacuum of fourdimensional
Minkowski spacetime times the Ricciflat metric on the R3 bundle over S 4 .
1 8/5 2
F(3) kinetic term
using a real 3form field in D = 7, but with the canonical 12 e
instead of the signreversed one in (56) that arose by dualising the 4form.
The upshot of the above discussion is that we can obtain real solutions in D = 7 that are
just like those in Sections 2 and 3, but for the opposite sign of c2 . These will be solutions
of the equations coming from the sevendimensional Lagrangian
7 = R 1 ()2 1 e 8/5 F 2 .
e1 L
(3)
2
12
(57)
180
The expression for F(3) for each solution will be given by F(3) = iF(4) , where F(4) is
the corresponding expression given in Section 2 or 3. The i factor in this relation between
F(3) and F(4) is precisely removed by the i factor that we acquire upon sending c ic.
A difference now arises when we consider the higherdimensional origin of the sevendimensional Lagrangian. We viewed (8) in Sections 2 and 3 as coming from the Kaluza
Klein reduction of D = 11 supergravity on the worldvolume of the M3brane. Instead, we
should now view (57) as coming from the KaluzaKlein reduction of type IIA, type IIB or
type I supergravity on the worldvolume of a 2brane. In other words, we obtain the 3form
in D = 7 as the direct worldvolume reduction of a 3form in D = 10. Accordingly, we
can then lift the D = 7 solutions of Sections 2 and 3, after sending c ic, to real solutions
of tendimensional supergravity. Thus there is a phase transition from one type of brane to
another, when we change the modulus parameter c of the solution from real to imaginary.
For the case of the S 2 bundle over S 4 in Section 2, we find that the corresponding
massless 2brane solution in D = 10 is given by
1
2
= H 3/2 dx dx + 2H 9/2U 1 dr 2 + r 2 H 3/2U Di Di
ds10
2
+ r 2 H 3/2 d42 ,
(58)
where
U =1
/4
,
r4
H = 1
c2 1/6
.
2r 12 U 2
(59)
For the S 3 bundle over S 3 of Section 3, the corresponding massless 2brane solution in
D = 10 is given by
4
2
= H 3/2 dx dx + 12H 13/2U 1 dr 2 + r 2 H 1/2 U i2 + r 2 H 1/2i2 ,
ds10
3
where
1/6
c2
/3
H = 1 + 12 3
.
U 1 3,
r
r U
(60)
(61)
We have written the solutions that come from reducing the NSNS 3form of the tendimensional supergravity. Of course in the case of type IIB we could instead use the RR
3form, in which case the lifted solutions in D = 10 would simply be the Sduals of those
we have just presented.
One can also, of course, further lift the above configurations, if viewed as solutions of
type IIA supergravity, to D = 11. For the case corresponding to the S 2 bundle over S 4 we
then find
1
2
ds11
= H 2 dx dx + H 4 dz2 + 2H 5 U 1 dr 2 + r 2 H U Di Di
2
+ r 2 H d42 ,
(62)
where z is the eleventh coordinate, and U and H are again given by (59). For the case
corresponding to the S 3 bundle over S 3 , we find
4
2
ds11
= H 2 dx dx + H 4 dz2 + 12H 6U 1 dr 2 + r 2 U i2 + r 2 i2 ,
3
where U and H are given by (61).
181
(63)
4
10
1
1
F(4) = gA(3) + (3) .
2
2
(66)
j
i
16 gij k Ai(1) A(1) Ak(1) . Domain wall and AdS7 black
Here, we have (3) Ai(1) F(2)
hole solutions in this theory have been constructed [3234], which can be viewed after
lifting back to Mtheory as distributed or rotating M5branes, respectively.
Here we consider a 3brane configuration, which is supported by one component of the
SU(2) YangMills gauge fields. We take the ansatz to be
ds72 = e2A dx dx + e2B dr 2 + d22 ,
(67)
3
= 2 ,
F(2)
F(4) = 0.
(68)
The resulting equations of motion can be derived from the Lagrangian L = T V , where
1
T = 12A 2 16A B 2B 2 + 2 ,
2
1 2 2B+ 2
8A+2B
2B
10
V =e
(69)
2e U e
,
2
together with the constraint T + V = 0. Here the dot denotes a derivative with respect to
, defined by d = e8A+2B dr.
182
Fig. 1. The flow defined by the 2vector (b , f ). The abscissa is b and the ordinate is f .
1
4A+2B 1
4A+2B+ 4
10 +
10 + ge
10 .
2g e
W = 2 2ge
2
(70)
The associated firstorder equations, after setting g = 1 without loss of generality, are given
by
4b2
a
= b 2f 4
+ 2f,
a
f
b
4b2
= b2f 4
8f,
b
f
f
4b2
= 4b2 f 4
+ 2f,
f
f
1
(71)
183
6 181
4 11
1
1
+ +
+ ,
b2
+
+ ,
10 5
30
10 5
15
638
+ 40 log 1120 3 log + ,
f 1 2 + 2
(72)
21
up to the first few orders in . It is easy to see that r 0, and so the metric
approaches
a2
1
(73)
dx dx + d22 + dr 2 .
2
r
Since r tends to zero here, this describes the largedistance asymptotic region.
Now consider the behaviour when f , with b approaching zero. In this case, we
have
ds72
1
f (1 )2/7 ,
2
with tending to 1 from below. The metric then has the form
ds72 = (r r0 )2/15 dx dx + d22 + dr 2 .
a 2 (1 )1/7 b2 ,
(74)
(75)
Since r r0 in this case, it clearly corresponds to the region at small proper distance. The
solution at r = 0 would be singular, but it is also a horizon
We can dimensionally reduce the solution on d22 , to obtain a domain wall in D = 5, of
the form
ds52 = e2C dx dx + dr 2 .
(76)
The radial coordinate r runs from r = 0, which is the AdS5 horizon, to r = r0 , which is a
null singularity. The conformal factor in these two regions is given by
r 0:
r r0 :
e2C
1
,
r 10/3
e2C (r r0 )2/9 ,
35
,
4r 2
V
1
.
12(r r0 )2
(77)
184
Here
= f 4 sin2 + f 1 cos2 ,
(2).
d = (2) +
(79)
(80)
(81)
(83)
Now let v X , so that X = v = (dv/dX)(dX/d) = v dv/dX, and hence (83) becomes
vX
dv
+ v 2 + 4vX3 + 10vX + 8X4 + 12X2 = 0.
dX
(84)
(85)
1
dw
+ zw + (z 1)(2z 5) = 0.
(86)
dz
2
This is a particular case of Abels equation, but unfortunately it appears to be difficult to
obtain the solution in closed form.
w
A2(1) = v d,
A3(1) = cos d,
(87)
where v is a function of r, and (, ) are the coordinates on the 2spheres foliating the
transverse 3space. (This ansatz was used, for example, in [38].) The Hamiltonian H =
185
(88)
7. Conclusions
In the context of fourdimensional field theories, it is of considerable interest to construct
3brane configurations in Mtheory. One class of such solutions has been obtained by
wrapping M5branes on a certain supersymmetric two cycles, such as Riemannian surfaces
[30]. These solutions are deformations of an AdS5 H 2 S 4 vacuum, where H 2 is the
hyperbolic plane.
In this paper we have constructed two new types of 3brane configuration. In the first
type, we exploit the fact that the transverse space of the 3brane in D = 11 is sevendimensional, and that there exist nontrivial sevendimensional Ricciflat manifolds with
G2 holonomy. It has been proposed that compactifications of Mtheory on G2 are related to
N = 1, D = 4 YangMills theory [810]. Three explicit complete noncompact manifolds
of G2 holonomy are currently known [16,17], and they can be used to smooth out the
singularities of compact G2 orbifolds. Each of them admits a harmonic 4form [18,
19], which suggests the possibility of turning on the 4form field strength of D = 11
supergravity in solutions that correspond to deformations of fourdimensional Minkowski
space times a Ricciflat G2 manifold. We have indeed managed to obtain exact solutions
of this type, which can be viewed as 3branes in Mtheory.
The general solutions contain two continuous parameters; /, which measures the size of
the gravitional instanton, and c, which measures the strength of the 4form. The solutions
are massless, and carry no 4form charge. In common with all other known massless brane
solutions, there are naked singularities at short distance. At large distance the solution
approaches the product of fourdimensional Minkowski spacetime and the original Ricciflat G2 manifold. The solution can be viewed as a supergravitional monopole, involving
both the metric and the 4form in the supergravity multiplet.
We obtained the solutions by deriving firstorder equations from a superpotential, and
we showed that these are precisely the integrability conditions for the existence of a
Killing spinor. Thus the two M3brane solutions that we have constructed in this paper
are supersymmetric.
186
Note added
In an earlier version of this paper it was claimed that the M3brane solutions were not
supersymmetric, but instead were pseudosupersymmetric with respect to a modified
D = 11 supersymmetry transformation rule. This incorrect conclusion resulted from a
systematic error in a computer program that we used for calculating the Killing spinors.
We are grateful to Jim Liu for calculations that encouraged us to recheck the computer
programs and discover the error.
Acknowledgements
We are grateful to Mike Duff, Gary Gibbons, James Liu and Jianxin Lu for useful
discussions. C.N.P. is grateful to the Michigan Center for Theoretical Physics for
hospitality during the completion of this work.
References
[1] J. Maldacena, The large N limit of superconformal field theories and supergravity, Adv. Theor.
Math. Phys. 2 (1998) 231, hepth/9711200.
[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.
[3] E. Witten, AntideSitter space and holography, Adv. Theor. Math. Phys. 2 (1998) 253, hepth/980215.
[4] M.J. Duff, H. L, C.N. Pope, E. Sezgin, Supermembranes with fewer supersymmetries, Phys.
Lett. B 371 (1996) 206, hepth/9511162.
[5] I.R. Klebanov, E. Witten, Superconformal field theory on the threebranes at a CalabiYau
singularity, Nucl. Phys. B 536 (1998) 199, hepth/9807080.
[6] I.R. Klebanov, A.A. Tseytlin, Gravity duals of supersymmetric SU(N) SU(N + m) gauge
theories, Nucl. Phys. B 578 (2000) 123, hepth/0002159.
[7] I.R. Klebanov, M.J. Strassler, Supergravity and a confining gauge theory: duality cascades and
(chi)SBresolution of naked singularities, JHEP 0008 (2000) 052, hepth/0007191.
[8] B.S. Acharya, On realising N = 1 superYangMills in Mtheory, hepth/0011089.
[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]
[34]
187
188
Abstract
We give the explicit form of the fourdimensional effective supergravity action, which describes
lowenergy physics of the RandallSundrum model with moduli fields in the bulk and charged chiral
matter living on the branes. The relation between 5d and 4d physics is explicit: the lowenergy action
is derived from the compactification of a locally supersymmetric model in five dimensions. The
presence of odd Z2 parity scalars in the bulk gives rise to effective potential for the radion in four
dimensions. We describe the mechanism of supersymmetry breaking mediation, which relies on nontrivial configuration of these Z2 odd bulk fields. Broken supersymmetry leads to stabilization of the
interbrane distance. 2001 Elsevier Science B.V. All rights reserved.
1. Introduction
There is a growing theoretical evidence that brane models with warped geometries
may play a prominent role in understanding the variety of field theoretical incarnations
of the hierarchy problem [13]. Nonetheless, the initial hope that the mere presence of
extra dimensions would be a natural tool to control mass scales in gauge theories coupled
to gravity turned out to be premature. Various hierarchy issues have resisted numerous
proposals, and revealed another, slightly disguised, face of the finetuning, wellknown
from four dimensions (an example is the cosmological constant problem, see [4] and
references therein). The most stumbling observation is that whenever one finds a flat 4d
foliation as the solution of higherdimensional Einstein equations, which seems to be
necessary for the existence of a realistic 4d effective theory, it is accompanied by a special
choice of various parameters in the higherdimensional Lagrangian. The finetuning seems
to be even worse in 5d than in 4d, since typically one must correlate parameters living
Email address: zygmunt.lalak@cern.ch (Z. Lalak).
05503213/01/$ see front matter 2001 Elsevier Science B.V. All rights reserved.
PII: S 0 5 5 0  3 2 1 3 ( 0 1 ) 0 0 3 7 6  5
190
on spatially separated branes. Then there appears immediately the problem of stabilizing
these special relations against quantum corrections.
This situation has prompted the proposal [57], that it is a version of branebulk
supersymmetry that may be able to explain apparent finetunings and stabilize hierarchies
against quantum corrections. And indeed, the branebulk supersymmetry turns out
to correlate in the right way the brane tensions and bulk cosmological constant in
the supersymmetric RandallSundrum model. In addition, supergravity is likely to be
necessary to embed brane worlds in string theory.
Hence, there are good reasons to believe that supersymmetry is an important ingredient
of the higherdimensional unification and the quest for consistent supersymmetric versions
of brane worlds goes on, see [59,1117,1925]. In earlier papers [1215,26] the attention
has been focused on models with solitonic (thick) branes and several nogo theorems were
established (but see [17,19,20,24]). Finally, in papers [5,7,9,11] explicit supersymmetric
models with deltatype (thin) branes were constructed.
The distinguishing feature of the pure supergravity Lagrangians proposed in [7] is
imposing the Z2 symmetry, such that gravitino masses are Z2 odd. An elegant formulation
of the model is given in Ref. [11], where additional nonpropagating fields are introduced
to independently supersymmetrize the branes and the bulk. In the onshell picture for these
fields the models of Ref. [11] and Refs. [7,9] are the same.
In Ref. [11] an extension of the model to include vector multiplets has been worked out.
On the other hand, in Refs. [7,9] it has been noted that supersymmetric RandallSundrumtype models can be generalized to include the universal hypermultiplet and gauge fields and
matter on the branes. The Lagrangian of such a construction has been given in [7,9]. This
opens up a phenomenological avenue, which we follow in the present paper, and allows us
to study issues such as supersymmetry breaking and its transmission through the bulk. We
want to stress that it is impossible to perform a trustworthy research of these issues without
having a complete, explicit, locally supersymmetric model including matter on the branes
embedded in extra dimensions. This is the main drawback of the phenomenological studies
of supersymmetric brane worlds published so far. Further, it is our opinion that one should
study thoroughly the classic, and in a sense minimal, version of the 5d brane worlds where
charged matter and observable gauge interactions are confined to branes. The point is that
putting matter and gauge fields into the 5d bulk amounts extending in a nontrivial way the
very attractive MSSM. Hence, for the time being we prefer to clarify the situation in the
minimal models with just neutral fields living in the bulk. On the other hand, populating
the bulk with 5d gravitational fields alone does not seem to be realistic. In typical string
compactifications, just on the basis of simple dimensional reduction of higherdimensional
supergravities, one expects neutral (with respect to the SM group) matter, namely moduli
fields, to coexist in the 5d bulk with supergravity multiplet. To represent such bulk matter,
we choose to work in the present paper with bulk hypermultiplets, which couple also to
the branes. In the explicit calculations which shall lead us to a consistent 4d supergravity
model in the final section of this paper, we shall employ a universal hypermultiplet, which
reduces to the dilatonic chiral multiplet in 4d.
191
Still, it should be clear that putting gauge fields (and more) in the bulk is a viable
alternative to our models, and as shown in [23] it may also provide a mechanism to stabilize
the extra dimensions.
The final goal of this paper is to formulate the effective lowenergy theory that
describes properly the physics of the warped fivedimensional models with gauge sectors
on the branes. On the way to fourdimensional theory we investigate supersymmetry,
supersymmetry breakdown and moduli stabilization using fivedimensional tools. In
particular, we show that in the class of models that we consider, i.e., models without nontrivial gauge sectors in the bulk, unbroken N = 1 local supersymmetry (classical solutions
with four unbroken supercharges) implies vanishing of the effective cosmological constant.
We demonstrate the link between vanishing of the 4d cosmological constant, minimization
of effective potentials in 5d and 4d, and moduli stabilization in fivedimensional
supersymmetric models presented in this paper. We also discuss supersymmetry breaking
due to a global obstruction against the extension of bulk Killing spinors to the branes,
which is a phenomenon observed earlier in the HoravaWitten model in 11d and 5d.
First steps towards the 4d effective theory were made in [22,23] (where the Khler
function for the radion field was identified). In the setup considered in this paper,
supersymmetry in 5d is first broken from eight down to four supercharges by the BPS
vacuum wall, and then again broken spontaneously down to N = 0 by a switching on
of expectation values of sources living on the branes. The general strategy follows the one
[2730] that led to the complete and accurate description of the lowenergy supersymmetry
breakdown in the HoravaWitten models, see [27,29]. We are able to find maximally
symmetric solutions to the 5d equations of motion within our supersymmetric model,
and deduce the Khler potential, superpotential and gauge kinetic functions describing
physics of corresponding vacua in four dimensions. It turns out that the warped background
modifies in an interesting way the kinetic terms for matter fields and the gauge kinetic
function on the warped wall. There also appears a potential for the radion superfield, its
origin being a modulusdependent prefactor multiplying the superpotential on the warped
wall in the expression for the 4d effective superpotential. We do not need to introduce
any nontrivial gauge sector in the bulk to generate a potential for the T modulus. It
is interesting to note that the structure of the effective 4d supergravity is completely
different from that of the noscale models. In these the 4d cosmological constant vanishes,
while FT is undetermined and sets the supersymmetry breaking scale. In our model FT
vanishes, supersymmetry is broken by FS and nonzero cosmological constant is induced.
The complete and phenomenologically relevant 4d N = 1 supergravity model which we
managed to construct in this paper, should finally facilitate a detailed investigation of the
lowenergy physics of warped compactifications.
192
S = M3
d 5x
1
g R + 6k 2 + d 5 x gi 1 x 5 2 x 5 .
2
(1)
Threebranes of nonzero tension are located at Z2 fixed points. The ansatz for the vacuum
solution preserving 4d Poincare invariance has the warped product form:
2
ds 2 = a 2 x 5 dx dx + R02 dx 5 .
(2)
The breathing mode of the fifth dimension is parametrized by R0 . The solution for the warp
factor a(x 5 ) is:
a x 5 = exp R0 k x 5 .
(3)
It has an exponential form that can generate a large hierarchy of scales between the branes.
Matching delta functions in the equations of motion requires fine tuning of the brane
tensions:
1 = 2 = 6k.
(4)
With the choice (4) the matching conditions are satisfied for arbitrary R0 , so the fifth
dimension is not stabilized in the original RS model. Thus R0 enters the 4d effective theory
as a massless scalar (radion), which couples to gravity in manner of a BransDicke scalar.
This is at odds with the precision tests of general relativity, so any realistic model should
contain a potential for the radion field.
Relaxing the condition (4) we are still able to find a solution in the maximally symmetric
form, but only if we allow for nonzero 4d curvature (AdS4 or dS4 ) [31]. In such a case
radion is stabilized and its vacuum expectation value is determined by the brane tensions
and the bulk cosmological constant.
The RandallSundrum model can be extended to a locally supersymmetric model [5
7]. The basic setup consists of 5d N = 2 gauged supergravity [32,33] which includes the
gravity multiplet (em , A , A ), that is the metric (vielbein), a pair of symplectic Majorana
gravitinos, and a vector field called the graviphoton. The U (1) subgroup of the Rsymmetry
group is gauged, the gauge charge g being constant between two branes but antisymmetric
in the x 5 coordinate. The form of the gauging is fully characterized by the SU(2) valued
prepotential. Choosing the prepotential along the 3 direction:
3
gP = k" x 5 i 3
2 2
(5)
reproduces the bosonic part of the RS bulk action (1). Moreover, because of the
antisymmetric function "(x 5 ), the supersymmetry variation of the action contains terms
proportional to the delta function, that cancel if the brane tensions satisfy the relation (4).
Thus the finetuning present in the original RS model can be explained by the requirement
of local supersymmetry [7].
New bosonic and fermionic fields do not affect the vacuum solution so the equations
of motion for the warp factor are the same as in the original, nonsupersymmetric RS
model. The RS solution satisfies the BPS conditions and preserves one half of the
193
(7)
i 1 +
(8)
as was pointed out in [9]. This relation is necessary to cancel the variation of the action
which arise because of the presence of "(x 5 ) in the gauge coupling and is an equivalent
of (4) in a model with hypermultiplets.
The relevant part of the supersymmetry transformation laws is (we use the normalization
of [28]):
A
2
1
A
A
ab A
gP3 3 B " B ,
= " + ab " +
4
3
1 u Aa
a
Aa
u 5
= iVu 5 q "A + g k Vu "A .
(9)
2
Our objective is to show that, in the class of warped compactification, unbroken
supersymmetry implies flat 4d spacetime. To achieve this, we will derive the effective
theory of the 4d metric degrees of freedom. It will turn out that supersymmetry requires
194
(10)
which describes oscillations of the 4d metric g about some vacuum solution. Using the
ansatz (10) the action (6) can be rewritten in the form which reveals the BPS structure:
2
5
6 a
1
3
5
4
2
+ 4kR0 " x
S=M
R+ 2
P
d x g R0 a
2a 2
R0 a
+ 9k 2 huv 4 u P 2 v P 2 k u k v
1
2 huv 5 u 6kR0 " x 5 u P 2 5 v 6kR0 " x 5 v P 2
R0
5
4 a
2
+ 3k" x
45 a
P
a
4
1
5
4
2
5
+ a 24k P 2 x + a 24k P 2 x .
(11)
To show that the effective potential vanishes when supersymmetry is preserved requires
some calculations. First, A = 0 conditions can be solved in terms of the warp factor
yielding:
a
= 4kR0 " x 5 P 2 .
(12)
a
Furthermore, the a = 0 condition yields:
12 2kR0 u 1a
1a
u
k Vu P3 + Vu2a (P1 iP2 ) ,
Vu 5 q = i
4 P 2
12 2kR0 u
2a
u
Vu 5 q = i
(13)
k Vu2a P3 + Vu2a (P1 + iP2 ) .
4 P 2
Using the Killing prepotential equation it is possible to eliminate k u from (13). After some
calculations we get:
huw 5 q w = 6kR0 u P 2 .
(14)
Another manipulations of (13) yield the relation huv 5 q u 5 q v = 9R02 k 2 huv k u k v .
Summarizing, BPS conditions imply the following relations between the scalars, the
warp factor and the Killing vector and prepotential:
a
= 4kR0 P 2 ,
a
huw 5 q w = 6kR0 u P 2 ,
huv 5 q u 5 q v = 9k 2 R02 huv k u k v .
(15)
195
Plugging the above formulae into the action (11) and integrating over x 5 , we obtain the 4d
effective action for the metric g :
2
Veff ,
S4 = MPL
d 4 x 12 R
2
MPL
Veff = M 3 a 4 (0) 24k P 2 (0) 1 + a 4 () 24k P 2 () 2 ,
2
MPL
(16)
M 3 R0 dx 5 a 2 .
In Section 3 we explain in detail that if the P1 and/or P2 components of the prepotential
are nonzero at the Z2 fixed point then supersymmetry is broken. The reason is that
in such
a case the BPS solution is not global. Thus unbroken supersymmetry requires P 2 (0) =
P3 (0) (the same at x 5 = ) and, in consequence, the 4d effective potential vanishes (recall
that the brane tensions satisfy (8)).
Similar conclusions hold also in the RS model, with vector multiplets constructed in
Ref. [11]. The proof goes the same way as in the hypermultiplet case; the ultimate reason
for the vanishing of the effective potential being the supersymmetric tuning of the brane
tensions.
Summarizing this section, unbroken local supersymmetry in 5d RStype scenarios
implies flat space solutions. This result is somewhat unexpected, since within the
framework of 4d supergravities we can a priori obtain an antide Sitter solution and
preserve supersymmetry at the same time. This means that compactifications of the
supersymmetric RS scenarios yield a very special subclass of 4d supergravities. This
should be kept in mind when phenomenological models are constructed.
Another consequence of the vanishing 4d potential is that RStype models with
unbroken supersymmetry cannot incorporate a mechanism of radion stabilization. The
BPS conditions together with the supersymmetric tuning of the brane tensions yield the
4d effective potential which is identically zero. Thus the 4d equation of motion for the
M 2
= 0.
= 0, which can be satisfied by any value of R0 , since R
radion field is PL R
R0
196
197
Im( )
P2 = ,
2 V
P3 =
 2
1
1
.
4
V
(18)
We see that P1 and P2 are nonzero in this model. Thus, inducing nonzero vev of the
field will break supersymmetry. To this end, we couple (in a supersymmetric way) the fifth
derivative of to sources located on the boundaries. The rest of the bosonic action we
consider is that given in (6) rewritten for our special choice of the quaternionic manifold
and gauging:
1
1
R 3
( A )2 D 2
S = M 3 d 5 x e5
( V )2
2
4
V
4V 2
1
1
2
2
4
 
+ 6k 1 +
 
2V
2V 2
e5
 2
M 3 d 5 x 5 x 5 x 5 6k 1
V
e5
1
W
2
x 5 W1 5 + 2 x 5
d 5 x e5
V g55
M3
2
W
+ (x 5 )W2 5 + 2 x 5
(19)
+
h.c.
.
M3
In the first line we displayed the relevant kinetic terms (we neglected the field which does
not play any role in the following). The
covariant derivative acting on is D = +
ig A with the gauge charge g = 3k 2 "(x 5 ). The boundary terms in the second line are
proportional to P3 and are necessary for local supersymmetry of the 5d action [9]. The third
line contains derivative coupling of the odd field to the boundary sources W ; the presence
of singular 2 terms is commented on in Appendix B. For the sake of concreteness we have
concentrated on the case, where sources of supersymmetry breakdown are represented by
expectation values of brane superpotentials (gaugino condensates will be discussed later in
the paper).
Locally in the bulk, it is possible to find a flat BPS solution which preserves
half of the supersymmetry (see Appendix A for details), but as soon as we switch
on nonzero sources W the BPS solution does not satisfy the matching conditions
at the Z2 fixed points. In such case we search for maximally symmetric, nonBPS
solutions. We take the ansatz that allows for AdS4 foliation of the 4d metric, g =
3
3
3
diag(a 2 (x 5 )e2Lx , a 2 (x 5 )e2Lx , a 2 (x 5 )e2Lx , a 2 (x 5 ), R02 ) (the generalization to the de
Sitter foliation is straightforward); the size of the 5th dimension is parametrized by R0 . Our
vacuum has the form of a constant curvature foliation (otherwise it could lead to violation
of Lorentz invariance after integrating out the extra dimensions [41]). The ansatz for the
198
A = 0,
A5 = const.
(20)
We start with the RS solution a = ekR0 y , = 0 and treat the boundary sources as a
perturbation. This procedure is justified since we expect that the parameters of the bulk
Lagrangian, which determine the zerothorder solution, are close to the Planck scale, while
the boundary sources should be of the order of supersymmetry breaking scale. Since the
sources set the boundary value of we expect that will get (W/M 3 ) corrections. On the
other hand, the enters quadratically into the equation for the metric, so the metric will get
the correction only at the order (W/M 3 )2 . We will be able to find a solution valid to the
order (W/M 3 )2 . First we consider the linearized equation of motion for the field which
has the form:
a
+ 4 + 2igA5 + 3k 2 R0 4igkR0 A5 g 2 A25
a
1
2
W
W
= 2 (x 5 ) 3 2 (x 5 ) 3 .
(21)
M
M
The solution is:
= C"(x 5 )ek(R0 3
2i A5 )y
(22)
Matching the in the equation of motion yields the boundary conditions: (0+ ) =
1 /M 3 , ( ) = W
2 /M 3 . As a consequence:
W
2
1
W
W
(R0 3 2 i A5 )k
,
Ce
=
.
(23)
M3
M3
In the absence of supersymmetry breaking the moduli R0 and A5 could have arbitrary constant values. When we switch on the sources for the odd fields and break supersymmetry,
the expectation value of the moduli is determined by the boundary sources Wi . In other
words, the moduli are stabilized. Hence, in our model the supersymmetry breaking indeed
leads to stabilization of the fifth dimension. In the 4d effective theory R0 and A5 enter
the socalled T supermultiplet. In the next section we show, that the configuration of the
field which we have found gives rise to the 4d effective potential which is able to stabilize
T modulus when the 4d dilaton is frozen.
The nonzero value of of order W will produce a backreaction on the metric and the
scalar V of order W 2 . Keeping only terms relevant to the order W 2 , the Einstein equations
read:
2
a
1
L2
1 2
a
2
2
 
+ 3 2 + D5  6k 1 +
3 +3
a
a
V
2V
a
 2
,
= x 5 x 5 6k 1
V
2
a
L2
1 2
1
2
6
(24)
 2 = 0.
+ 6 2 D
5  6k 1 +
a
a
V
2V
C =
199
The terms involving boundary sources W gather nicely into the full square with the kinetic
term of . The boundary conditions (23) ensure that the hatted derivative contains no delta
functions. Thus the only boundary terms we have to consider are those displayed explicitly
in (24). Matching the delta functions in the first equation yields the boundary conditions
for the warp factor:
a
a
(25)
(0) = 6k 1 (0)2 ,
() = 6k 1 ()2 .
a
a
Plugging in our solution for we can satisfy the boundary conditions if the warp factor has
the form:
C2 kR0 x 5 
5
e
.
a x 5 = ekR0 x  +
2V
(26)
k C
Away from the branes the Einstein equations can be satisfied if we choose L2 = 16
6
V .
2
2
k C
, hence this solution corresponds to
This means that the 4d curvature is R = 32
2
= C"(x 5 )ek(R0 3
a = ekR0 x  +
5
2 ik A5 )x 5 
C2 kR0 x 5 
e
,
2V0
V = V0 C2 e2kR0 x  ,
5
1
2
W
W
k(R0 3 2 ik A5 )
,
Ce
= 3,
3
M
M
2 C2
k
8
L2 =
,
3 V0
C=
= 0 .
(28)
In the above solution V0 and 0 are arbitrary constants. Hence these moduli are not
stabilized. In fact, matching conditions in the equation of motion for V require V0
. This means that V0 exhibits the runaway behaviour. To achieve stabilization of all the
moduli we need to complicate our model by adding boundary sectors that couple to V .
As a crosscheck of the above results we can calculate the 4d effective potential, obtained
by integrating out the 5d bosonic action in the background (28). The result (to the order
W 2 /M 3 ) is:
M3
1
k 2 C2
2kR0
L4 = g
(29)
.
1e
R+8
k
2
V0
We denoted by g the oscillations of the 4d metric around the vacuum solution. Solving the
= 32k 2 C2 /V0 which is consistent
Einstein equations in the 4d effective theory yields R
1
2
with the value of L 12 R in (28). We also see that V0 enters the denominator of the
effective potential, which explains its runaway behaviour commented on earlier.
200
Before closing the discussion of the 5d classical solutions and supersymmetry breakdown, let us comment on proposals [17,18,36] of solving the cosmological constant problem due to supersymmetry of the bulkbrane system. To put the issue into the perspective,
let us note that the second equation in (24) does not contain second derivatives of fields,
hence it acts as a sort of constraint on the solutions of the remaining equations. This becomes more clear in the Hamiltonian approach towards the flow along the fifth dimension,
where this equation arises as the Hamiltonian constraint H = 0, and is usually used to illustrate the way the conservation of the 4d curvature L2 is achieved through the compensation
between gradient and potential terms along the classical flow. However, this classical conservation hinges upon fulfilling certain consistency conditions between brane sources, or
between boundary conditions induced by them, as illustrated by the model above. When
one perturbs the boundary terms on one wall, then to stay within the family of maximally
symmetric foliations one of two things must happen. Either the distance between branes
must change, or the source at the distant brane must be retuned. In the class of models
which we constructed, if the 4d curvature is present then supersymmetry is broken, and
does not take care of such a retuning. Furthermore, even if retuning takes place, the size
of 4d curvature, i.e., of the effective cosmological constant, does change as well; moreover, the magnitude of the effective cosmological constant has quadratic dependence on
the boundary terms which induce supersymmetry breakdown. Hence, any perturbation of
the boundary, instead of being screened by the bulk physics, contributes quadratically to
the effective cosmological constant. Of course, we are talking about perturbations that can
be considered quasiclassical on the brane. Thus we do not see here any special new effect
of the extra dimension in the cancellation of the cosmological constant. The positive aspect
of supersymmetry is exactly the one we know from 4d physics. Supersymmetry, even the
broken one, limits the size of the brane terms inducing supersymmetry breakdown, thus
limiting the magnitude of the 4d cosmological constant, since the two effects are strictly
related to each other.
201
complete kinetic terms for moduli, gauge and matter fields, which are of order (W1 )0 and
are sufficient to read off the Khler potential for moduli and matter fields. The complete
procedure consists of solving to the given order for all background fields, including the Z2 odd ones, substituting the solutions back to the 5d Lagrangian and integrating over the fifth
dimension. This procedure can be carried out to the full extent, however here, taking the
existence of the effective 4d supergravity for granted, we shall perform the integration only
for certain relevant terms the ones which give direct information about K, W and H .
4.1. Khler function for the radion and 4d dilaton
To the lowest order in field fluctuations the ansatz defining the radion modulus in the
RandallSundrum background, which is also the lowest order solution in W1 expansion, is
ds 2 = ekR0 (x
)x 5 
2
g dx dx + R02 x dx 5 .
S = V0 + i0 ,
T = k(R0 + i 2 A5 ),
(30)
(31)
where we defined
MP2 =
with =
M3
1 e2kR0 ,
k
M3
kMP2
f = 1 e(T +T )
. The form of the Khler potential for the multiplet T was previously derived
in [22,23].
To see in more detail how the argument goes, let us summarize those terms in the 5d
action that are most relevant to the forthcoming discussion:
1 A
1
3
5
5
A B
D A 3ik"(x ) PAB ,
d x e5 R
Sgrav = M
2
2
1
1
1 a D a
D
Shyp = M 3 d 5 x e5 2 V V + D
V
2
4V
1
5 q u ,
A B uAB D
(32)
2
where
5 = 5 + 3i 2 k"(x 5 ) A5 + 2(xi )W
i ,
D
the prepotential is given by
1 x2 + y2
x
y
P=
i 1 +
i 2
i 3
1/2
4
4V
2V
2V 1/2
and the SU(2) spin connection is given in Appendix A.
202
The goal is to reduce various terms in this action down to the Einstein frame in four
dimensions, where the mattersupergravity action is of the standard form of Cremmer et al.
[37,38]. After inserting the ansatz (30) into the 5d gravitational action, substituting the 5d
1 = a 1/2 ( ) , 2 = a 1/2 ( )
gravitino with the 4d zeromodes R
R 4d
L 4d (see [9,39]),
L
and integrating over x 5 one finds the following graviton and gravitino kinetic terms in 4d:
1 2
(4) 1 MP2 e4 (1 e2kR0 ) D + , (33)
MP e4 (1 e2kR0 )R
2
2
where the R0 dependent bracket is simply the scalar part of the real vector superfield f (in
what follows we shall use the symbol f for this scalar function as well). After performing
the Weyl rescaling of the graviton and gravitino
ea ea f 1/2 ,
f 1/4 ,
(34)
one arrives at the canonical action for graviton and gravitino in four dimensions.
Substituting the ansatz (30) into the 5d
action results also in (a part of) the kinetic terms for
R0 Skin = 3M 3 k()2 e2kR0 d 4 x e4 R0 R0 ; together with another term with
two derivatives on R0 generated by the Weyl rescaling, this gives exactly the kinetic
energy reproduced by the Khler potential K given in (31). A procedure of rescalings and
integrating over x 5 applied to the kinetic term of V (and ) also produces 4d kinetic terms
which are immediately seen to be exactly given by (31). It is worth noticing at this point
that, contrary to assumptions usually made, the natural 5d frame where we have coupled
bulk gravity and moduli with general gauge sectors on branes does not directly give us the
superspace frame of the 4d supergravity. This is seen from the fact, that the Weyl rescaling
in 4d is not given by the complete Khler function, but only by the part of it that depends
on the radion. Fortunately, radion and hypermultiplet moduli do not have a kinetic mixing
in the canonical 5d frame.
4.2. Effective superpotential
Knowing the Khler function and performing the reduction of the gravitino mass terms
(the ones which in five dimensions couple Z2 even components with even components)
one can identify the effective 4d superpotential
W = 2 2 W1 + e3T W2 .
(35)
One can see that the contribution to the effective superpotential given by the second
brane at x 5 = is suppressed by the factor a03 with respect to the contribution from the
Planck brane. This fits nicely the notion of a universal downscaling of all the mass scales
on the visible brane.
The derivation of the superpotential goes as follows. There are three contributions to the
4d gravitino mass terms:
1
1
3 5 1 2
1 5 2
D
Lgm = M 3 e5
5 k" x
2
2
V
3
+ iA5 k"(x 5 ) 1 2 + h.c. .
(36)
2 2
203
The first contribution comes from the term in (32) involving the part of the spin
connection , the second from the 5d gravitino mass term involving the prepotential P.
Finally the term with the graviphoton A5 comes from the SU(2) covariant derivative in
the gravitino kinetic term: D A = A + gA P A B B . Plugging in the gravitino zero
modes in the solution (28), integrating over x 5 and Weyl rescaling, we get:
1
1 1 ek(2R0 +3 2 i A5 ) R L + h.c.
L4dgm =
e
W
4
2
3/2
2 V0 f MP
1
1 + e3kR0 W
2 R L + h.c.
=
e4 W
2
3/2
2 V0 f MP
(37)
In the last step we used the boundary condition (23). In the standard 4d supergravity
formulation [37] this term has the form e24 eG/2 R L + h.c. with G = K + ln W 2 .
The form of the Khler potential (31) and the holomorphicity of the superpotential requires
the superpotential of our 4d model to be exactly (35).
The nontrivial test of the consistency of the superpotential (35) comes from minimizing
the 4d effective scalar potential derived from (31) and (35). In the standard formulation it
4
MP2 V0 3 (1 e2kR0 )3
W1 2 3e2kR0 2 + W2 2 3e4kR0 2e6kR0
2 e3kR0 i 2A5 + W2 W
1 e3kR0 +i 2A5 .
+ W1 W
(38)
Minimizing the above scalar potential with respect to R0 and A5 yields the relation (23),
consistently with the 5d picture. Also, the value of the prepotential at the minimum (that
is the cosmological constant) is consistent with the solution (28). Note also that, the 4d
effective potential is of the runaway type with respect to V0 .
Just as a remainder one should mention that the usual redefinition of the 4d components
of the gravitino leading to removal of the kinetic mixing terms between A and 5B is
needed A A + 15 5 5A . Inspecting more closely the fermionic mass matrix one
2e5
notices readily the mass terms proportional to 5 and which mix 4d gravitini with 5
and hyperini . The origin of these mass terms is analogous to that of gravitini masses, and
their presence signals that the superHiggs mechanism is at work (as expected). To see that
the higgsino is a mixture of hyperino and modulino 5 one can inspect the supersymmetry
transformation laws of these fermions:
2ie5
1
51 = 5 " 2 + i x 5 5 W " 2
V
V
1B Im( ) 2 1B
5
Re( )
5 "B ,
1
+ i2k" x
+
2 V
2 V
2ie5
1
"1
52 = + 5 " 1 i x 5 5 W
V
V
204
Re( ) 1B Im( ) 2 1B
+ i2k" x 5 1
+
5 "B ,
2 V
2 V
5 2 V
i
1
5
2
" 2 + 3k" x 5 VuA1 k u "A ,
W
= + 5 " i x
a
2V
V 1
i
2
2 = + 5 5 " 1 i x 5
W " + 3k" x 5 VuA2 k u "A
a
2V
(39)
just repeating the procedure given in [7] for the HoravaWitten model. In the above VuAb
are SU(2) vielbeins given in Appendix A. Since the theory has a mass gap <m = mKK
Ma(), to find the lowenergy goldstino one needs to substitute into above equations the
vacuum solutions for bulk scalars and to project resulting expressions onto their zeromode
components.
Another way to identify the 4d goldstino is to derive FS and FT with the help of the
4d effective Lagrangian and we shall give the result at the end of this chapter.
4.3. Gauge kinetic functions
To arrive at a realistic model, one should introduce gauge and charged matter fields. One
option with gauge fields living in fivedimensional vector multiplets was studied
in [23,39] and yields the gauge kinetic function H T . Here we present an alternative, a
model with gauge and matter fields confined on the boundaries. The action of the gauge
sector is:
e5
V
V a a 1
a
a
F
F
aD
/ a
SYM = d 5 x 5 x 5 F
F
4
4
2
e5
a
V
a F
+
4
3i V
1 a a
i
+ 5 a 5 a F5 (
)F a 5 a
4
8
4 2 e5
V a a
a
a
5 L R 5 + R L 5 + (4 fermi)
(40)
2e5
and similarly on the visible brane at . In the above the bulk fermions appear in their
even (and Majorana in the 4d sense) combinations defined as:
2
L
i1L
=
(41)
,
=
,
2
V
1
i2R
R
and brane supersymmetry is generated by the Majorana fermion
2
"
" = L1 .
"R
The treelevel gauge kinetic function turns out to be universal and equal:
H (S) = S
(42)
on either brane. However, we expect corrections to this universality (see the forthcoming
discussion).
205
2
we need to redefine the real part of the T modulus: Re T = R0 2  .
The additional check comes from the first term in (43) which after rescalings
In canonical supergravity it should equal
takes the form e4 f 1 e2kR0 D D .
2K
e4 D D and indeed it is for our choice (45) with = k/3M 3 (one should
notice that fT =0 = e2kR0 ).
The above discussion applies to matter living on the warped brane (whenever necessary
we should denote it by 2 ). To include matter living on the Planck brane (1 ) we need
to improve the Khler potential further. The expression which reproduces properly also
matter Lagrangian on the first, unwarped, brane is
3MP2 log f T + T 2 2 1 2 , (46)
K(1 , 2 ) = MP2 log(S + S)
2
k
where as before = 3M
3.
To summarize, we have deduced the zerothorder approximations to the Khler function,
superpotential and gauge kinetic functions of the 4d effective supergravity for 5d warped
RandallSundrum model with general gauge and matter sectors on the branes.
206
It turns out that the leading effects of the RandallSundrum brane tensions are encoded
in the exponential dependence of the effective Khler function on radion and matter fields,
and in the exponential suppression of the contributions to the effective superpotential borne
on the warped brane.
4.5. Gaugino condensates
In this context it is interesting to ask the question about the proper immersion of the
gaugino condensates into the effective supergravity picture. Let us discuss this issue at
the level of the 4d model we have just constructed. The basic expression for the effective
potential including the contribution from the gaugino condensates we start with is
2
1
V = eK g S S DS W + eK/2 .
4
(47)
Usually, for canonically normalized gauge and gaugino fields, one replaces the condensate
3
by 3c = MGUT
e3 Re(S)/2b0 . Using holomorphicity and Rsymmetry of the gauge sector
one often promotes this contribution to the one generated by an effective superpotential
for the dilaton superfield S. However, the question arises of what we shall substitute for
MGUT in the warped case. To answer this we should carefully recompute the condensation
scale c using the oneloop renormalization group equation for the gauge coupling and
watching the rescalings we make before reaching the 4d canonical frame. The point is
that, to achieve canonical normalization of gauginos, we must perform the rescaling
a 3/2f 3/4 . These rescalings are anomalous and amount to threshold corrections at the 4d
upper scale of running which is MP
1
1
Re(S) +
log f log a 2b0 .
(Re(S) =) 2
(48)
g (MP )
2
Now the renormalization group running looks like
1
1
MP
= Re(S) +
log f log a 2b0 2b0 log
.
g 2 (p)
2
p
(49)
(50)
Thus we see that the effective MGUT at the warped brane equals essentially MP a() (but
MGUT = MP on the Planck brane). We can substitute the condensation scale that we have
just derived into the potential (47) to obtain
2
1
2
3 3kR0 3 Re(S)/2b0
e
V = e (S + S) DS W + (S + S)MP e
4
K
(51)
207
(52)
(the factor Re(S) multiplying the exponential dependence on S can safely be neglected).
This would fit nicely with the formula for the perturbative 4d superpotential. However, one
should notice that plugging such a superpotential into the standard SUGRA expression
would generate in the potential new terms which are not present in the expression (51),
namely these containing DT Wnpert , which would be nonvanishing. This inconsistency
is alleviated if one notices that together with threshold corrections we have used, there
are suitable oneloop corrections to gauge kinetic function on the warped brane. The
threshold correction on that brane may be split into two pieces. The first is the same as on
the Planck brane and comes from the Weyl rescaling by the power of f . The f became
identified with a scalar component of a real superfield defining the Khler potential, hence
this part of the 1loop corrections is fully analogous to result of the Weyl rescaling leading
from superspace to canonical frame given by Bagger et al. [42]. The second piece is nonuniversal, and is associated with additional powers of the warp factor a multiplying the
gaugino terms on the warped brane. Since already at the level of the effective perturbative
superpotential we have found it consistent to promote a to a chiral superfield, a eT ,
then also here this part of the correction should be understood as a correction to the gauge
kinetic function on the warped brane, which now becomes
Hwarped (S, T ) = S + 2b0 T .
(53)
The supergravity model defined with (52) and (53) gives effective potential suitable to
study moduli stabilization and supersymmetry breakdown due to gaugino condensation in
the effective fourdimensional theory.
T , T; , )
=
The fourdimensional supergravity model defined by K(S, S;
k
k
2
2
2
2
MP log(S + S) 3MP log(f (T + T 3M 3 2  ) 3M 3 1  ), W = 2 2 (W1 +
e3T W2 ), Hwarped (S, T ) = S + 2b0 T and Hplanck(S) = S has been constructed as a small
perturbation around the generalized RandallSundrum background. However, it is very
likely that extrapolating this model away from the original vacuum and using it on its
own makes sense as a tool to explore phenomenology of warped compactifications at TeV
energies.
Let us summarize basic features of our model. The Fterms take at the minimum the
expectation values
2 a 2 ()W2 2 1 a 2 () 2 = 0,
F S 2 = 8eK (S + S)
F T 2 = 0
which means that supersymmetry is broken along the dilaton direction. The potential
energy at this vacuum is negative:
Vvac =
a 2 ()
8W2 2
.
V0 (M 3 /kMP2 )3 MP2 1 a 2 ()
(54)
208
a 2 ()W2 2 1
24
V0 (M 3 /kMP2 )3 (1 a 2 ()) MP4
(55)
(56)
5. Summary
The main result of this paper is the fourdimensional effective supergravity action which
describes lowenergy physics of the RandallSundrum model with moduli fields in the bulk
and charged chiral matter living on the branes.
The relation between 5d and 4d physics has been made explicit; the lowenergy action
has been read off from a compactification of a locally supersymmetric model in five
dimensions. The exponential warp factor has interesting consequences for the form of
the effective 4d supergravity. The asymmetry between the warped and unwarped walls is
visible in the Khler function, gauge kinetic functions and in the superpotential. Roughly
speaking the contributions to these functions which come from the warped wall are
suppressed by an exponential factor containing the radion superfield. This is the way the
warp factor and (and RS brane tensions) are encoded in the lowenergy Lagrangian.
We have described the mechanism of supersymmetry breaking mediation which relies
on nontrivial configuration of the Z2 odd fields in the bulk. We point out that the oddZ2 parity fields can be an important ingredient of 5d supersymmetric models. They play a
crucial role in communication between spatially separated branes.
Moreover, we have demonstrated that, after freezing the dilaton, it is possible to stabilize
the radion field in the backgrounds with broken supersymmetry and excited oddparity
fields. To achieve this we do not need to add vector fields and/or exotic charged matter in
the bulk.
We believe that the class of models we have constructed in this paper provides the proper
explicit setup to study lowenergy phenomenology of the supersymmetric brane models
with warped vacua.
Acknowledgements
This work has been supported by RTN programs HPRNCT200000152 and HPRNCT200000148. Z.L. and S.P. are supported by the Polish Committee for Scientific Research
under grant 5 P03B 119 20 (20012002).
209
hV V = 4V 2 ,
hxx = V ,
h x = 2V y,
1
,
4V 2
x2
1
hyy = + 2 ,
V
V
x
hy =
.
2V 2
h =
h = 4V 2 +  2 ,
x2
hyy = V + 2 ,
V
hy = 2V x.
h V = 0,
hxy =
xy
,
V2
(A.1)
h V = 0,
hxy = 0,
(A.2)
dq u .
(A.3)
where " and are totally antisymmetric and we choose "12 = 21 = 1. This formula
determines the vielbein (up to SU(2) transformation) to be:
1
1
y
1
1
2
3
2
dV +
d + +
dx
V=
2 2V
2 2V
2V
2V
1
x
2 dy
+ iI
(A.4)
2V
2V
or more explicitly
1
0
0
i 1
2 2V
2 2V
VV =
=
,
V
,
1
0
i 1
0
2 2V
2 2V
1
1
i
i y
i x
2V
2V
2V
2V
,
V
.
=
Vx =
(A.5)
y
i y
1
i x
i 1
2V
2V
2V
2V
0
2V
V
V =
,
2V
0
(A.6)
V =
i 2V
i 2 V
i 2 V
,
i 2V
210
V =
x
V
2
0
V
2
V =
y
V
2
0
V
2
(A.7)
1
i 3 ,
4V
x
1
i 3 1/2 i 1 ,
y =
2V
V
y
1
i 3 1/2 i 2 ,
2V
V
and the Sp(1) spin connection :
x =
(A.8)
3
i 3 ,
4V
3y 3
3x 3
i ,
i .
x =
(A.9)
y =
2V
2V
The Khler form (which for quaternionic manifolds that can occur in supersymmetric
theories is minus the curvature form) is:
V = 0,
1
1
y
i 3 ,
KV x = 3/2 i 2 +
i 3 ,
8V 2
4V
4V 2
1
x
KV y = 3/2 i 1
i 3
4V
4V 2
1
1
K x =
i 1 ,
Ky = 3/2 i 2 ,
3/2
4V
4V
1
x
y
3
1
Kxy =
(A.10)
i +
i
i 2 .
3/2
2V
2V
2V 3/2
To illustrate our discussion of supersymmetry breaking let us study the explicit
model presented in [9]. It is based on 5ddimensional supergravity with the universal
hypermultiplet and the U (1) symmetry ei gauged. The Killing vector and
prepotential corresponding to our gauging have components:
KV =
k x = y,
k y = x,
k V = k = 0,
x
y
P2 = ,
P1 = ,
2 V
2 V
1
1
 2
 2
P3 =
P 2 =
1
,
1+
.
4
V
4
V
(A.11)
Locally, even with nonzero sources for the odd field it is possible to find a BPS
solution, which preserves half of the supersymmetry. Plugging in the explicit metric and
the prepotential into the BPS conditions (15) we get:
(V )
V = 6kR0  2 ,
( )
2(yx xy ) = 0,
y
2(yx xy ) = 3kR0 x,
x
V
x
2(yx xy ) = 3kR0 y,
y +
V
(x)
(y)
 2
a
= kR0 1 +
.
a
V
It is straightforward to find the solution:
x = C1 exp 3kR0 x 5 " x 5 ,
y = C2 exp 3kR0 x 5 " x 5 ,
V = V0 C12 + C22 exp 6kR0 x 5  ,
211
(A.12)
(A.13)
= 0 .
Integrating equation for the warp factor we find:
5 1/6
a x 5 = exp kR0 x 5 V0 C12 + C22 e3kR0 x 
.
The matching condition in the Einstein equation yields:
a
a
 2
 2
(0) = kR0 1
(0) ,
() = k 1
() .
a
V
a
V
(A.14)
(A.15)
If C1 = C2 = 0, the above set is satisfied and the flat BPS solution satisfies all boundary
conditions. But as soon as we induce a nonzero value of (by coupling 5 to the
boundary sources) it is impossible to find any solution to (A.15) and supersymmetry is
broken.
(B.2)
The odd fields that are nonzero at the Z2 fixed points must have a jump there. As
a consequence their first derivatives have delta functions singularities and their second
derivatives have derivatives of the delta. The equations of motion are second order, thus we
must somehow cancel singularities by a suitable choice of the brane action. One way to
achieve this is to couple the odd fields to the sources on the boundary through their fifth
derivative. Such couplings naturally arise in Mtheory [40] and its compactification to 5d
212
[28] where the field from the universal hypermultiplet couples to gaugino bilinears and
to the superpotential of the boundary scalars.
Local supersymmetry considerably restricts the possible form of such derivative
couplings to the brane sources. We will show that couplings of the fifth derivative of an odd
field to the boundary sources r has to enter the bosonic action (6) in a form of the full
square. That is we have to replace 5 q r 5 q r 5 q r + (x 5 )( r )1 + (x 5 )( r )2 ,
in the kinetic terms of the odd fields (note that it implies that 2 singularities appear in the
Lagrangian).
Let us introduce the following coupling of an odd scalar q r to the source located on,
say, the hidden brane:
e4
L = x 5 5 5 q r r (qu ).
(B.3)
e5
We allow for the dependence of the source on the even hypermultiplet scalar (it will
turn out crucial for the consistency). Now we want to supersymmetrize (B.3). Consider the
variation of the (odd) scalar q in L :
i r A a
e4 i
r A a
" L = 5 5 (VAa
" )r .
q r = VAa
2
e5 2
(B.4)
First consider the fifth derivative acting on ". Using the fact that 5A = 5 " A this variation
can be cancelled by adding a new term to the boundary Lagrangian:
e4 i r a A
5 r .
L5 = x 5 5 VAa
e5 2
(B.5)
The variation with the 5th derivative acting on can be cancelled by the variation of the
hyperino kinetic term provided we modify the hyperino transformation law in the following
way:
a =
(x 5 ) i ab s 5 B
VbB " s .
e55 2
(B.6)
This modification has an immediate consequence that 2 terms are necessary in the brane
Lagrangian because is already present in the brane Lagrangian. Namely, varying (B.5)
we get:
2 e4 1 r ab s
V VbB 5 A " B r s
L5 = x 5
5 2 4 Aa
(e5 )
5 2 e4 1 A 5 rs
= x
5 "A h r s .
(e55 )2 8
(B.7)
Recalling, that e55 = 12 (5 A 5 "A ) the above can be cancelled by adding a singular term
to the brane Lagrangian:
2 e4 1 rs
h r s .
L = x 5
e55 4
(B.8)
This is exactly what we need to cancel the 2 singularities in the equations of motion.
213
Note that (B.3), (B.8) and the scalar kinetic term can be gathered in a full square:
u qw
LF S = e5 g huw
q
(B.9)
where we defined:
5
u
5 1 ur
u
q = q h r x .
2
(B.10)
Thus we can guess that all the terms needed to supersymmetrize the action can be found
by exchanging in the bulk Lagrangian the fifth derivative of the scalars by the hatted one.
This procedure leads to appearance of the following new terms in the brane Lagrangian:
5
i
1
r
rs
LB = e4 x
a 5 A VAa
+ A 5 B (r )A
B h s
2
4
1
a 5 b (r )ab hrs s
(B.11)
4
and in the SUSY transformation laws:
5
1
B rs
5A = (r )A
B " h s x ,
2
(x 5 )
i
s
a = 5 " B ab VbB
s 5 .
2
e5
(B.12)
Note that this procedure correctly reproduces the terms we determined before, e.g., the
correction to the hyperino transformation laws.
Nevertheless, exchanging the derivatives with their hatted counterpart does not ensure
that the action is supersymmetric. If we make the supersymmetry transformation of the
modified Lagrangian we get:
L
L u
L
u
u qw =
u qw
+
L ,
q + 2
5 q ,
5 q
5
5 q
5
u 5
w
q
5
5 q u
5q
L
L
L
w
+
5 q u + 2
5 q u
5q
u
u
w
5 q
q
q
5
5
1 ur 5
L
L
w
h r (x )
+2
5 q .
u
u
w
2
5q
5 q
5q
(B.13)
The first three terms vanish here if they vanish in the unmodified 5d supergravity. But
the last term does not vanish automatically. The easiest way to nullify this contribution
is to assume that the supersymmetry variation (hur r ) is identically zero. This can be
achieved if the function which determines the coupling of odd scalars to the boundary
can be expressed as:
r = hrs As ,
where As are constants (which determine the boundary values of the odd fields).
(B.14)
214
Let us turn to the example of the universal hypermultiplet. Eq. (B.14) implies that the
coupling of the odd scalars to the brane has the form:
y
x
y 2 xy A
x 2 xy A
Ax
Ay
Ax
Ay
1+
5 x +
1+
5 y.
L =
V
V
V
V
(B.15)
Sbulk = d x e5
V V +
2
4V 2
V
1
1
2
2
4
 
 
+ 6k 1 +
2V
2V 2
e5
V
V a a 1
a
a
F
F
aD
/ a
SYM i = d 5 x 5 x 5 xi5 F
F
4
4
2
e5
a
V
a F
+
4
a
3i V
1
i
+ 5 a 5 a F5 a F
a 5 a
4
8
4 2 e5
V
5 a L Ra 5 + a R La 5
2e5
V 3/2 a 2 2
+ (4 fermions)
+ (0)
8
5
2
 2
5 e5
55
i
W + W
Smatter i = d x 5 x xi "i 6k 1
g
V
V
e5
i
4 Wi W
+ V 3/2 W
a a
i
D i D
+ (0) 4W W
R L
i
V i
1
W
L R 2/3 W L L
V
V
i 5
(C.1)
where i = 1, 2 labels branes and "1,2 = +1, 1. In the above Z2 even 4d Majorana
fermions are defined as:
2
L
i1L
=
(C.2)
2
V
,
=
,
1
i2R
R
215
i "R ) x 5 x 5 .
= 2 V (Wi "L W
i
(C.3)
8V
p
V
for , boundary operators appear also in bulk parts of the modulini supersymmetry
transformations (these are given in the main text).
(D.90)
216
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]
[34]
[35]
[36]
[37]
[38]
217
Abstract
We comment on the recent papers by Costa et al. and Emparan, which show how one might
generate supergravity solutions describing certain dielectric branes in ten dimensions. The basic
such solutions describe either N fundamental strings or N D4branes expanding into a D6brane,
with topology M2 S 5 or M5 S 2 , respectively. Treating these solutions in a unified way, we note
that they allow for precisely two values of the radius of the relevant sphere, and that the solution with
the smaller value of the radius has the lower energy. Moreover, the possible radii in both cases agree
up to numerical factors with the corresponding solutions of the D6brane worldvolume theory. We
thus argue that these supergravity solutions are the correct gravitational description of the dielectric
branes of Emparan and Myers. 2001 Elsevier Science B.V. All rights reserved.
PACS: 04.65.+e
1. Introduction
The higherdimensional counterparts of the Melvin universe [1], which describe
gravitating magnetic fluxbranes, have appeared in various string and Mtheoretic guises
over the last few years (see, e.g., [215]). Within the KaluzaKlein context, the
most natural way to generate such fluxbranes is via a twisted compactification of flat
space [1618]. Moreover, starting instead with the Euclidean Schwarzschild solution
cross a trivial time direction, one can generate spherical branes via just such a twisted
compactification [18]. In slightly different language, one can think of this as follows:
take a Ddimensional uncharged black string; analytically continue the worldvolume
coordinates; and perform a doubledimensional reduction along a twisted direction. The
Email addresses: dominic.brecher@durham.ac.uk (D. Brecher), p.m.saffin@durham.ac.uk (P.M. Saffin).
05503213/01/$ see front matter 2001 Elsevier Science B.V. All rights reserved.
PII: S 0 5 5 0  3 2 1 3 ( 0 1 ) 0 0 3 8 2  0
219
220
the presence of this dissolved Dpbrane charge which allows for spherical solutions of the
D(p + 2)brane theory.
The existence of such spherical D(p + 2)branes with dissolved Dpbrane charge was
anticipated by Emparan in [21]. The main point of this work, however, was to analyse
spherical D(p + 2)brane solutions with dissolved F string charge. The interpretation of
such solutions was that of N BornInfeld strings [26,27] expanding into a sphere under
the influence of an external RR potential. Various arguments can then be used to identify
the stringlike solution of BornInfeld theory as a fundamental string, but it is clear that
one lacks a proper description of this process from the point of view of the F strings
themselves.
Many features of the supergravity solutions discovered by Costa et al. [19] and
Emparan [20] suggest that they should, indeed, be thought of as the proper gravitational
description of these dielectric branes for p = 4. 3 We will provide new evidence which
significantly strengthens the case for such a connection, and clarifies the relationship
between these geometries and the unstable solutions of [18].
In Sections 2 and 3 we review possible spherical solutions of the D6brane worldvolume theory due to the presence of dissolved D4brane and F string charge, respectively.
We comment briefly on the more general case of the spherical Dpbrane with dissolved
F string charge, for arbitrary p, the p = 2 case being qualitatively different to the other
examples. In Section 4, we turn to the supergravity solutions of [19,20]. To comment on
them, a brief review of their construction is necessary. We treat both the F string and
D4brane solutions in a unified way, since they have the same structure, and then compare
directly with the worldvolume analysis of Sections 2 and 3. In both cases, the solution is
consistent for precisely two values of the radius of the five or twodimensional sphere,
a point missed in the treatment of [19] because the limiting procedure used therein sends
one of the radii to infinity. Moreover, the functional form of these radii as a function of N ,
the number of D4branes or F strings, precisely matches that of the world volume calculation. That this should be the case is rather unexpected. Moreover, a brief consideration
of the smeared solutions of Emparan [20] shows that the atypical structure of the D2brane
with dissolved F string charge is also captured by the corresponding supergravity solution.
In Section 4.3, we comment on the energetics of these two consistent solutions, and
show that that with the smaller value of the radius has the lower energy, in agreement with
the worldvolume picture. We then turn to a consideration of offshell configurations,
those with arbitrary values of the radius of the sphere. In general, such solutions exhibit
conical deficits which can be viewed as deficit branes providing the tension necessary
to hold the dielectric brane in equilibrium. We then argue that the tension of these deficit
branes gives us a handle on the stability of the supergravity solutions, with the results that
the solution at the smaller radius is in fact stable at least to radial perturbations and
that the second solution is unstable. 4 These additional pieces of evidence exhibit the close
3 Corresponding, unsmeared supergravity solutions for different values of p should exist, but they cannot be
221
(2.4)
(2.5)
(2.6)
D6
where we have set the KalbRamond 2form to zero, F is the Abelian BornInfeld 2form
field strength, P [ ] denotes a pullback to the worldvolume of the spacetime fields,
T6 = 6 = ((2)6 7/2 )1 and = 2 . The ChernSimons action is a sum of terms
containing all the RR potentials, C[n] , for 1 n 7:
T6
SCS =
(2.7)
P C[7] + P C[5] F + ,
gs
D6
5 Approximation in the sense that it is not a consistent supergravity background.
222
which are the relevant couplings of these potentials to Dpbranes for 0 p 6. The Born
Infeld field strength is then chosen such that the second term in (2.7) mimics the coupling
of N D4branes to C[5] . With
F=
N 2
S ,
2
we have
(2.8)
P [C[5] ] F = NT4
T6
M5 S 2
P [C[5] ],
(2.9)
M5
R 4 + 14 2 N 2 23 f R 3 ,
V(R) =
(2.10)
gs
where we have taken the dilaton to vanish, and where Vn denotes the volume of En . If we
introduce the dimensionless worldvolume quantities
D = f R,
1
ND = f 2 N,
2
gs
4 + N 2 2 3 ,
VD =
f 2 V = D
D
4T6 V4
3 D
(2.11)
(2.12)
(2.13)
then the dimensionless potential, VD , has extrema when the following condition holds:
2
2
4
ND
= D
D
.
For ND <
(2.14)
1
2
2
2 ,
D
(2.15)
= 1 1 4ND
2
as shown in Fig. 1. There are thus two solutions describing spherical D6branes, one a
maximum of the potential, the other a local minimum. In Fig. 2(a) we show how the
stationary points of the potential change as the number of D4branes change. Fig. 2(b)
shows the value of the potential at these extrema.
2 in which case
Myers [22] considers the case where ND < 1/2 in the limit ND D
4
3
one may expand the potential (2.13) as VD ND + 2NDD 23 D
+ . The single nonzero
extremum is then at D ND so, in this limit, we lose the information about the maximum.
A similar limit was taken in the supergravity solution of [19] which again obscured the fact
that there is a second consistent solution. We shall see below that, prior to any limit being
taken, the allowed radii of the supergravity solution have precisely the same functional
form as (2.14). We believe this to be strong evidence that the worldvolume and supergravity
pictures are describing the same phenomenon.
223
Fig. 1. This plot shows the potential energy, (2.13), of the spherical D6brane for different values of
dissolved D4brane charge, ND (2.12). The top curve is for ND = 0 and the lower one for ND = 12 .
The potential has been shifted by a constant such that the energy at D = 0 is zero.
Fig. 2. Here we show how the charge, ND (2.14), and energy, VD (2.13), at the stationary points vary
with the radius of the spherical D6brane with dissolved D4brane charge.
224
(3.17)
where E is a constant electric field, and we take the constant RR 8form field strength and
corresponding 7form potential to be
F[8] = 5h M2 dr r 5 (5 ),
(3.18)
5
C[7] = h M2 r 6 (5 ).
6
(3.19)
T6 5 l
gs
5h
dt R 5 1 2 E 2 R 6 dt L,
6
(3.20)
where l is the length of the string and n denotes the volume of a unit nsphere.
Introducing the conjugate momentum, D = L/E, we find that the Hamiltonian,
(ED L), is given by [21]
T6 5 l
H=
gs
R 10
gs
+
T6 5
2
D2
5h 6
R .
6
(3.21)
To relate D to N , the number of dissolved F strings, we take R to zero to obtain a stringlike state with energy per unit length D/ = DTF (TF
being the F string tension), so we
simply associate D = N [21]. Now we use R11 = gs , 5 = 3 , and introduce the
following dimensionless quantities
F = hR,
(3.22)
NF = 8h5 R11 2 N,
H
5
VF = 83 R11 h5
= F10 + NF2 F6 .
V1
6
(3.23)
(3.24)
(3.25)
Again there is a maximum charge, above which there are no stable solutions. To calculate
it, we use the fact that at this maximum charge, the two extrema
merge into a point
of inflection, so VF = VF = 0 at the extrema. This gives NF (4/5)4 (4/5)5 . In
Fig. 3 we show how the potential changes with the number of F strings dissolved in the
D6brane. How the location of the extrema changes with NF is shown in Fig. 4(a), there
clearly being two extrema for NF less than the maximum value. This system is seen to be
qualitatively the same as the one considered in Section 2, the D4brane blowing up into a
D6brane.
225
Fig. 3. This plot shows the potential energy, (3.24), of the spherical D6brane for different values
of dissolved F string charge,
NF (3.23). The top curve is for NF = 0 and the lower one for the
maximum charge NF = (4/5)4 (4/5)5 . The potential has again been shifted by a constant to
ensure that the zero of energy is at F = 0.
Fig. 4. Here we show how the charge, NF (3.25), and energy, VF (3.24), at the stationary points vary
with the radius of the spherical D6brane with dissolved F string charge.
226
VF =
2(p1)
+ NF2
p1 p
F ,
p
(3.26)
in terms of dimensionless quantities analogous to those defined in (3.22) and (3.25) above.
The extrema of this potential are given by solutions of
2(p2)
NF2 = F
2(p1)
(3.27)
so for p > 2 there are always two extrema. The p = 2 case is qualitatively different,
however. There is no minimum in this case, the single nontrivial extremum being at
F = 1 NF2 ,
(3.28)
and this is a maximum of the potential. We will see that the relevant (smeared) supergravity
solution describing this configuration has a similar structure.
sin d
+ f d +
ds = H
ds M
r d f
2
f 2
2/d dr
2
2
2
2
2
f
f
= (3, 6) and (6, 3) for the M2 and M5branes, respectively, and where
where (d, d)
H =1+
k sinh2
r d
f =1
k
r d
l 2 cos2
r d l 2 r d2 k
,
=
1
.
f =
r2
r d
For the M2brane, the 3form potential is given by
k sinh 2
M cosh d l sin2 d ,
r d H
whereas for the M5brane it is
k sinh
l2
l
cos3 1 2 cosh d 2 d S 2 .
A[3] =
r
r
A[3] =
(4.30)
(4.31)
(4.32)
The location, rH , of the Euclidean horizon or bolt is given by the zero of f, so that
2
d2
rH l 2 ,
k = rH
(4.33)
6 The standard versions of which are to be found in [31,32].
227
l
.
2 l2)
cosh (rH
(4.34)
2k cosh
R11 =
(4.35)
= gs .
d1 (d 2)l 2 r d3
dr
H
H
Finally, we must also consider the quantization of M2 and M5brane charge which gives
(4.36)
where wehave used the fact that the elevendimensional Planck length, lP , is given by
1/3
lP = gs
. N denotes the number of M2 or M5branes, and the constant cd is
cd =
(2)2d/3
,
d
(4.37)
d+1
+B ,
(4.38)
the fixed point set of the corresponding isometry being {r = rH , = 0}. To dimensionally
reduce, one identifies points separated by a distance of 2R11 along integral curves of K.
Introducing a new angular coordinate = B , which has standard 2 periodicity and
is constant along orbits of K, gives K = / . Then the tendimensional string frame
metric is
2
dr
2
ds 2 = 1/2 H 3/d ds 2 Md1 + 1/2 H 13/d
+ r 2 d 2 + cos2 dd1
f
+ 1/2 H 13/d fr 2 sin2 d 2 ,
(4.39)
where
f
Bkl cosh sin2 2
(Br sin )2 ,
+H
=f 1+
f
r d f
(4.40)
(4.41)
H (d 2)l 2 rH 1
1 dr
.
2 l2)
2 l + B cosh (rH
(4.42)
7 We will not write down the dimensional reduction of the 3form potentials (4.31) and (4.32) since they are
superfluous to our discussion.
228
Then by imposing
B=
H + (d 2)l
1
dr
=
,
R11
2rH (rH + l) cosh
(4.43)
sin
ds 2 = H 2/3 ds 2 M2 + f d +
r p f
6p dr 2
f 2
1/3
2
2
2
2
2
2
(R11 )2
,
V6p
(4.45)
where
cp =
(2)4
.
p1+p
(4.46)
Dimensional reduction proceeds as above, and gives rise to a D6brane with topology
M2 E6p S p1 , so that T duality along the flat transverse directions gives the desired
solution describing N F strings expanding into a Dpbrane with topology M2 S p1 [20].
Again, the potential conical singularity is avoided if B is given by (4.43), up to the
replacement of d with p.
229
d1
d2
lrH
l 2 rH
+ kB cosh rH
1
(d 2)2 (1 BR11 )2 k 2
4
d
2d
+ (d 2)(1 BR11 )2 + 1 rH
k + (1 BR11 )2 1 rH
= 0.
(4.48)
We may now solve this quadratic for k and substitute into (4.36). We note that the
requirement k 0 gives the restriction
0 BR11 1,
(4.49)
so that B = 1/R11 is the maximum magnetic field [18]. However, for the tendimensional
KaluzaKlein picture to be valid, we need all tendimensional length scales to be larger
than the compactification scale, R11 . The magnetic field gives just such a length scale, 1/B,
so the solution should have a reasonable tendimensional description only for BR11 1.
A more detailed examination of the region in which (the nearcore limit of) these solutions
is tendimensional can be found in [19]. At any rate, we now find
d
k = rH
,
(4.50)
d 2
1
1+
,
A=
we have
2 cd2
1
A2d
2
R11
2d/3
(4.52)
rH
N +
AR11
2
2d2
rH
AR11
2d4
= 0.
(4.53)
230
Fig. 5. In order to find rH for a given set of parameters, R11 , B and N we need the dimensionless
functions A(BR11 ), (BR11 ). (We plot A1 (BR11 ) for clarity.) The scales on the left of the graphs,
and the solid lines, correspond to d = 3 (D4 D6) and the righthand scales, and dashed lines,
correspond to d = 6 (F1 D6).
The functions A(BR11 ) and (BR11 ) which control this equation, are plotted in Fig. 5. By
introducing the dimensionless quantities
rH
,
AR11
cd
d/3
NS =
N,
2
Ad R11
S =
(4.54)
(4.55)
2(d2)
NS2 = S
2(d1)
(4.56)
which is of precisely the same form as (2.14) for d = 3, and (3.25) for d = 6. Although
we cannot reproduce the correct numerical factors for any value of BR11 , it is surprising
enough that the form of the equations are the same when one considers that the
worldvolume calculation did not make use of a consistent supergravity background.
In Fig. 6 we show how the radius, S , of the (d 1)sphere depends on NS , the
dimensionless quantity defining the (d 2)brane charge of the solution. There are various
noteworthy points. For NS less than some critical value, there are always two values of rH ,
which we denote by r+ and r (where r+ > r ). As long as the F string or D4brane
charge is not too large, then, there are two distinct supergravity solutions of the form
(4.39), for which the spacetime has no conical singularities. In the following subsection,
we will see that the solution with rH = r has the lower energy of the two. As we decrease
the charge, r+ increases, and r decreases. For zero charge essentially the situation
considered by Dowker et al. [18] there is only a single nontrivial value of rH , this
231
Fig. 6. Analogous plots to Figs. 2 and 4 for the supergravity case without conical singularities. They
show how the charge, NS (4.56), and energy, ES (4.64), vary as the radii of the static solutions
change. The solid line is for d = 3 (D4 D6) and the dashed line is for d = 6 (F1 D6).
NS2 = S
2(p1)
(4.57)
which is of the same form as (3.27). The p = 2 case is simpler since the equation analogous
to (4.48) is then only linear in k. We have
2
,
k = 1 (1 BR11 )2 rH
(4.58)
so that, in terms of the relevant dimensionless quantities, we find
232
S =
1 NS2 ,
(4.59)
which is of the same form as (3.28). We note that this single consistent solution corresponds
to the larger of the two radii, rH = r+ , in the case of general p. As we will see below, it is
only the smaller radius which is stable, so it would seem that F strings cannot expand into
a stable spherical D2brane. At any rate, the atypical properties of the spherical D2brane
with dissolved F string charge are precisely captured by the corresponding (smeared)
supergravity solution. We expect that a localized solution should exist, and that it should
also exhibit the same behaviour.
4.3. Energetics and stability
Following [19], to compute the energy of the tendimensional solutions we make use
of the background subtraction method [33], where the relevant background is the k = 0
solution: the standard Melvinlike flux 7brane solution [9] written in spherical oblate
coordinates. That is to say we define the zero of energy to be the k = 0 solution. We
also need to transform the metric (4.39) to the Einstein frame for this calculation, using
dsE2 = e/2 dsS2 . Specifically [33]
1
8
8
E = 2
(4.60)
d x h N K d x h0 N0 K0 ,
210
where hij is the metric induced on a constant r slice of the constant time slice, h is
its determinant and the integration is performed as r . The lapse function, N , and
extrinsic curvature, K, of the constant r slice, are given by
N=
gt t ,
1
K = hij r hij ,
grr
(4.61)
d+1
Vd2
2 3
(2)4 R11
d sinh2 + d + 2 k.
(4.62)
ES =
cd (2)2(2d/3) 3
d2
Ad Vd2 R11
gives
ES = Sd2
E,
2 2
1 + S ,
d
(4.63)
(4.64)
8 The energy of these dielectric branes is thus slightly higher than the corresponding energy of the standard
black F string and D4brane solutions, which have a factor of (d + 1) rather than (d + 2) [34].
233
2
d1
rH
l + B cosh rH
2
(2)4 R11 3
d2
d+1
Vd2 R11
TS
(2)4 3
d+1
Vd2
l 2 (1 a)
(4.65)
which, as promised, is proportional to the deficit angle. TS is the mass of the deficit brane
in terms of the dimensionless quantities. It is interesting to note that we can derive this
expression using different techniques. In particular, we have considered methods analogous
to those used in [36] to compute the effective energymomentum tensor of the conical
singularity between two collinear black holes in four dimensions; and these methods give
exactly the same result for the mass T .
To analyse the form of (4.65) further, we fix B and N , and substitute for l, and
k using the expressions (4.33), (4.35) and (4.36). Of course, we no longer have access
to the expression (4.43) for B, since this was derived to ensure the absence of conical
singularities. Indeed, we can no longer solve for or k analytically, and must proceed
using numerical methods. The plot of the mass, TS , as a function of S is shown in Fig. 7,
for both the F string and D4brane solutions.
As argued previously, a deficit brane with positive tension indicates that the dielectric
sphere wants to expand, and negative tension implies that it wants to contract. We see from
Fig. 7, then, that at rH = r+ the sphere is unstable to both expansion and contraction,
whereas at rH = r the sphere is stable. The sphere at rH = r+ would decay either into
the sphere at rH = r , or into a sphere which expands forever. We should note that these
arguments apply only to radial perturbations.
234
Fig. 7. This plot shows the dimensionless mass (4.65) of the deficit brane for d = 3 (solid line,
lefthand scale), d = 6 (dashed line, righthand scale), and for some specific values of B and N . The
inset is a magnification of the region around rH = r . The arrows show whether the sphere wants to
expand or contract as one moves away from the zeros of TS .
5. Discussion
We have provided strong evidence that the supergravity solutions of Costa et al. [19]
and Emparan [20] are describing the dielectric effect whereby F strings or D4branes
expand into spherical D6branes. We have presented the two supergravity solutions in a
unified manner showing that, in both cases, there are two distinct static solutions which
are free of conical singularities. Moreover, the radius of these solutions has precisely the
same functional form as that implied by the worldvolume analyses of Emparan [21] and
Myers [22]. Of course, since the solutions in question do not preserve any supersymmetry,
one should ask whether they would be modified by higher derivative and/or string loop
corrections. It seems likely that they would be so modified, and presumably the spherical
worldvolume solutions discussed above would also be affected by such corrections.
Certainly one cannot trust the solutions in regions of very large curvature.
On the supergravity side we have seen that the energy of the rH = r solution was
lower than that with rH = r+ , although this alone does not guarantee stability. To address
the latter issue, we considered the tension of the deficit brane formed by static solutions
away from rH = r , which indicated that the rH = r solution was stable and that with
rH = r+ was unstable. Thus we see that the supergravity solutions capture the worldvolume
properties with remarkable accuracy.
To further strengthen the interpretation of the supergravity solutions, we considered the
special case of N F strings expanding into a D2brane, for which the worldvolume analysis
indicates that no stable dielectric solution should exist. Although the supergravity solution
for this configuration was smeared in certain transverse directions, we have seen that indeed
there was no stable solution.
235
Acknowledgements
We would like to thank Miguel Costa, Bert Janssen and Simon Ross for useful
discussions and comments on a draft of this paper. We are also indebted to Roberto
Emparan for explaining the stability arguments and urging us to look at the smeared
solutions. D.B. is supported in part by the EPSRC grant GR/N34840/01 and PMS by
PPARC.
References
[1] M.A. Melvin, Pure magnetic and electric geons, Phys. Lett. 8 (1964) 65.
[2] G.W. Gibbons, K. Maeda, Black holes and membranes in higherdimensional theories with
Dilaton fields, Nucl. Phys. B 448 (1988) 741.
[3] J.G. Russo, A.A. Tseytlin, Constant magnetic field in closed string theory: an exactly solvable
model, Nucl. Phys. B 448 (1995) 293, hepth/9411099.
[4] J.G. Russo, A.A. Tseytlin, Magnetic flux tube models in superstring theory, Nucl. Phys. B 461
(1996) 131, hepth/9508068.
[5] A.A. Tseytlin, Closed superstrings in magnetic flux background, Nucl. Phys. Proc. Suppl. 49
(1996) 338, hepth/9510041.
[6] J.G. Russo, A.A. Tseytlin, GreenSchwarz superstring action in a curved magnetic Ramond
Ramond background, JHEP 9804 (1998) 014, hepth/9804076.
[7] R. Emparan, Composite black holes in external fields, Nucl. Phys. B 490 (1997) 365, hepth/9610170.
[8] C. Chen, D.V. Galtsov, S.A. Sharakin, Intersecting Mfluxbranes, Grav. Cosmol. 5 (1999) 45,
hepth/9908132.
[9] M.S. Costa, M. Gutperle, The KaluzaKlein Melvin solution in Mtheory, JHEP 0103 (2001)
027, hepth/0012072.
[10] G.W. Gibbons, C.A.R. Herdeiro, The Melvin universe in BornInfeld theory and other theories
of nonlinear electrodynamics, Class. Quantum Grav. 18 (2001) 1677, hepth/0101229.
[11] P.M. Saffin, Gravitating fluxbranes, grqc/0104014.
[12] M. Gutperle, A. Strominger, Fluxbranes in string theory, hepth/0104136.
[13] J.G. Russo, A.A. Tseytlin, Magnetic backgrounds and tachyonic instabilities in closed
superstring theory and Mtheory, hepth/0104238.
[14] J.F. Sparks, KaluzaKlein branes, hepth/0105209.
[15] P.M. Saffin, Fluxbranes from pbranes, hepth/0105220.
[16] F. Dowker, J.P. Gauntlett, D.A. Kastor, J. Traschen, Pair creation of dilaton black holes, Phys.
Rev. D 49 (1994) 2909, hepth/9309075.
236
[17] F. Dowker, J.P. Gauntlett, G.W. Gibbons, G.T. Horowitz, The decay of magnetic fields in
KaluzaKlein theory, Phys. Rev. D 52 (1995) 6929, hepth/9507143.
[18] F. Dowker, J.P. Gauntlett, G.W. Gibbons, G.T. Horowitz, Nucleation of pbranes and
fundamental strings, Phys. Rev. D 53 (1996) 7115, hepth/9512154.
[19] M.S. Costa, C.A.R. Herdeiro, L. Cornalba, Fluxbranes and the dielectric effect in string theory,
hepth/0105023.
[20] R. Emparan, Tubular branes in fluxbranes, hepth/0105062.
[21] R. Emparan, BornInfeld strings tunneling to Dbranes, Phys. Lett. B 423 (1998) 71, hepth/9711106.
[22] R.C. Myers, Dielectric branes, JHEP 12 (1999) 022, hepth/9910053.
[23] W. Taylor, M. Van Raamsdonk, Multiple Dpbranes in weak background fields, Nucl. Phys.
B 573 (2000) 703, hepth/9910052.
[24] S.P. Trivedi, S. Vaidya, Fuzzy cosets and their gravity duals, JHEP 0009 (2000) 041, hepth/0007011.
[25] M.R. Douglas, Branes within branes, hepth/9512077.
[26] C.G. Callan, J.M. Maldacena, Brane dynamics from the BornInfeld action, Nucl. Phys. B 513
(1998) 198, hepth/9708147.
[27] G.W. Gibbons, BornInfeld particles and Dirichlet pbranes, Nucl. Phys. B 514 (1998) 603,
hepth/9709027.
[28] R.G. Leigh, DiracBornInfeld action from Dirichlet sigma model, Mod. Phys. Lett. A 4 (1989)
2767.
[29] M. Li, Boundary states of Dbranes and Dystrings, Nucl. Phys. B 460 (1996) 351, hepth/9510161.
[30] M. Green, J.A. Harvey, G. Moore, Ibrane inflow and anomalous couplings on Dbranes, Class.
Qauntum Grav. 14 (1997) 47, hepth/9605033.
[31] M. Cvetic, D. Youm, Rotating intersecting Mbranes, Nucl. Phys. B 499 (1997) 253, hepth/9612229.
[32] C. Cski, J. Russo, K. Sfetsos, J. Terning, Supergravity models for (3 + 1)dimensional QCD,
Phys. Rev. D 60 (1999) 044001, hepth/9902067.
[33] S.W. Hawking, G.T. Horowitz, The gravitational Hamiltonian, action, entropy, and surface
terms, Class. Quantum. Grav. 13 (1996) 1487, grqc/9501014.
[34] M.J. Duff, H. Lu, C.N. Pope, The black branes of Mtheory, Phys. Lett. B 382 (1996) 73,
hepth/9604052.
[35] A. Vilenkin, E.P.S. Shellard, Cosmic Strings and other Topological Defects, CUP, 1994.
[36] M.S. Costa, M.J. Perry, Interacting black holes, Nucl. Phys. B 591 (2000) 469, hepth/0008106.
Abstract
We calculate the integrated trace anomaly for a real spin0 scalar field in six dimensions in a
torsionless curved space without a boundary. We use a pathintegral approach for a corresponding
supersymmetric quantum mechanical model. Weyl ordering the corresponding Hamiltonian in phase
space, an extra twoloop counterterm 18 (R + g ij kil ljk ) is produced in the action. Applying a
recursive method we evaluate the components of the metric tensor in Riemann normal coordinates in
six dimensions and construct the interaction lagrangian density by employing the background field
method. The calculation of the anomaly is based on the endpoint scalar propagator and not on the
string inspired centerofmass propagator which gives incorrect results for the local trace anomaly.
The manipulation of the Feynman diagrams is partly relied on the factorization of fourdimensional
subdiagrams and partly on a brute force computer algebra program developed to serve this specific
purpose. The computer program enables one to perform index contractions of twelve quantum fields
(10 395 in the present case) a task which cannot be accomplished otherwise. We observe that the
contribution of the disconnected diagrams is no longer proportional to the twodimensional trace
anomaly (which vanishes in four dimensions). The integrated trace anomaly is finally expressed in
terms of the 17 linearly independent scalar monomials constructed out of covariant derivatives and
Riemann tensors. 2001 Elsevier Science B.V. All rights reserved.
PACS: 04.62.+v; 31.15.K; 02.40.K; 02.70
Keywords: Weyl anomaly in six dimensions; Riemannian geometry; Computational techniques; Path integral
methods
1. Introduction
The trace anomaly, namely the breaking of classical conformal invariance of gravity
actions under Weyl rescaling of the metric:
Email addresses: ahatzini@tem.uoc.gr (A. Hatzinikitas), portugal@lncc.br (R. Portugal).
05503213/01/$ see front matter 2001 Elsevier Science B.V. All rights reserved.
PII: S 0 5 5 0  3 2 1 3 ( 0 1 ) 0 0 0 9 8  0
238
(1)
has a long history with numerous applications and implications to highenergy physics,
general relativity and statistical mechanics. The literature on this subject is vast and in
this brief review we will only concentrate on those aspects which are associated with the
present problem. For a historical review the reader is advised to consult [1].
Interesting different methods have been developed to investigate and calculate Weyl
anomaly in four and higher dimensions. The authors in [2] were able to express the
integrated trace anomaly in four dimensions as a linear combination of two invariants,
the square of Weyl tensor
Cij kl C ij kl = Rij2 kl 2Rij2 + 13 R 2
(2)
(3)
239
1 l
i1 i2 ik P i1 i2 ik s k ,
k!
(5)
k=2
where (il1 i2 ik )P are the generalized Christoffel symbols at the point P and the geodesics
through P which are straight lines are defined in terms of the arc length s by
y l = l s.
(6)
240
(7)
Paraphrasing Eq. (7), one can state that all symmetric derivatives of the affine connection
vanish at the origin in RNC.
In general a covariant secondrank tensor field on a manifold can be expanded according
to
1
Tk1 k2 ()i1 i2 in .
) = Tk1 k2 () +
Tk1 k2 (
(8)
n! i1 i2
in
n=1
The coefficients of the Taylor expansion are tensors and can be expressed in terms of the
l
components Rmnp
of the Riemann curvature tensor and the covariant derivatives Dk Tlm
l
and Dk Rmnp . Without much effort one can prove that
(i1 i2 in1 iln )k
n1
l
l
D(i1 Di2 Din2 R
=
in1 kin ) + (i1 i2 in2 in1 k in )
n+1
l
(i1 i2 in3 in2
Rin1 kin ) l , k
l
(i1 i2 in4 in3
Din2 Rin1 kin ) l , k
..
.
(i1 i2l Di3 Din2 R
in1 kin ) l , k ,
(9)
where the interchange of covariant and contravariant indices act independently and
symmetrization acts only on i indices. Expression (9) reproduces for various values of
n 1 the following results:
l
(i1 i2l )k = 13 R
(i1 ki2 ) ,
l
(i1 i2 i3l )k = 12 D(i1 R
i2 ki3 ) ,
2 l
l
(i1 i2 i3 i4l )k = 35 D(i1 Di2 R
i3 ki4 ) + 9 R(i1 i2 Ri3 i4 )k ,
l
l
(i1 i2 i3 i4 i5l )k = 23 D(i1 Di2 Di3 R
i4 ki5 ) D(i1 Ri2 ki3 Ri4 i5 ) ,
l
(i1 i2 i3 i4 i5 i6l )k = 57 D(i1 Di4 R
i5 ki6 )
1
l R
l
5 7D(i1 Di2 R
i3 i4 i5 i6 )k + D(i1 Di2 Ri3 ki4 Ri5 i6 )
16 l
iki Di4 R
l
+ 32 D(i1 R
i5 i6 ) 45 R(i1 i2 Ri3 i4 Ri5 i6 )k ,
2 3
..
.
(10)
The coefficients of (8) can be rewritten as:
241
+ (i1 i2 in2 ) in1
k1 Din ) Tk2 + k1 k2
..
.
(11)
Expressions (9) and (11) compose the building blocks of the current recursive method
which produces the following results for different values of n:
(i1 i2 ) Tk1 k2 = D(i1 Di2 ) Tk1 k2 13 R
(i1 k1 i2 ) Tk2 + k1 k2 ,
(i1 i2 i3 ) Tk1 k2 = D(i1 Di2 Di3 ) Tk1 k2
+ k1 k2 ,
(i1 i2 i3 i4 i5 ) Tk1 k2 = D(i1 Di2 Di3 Di4 Di5 ) Tk1 k2
10
3 R(i1 k1 i2 Di3 Di5 ) Tk2 + 5D(i1 Ri2 k1 i3 Di4 Di5 ) Tk2
ik i Di5 ) Tk2 + 2 D(i1 Di2 Di3 R
+ 3D(i1 Di2 R
i4 k1 i5 ) Tk2
3
3 1 4
R
23 D(i1 R
i2 k1 i3 i4 i5 ) Tk2
ik i R
+ D(i1 R
i4 i5 ) Tk2 + k2
2 1 3
Tk2 + k2 R
+ 23 D(i1 R
i4 i5 )k1 k1 k2 ,
i2 i3
(i1 i2 i3 i4 i5 i6 ) Tk1 k2
= D(i1 Di6 ) Tk1 k2
+ (i1 i5 i6)k1 Tk2 + 6(i1 i4 i5k1 Di6 ) Tk2
+ 15(i1 i3 i4k1 Di5 Di6 ) Tk2 + 20(i1 i2 i3k1 Di4 Di6 ) Tk2
+ 10(i1 i2 i3k1 i4 i5 i6 ) Tk2 + k2
+ 5(i1 i2k1 i3 i5 i6 ) Tk2 + k2
242
+ 15 (i1 i2k1 i3 i4 i5 i6) T k2 + k2 + k2
(12)
k1 i4 i5 k2 + 2 Di1 R
k1 i2 i3 R
(i1 i5 ) gk1 k2 = 43 Di1 Di3 R
(13)
i4 i5 k2 + k1 k2 .
On the other hand, for n = 6 one gets:
(i1 i2 i3 i4 i5 i6 ) gk1 k2 =
k1 i5 i6 )k2
Di4 R
+ 34
7 D(i1 Di2 Rk1 i3 i4 Ri5 i6 )k2 + k1 k2
10
7 D(i1
55
7 D(i1 Rk1 i2 i3 Di4 Ri5 i6 )k2
16
l
7 Rk1 (i1 i2 Ri3 i4 l Ri5 i6 )k2 .
(14)
Thus, plugging into (8) expressions (13) and (14) we end up with the following expansion
of the metric tensor in RNC:
k1 i1 i2 k2 i1 i2 + 1 Di1 R
k1 i2 i3 k2 i1 i3
gk1 k2 = gk1 k2 + 2!1 23 R
3!
i
i4
1
k1 i3 i4 k2 + 8 R
m
+ 4!1 65 Di1 Di2 R
9 k1 i1 i2 m Ri3 i4 k2
i
i5
1
k1 i4 i5 k2 + 2 Di1 R
k1 i2 i3 R
+ 5!1 43 Di1 Di3 R
i4 i5 k2 + k1 k2
17
+ 6!1 10
7 Di1 Di4 Rk1 i5 i6 k2 + 5 Di1 Di2 Rk1 i3 i4 Ri5 i6 k2 + k1 k2
i
8
i6
1
l
+ 11
2 Di1 Rk1 i2 i3 Di4 Ri5 i6 k2 + 5 Rk1 i1 i2 Ri3 i4 l Ri5 i6 k2
+ O i1 i7 .
(15)
The expression (15) will play a crucial role in calculating the contribution steming from
the quadratic Christoffel symbol term in the interaction Lagrangian. It is also in perfect
agreement with the closed formula for these coefficients which are encoded in the integral
representation of [15].
2 For a manifold having odd dimensionality, one cannot form a scalar out of an odd number of derivatives.
3 We perform the usual redefinition of the scalar fields
= g 1/4 which leaves the Jacobian invariant under
the similarity transformation g 1/4 .
243
for
(x)/
(y) and the regulator R
where J is the Jacobian
(y) = f (x)
of the fields
consistent anomalies is uniquely determined [17]. For local Weyl anomalies and for real
scalar fields one finds that
=H
1 h 2 R(x)
R
= 12 g 1/4 (x)
p i g ij (x)g
1/2 (x)
pj g 1/4 (x)
12 h 2 R(x),
g(x)
= det gij (x).
(17)
The second term in (17) is the wellknown improvement potential term and the
dimensionless coefficient
(n 2)
=
4(n 1)
takes the value (n = 6) = 15 in the present case. The Hamiltonian is Einstein invariant
commutes with the generator
which means that H
) = 1 p i i (x)
G(
+ i (x)
p i
2 ih
1
= lim
0 (2 h
)n/2
dx0i
n
i i S int /h
g x0 f x0 e
,
(18)
i=1
1
=
2 h
1
=
h
gij (x0 + q) gij (x0 ) q i q j + bi cj + a i a j d
6 Rii1 i2 j q
+
+
+
q +
i1 i2
1
i1 i2 i3
12 Di1 Rii2 i3 j q q q
i1
i4
3
8
k
5! Di1 Di2 Rii3 i4 j + 9 Rii1 i2 k R i3 i4 j q q
i
i5
4
1
k
6! Di1 Di3 Rii4 i5 j + 2 Di1 Rii2 i3 k R i4 i5 j + i j q q
5
17
k
7! Di1 Di4 Rii5 i6 j + 5 Di1 Di2 Rii3 i4 k R i5 i6 j + i j
i
i6
8
1
k
k
l
+ 11
2 Di1 Rii2 i3 k Di4 R i5 i6 j + 5 Rii1 i2 k R i3 i4 l R i5 i6 j q q
4 The expectation value means that all quantum fields must be contracted using the appropriate
propagators. A bar over the various geometrical quantities indicates that they depend exclusively on x0i .
244
+ O q i1 q i7 q i q j + bi cj + a i a j d,
S2int
1
= h
8
1
= h
8
(1 4 )R + g ij l ki k lj (x0 + q) d
+ q i1 Di1 R
+
(1 4 ) R
(19)
0
1
1 i1
3! q
1 i1 i2
2! q q Di1 Di2 R
+
q i3 Di1 Di3 R
1 i1
4! q
q i4 Di1 Di4 R
+ g ij i1 l ik i2 k lj q i1 q i2
+ 2!1 g ij i1 i2 l ik i3 k lj + i1 l ik i2 i3 k lj q i1 q i3
i j
i1 i2 i3 l ik i4 k lj 1 g ij i1 i2 l ik i3 i4 k lj
13 R
4
ij 1
l
g 3! i1 i3 ik i4 k lj + i1 l ik i2 i4 k lj
q i1 q i4 + O q i1 q i5 d.
(20)
As emphasized in [12], the above expression is the continuum limit of a rigorous discrete
result.
In the action (19), (bi , cj ) and i is a set of anticommuting and commuting LeeYang
ghosts, respectively [11]. Their existence is imposed by the integration over the momenta
l i2 ki3 ) ,
(i1 i2 i3 )k = 12 D(i1 R
l i ki )
(i1 i2 i3 l i4 )k = 35 D(i1 Di2 R
3 4
(21)
l (i i R
i i )k .
+ 29 R
1 2
3 4
(22)
(23)
The symmetrization of the various indices in the above identities is understood with the
inclusion of a 1/n! factor. All other extra terms in (19) and (20) are produced by expanding
the metric in Riemann normal coordinates with the help of (15).
Propagators are derived in closed form through the discretised configuration space path
integrals via a midpoint rule. In this way ambiguities arising from products of distributions
are resolved. Taking the continuum limits of the propagators one can read off the Feynman
rules. One then finds that ( ) is to be considered as a Kronecker delta ij and the
245
propagators depend on the discrete Heaviside function ii = 1/2. The continuous twopoint Green function may also be determined by the equation
2
( ) = ( ),
2
subjected to the boundary conditions on the interval [1, 0] (endpoint approach):
(0, ) = (1, ) = (, 0) = (, 1) = 0.
(24)
(25)
(26)
The propagators of the various fields are proportional to h and given by:
q i ( )q j ( ) = h g ij (, ),
q i ( )q j ( ) = h g ij + ( ) = h g ij (, ),
q i ( )q j ( ) = h g ij + ( ) = h g ij (, ),
q i ( )q j ( ) = h g ij 1 ( ) = h g ij (, ),
(27)
(28)
(29)
(30)
b ( )c ( ) = 2 h g ( ),
(31)
a ( )a ( ) = + h g ( ).
(32)
ij
ij
All other Wick contractions of the fields vanish. Regarding the vertices they may be read
directly from the continuum S int given by (19) and (20).
1.
1 5
I1 =
h 7!
0
3 1
5
) 120
7! ( h
ii5 i6 j +
Di1 Di4 R
17
5
ii3 i4 k R
k i5 i6 j + i j
Di1 Di2 R
ii1 i2 k R
k i3 i4 l R
l i5 i6 j
+ 85 R
q i1 q i6 q i q j + bi cj + a i a j ( ) d
+
11
k
2 Di1 Rii2 i3 k Di4 R i5 i6 j
2
+ 2Da D a R
R
5 The recipe the diagrams have been drawn is: a line closing on a single vertex is represented by a Green
function G(, ) with one integration variable while a line connecting two vertices stands for G(, ) and
indicates two integration variables.
246
1
60
ab + Da Db D a Dc R
ab + Da Db R
bc
Da Db R
1
120
1
120
1
120
ab + Da Db Dc D b R
ac + Da Db Dc D a R
bc
+ Da Db R
34 ab
3 abcd
dacb
ab D c D d R
Rabcd + 2R
5 R Rab + 2 R
abcd Dd Db R
abcd D(e Dd) R
ac + 12R
e abc
+ 2R
2
11
3
2
d
2 (Da Rbc ) + 4Da Rbc D Rdcab + 2 (Da Rbcde )
abcd D e R
ebcd + 3D a R
bcde De R
bcda
+ 3Da R
8 b ac
a edc R
bcde
ab R
+ 9R
5 Rab Rc R
e a f c R
bedf + 7 R
ab ef cd
abcd R
R
2 abcd Ref R
1
= ( h )3 7!5 120
12K4 + 12K5 + 4K6 2K7 + 8K8
34
5 (2K4 + 2K5 + 4K6 2K7 + 8K8 + 3K10 + 5K11 K12 )
11
2
2 (11K13 10K14 + 3K15 ) + 5 (2K4 + 9K6 + 7K7 + 2K8 ) ;
(33)
2.
1
I2 = h
8
0
4! (1 4 )Di1
= ( h )3 18
1 R
i i1 i2 j i3 l ik i4 k lj
Di4 R
3
+ 14 g ij i1 i2 l ik i3 i4 k lj
+ 3!1 g ij i1 i3 l ik i4 k lj + i1 l ik i2 i4 k lj
i1
i4
q q ( ) d
1
720 (1 4 )
2
2Da R
ab Db R
2R
ab Da Db R
3 R
1 1 ab cde
Rbcde
540 2 R Ra
abcd R
a e c f R
bedf
+ 13 R
abcd R
ab ef R
cdef
+ 13
12 R
1
bc Dd R
ab d c + 1 D a R
bc 2
Da R
1728
17280
5
abcd De R
e bcd
2304 Da R
1
480
1259 abcd ef
Rab Rcdef
518400 R
1 abcd
ac
Dd Db R
7200 R
7 abcd
Rabcd
3200 R
7 abcd
e abc
De Dd R
1600 R
209 abcd e f
Ra c Rbedf
129600 R
bcde
Da R
2
a bc
1
5760 D R Dc Rab
a bcde
1
bcda
De R
360 D R
47 ab cde
Rbcde
86400 R Ra
23 abcd
e abc
Dd De R
4800 R
= ( h )3 18
163
43200 K6
1
1800 K12
91
43
259200 K7 8100 K8
7
11
1440 K13 + 2160 K14
13
2880 K11
1
3600 K16
1
1200 K17
(34)
3.
247
0 0
1 2
1 1 2
ii2 i3 j q i1 q i3 q i q j + bi cj + a i a j ( )
I3 =
Di1 R
12
2 h
1 1
ki5 i6 l q i4 q i6 q k q l + bk cl + a k a l ( )
Di4 R
ii1 i2 j q i1 q i2 q i q j + bi cj + a i a j ( )
+ 5!1 R
i
i6
3
ki5 i6 l + 8 R
m
Di3 Di4 R
9 ki3 i4 m R i5 i6 l q q
q k q l + bk cl + a k a l ( ) d d
= 12
3 1
1
bcde D e R
bcda
) 10 Da R
144 ( h
2
abcd
15 Da Rbc Dd R
1
ab c
48 Da R D Rcb
5!
1 108
1
abcd D e R
ebcd
10 Da R
1
2
30 (Da Rbc )
1
ab
48 Da R Db R
1
2
192 (Da R)
15
abcd +
abcd R
R
72
ab
15 Rab R
b acde
432
Rbcde
30 Da D R
144
bcad R
cd
15 Da Db R
64
b d eaf c
15 Rabcd R e f R
64 b ca
15 Rab Rc R
R
32 R
432 a
bcde R
acde
30 D Db R
144
adbc
15 Da Db Rcd R
a Db R
ab + 4 R(
R
ab )2
3RD
3
192
bcde R
ea
5 Rabcd R
224
abef R
cd ef
15 Rabcd R
R
abcd )2
+ 2R(
1 11
1
1
1
= ( h )3 12 144
30 K13 3 K14 + 10 K15 48 K16
1
1
1
1
23
+ 5!1 108
K2 + 72
K3 + 27
K4 + 15
K5 135
K7 +
1
1
+ 10
K10 + 16 K11 30
K12 ;
4.
1
I4 =
8
0 0
+
1
12 (1 4 )
2
1
20 (Da Rbcde )
32
135 K8
ii1 i2 j q i1 q i2 q i q j + bi cj + a i a j ( )
R
1 1
i3 q i4 ( )
Di3 Di4 Rq
1
48 K9
(35)
248
ii1 i2 j q i1 q i2 q i q j + bi cj + a i a j ( )
R
g mn i3 l mk i4 k ln q i3 q i4 ( )
1
ii2 i3 j q i1 q i3 q i q j + bi cj + a i a j ( )
(1 4 ) Di1 R
+ 12
i4 ( )
Di4 Rq
mi3 i4 n + 8 R
k
+ 5!3 (1 4 ) Di1 Di2 R
9 mi1 i2 k R i3 i4 n
1
6
d d
q i1 q i4 q m q n + bm cn + a m a n ( )R
= ( h )3 18
2
1 1
1 1
60 24 RR 6 144 R(Rabcd )
1 1
ab 1
2
60 24 Da R Db R + 48 (Da R)
1
1
1
ab 1 R(
R
ab )2
200 24 RR + 12 RDa Db R
27
1
31
1
1
= ( h )3 18 5400
K2 21600
K3 + 900
K9 + 1440
K16 ;
5.
2
1
18 R(Rabcd )
(36)
0 0
1 h 2
i1 ( ) Di2 Rq
i2 ( )
I5 =
(1 4 )2 Di1 Rq
2 8
1 1
( )R
+ q i1 q i2 Di1 Di2 R
g ij i1 l ik i2 k lj q i1 q i2 ( ) d d
+ 2(1 4 )R
1 2 1
2 1
2
1
a 1
= ( h )3 128
5
12 Da RD R + 6 RR 5 36 R(Rabcd )
1
7
1
= ( h )3 8!1 63
5 6 K9 + 12 K16 2 K3 ;
3
+
+
+
6.
+
+
(37)
3 0 0 0
1
1
ii1 i2 j q i1 q i2 q i q j + bi cj + a i a j ( )
I6 =
R
1296 h
111
ki3 i4 l q i3 q i4 q k q l + bk cl + a k a l ( )
R
m n
m
n
m
n
mi5 i6 n q q q q + b c + a a () d d d
R
1
= ( h )3 1296
1 3
64 R
37 ab cd
360 R R Racbd
R
ab )2
18 R(
i5 i6
2
3
40751 ab cde
a Rbcde
16 (Rabcd ) R 10080 R R
20911 abcd ef
cdef
R ab R
10080 R
137 ab c
2016 R Rb Rca
409 abcd e f
R a c Rbedf
280 R
2
7.
1
1
64 K1 8 K2
409
280 K8 ;
3
16 K3
137
2016 K4
37
360 K5
40751
20160 K6
20911
10080 K7
(38)
1 1
(1 4 )
I7 =
576 h
249
0 0 0
111
ii1 i2 j q i1 q i2 q i q j + bi cj + a i a j ( )
R
ki3 i4 l q i3 q i4 q k q l + bk cl + a k a l ( )
R
d d d
R
2
2
1 1 1 2
1
1
= ( h )3 576
5 R 16 R 6 (Rab ) 4 (Rabcd )
1
1
1
= ( h )3 14
8! 16 K1 6 K2 4 K3 ;
(39)
8. [ ]2
0 0 0
3 1 21
2
ii1 i2 j q i1 q i2 q i q j + bi cj + a i a j ( )
(1 4 )
R
I8 = h
3! 8 6
1 1 1
R
d d d
R
= ( h )3 8!1 21
10 K1 ;
(40)
9. [ ]3
0 0 0
1 h 3
3
R
R
d d d
I9 =
(1 4 )
R
3!
8
1 1 1
21
= 18 200
K1 .
(41)
2
1
16 R(Rabcd )
1 2
24 R(Rab )
1
+ 12
R.
250
1
R.
24
(42)
1
ki1 i2 l R
mi3 i4 n
R
+
72( h )2
q q
i3 i4
0 0
q i1 q i2 q k q l + bk cl + a k a l ( )
11
( h )2 2
R ,
q q + b c + a a ( ) d d +
1152
m n
m n
m n
(43)
1 2 2
1 2
R 432
= ( h )2 12 24
(Rab ) ,
1 2 2
,
= ( h )2 12 24
R
produce a vanishing result.
terms created by the interaction
The same observation holds for the R
0
1 1
1
i1 i2 i j
i j
i j
IR =
q q q q + b c + a a d h
Rii1 i2 j
R
h 6
24
(44)
1 1 .
+ = h R
24
24
One making use of the identities (9)(24) of Appendix A and the 17 linearly independent
terms listed below:
K1 = R 3 ,
K4 = Rab Rcb R ac ,
K7 = Rabcd R abef R cd ef ,
K10 = Rab R ab ,
K13 = (Da Rbc )2 ,
K16 = (Da R)2 ,
K2 = R(Rab )2 ,
K5 = R ab R cd Rcabd ,
K8 = R abcd Rb ef c Reaf d ,
K11 = Rabcd R abcd ,
K14 = D a R bc Db Rac ,
K17 = 2 R,
K3 = R(Rabcd )2 ,
K6 = Rab R acde R b cde ,
K9 = RR,
K12 = R ab Da Db R,
2
K15 = D a R bcde ,
(45)
(46)
251
where
I (x0i )
=
( h )3
1 7
8! 225 K1
98
45 K2
14
15 K3
661
108
9 K8 5 K10
1
6
5 K16 5 K17 .
2701
324 K4
191
12 K11
6047
405 K5
16
3 K12
18413
648 K7
98
3 K14
2K15
159421
3240 K6
127
3 K13
(47)
L := q i1 ( ), q i2 ( ) ,
qdot1 ( ), qdot1 ( ) , b1 ( ), c1 ( ) , a 1 ( ), a 1( ) ,
252
q i3 ( ), q i4 ( ) ,
Next command finds all independent 6 index configurations that contribute to the final
result.
> Lic := ind_config([i1,i2,mu1,nu1,i3,i4,mu2,nu2]):
> N := nops(Lic);
N := 105
In the case of 8 field indices, there are 105 independent index configurations. We show the
first 3 ones.
> for i to 3 do evaln(Lic[i])=Lic[i] od;
Lic1 = [i1, i2, 1, 1, i3, i4, 2, 2]
Lic2 = [i1, i2, 1, 1, i3, 2, i4, 2]
Lic3 = [i1, i2, 1, 1, i3, 2, 2, i4]
Next command is a loop over the 105 index configurations. The results are stored in a table
of results called res.
> for i to N do
> expr := WickContractions(L, Lic[i]):
> res[i] := Vertex(tensor_coeff*expr):
> od:
Command WickContractions performs the Wick contractions given by Eqs. (27)
(32). We show an example with the fourth index configuration.
> expr := factor(WickContractions(L, Lic[4]));
2
expr := 4 hbar4 g i1 i2 g 2 2 g 1 i3 g 1 i4 (1 + ) + Heaviside( )
Command Vertex simply multiplies the result of Wick contractions to the tensor
coefficient, simplifies the tensor expression, and isolates the terms to be integrated, since
we noticed that, for the most complicate vertices, Maple spends more time integrating
Dirac delta and step functions than simplifying the tensorial terms.
> Vertex(tensor_coeff*expr);
2
4
hbar4 R 2 1 R2 1 , (1 + ) + Heaviside( )
Next command adds the results of all index configurations after integrating over variables
and . Functions int_tau and int_sigma replace the usual Maple integrator, since
Maple int command fails to return the correct value for complicated vertices.
> final_value := simptensor(add(res[i][1]*
6 The number of independent index configurations is given by n!/((n/2)!2n/2 ) where (n/2)! represents the
possible permutations of index pairs and 2n/2 the permutation of indices within the pairs.
253
> int_tau(int_sigma(res[i][2])),i=1..N));
final_value :=
4
4 1 2
1 2 4
1 4
R1 2
16 R hbar 6 hbar R
1 4
4 2 i1 _b0 _c0
4 hbar R
R2 i1 _b0 _c0
1 2
2 i1 _b0 _c0
hbar4 4
For vertices I2 and I6 , the method described above has some extra complications. Vertex
I6 has a large number of index configurations (N := 10395) and vertex I2 uses huge side
identities obtained in Riemann normal coordinates as described in Appendix A.3.
This computer algorithm can easily be extended to higher dimensions to predict the
trace anomalies in 8 and 10 dimensions. Of course there are strong limitations implied by
the exponential growth of the problem which can be relaxed by increasing the available
computing power. The tensorial upper bound of the program is reached by the product of
eight Riemann curvature tensors but such a case is beyond the scope of the present work.
6. Conclusions
In this article we calculate the integrated trace anomaly for a real spin0 scalar field
living on a curved sixdimensional manifold. This is achieved by relying on a recursive
computation of the metric tensor components in Riemann normal coordinates. One can use
the general formulae (9) and (11) to reach the desired order of metric expansion induced
by the dimensionality of the manifold. Adopting the path integral formalism of quantun
mechanical nonlinear sigma models, we evaluate all the vertices and the corresponding
Feynman diagrams that contribute to the present anomaly. A computer based program
is used to perform the otherwise inevitable task of integration over distributions and
contractions of the various tensors involved. The final result derived in this process involves
17 scalar monomials consisting of covariant derivatives and/or Riemann tensors.
Acknowledgements
We are grateful to F. Bastianelli for his critical comments, S. Frolov for bringing
to our attention their work [19] and K. Lake for advising A.H. to contact R.P. for the
computational manipulation of the problem.
Appendix A
A.1. Useful identities
We consider a riemannian (or pseudoriemannian) manifol equipped with its standard
(metric compatible, torsion free) connection. The Riemann curvature is defined as:
R a bcd = c a bd + e bd a ec c d,
(A.1)
254
(A.2)
pair symmetry:
Rabcd = Rbadc ;
(A.3)
cyclic symmetry:
Rabcd + Radbc + Racdb = 0;
(A.4)
(A.5)
(A.6)
(A.7)
R = Rq q .
(A.8)
In this paper the following identities have been exploited to simplify our expressions:
2
Rabcd R acbd = 12 Rabcd
,
(A.9)
= 12 Rabcd R ebcd ,
Rabcd R
(A.10)
e c b f 7 abef
e(af )b (d c
cd
1
R e) = 8 Rabcd R af R ed + 2 R
Ref ,
Ra(bc)d R
(A.11)
ab
1 2
Da Db R = 2 R,
(A.12)
Da Db R ab = 12 2 R 12 (Da R)2 2Rab D a D b R + 2Rab Rcb R ac 2R ab R cd Racbd
4Da Rbc D b R ac + 3(Da Rbc )2 + 12 Rabcd R abcd
2Rabcd R e af c R be df + 2R ab R cde a Rbcde
ecbd
ac
bc
Da Db Dc D R
Da Db Dc D R
(A.13)
(A.14)
1 2
2 R
2
a b
b ac
1 2
1
2 R 2 (Da R) 2Rab D D R + 2Rab Rc R
2R ab R cd Racbd 3Da Rbc D b R ac + Rab R ab + 2(Da Rbc )2
(A.15)
(A.16)
255
R D Dd Rca
b =R
ab
Rab
(A.17)
ab c
1 ab
2 R Da Db R + R Ra Rcb
R R Racbd , (A.18)
ab
cd
abcd
(A.19)
(A.20)
1 abcd
Rabcd ,
4R
1
2
2 (Da Rbcde ) ,
(A.21)
(A.22)
(A.23)
2R R Racbd + 2R
ab
cd
ab
Rac Rbc .
(A.24)
(A.25)
(A.26)
Solve for the desired object. In the same vein we can prove (A.10) as well as
Da R abcd D e Recbd = 12 (Da R abcd )2 and R abcd Racbd = 12 R abcd Rabcd .
Proof of (A.11). Expanding out the products, the resulting sum can be expressed in terms
of the invariants:
L1 = Rabcd R abef R c e c f ,
L3 = Rabcd R e f a c R f de b ,
L4 = Rabcd R e af c R b ed f ,
(A.27)
L3 = 14 L2 ,
L4 = L3 + Rabcd R e af c R b f d e ,
(A.28)
= R abcd Rcdef R ef ab
256
16(d 1) ab
24
R Rbc Rac +
R abcd Rac Rbd ,
3
(d 2)
(d 2)2
(A.29)
(d 2 + d 4)
3(d 2 + d 4)
3
3
2
2
RRabcd
R
RRab
+
(d 1)2 (d 2)3
(d 1)(d 2)3
2(d 1)(d 2)
2(3d 4)
3d
3
Rabcd R abce Red
+
Rab R bc Rca +
Rabcd R ac R bd
d 2
(d 2)3
(d 2)2
+ Rabcd R ebcf R a ef d .
(A.30)
Expressions (A.29) and (A.30) reproduce in d = 6 the unique already known Weyl
invariant polynomials of [5] constructed only out of Weyl tensors. In higher dimensions
the independent ways to contract indices among a number of Weyl tensors increases.
Another invariant made out of covariant derivatives of the Weyl tensor is
Inv3 = R 3 + 8RR ab Rab + 2RR abcd Rabcd 10R ab Rbc Rca 10Rab R acdb Rcd
+ 12 RR 5R ab Rab + 5R abcd Rabcd .
(A.31)
The final nontrivial 7 invariant (typeA) we would like to consider stems from the Euler
form which exists in any even dimension d = 2n:
1
:a1 a2 a2n R a1 a2 R a2n1 a2n
E2n =
(4)n n!
M
1
1
=
:a1 a2 a2n : b1 b2 b2n R a1 a2 b1 b2 R a2n1 a2n b2n1 b2n dV
(4)n n! 2n
M
1
=
(A.32)
E2n dV ,
(4)n n!
M
where
1
:a a a : b1 b2 b2n R a1 a2 b1 b2 R a2n1 a2n b2n1 b2n
2n 1 2 2n
1
= n det abii R a1 a2 b1 b2 R a2n1 a2n b2n1 b2n , i = 1, . . . , 2n,
(A.33)
2
is the Euler number which is a total divergence, vanishes in all lower integer dimensions
and dV = e1 e2n = g d2n x is the volume form. In six dimensions the Euler
number becomes:
E2n =
Inv4 = E6 = R 3 12RR ab Rab + 16R ab Rbc Rac + 24R ab R cd Racbd + 3RR abcd Rabcd
+ 4R abcd Rab ef Rcdef 24R ab Ra cde Rbcde 8R abcd Ra e c f Rbedf .
(A.34)
7 The terminology trivial invariants is adopted to justify the existence of covariant total derivatives of
polynomials over Riemann tensor and its covariant derivatives. These invariants coincide with variations of all
independent local counterterms to effective action.
257
i i1 i2 j i3 ikl i4 ljk = 4 R
i i1 i2 j R
l
k
13 R
(ik)i3 R(lj )i4 .
27
(A.35)
e
e
= 12 D(a R
Expanding out the identity (a b c)d
bdc) we get:
l + Di2 R
l + Dk R
l
i1 i2 kil + i1 k i2l i + i2 k i1l i = 14 Di1 R
i2 ik
i1 ik
i1 ii2
l + Di1 R
l + Di2 R
l .
+ Dk R
i2 ii1
kii2
kii1
(A.36)
Notice that the third term of I2 is symmetric under the interchange of the lower indices
of the Christoffel symbols namely k and l . So in (A.36) we interchange
these indices and add the two expressions together. We then obtain:
2 i1 i2 kil + i1 (k i)il 2 + i2 (k i)il 1
l + Di2 R
l + Di1 R
l + Di2 R
l + Dk R
l + Dk R
l
= 14 Di1 R
kii2
kii1
iki2
iki1
i2 ii1
i1 ii2
(A.37)
l + Di R
l
.
+ Di R
i2 ki1
i2 ki1
We would like now to express the second and third terms of the lefthand side
l
of (A.37) in terms of i1 i2 k
and covariant derivatives of the Riemann curvature.
Starting from
l = i1 i2 l i1 i l
Di1 R
ii2 k
ik
ki2
l
l
l
i1 (i k)i
= i1 i2 (ik)
Di1 R
(ii2 k) ,
2
(A.38)
one has
l
l
Di2 R(ii
.
i1 (k i)il 2 + i2 (k i)il 1 = 2i1 i2 kil Di1 R(ii
2 k)
1 k)
(A.39)
258
Finally the expression we are going to use for the contribution of the third term in the
action I2 is:
1
l + Di2 R
l + Di1 R
l + Di2 R
l + Dk R
l
i1 i2 kil = 24
Di1 R
kii2
kii1
iki2
iki1
i2 ii1
l
l
l
+ Dk Ri1 ii2 + Di Ri2 ki1 + Di Ri2 ki1
l
l
1
+ Di2 Riil 1 k + Di2 Rki
+ 6 Di1 Riil 2 k + Di1 Rki
(A.41)
.
2i
1i
and a similar relation holds for the term i3 i4 lik .
References
[1] M.J. Duff, Twenty years of the Weyl anomaly, Class. Quantum Grav. 11 (1994) 1387.
[2] S. Deser, M.J. Duff, C.J. Isham, Nonlocal conformal anomalies, Nucl. Phys. B 111 (1976) 45;
M.J. Duff, Comment on quantum gravity and world topology, QMC7629, 1976;
M.J. Duff, Observations on conformal anomalies, Nucl. Phys. B 125 (1977) 334.
[3] B.S. DeWitt, Dynamical Theory of Groups and Fields, Gordon and Breach, New York, 1965,
first published, in: Relativity, Groups and Topology, Les Houghes Summer School, 1964;
B.S. DeWitt, Phys. Rep. C 19 (1975) 297;
L.S. Brown, Stress tensor trace anomaly in a gravitational metric: scalar field, Phys. Rev. D 15
(1977) 1469;
L.S. Brown, C.J. Cassidy, Stress tensor trace anomaly in a gravitational metric: general theory,
Maxwell field, Phys. Rev. D 15 (1977) 2810;
A. Duncan, Conformal anomalies in curved spacetime, Phys. Lett. B 66 (1977) 170;
J.S. Dowker, R. Critchley, The stresstensor conformal anomaly for scalar and spinor fields,
Phys. Rev. D 16 (1977) 3390;
A.E.M. van de Ven, Indexfree heat kernel coefficients, hepth/9708152.
[4] A. Hatzinikitas, K. Schalm, P. van Nieuwenhuizen, Trace and chiral anomalies in string and
ordinary field theory from Feynman diagrams for nonlinear sigma models, Nucl. Phys. B 518
(1998) 424.
[5] L. Bonora, P. CottaRamusino, C. Reina, Conformal anomaly and cohomology, Phys. Lett.
B 126 (1983) 305;
L. Bonora, P. Pasti, M. Bregola, Weyl cocycles, Class. Quantum Grav. 3 (1986) 635.
[6] S.A. Fulling, R.C. King, B.G. Wybourne, C.J. Cummins, Normal forms for tensor polynomials:
the Riemann tensor, Class. Quantum Grav. 9 (1992) 1151.
[7] P.B. Gilkey, The spectral geometry of a Riemann manifold, J. Differential Geom. 10 (1975)
601.
[8] I.G. Avramidi, A covariant technique for the calculation of the oneloop effective action, Nucl.
Phys. B 355 (1991) 712.
[9] S. Deser, A. Schwimmer, Geometric classification of conformal anomalies in arbitrary
dimensions, Phys. Lett. B 309 (1993) 279.
[10] L. AlvarezGaum, E. Witten, Gravitational anomalies, Nucl. Phys. B 234 (1983) 269.
[11] F. Bastianelli, P. van Nieuwenhuizen, Trace anomalies from quantum mechanics, Nucl. Phys.
B 389 (1993) 53.
[12] J. de Boer, B. Peeters, K. Skenderis, P. van Nieuwenhuizen, Loop calculations in quantum
mechanical nonlinear sigma models with fermions and applications to anomalies, Nucl. Phys.
B 446 (1995) 211, Nucl. Phys. B 459 (1996) 631.
[13] L. Eisenhart, Riemannian Geometry, Princeton Univ. Press, Princeton, NJ, 1965;
V.I. Petrov, Einstein Spaces, Pergamon, Oxford, 1969;
[14]
[15]
[16]
[17]
[18]
[19]
259
A.D. Dolgrov, L.B. Khriplovich, Normal coordinates along a geodesic, J. Gen. Rel. Grav. 15
(1983) 1033.
L. AlvarezGaum, D.Z. Freedman, S. Mukhi, The background field method and the ultraviolet
structure of the supersymmetric nonlinear model, Ann. Phys. 134 (1981) 85.
U. Mller, C. Schubert, A.M.E. van de Ven, A closed formula for the Riemann normal
coordinate expansion, J. Gen. Rel. Grav. 31 (1999) 1759.
K. Fujikawa, Pathintegral measure for gaugeinvariant fermion theories, Phys. Rev. Lett. 42
(1979) 1195;
K. Fujikawa, Comment on chiral and conformal anomalies, Phys. Rev. Lett. 44 (1980) 1733;
K. Fujikawa, Energy momentum tensor in quantum field theory, Phys. Rev. D 23 (1981) 2262;
K. Fujikawa, Phys. Rev. Lett. 42 (1979) 1195.
A. Diaz, W. Troost, A. Van Proeyen, P. van Nieuwenhuizen, Int. J. Mod. Phys. A 4 (1989) 3959.
R. Portugal, The Riegeom package: abstract tensor calculations, Comput. Phys. Commun. 126
(2000) 261268.
F. Bastianelli, S. Frolov, A.A. Tseytlin, Conformal anomaly of (2, 0) tensor multiplet in six
dimensions and AdS/CFT correspondence, JHEP 013 (2000) 0002;
F. Bastianelli, O. Corradini, 6D trace anomalies from quantum mechanical path integrals, Phys.
Rev. D 63 (2001) 065005.
Abstract
We investigate the 1/Q2 powersuppressed corrections to structure functions in the flavoursinglet
channel of deep inelastic lepton scattering arising from renormalon insertions into an initialstate
gluon, as obtained using the dispersive approach. The pinchtechnique is used as a convenient tool in
the separation of contributions. 2001 Elsevier Science B.V. All rights reserved.
PACS: 12.38.t; 12.38.Cy; 13.60.Hb
1. Introduction
We have learned much about the strong interaction of particle physics from deep
inelastic scattering (DIS) experiments over many years. This process is continuing, with
present and future experiments generating more data over larger regions of phase space.
Alongside this ongoing experimental work, theoretical developments continue to be
made. QCD, the established theory of strong interactions, continues to pose challenges:
perturbation theory has been relatively successful at high energies, but the series expansion
even here is at best asymptotic. In the nonperturbative rgime lattice techniques are
making advances. Yet a clean distinction between perturbative and nonperturbative
cannot be made: all QCD observables involve some interplay between them.
Although the structure functions (for example) cannot be calculated using perturbative
QCD, the general shape of their asymptotic Q2 behaviour is well known: the observed
Bjorken scaling at high Q2 is violated by additional smaller terms. The dominant scaling
violation is a logarithmic Q2 dependence, originating in the scale dependence of the parton
Email address: graham.smye@mib.infn.it (G.E. Smye).
1 Research supported by the EU Fourth Framework Programme, Training and Mobility of Researchers,
Network Quantum Chromodynamics and the Deep Structure of Elementary Particles, contract FMRXCT980194 (DG12  MIHT).
05503213/01/$ see front matter 2001 Elsevier Science B.V. All rights reserved.
PII: S 0 5 5 0  3 2 1 3 ( 0 1 ) 0 0 3 9 3  5
261
density functions of the incident hadron. This can be calculated using perturbative QCD
and used to measure the strong coupling s .
In addition to the logarithmic scaling violations, there are known to be contributions
behaving as inverse powers of the hard scale, i.e., as 1/Qn . These include both corrections
due to the nonzero hadron mass M, which are suppressed by a factor M 2 /Q2 ,
and nonperturbative powersuppressed terms arising from highertwist operators in the
operator product expansion. Such contributions are not included in fixedorder perturbative
calculations, yet they are known to be important over the wide Q2 range of available data.
Over a number of years the highertwist powersuppressed terms of a wide variety of
observables have been estimated using two related approaches, the renormalon (see [1]
for a review) and dispersive [2] models, although it is the dispersive approach that is
used here. If we consider graphs with an arbitrary number of loop insertions in a gluon
propagator, we assume that we can reconstruct a welldefined effective strong coupling
at the scale of the gluon virtuality. The difference between this true coupling and
that reconstructed using fixed order perturbation theory gives rise to nonperturbative
corrections, which are typically powerbehaved. For a given observable, the shape of the
leading correction can be found; an estimation of its magnitude requires the additional
assumption of universality. Thus, starting from perturbative QCD, we aim to investigate
the transition to the nonperturbative region.
These approaches have been applied to various QCD observables, with the assumption
of universality approximately holding [3]. In flavour nonsinglet DIS there are results for
structure functions [4,5], fragmentation functions [6] and event shape variables [7], while
studies have been made of power corrections to structure functions [8,9] and fragmentation
functions [10] in the flavour singlet channel.
The flavour singlet contribution to DIS is that involving the interaction of the gluons
within the hadron. (We do not study the fermionic singlet, arising from the total sum of
quark distributions, since these behave just like the nonsinglet contribution [4,5].) Power
corrections are calculated using the renormalon or dispersive models by making insertions
into the gluon propagators as shown in Fig. 1. Here the lower part of the diagram represents
the finding of a virtual gluon within the proton, and the upper part represents photongluon
fusion.
In the normal perturbative treatment of DIS, the asymptotic freedom of QCD enables
us to treat the initialstate partons as free particles confined within the nucleon; so in
a singlet calculation we would start with a free gluon and convolute the perturbative result
with the gluon distribution function g(x). We cannot however do this in a calculation
of power corrections, since the models we use consider modifications to the gluon
propagator (loop insertions in the renormalon model, or, equivalently, a mass in the
dispersive approach). We therefore consider the initialstate gluon to be generated by some
perturbative mechanism, the simplest of which is by radiation from an onshell parton.
We immediately encounter a very serious problem: if the gluon is not on shell, both
the upper and lower parts of this diagram are gaugedependent. Gauge independence is
achieved only when we include all the elements of some subset of diagrams of a given order
in s , and this includes diagrams that cannot be separated into the two halves of Fig. 1:
262
diagrams which do not have analogous gluon propagators in which to make insertions or
from which to take a scale for the running coupling.
This is precisely the same problem encountered in the twoloop calculations of power
corrections to event shapes in e+ e annihilation, as discussed in [11], except that the
gluon is now in a different channel. We might thus expect a similar solution: namely,
the use of the pinch technique [12] to generate the running coupling at the scale of the
gluon virtuality [13] in squared diagrams with two exchanged gluons, and the remainder
of diagrams contributing to the power correction via some asyet unknown mechanism.
The first studies [810] of these singlet quantities considered the incoming virtual gluon
to be radiated from a quark line, as in Fig. 2. This is the simplest case, since the lower half
of the diagram is gauge independent, as is the upper half to leading order, O(s ).
There is however another problem to be overcome, concerning the interpretation of the
lower half of the diagram. We may try to recover the singlet contribution to the power
corrections by deconvoluting the full result with the quarktogluon splitting function (this
is the approach taken in [8]), or we may leave the result as it is and interpret it as a genuine
secondorder contribution. These two interpretations give very different predictions for the
magnitudes of the powersuppressed corrections. I argue below (in Section 6) that the latter
approach should be adopted.
In Section 2 we discuss the flavour singlet contribution to deep inelastic scattering. The
dispersive approach to power corrections is then briefly reviewed in Section 3. Section 4
examines in detail the virtual gluon production represented by the bottom half of Fig. 1,
and the application of the pinch technique to this, while in Section 5 we calculate the
contribution to the structure functions from photongluon fusion. Section 6 investigates
the distribution of virtual gluons in the proton, paying particular attention to the 1/Q2
power corrections. Finally a summary is given in Section 7.
263
,
FL (x) = 0,
q
are the quark and antiquark distributions in the target hadron. Thus
at this level there is no contribution arising from the gluon distribution in the hadron.
The O(s ) contributions are most easily given as the distribution in the finalstate
variable = P r/P q, (0 1):
d
s 2 d
Fi (x) =
eq
)
CF Ci,q (, ) q(x/ ) + q(x/
d
2 q
x
+ TR Ci,g (, )g(x/ ) ,
1
(2.3)
264
where g(x) is the gluon distribution in the target hadron, CF = 4/3, TR = 1/2 and the
coefficient functions Ci,j (, ) [14] are
2 + 2
+ 2 + 2,
(1 )(1 )
CL,q (, ) = 4 ,
2 + (1 )2
,
CT ,g (, ) = 2 + (1 )2
(1 )
CL,g (, ) = 8(1 ).
CT ,q (, ) =
(2.4)
(2.5)
(2.6)
(2.7)
d2
s 2 ,
2
+k
2
0
(3.1)
where the spectral function s represents the discontinuity across the cut:
1 2 i
1
s e s 2 ei =
Disc s 2 .
s 2 =
2i
2i
(3.2)
0
To lowest order in perturbation theory we have s (2 ) = 4
s (2 )2 .
Nonperturbative effects at long distances are expected to give rise to a nonperturbative
modification to the perturbativelycalculated strong coupling at low scales, s (2 ) =
s (2 ) sPT (2 ), where sPT (2 ) is the perturbativelycalculated running coupling [2].
Note that here sPT (2 ) refers only to the contribution to the running coupling from a fixed
(nexttoleading) order perturbative calculation, and so is itself wellbehaved down to low
scales, without any divergences: the Landau pole appears when we include an arbitrary
number of loop insertions in the propagator. Hence s (2 ) is assumed to be welldefined
for all positive 2 .
We now consider the calculation of some observable F in an improved approximation
which takes into account fixedorder contributions plus those higherorder terms that lead
to the running of s . As is well documented, for processes involving a single gluon, it
is required to calculate the relevant contributions as though the gluon had a small mass
265
d2 n 2
s .
2
(3.3)
Numerical values for these parameters must be obtained from data: fits of the 1/Q2
corrections to DIS structure functions suggest that A2 0.2 GeV2 [4].
However, DIS in the singlet channel involves two such gluons. Both gluons have an
associated dispersive variable, so we obtain a characteristic function F ($1 , $2 ), where
$i = 2i /Q2 , which is simply the observable F calculated as though the gluons had masses
1 and 2 , and without any factors of s .
Since both gluons are constrained to have the same 4momentum s, we can simplify
this to require only one dispersive variable. By defining = s 2 /Q2 , we see that the
dependence of the characteristic function F on $1 and $2 is given by
d f ()
,
F ($1 , $2 ) =
(3.4)
( + $1 )( + $2 )
where the integration limits and the function f depend on the particular calculation. This
may be reexpressed in the form
F ($1 , $2 ) =
where
($1 ) $2 F
($2 )
$1 F
,
$1 $2
d
f ().
( + $)
So it is sufficient to perform the calculation with one mass set equal to zero, the
giving us the form of the characteristic function.
The alternative expression
1
1
d f ()
F ($) =
d f ()
$
$
+$
found in [9,10]. In
shows how this relates to the slightly different definition of F
($) =
F
(3.5)
(3.6)
other
(3.7)
these
references it is shown that a characteristic function of the form (3.5) leads to a nonperturbative correction to the observable F given by
F =
d2
2 2 2 2 2 2
2s s s
G /Q ,
2
(3.8)
266
F = a1
D1
,
Q2
(3.10)
and
1 a2 $ log2 $
F
2
F = a2
D1
D2
log 2 ,
2
Q
Q
(3.11)
d2 2
2 2 2 2
,
2s s s
2
(3.12)
1
log D2
D1
d2 2
2
2
2s 2 s 2 s 2
.
log
(3.13)
s2
=
= ,
2
Q2
(4.1)
and the azimuthal angle in the Breit frame. Then the relevant 4momenta are, in the Breit
267
frame,
1
q = Q(0, 0, 0, 2),
2
1
p = Q(1/, 0, 0, 1/ ),
2
1
s = Q(s0 , s cos , s sin , s3 ),
2
where
(4.2)
(4.3)
(4.4)
1 + (1 2/ )
,
4(1 / )(1 / )
2
s
=
,
2
1+
s3 = .
s0 =
(4.5)
(4.6)
(4.7)
The gluon is therefore produced with three degrees of freedom: determines the
longitudinal momentum fraction x/ of the gluon, is proportional to the gluons
virtuality, and gives its azimuthal angle. If we are concerned only with the produced
gluon, and not with the parton that emitted it, then is also free. Note that in the
limit 0, the gluon 4momentum s becomes xP / , the usual parton model result,
independent of and . This limit gives the standard photongluon fusion contribution
to the observable under consideration. However in Section 3 we saw that it is terms of
higher order in that generate power corrections: the transverse momentum of the gluon
is crucially important. Consequently we might like the gluon to have a 4dimensional
4momentum distribution within the proton: this will be discussed in Section 6.
With these definitions, the integration measure for the splitting becomes
1
d 3 p
=
(2)3 2p 0 (4)2
1
d
2
0
d
2
d
(1 ) s 2 .
(4.8)
Let us also introduce polarisation vectors $i , (i = 1, 2, 3), for the radiated gluon. We
$i $i = g +
ss
.
s2
(4.9)
For those contributions to Fig. 1 where both the top and bottom halves are gaugeindependent (such as the lowestorder contribution to Fig. 2), we can use (4.9) to replace
the gluon propagators with sums over polarisation vectors, thereby detaching the top and
bottom halves of the diagram. Where it is not gauge independent, this gives the contribution
in Landau gauge.
For the purpose of later discussion, let us introduce two such sets of polarisation vectors.
Firstly, let us define the natural polarisation vectors $i by imposing q $2 = q $3 =
268
$1 =
$2 =
(1 + ) s + 2q
,
iQ(1 )
[(1 + ) 2 ]s + [2 (1 + )/ ]q (1 )2 p
Q(1 ) (1 / )(1 / )
(4.10)
describing gluon production, and is formed by mixing $1 and $2 to give the new basis
vectors
$1 =
$2 =
2 p s
,
iQ
[2 (1 + ) ]p s (/ )q
,
Q (1 / )(1 / )
$3 = $3 .
(4.11)
The disadvantage of this basis is that $1 and $2 are neither transverse nor longitudinal, and
therefore are not suitable for a discussion of parton splitting or of splitting functions. The
use of this basis will be denoted by a tilde on appropriate symbols.
4.2. Radiation from quark line
The matrix element for the process shown in Fig. 2, integrated using the measure (4.8),
gives the following partonlevel contribution to some observable F :
1
2
2(p $i )(p $j )
1
d
d
s CF d
q
(1 ) ij
F =
2
2
2
s2
i,j
Fij (, , , ),
(4.12)
where Fij is the contribution due to photongluon fusion with the appropriate gluon
polarisations.
269
1
2
d
d q
s CF d
q
M (/ , )Fij (, , , ),
F =
2
2
ij
i,j
where
q
Mij (z, ) =
(4.13)
2(1z)(1z)
(1)z
(2zz) (1z)(1z)
i(1)z
(2zz) (1z)(1z)
i(1)z
(2zz)2
2(1)z
(1)z
2
ij
(4.14)
(4.15)
the fact that as 0, s xP / and $1 s / s 2 , so F11 , F12 and F21 all become
subleading and (for an unpolarised observable) F22 and F33 become equal and functions
of only. Thus the leading logarithmic divergence in F is
s CF
F leading( ) =
2
1
d
d
Pqg (/ )Cg ( ),
(4.16)
1
d
2
0
d
2
d (1 )
33 .
F22 + F
2
(4.17)
In other words, only two gluon polarisations, $2 and $3 , are permitted, but they are not
the transverse polarisations assumed in the parton model. Since this basis fails to make
270
apparent the form of the leading divergence and does not naturally give rise to the splitting
function, it is clear that any attempt to remove the initial quark line from the power
corrections calculated in [9] and [10] cannot be simply a matter, as with the logarithmic
divergences, of deconvoluting with the quark to gluon splitting function. Also, since all the
elements of (4.14) contribute to the power correction we would need to be able to include
the unphysical polarisations in the gluon distribution.
4.3. Radiation from gluon line
The radiation of the virtual gluon from a gluon line is shown in Fig. 4. This diagram is
gaugedependent, and while its logarithmically divergent piece is gaugeindependent and
given by the g g splitting function, the remainder of the diagram, and in particular the
pieces generating power corrections, are not.
The matrix element for the process shown in Fig. 4, integrated using (4.8) and using
the natural basis of polarisation vectors (4.10), gives the following contribution to the
observable F :
1
2
d
d g
s CA d
F =
M (/ , )Fij (, , , ),
2
2
ij
g
i,j
(4.18)
where
(2z)2
2z
2)
(2z)(22z+z
2iz 1z
2iz 1z
(2z)(22z+z2)
g
Mij (z, ) =
z(22z+z2 )
1z
+ O().
ij
(4.19)
271
the use of the basis $i does not diagonalise the matrix, even to leading order.
g
g
(iii) The transverse elements M22 and M33 yield polarised splitting functions, and their
sum gives
g
2(1 z + z2 )2
g
= Pgg (z).
lim M22 + M33 =
0
z(1 z)
Thus the leading divergence is the wellknown result
s CA
F leading( ) =
2
1
d
(4.20)
d
Pgg (/ )Cg ( ).
(4.21)
invariance. We could add in all O(s2 ) diagrams that contribute to the process g q qg,
as in Milan factor calculations for event shape variables (see [11]), but that would force
us to include diagrams unlike that of Fig. 1: it is not clear what relation (if any) these
diagrams have to the scale chosen for the running coupling.
First we notice that Fig. 4 is related by crossing to the e+ e annihilation diagram shown
in Fig. 5. So, just as in e+ e annihilation in [11], we can apply the pinch technique to
272
the diagram to define a gaugeinvariant contribution. But while the pinch technique has
previously been applied to internal gluon loops [13] and cut loops containing outgoing
partons [11], it is now being applied to a cut loop where one of the particles is incoming
and the other outgoing. This difference is irrelevant since the algebra is identical.
The matrix element for Fig. 5, on applying the pinch technique, is given in Eq. (A.9)
of [11]: in our notation it may be written
1
M2 = g 2 CA 4 8 s 2 g s s + 2(p + p )(p + p )
$i $j Fij (s),
s
i,j
(4.22)
where Fij is the contribution from the shaded blob. Using the crossing relation and
integrating using (4.8), we obtain the following form for the observable F :
1
2
2(p $i )(p $j )
d
d
s CA d
g
(1 ) 2ij
F =
2
2
s2
i,j
Fij (, , , ).
(4.23)
Note that this differs from the incoming quark case only by having a different constant
multiplying ij . In fact the only linearly independent terms allowed inside the square
bracket, consistent with gauge invariance and having the correct dimensions, are the two
shown. This was also seen in the e+ e annihilation calculation of [11].
Using the natural basis (4.10), we find
1
2
d
d g
s CA d
M (/ , )Fij (, , , ),
F =
2
2
ij
g
i,j
where
g
Mij (z, ) =
(4.24)
(23z+z)(2+z3z)
2(1)z
(2zz) (1z)(1z)
i(1)z
(2zz) (1z)(1z)
i(1)z
(2zz)2
3(1)z
2(1)z +
2
2(1 )z
ij
(4.25)
As previously, there are several comments to be made:
(i) Although this matrix is gauge invariant, it does not reduce to (4.19) in the limit
0. One consequence of this is that the g g splitting function is no longer trivially
g
g
recovered from the total transverse piece M22 + M33 in the limit 0:
g
2(1 z + 2z2 )
2z3
g
= Pgg (z)
,
lim M22 + M33 =
0
z
1z
(4.26)
2(1 z + z2 )2
.
z(1 z)
(4.27)
273
The additional piece arises from the fact that the pinch part of the diagram was removed (or
alternatively that the pinch parts of other diagrams were added) in order to secure gauge
invariance. This can be understood as follows: the factor 1/(1 z) in the splitting function
can only arise from the collinear limit of the gaugedependent piece (n p)/(n p ) in the
sum over polarisations of the p gluon in Fig. 4. In order to achieve gauge invariance we
must add or remove terms so as to cancel all ndependence, not just in the collinear limit
but identically for all p and p , and so terms with the factor 1/(1 z) must necessarily
disappear.
This happens even though the remaining diagrams themselves have no logarithmic
divergence associated with the gluon splitting (although of course they may have
divergences arising elsewhere): applying the pinch technique separates the diagrams into
a pinch part and a remainder, which have equal and opposite logarithmically divergent
pieces. Thus this is not a sensible way to study splitting functions. Also, since the only two
linearlyindependent gaugeinvariant terms with the correct dimension are those in (4.23),
and no linear combination of these can reproduce the splitting function, we see that we
cannot have a fully gauge invariant expression for the splitting g g + g (with some
suitable definition for what is meant by g , e.g., use of pinch technique) while retaining
the full splitting function.
We also notice that the contribution (4.26) is no longer invariant under z 1 z: this is
because the two daughters of the splitting are no longer identical, one being a real onshell
gluon and the other a virtual gluon with a modified propagator.
(ii) The difference between the production matrix employing the pinch technique and
that using standard perturbation theory is
2z
(2z)z
g
Mij (z, ) =
2i 1z
(2z)z
2i 1z
z3
1z
0
+ O()
z3
1z
(4.28)
ij
(where O() terms are gaugedependent). This is the contribution that gives the additions
to the splitting function. It becomes zero as z 0, which is as the interacting virtual gluon
becomes soft.
(iii) As in the quark and conventional gluon cases, there are both on and offdiagonal
g
elements. Now, however, the matrix Mij is diagonalised using the diagonal basis (4.11),
just as in the case where the emission is from a quark line. (Indeed one can trivially see
that any linear combination of the two terms in (4.23) is diagonalised using this basis.) One
eigenvalue is 3(1 )z/2, and the other two are both 2(1 )z. Then the decomposition
(4.23) simplifies considerably to:
s CA
Fg =
2
1
d
2
0
d
2
0
d 2(1 ) 3
F11 + F22 + F33 .
4
(4.29)
274
5. Structure functions
Consider now the part of the process involving the interaction of the virtual gluon
within the proton with the virtual photon. This is the process shown in diagram 3, which
corresponds to the quantities Fij above.
Working in the Breit frame, the 4momentum of the incoming photon and gluon are
given by (4.2) and (4.4); in this part of the process the variables , and are considered
fixed. Let us introduce the variables = P r/P q, = P k/P q, the azimuthal angle
between r and s, and the azimuthal angle between k and s. The kinematics are then
given by:
r = 12 Q z0 , z cos( + ), z sin( + ), z3 ,
(5.1)
1
k = 2 Q z 0 , z cos( + ), z sin( + ), z 3 .
(5.2)
The definitions of and along with the onshell conditions for the outgoing particles
require that
2
z
,
4
z 2
z 0 = + ,
4
z0 = +
2
z
,
4
z 2
z 3 = ,
4
z3 =
(5.3)
(5.4)
whence conservation of the 0th and 3rd components of 4momentum give the conditions
+ = 1 / 2 ,
2
z 2
z
(1 ) + (1 / )
+ =
,
4 4
(5.5)
(5.6)
while conservation of transverse momentum requires that s , z and z satisfy the triangle
inequalities
z s  z ,
(5.7)
z s  z .
(5.8)
(1 )( )
,
2 (1 / 2 )2
b=
(1 / )(1 / )
.
2 (1 / 2 )2
(5.11)
Given s , z and z , the angles and are determined up to a sign. We may then
choose and as the independent variables, with phase space
0 1 / 2 ,
0 .
(5.12)
275
1
z dz d d (A B 2s z cos )
(5.14)
,
4(2)2
B 2s z cos
where A and B do not depend on .
Next we integrate over . There are two values satisfying the integration condition, and
they differ by a sign. This gives
2
dz
1
d
.
(5.15)
2
4(2)
s z  sin 
Applying the parametrisation in terms of , we find that
2
z
(5.16)
= 8 sin ab,
2
1/
d
0
d.
(5.18)
(5.19)
q
where the sum over represents the outgoing quarkantiquark flavours. We first perform
the integral: the contributions to FL have no dependence, while the contributions to FT
have terms proportional to cos and cos2 . Then the integration over can be performed
by writing t = tan(/2), and that over by carefully expanding in powers of = 2 as
described in [9]. We obtain, for the longitudinal structure function,
CL11 = 32(1 ) (3 + 4 ) + (1 ) log + O 2 ,
(5.20)
CL12 = CL21 = 32i 1 / (1 )(1 3 log ) + O 2 ,
(5.21)
22
2
CL = 8 (1 ) + 8 1/ + 2/ 10 8 + 23 + 9 17
+ / 3 2 + 8 + 2 6 2 log + O 2 ,
(5.22)
33
2
CL = 8 (1 ) + 8 1/ + 2/ 10 8 + 23 + 7 15
+ / 3 2 + 8 + 2 6 2 log + O 2 .
(5.23)
Hence in the limit 0 we recover the wellknown results CL11 = CL12 = CL21 = 0 and
CL22 = CL33 = 8 (1 ).
276
1 /
18 2 / + 68 52 + 16 2 48 + 34 2
+ 2 / 2 1/ + 2 / 2 7/ + 6 3 2 / + 13 11
+ 3 2 10 + 8 2 log + O 2 .
(5.27)
As 0 the pieces CT22 and CT33 diverge logarithmically: this is the piece associated with
the collinear splitting of the gluon into a quarkantiquark pair.
6. Power corrections
We noted above that the interaction of the produced virtual gluon is given independently
of the production mechanism by the function Fij (, , , ), where i and j represent
polarisations and , , and parametrise the gluons 4momentum. In general,
calculations of power corrections will depend on all four of these quantities.
Thus a natural and intuitive way to combine this with a gluon distribution function might
be to write
1
2
1
d d d d
gij (x/ , , / )Fij (, , , ),
F (x) =
2
i,j x
(6.1)
where gij (x/ , , / ) is the relevant gluon distribution function which by symmetry
does not depend on . Of course this needs to be treated with some care, since both halves
of the diagram in Fig. 1 are gaugedependent and so therefore are all these distributions
when the gluon is offshell, i.e., = 0. Nevertheless, we can still make progress: we
277
present an intuitive argument that, although mathematically not totally rigorous, is helpful
in understanding the underlying physics.
For the diagonal elements we can view this probabilistically: the expected number of
gluons of polarisation i in an element of parameter space du d dz at (u, , z), where
u = x/ is the longitudinal momentum fraction of the proton carried by the gluon, and
z = / , is:
gii (u, , z) du
d dz
.
z
(6.2)
This is intended to be schematic only 2 clearly any gluon not collinear with the proton,
i.e., with = 0, will experience the confining effect of the QCD potential, and so cannot
exist for more than a short time (where short in this context means 1/QCD ). Such
particles are not asymptotically free: they can only be resolved at high momentum scales,
and it is precisely this resolution of gluons with nonzero transverse momentum that gives
rise to the running of the parton distributions. Power corrections are also known to arise
from the interactions of gluons with nonzero transverse momentum (see, e.g., [15]), and
so are also generated by the resolution of shortlived virtual gluons within the proton. Thus
we expect power corrections in singlet DIS to be intimately related to the running of parton
distributions.
Consider first the transverse polarisations: these are the physical polarisations. Speaking
schematically (i.e., not being too precise about our choice of factorisation scheme), we
may view the polarised gluon distribution function gi (u, M 2 ) as representing the gluons
with transverse momentum less than M. So
gi u, M 2 =
for i = 2, 3, where
Therefore,
gi (u, M
M2
M 2
1
dz
z
(M
2 )
d
gii (u, , z),
(6.3)
2)
(M
2)
1
=
u
dz M 2
gii (u, ,
z),
z M 2
(6.4)
= 1 + O().
=
2
M
1 2z
2)
s
=
2
1
dz
i
i
CF qtot(u/z)Pqg
(z) + CA g(u/z)Pgg
(z) ,
z
(6.5)
(6.6)
u
2 It could of course be more rigorously formulated in terms of welldefined unintegrated parton distributions,
but that is unnecessary for the purposes of the argument given here.
278
so for the sake of illustration let us make the simplest consistent assignment for the
differential gluon distribution gii (u, ,
z), which is
s
i
i
CF qtot (u/z)Pqg
z) =
(z) + CA g(u/z)Pgg
(z) + O().
gii (u, ,
(6.7)
2
Here qtot(u) = q [q(u) + q(u)]
(6.8)
which is independent of , and polarisation i. Let us separate off the piece that generates
the collinear divergence by writing
2
Fii (, , , )
d
= Cg ( ) M 2 + Fii (, , ),
2
(6.9)
1
2
3 1
d
d
d
d
gii (x/ , , / )Fii (, , , )
Ftrans (x) =
2
i=2 x
1
=
(6.10)
d
g x/ , M 2 Cg ( )
x
1
3 1
d
d
d
gii (x/ , , / )Fii (, , ).
+
i=2 x
(6.11)
The first term on the righthand side of (6.11) is simply the standard contribution to
photongluon fusion, evaluated with an onshell initial gluon and convoluted with the
gluon distribution g(x) = 3i=2 gi (x). This contains logarithmic scaling violations given
by the running of the gluon distribution, but contains no power corrections associated with
the virtuality of this gluon.
The second term gives us the required power corrections, since this is the term with an
integral over the gluons virtuality. Let us now substitute the differential gluon distribution
with expression (6.7), thus obtaining
1
3 1
d
d
d
gii (x/ , , / )Fii (, , )
Ftrans (x) =
i=2 x
0
1
3 1
d
d
s d
i
(/ )
=
CF qtot (x/ )Pqg
2
i=2 x
i
(/ ) + O()
+ CA g(x/ )Pgg
Fii .
(6.12)
279
Although the above argument has been restricted to the contributions from transverse
polarisations, and even these have been treated only schematically, we are now in a position
to be able to see what is happening physically. We have gone round in a big circle: starting
with Figs. 2 and 4, we have detached the lower parts of the diagrams corresponding to gluon
production, to give the piece corresponding to photongluon fusion. But then in order to
convolute with the correct differential gluon distribution within the proton we had to make
use of the DGLAP equation, which reintroduced those lower legs of the diagrams. Thus for
the purposes of power corrections we find that the lower parts of those diagrams genuinely
are important and cannot be removed. This also explains the presence of the longitudinal
gluon polarisation: we are not dealing with a real incoming gluon and so we need not be
restricted to physical polarisations. Eq. (6.12) is intended to be illustrative only it was
not rigorously derived. Yet it is clear that while the leading perturbative contribution to
singlet DIS, given by the first term in Eq. (6.11), is O(s ), the 1/Q2 power corrections
as well as the logarithmic scaling violations are O(s2 ), the additional factor of s arising
from the DGLAP equation. (Contrast this with flavour nonsinglet DIS, whose lowest order
is O(s0 ) but which has leading power corrections at O(s ).)
Thus the leading singlet power corrections are from the diagrams 2 and 4 as they appear:
we have two independent renormalon chains, and the magnitude of the power corrections
is given by the quantities D1 and D2 of Eqs. (3.12) and (3.13).
($) generating power corrections to structure functions is
The characteristic function F
thus given by
1
d TR CF
TR CA
2
q
g
eq
qtot(x/ )C (, $) +
g(x/ )C (, $) ,
F ($) =
(2)2
(2)2
q
(6.13)
where
1
C q/g (, $) =
d
1/
0
d q/g
Mij / , 2 C ij , , 2 ,
+$
(6.14)
i,j
and the C ij were defined in Eq. (5.19) and calculated in (5.20) to (5.27). The sum over
q represents outgoing quarkantiquark flavours in the photongluon fusion. We retain
($) up to O($) that are nonanalytic as $ 0. In order to define a gaugeterms in F
invariant quantity with a natural scale for the running coupling, we use the matrix (4.25)
evaluated using the pinch technique the remaining diagrams are also expected to give
power corrections, but as in [11] they are expected to have a different structure.
Using (4.13) and (4.24) we find
q
CT = 29 2 63 + 63 2 2 3 + 12 log 27 log
27 2 log + 12 3 log log $
+ 23 4 + 3 3 2 4 3 + 6 log + 6 2 log log 1 + 12 log $ log $
+ 25 2 + 25 2 25 3 2 5 + 15 2 log + 15 3 log $ log $
2 5 2 5 3 + 2 2 log + 2 3 log log 1 + 12 log $ $ log $, (6.15)
280
q
CL = 83 1 3 + 2 3 3 2 log log $
8
17 + 75 2 125 3 + 33 5 + 30 log
225
+ 75 3 log 45 5 log $ log $
8
2 15 2 + 10 3 + 3 5 15 3 log log 1 + 12 log $ $ log $,
+ 15
(6.16)
which is the result given in [9], and
g
CT = 49 1 99 + 99 2 3 + 6 log 27 log
54 2 log + 33 3 log log $
+ 43 2 + 6 + 3 2 11 3 + 3 log + 12 2 log log 1 + 12 log $ log $
+ 45 1 + 40 2 35 3 6 5 + 15 2 log + 30 3 log $ log $
8 5 2 5 3 + 2 2 log + 2 3 log log 1 + 12 log $ $ log $, (6.17)
g
CL = 83 1 3 9 2 + 11 3 12 2 log log $
8
17 + 1275 2 1475 3 + 183 5 + 30 log + 450 2 log
225
+ 750 3 log 270 5 log $ log $
2
3
5
3
1
+ 16
15 1 15 + 5 + 9 30 log log 1 + 2 log $ $ log $.
(6.18)
Thus we obtain the
quark distribution are
q
Fi (x) =
1/Q2
1
TR CF 2
d
q
qtot(x/ )Ci ( ),
eq
(2)2
q
(6.19)
where
D1 2(2 + 25 2 25 3 2 5 + 15 2 log + 15 3 log )
Q2
5
D2
2
2
2 5 5 + 2 log + 2 log log
(6.20)
,
eQ2
D1
8(17 + 75 2 125 3 + 33 5 + 30 log + 75 3 log 45 5 log )
q
CL ( ) = 2
Q
225
2
3
5
3
D2
8(2 15 + 10 + 3 15 log )
log
+
(6.21)
15
eQ2
q
CT ( ) =
1
d
TR CA 2
g
g(x/ )Ci ( ),
eq
(2)2
q
(6.22)
281
where
D1 4(1 + 40 2 35 3 6 5 + 15 2 log + 30 3 log )
Q2
5
D2
,
8 5 5 2 + 2 log + 2 2 log log
eQ2
8
D1
g
CL ( ) = 2
17 + 1275 2 1475 3 + 183 5 + 30 log
225
Q
+ 450 2 log + 750 3 log 270 5 log
D2
16(1 15 2 + 5 3 + 9 5 30 3 log )
log
+
.
15
eQ2
g
CT ( ) =
(6.23)
(6.24)
FT /L (x)
(0)
FT (x)
D1 q/g
K (x),
Q2 T /L
(7.1)
(0)
where FT (x) is the Bornlevel result for the transverse structure function given in
Eq. (2.2). These were calculated at Q2 = 500 GeV2 , using the corresponding MRST
Fig. 6. Graph showing KT (x) and KL (x). The dotted lines represent the contributions from the
quark distribution; the dashed lines represent the gluon contributions. The totals are represented by
the solid lines.
282
Fig. 7. Graph showing K2 (x)/x. The dotted line represents the contribution from the quark
distribution; the dashed line represents the gluon contribution. The total is represented by the solid
line.
(central gluon) parton distributions [16], and assuming four flavours of outgoing quark
antiquark pairs. The unknown value of D2 /e was set to 0.06 GeV2 , i.e., approximately 2 ,
although the qualitative behaviour of these power corrections does not change provided we
keep D2 Q2 .
The dominant contribution to both of these comes from the gluon distribution, as may be
expected at these relatively low values of x. All the contributions tend to zero at large x, and
become large at small x. The contributions to KT are positive and those to KL negative,
but they have similar magnitudes.
The corresponding quantity related to F2 /x, which is K2 /x = KT + KL , is shown in
Fig. 7. The positive and negative contributions partially cancel, giving a power correction
smaller by a factor 3 or 4 than that for FT or FL .
The power corrections are multiplied by the unknown factor D1 , defined in (3.12). If
D1 is positive, we might expect FT (x) to show a positive 1/Q2 power correction and
FL (x) a negative one. These results are all qualitatively the same as those in [9], which
took into account only the quark contribution, but the inclusion of the gluon contribution
increases considerably the size of the correction, in the case of FT by more than an order
of magnitude.
8. Discussion
The calculation of power corrections using the renormalon model or the dispersive
approach in flavour singlet DIS is a nontrivial problem, due to the fact that Fig. 1 as it
283
stands is gauge dependent, both in the upper and lower halves of the diagram. These halves
may both be considered to be cut insertions into a single gluon propagator, albeit in another
channel, and so the pinch technique can be used to define a natural gaugeinvariant quantity.
The diagram 1 evaluated using the pinch technique gives only one gaugeindependent part
of the full O(s2 ) amplitude, and the expectation is that the other parts will also give power
corrections, although not taking the form of two renormalon chains with equal 4momenta.
Power corrections thus arise when a virtual gluon is radiated from a quark or gluon, and
this virtual gluon then interacts with the photon. We found that, as well as the physical
transverse polarisations, we also have contributions to power corrections from longitudinal
polarisations and interference terms. In addition, since the gluons 4momentum is not
simply some multiple of that of the proton, the contribution from the photongluon fusion
is a function of all four momentum components. In particular the kinematics of the virtual
gluon production are important: where it is radiated from an onshell parton, the gluon
remembers the momentum fraction of the parton that emitted it. This, along with the
discussion of the virtual gluon distribution in Section 6, indicates that we cannot detach
the two halves of Fig. 1 but rather that the power corrections arise from the full diagram
and are thus an O(s2 ) effect.
Calculations of power corrections to structure functions and fragmentation functions due
to gluon radiation by a quark are already published in the literature [810]. However the
contributions from radiation by a gluon dominate, because of the relative behaviours of
the quark and gluon parton densities. This larger correction was evaluated above for the
structure functions. Above x = 0.05 the power corrections to the structure functions are
small, but we predict that for x below 0.05 the corrections grow. So the correction to F2
at x = 0.01 and Q2 = 4 GeV2 might be around 2% (assuming a reasonable value of
D1 0.1 GeV2 ). The corrections to FT and FL are expected to be slightly larger. In any
case, we do not expect to see the large ( 50%) corrections predicted by [8].
It was seen that the use of the pinch technique lead to a failure to recover the usual g g
splitting function. This is not a problem the remaining pieces come from the remaining
diagrams. All the pinch technique has done is move certain terms from one diagram to
another. But the point of using this particular separation of terms is that one can define
a gaugeinvariant QCD effective charge, so the virtuality of the gluon enters as the natural
scale for the coupling; it is not clear that this is the case for the other terms.
Finally, there is clearly some similarity between the situation here and that of a
decaying outgoing virtual gluon in e+ e annihilation (Fig. 5), where we also have two
renormalon chains with equal 4momenta k, for some k. Again we have an integral over
the virtuality k 2 , in which the coupling appears as s (k 2 )2 . But in that case we use
Eq. (3.2) to convert s2 into the spectral function s (k 2 ), and the two chains became one,
with a cut bubble insertion. There are two reasons why we cannot do the same here. Firstly,
in the e+ e case, the virtual gluon is timelike, k 2 is positive and thus we have s (k 2 )
integrated over positive k 2 . But here the gluon is spacelike, so k 2 is negative, and while
it is quite possible to convert the two factors of s into a single s (k 2 ), it is integrated
over negative values of its argument. It is therefore not naturally manipulated into standard
singlechain form. Secondly, although the algebra of the Feynman diagrams is identical
284
(i.e., related by crossing), in the e+ e case we can simply integrate over the cut bubble
but in DIS we must include parton density functions for the incoming quark or gluon. This
makes it quite impossible to integrate out the crossed cut bubble that is the mechanism for
production of the virtual gluon. So, while there are interesting and useful parallels between
power corrections to singlet DIS and e+ e annihilation with outgoing gluon splitting, there
are also significant differences and the relationship between them is not as simple as may
naively have been supposed.
Acknowledgements
The author wishes to thank Yuri Dokshitzer, Mark Smith, David Summers, Jay Watson
and Bryan Webber for helpful discussions and comments.
References
[1] M. Beneke, Phys. Rep. 317 (1999) 1, hepph/9807443.
[2] Yu.L. Dokshitzer, G. Marchesini, B.R. Webber, Nucl. Phys. B 469 (1996) 93, hepph/9512336.
[3] O. Biebel, Contribution to the proceedings of the DIS 2000 conference, April 2000, hepex/0006020.
[4] M. Dasgupta, B.R. Webber, Phys. Lett. B 382 (1996) 273, hepph/9604388.
[5] E. Stein, M. MeyerHermann, L. Mankiewicz, A. Schfer, Phys. Lett. B 376 (1996) 177, hepph/9601356;
M. MeyerHermann, M. Maul, L. Mankiewicz, E. Stein, A. Schfer, Phys. Lett. B 383 (1996)
463, hepph/9605229;
M. MeyerHermann, M. Maul, L. Mankiewicz, E. Stein, A. Schfer, Phys. Lett. B 393 (1997)
487, Erratum;
M. Maul, E. Stein, A. Schfer, L. Mankiewicz, Phys. Lett. B 401 (1997) 100, hepph/9612300;
M. Maul, E. Stein, L. Mankiewicz, M. MeyerHermann, A. Schfer, hepph/9710392;
M. MeyerHermann, A. Schfer, hepph/9709349.
[6] M. Dasgupta, G.E. Smye, B.R. Webber, J. High Energy Phys. 04 (1998) 017, hepph/9803382.
[7] M. Dasgupta, B.R. Webber, Eur. Phys. J. C 1 (1998) 539, hepph/9704297;
M. Dasgupta, B.R. Webber, J. High Energy Phys. 10 (1998) 001, hepph/9809247.
[8] E. Stein, M. Maul, L. Mankiewicz, A. Schfer, Nucl. Phys. B 536 (1998) 318, hepph/9803342.
[9] G.E. Smye, Nucl. Phys. B 546 (1999) 315, hepph/9810292.
[10] G.E. Smye, Nucl. Phys. B 549 (1999) 263, hepph/9812251.
[11] G.E. Smye, JHEP 0105 (2001) 005, hepph/0101323.
[12] J.M. Cornwall, Phys. Rev. D 26 (1982) 1453;
J.M. Cornwall, J. Papavassiliou, Phys. Rev. D 40 (1989) 3474;
N.J. Watson, Nucl. Phys. B 552 (1999) 461, hepph/9812202;
J. Papavassiliou, Phys. Rev. Lett. 84 (2000) 2782, hepph/9912336.
[13] N.J. Watson, Nucl. Phys. B 494 (1997) 388, hepph/9606381.
[14] R.D. Peccei, R. Rckl, Nucl. Phys. B 162 (1980) 125.
[15] R.K. Ellis, W. Furmanski, R. Petronzio, Nucl. Phys. B 207 (1982) 1;
R.K. Ellis, W. Furmanski, R. Petronzio, Nucl. Phys. B 212 (1983) 29.
[16] A.D. Martin, R.G. Roberts, W.J. Stirling, R.S. Thorne, Eur. Phys. J. C 1 (1998) 463, hepph/9803445.
Abstract
In this work we update the constraints on tree level FCNC couplings in the framework of a theory
with n isosinglet vectorlike down quarks. In this context, we emphasize the sensitivity of the B
J /KS CP asymmetry to the presence of new vectorlike down quarks. This CP asymmetry, together
with the rare decays B Xs,d l l and K are the best options to further constrain the FCNC
tree level couplings or even to point out, in the near future, the possible presence of vectorlike quarks
in the low energy spectrum, as suggested by GUT theories or models of large extra dimensions at the
TeV scale. 2001 Elsevier Science B.V. All rights reserved.
PACS: 12.15.Ff; 12.60.i; 13.25.Hw; 13.25.Es; 12.15.Mm
1. Introduction
The study of flavor changing neutral currents (FCNC) in particle physics phenomenology has played a key role in the advance of high energy physics in the past decades. Already
in the early seventies the nonobservation of FCNC was used to predict the existence of the
charm quark through the GIM mechanism [1] well before its direct experimental discovery.
Subsequently, a precise prediction of its mass was made from FCNC kaon processes [2].
mass difference
Later on, the discovery of the bottom quark and the measure of the BB
indicated the presence of a heavy top quark. Due to this GIM mechanism, FCNC in the
SM arise only at higher loop level and are suppressed by powers of light quark masses and
small mixing angles. This strong suppression make FCNC phenomena a privileged ground
to search for signs of new physics beyond the SM. In many extensions of the SM there is no
analog of the GIM mechanism to protect FCNC processes and hence potentially large new
Email address: vives@ific.uv.es (O. Vives).
05503213/01/$ see front matter 2001 Elsevier Science B.V. All rights reserved.
PII: S 0 5 5 0  3 2 1 3 ( 0 1 ) 0 0 3 9 0  X
286
287
VLdQs in the low energy spectrum, and especially, to reach a rough understanding of
FCNC effects in models with large extra dimensions [5,6], we present our bounds in a
general framework with n additional VLdQs. A general analysis with both up and down
isosinglets and isodoublets will be presented elsewhere [7].
From the low energy point of view, the model we analyze here has been thoroughly
described in Ref. [8,9]. Its main feature is that the additional VLdQs mix with the three
ordinary quarks and consequently the mass matrix in the down sector is now (3 + n)
(3 + n). This matrix is diagonalized by two (3 + n) (3 + n) mixing matrices and only
the lefthanded rotation is observable giving rise to the CKM matrix. As we have already
announced in the introduction, the fingerprint of this model is that it allows tree level FCNC
in the Z and H vertices. To see how these FCNC couplings appear, we can work in the
basis where the up quark mass matrix is diagonal. Here the down quark mass matrix is
diagonalized by a (3 + n) (3 + n) unitary matrix, V . In this model charged current
couplings are unchanged except that VCKM is now the 3 (3 + n) upper submatrix of
V , and at low energies a nonunitary 3 3 mixing matrix appears at tree level. However,
the mixing of doublet and singlet weak eigenstates modifies the neutral current couplings
that in terms of flavor eigenstates will be V Diag(1, 1, 1, 0)V = I44 . So, neutral current
interactions are given by
g
g
u Li mui uLi dL U md dL ,
2MW
U =
3
Vi
Vi ,
(1)
i=1
where Uds , Ubs or Ubd = 0 would signal new physics and the presence of FCNC at tree
level.
In the following we analyze the effects of these new couplings in FCNC processes and
we update the phenomenological bounds on them. More specifically, in processes changing
flavor in one unit, %F = 1, that in the SM go through electroweak penguin diagrams
of order GF Vti Vtj , we simply consider the dominant tree level FCNC contributions,
order GF Uij . Similarly, in %F = 2 processes mediated by boxes in the SM of order
GF (Vti Vtj )2 we include two different additional contributions, a pure tree level diagram
with two FCNC vertices, order GF Uij2 and a mixed SM vertex contribution with a tree
level vertex, roughly order GF Vti Vtj Uij [12]. Other new physics contributions in this
framework will always be suppressed by additional loop factors or higher powers of the
FCNC couplings with respect to the contributions considered here [13]. Therefore, the
effective low energy theory we consider is then identical to the SM with a nonunitary
CKM matrix and the presence of tree level FCNC as shown in Eq. (1).
288
vertices. The minimal extension from the SM would consist in the addition of a single
vectorlike down quark. In this situation the mixing matrix can be parametrized as [14],
SM
V = R34 (34, 0)R24 (24 , 3 )R14 (14 , 2 )VCKM
(12 , 13 , 23, 1 ),
(2)
where
is 4 4 block diagonal matrix composed of the standard
CKM [15,16] and a 1 1 identity in the (4, 4) element, and Rij (ij , k ) is a complex
rotation between the i and j families. Note that, in the limit of small new angles, we
follow the usual phase conventions. In fact, this 4 4 mixing matrix can represent a good
approximation to more complete models with several vectorlike generations if the mixings
are hierarchical, similarly to the standard CKM matrix.
On the other hand, given that the deviations from unitarity of the CKM mixing matrix
are experimentally known to be small, it is possible to use an approximate parametrization
of the mixing matrix valid for an arbitrary number of vectorlike quarks [9]. This is an
extension of the usual Wolfenstein parametrization of the CKM [17] in terms of
Udd =
Vid 2 = 1 Dd2 ,
Usd =
Vis Vid ,
Vus = ,
SM ( , , , )
VCKM
12 13 23 1
i=u,c,t
Vcb = A ,
2
Uss =
i=u,c,t
2
= 1 Ds2 ,
Vis 
Vib 
= 1 Db2 ,
Ubs =
i=u,c,t
Vub = A e ,
3 i
Ubb =
Vib Vis ,
i=u,c,t
2
Ubd =
i=u,c,t
Vib Vid ,
(3)
i=u,c,t
n+4
2
8
16
2
4
D4
d + Re{Usd } + O 7 ,
8
2 1 + 4A2 4 1 4A2 + 16A2 cos 6
Vcs =1
2
8
16
D2 D2
D4
s s 2 + A2 Re{Ubs } s + O 7 ,
2
4
8
A2 4 A2 2 6 Db2 Ds4
+ O 7 ,
Vt b =1
2
2
2
8
2
2 D2
D
A
s
A2 ei 5 + d
+ Usd
Vcd = +
2
2
U
+ Ubd A sd 2 + O 7 ,
2
Vud =1
289
A
A
Vt d =A 1 ei 3 + ei 5 + Ds2 3 + Ubd
2
2
A 2
2
i 3
Db Dd 1 e
A2 Usd + O 7 ,
+
2
A
A 2
2
i
Vt s =A +
Ds Db2 2
4 + Ubs +
Ae
2
2
7
A 6
+ +O .
(4)
8
This is a completely general parametrization of the 3 3 submatrix of a complete (3 + n)
(3 + n) unitary matrix. 1 Hence, it includes also the simplest case of a single vectorlike
quark. In this last case, there are several relations among the 13 parameters of this general
matrix and only the nine independent parameters of a unitary 4 4 mixing matrix remain.
In fact, we have
Dd2 = V4d 2 ,
V4d ,
Usd = V4s
Ds2 = V4s 2 ,
Db2 = V4b 2 ,
Ubs = V4b
V4s ,
Ubd = V4b
V4d .
(5)
(6)
(7)
Naturally, the elements of the extended mixing matrix corresponding to CKM elements
are experimentally measured in low energy experiments. Some of these measurements
are obtained from treelevel processes, therefore not affected by new physics to a high
degree of approximation. Hence, they constrain directly the corresponding elements of our
extended mixing matrix. Specifically, we have [16,18], at 95% CL,
These constraints restrict the different parameters in Eq. (4), and so from the bounds on
Vud  and Vcs , we obtain Dd2 5.5 103 and Ds2 5.6 102 , respectively. Another
constraint [8,9,19] comes from the SU(2)L coupling of the Z 0 to bb. In the SM, this
290
Notice that, in the general case, the Di2 parameters are completely independent from
the FCNC couplings U . However, in a more definite model, as for instance the single
vectorlike quark model, due to the unitarity of the 4 4 matrix, these constraints have
a strong impact on all other elements of the extended mixing matrix and consequently
also on the tree level FCNC couplings, as shown in Eq. (6). In any case, even with these
restrictions, the FCNC couplings can have large effects on rare processes, where the SM
contributes at the 1 loop level and new physics is allowed to compete on equal footing. In
fact, most of the experimental results are well accommodated within a pure 3 generations
SM and hence these measurements provide additional constraints on the FCNC couplings.
In other observables, like B 0 CP asymmetries or B rare decays, the experimental results
may still differ from the SM prediction once the experimental accuracy is increased in the
near future. Therefore, we analyze these observables from a slightly different point of view,
and in the last section we show what are the possible deviations from the SM consistent
with the constraints discussed here.
In the first place we analyze the constraints associated with kaon physics and then the
constraints we get from the B system.
3.1. Kaon physics constraints
Rare kaon decays and CPviolating observables in the kaon sector can receive important
contributions from the FCNC coupling Usd . In fact, this new coupling is constrained
mainly by the decay (KL + )SD and the experimental value of /. Other
observables constraining this coupling are K + + and K [20].
The decay KL + is CPconserving and in addition to its short distance part,
given by Z penguins and box diagrams, receives important contributions from the
two photon intermediate state which are difficult to calculate reliably. Unfortunately,
the separation of the shortdistance part (similar to K + + , free of hadronic
uncertainties) from the longdistance piece in the measured rate is rather difficult.
Therefore, the full branching ratio is generally written as a sum of a dispersive and
absorptive contributions, of which the latter can be calculated using the data for KL
. The dispersive contribution can be decomposed as a long distance and a short distance
part. Following [20,21], we can write down
2
Br(KL )SD = 6.32 103 CU 2Z Re(Usd ) + c + Y0 (xt ) Re sd
t
2.8 109 ,
(8)
2 / 2 )1 92.7, ab = V V ,
c = 6.54 105 is
where CU 2Z = ( 2GF MW
ia ib
i
the charm quark contribution and Y0 is the InamiLim [22] function as defined in [23].
Experimentally, we must require this branching ratio to be BR(KL + )SD 2.8
109 [21].
A second observable constraining Usd is /. Within the SM there is a cancellation
between QCD and electroweak penguin contributions (dominated by Z penguins) which
suppresses this ratio. When new physics enters into the game and if, as it is expected on
general grounds, it does not affect considerably the QCD contributions but does so with the
291
Z penguins, the abovementioned cancellation does not take place and significant deviations
from the SM results (or strong bounds on the new parameters) can be expected. We now
decompose / as follows
= U CU 2Z Im(Usd ) + t Im sd
t ,
(3/2)
,
U = 1.2 Rs rZ B8
(1/2)
(3/2)
t = U C0 2.3 + Rs 1.1 rZ B6
+ (1.0 + 0.12 rZ )B8
,
(9)
where the first term comes from the Z piece and the other one contains all the remaining
ones [20]. It is worth noting that unlike previous observables, here the theoretical errors
overwhelm the experimental precision. The main sources of uncertainties lie in the
(1/2)
(3/2)
and B8 . The importance of these uncertainties is somehow increased
parameters, B6
because of the cancellation we have mentioned.
(3/2)
= 0.8 0.2 [20]. The value
Here, we take Rs = 1.5 0.5, rZ = 7.5 1 and B8
(1/2)
has caused some controversy in the literature because of the different values
of B6
obtained in different schemes. We take two different values in order to illustrate the
situation where new physics in the sd sector is needed or not needed to reproduce the
(1/2)
= 1.0 0.2 with the other parameters
experimental value of /. For set I, we use B6
as given above, as in Refs. [20,24,25] which comes from large Nc calculations [26] and
lattice analysis, and this tends to favor the presence of new physics in Uds . In set II we use
(1/2)
= 1.3 0.5 in order to incorporate the predictions of Refs. [27,28], where inclusion
B6
of the correction from finalstate interactions in a chiral perturbation theory analysis tends
(1/2)
and
to favor the SM range. Still, there are other schemes where different values for B6
(3/2)
B8
are obtained [2931]. In fact, we have explicitly checked that with the values in
(1/2)
(3/2)
= 2.5 0.4 and B8
= 1.35 0.20, the constraints on the FCNC couplings
[29], B6
U are still consistent with the results presented below.
Numerically, we have for the central values
6.2
= 7.8 CU 2Z Im(Usd ) +
Im sd
t
9.9
(10)
with the parameters in set I and set II, respectively, and we calculate its errors with a
Gaussian method. This observable has to reproduce the experimentally obtained value of
( /)exp = (2.12 0.46) 103 .
A theoretically cleaner constraint [32], although less restrictive than the previous two
constraints is provided by BR(K + + ),
2
BR(K + + ) = 1.55 104 CU 2Z Im{Usd } + X0 (xt ) Im sd
t
2
,
+ CU 2Z Re{Usd } 2.11 104 + X0 (xt ) Re sd
t
(11)
with X0 (xt ) = C0 (xt ) 4B0 (xt ) a gauge invariant loop function combination of boxes and
vertices [22,23]. This decay has already been measured in the experiment E787 at BNL
[33], however, so far, a single event has been found and this is not enough to provide a
292
definite value for the branching ratio. Hence we take here only the upper limit at 95% CL
BR(K + + ) 8.3 1010 . 2
Finally, we include also K , whose leadingorder expression is [12,23]:
ei/4 GF BK FK2 mK
K =
Im t t (Usd )2 +
6%mK
4 sin2 W
t
t
sd
sd sd
,
8
t i Y0 (xi )i Usd
ij S0 (xi , xj )i j
(12)
i,j =c
i=c
where S0 is another InamiLim function [23] and the QCD correction factors (which take
into account short distance QCD effects) are given as follows,
cc = 1.38 0.20,
t t = 0.57 0.01,
t c = 0.47 0.04.
(13)
Here, contrary to Ref. [20], the coefficient Y0 (x) of the linear term in Uds is characteristic
of the present model, therefore, in principle, the irrelevance of K to constraint Uds is not
fully guaranteed.
At this point, it is important to emphasize that Eqs. (8), (9), (11) and (12) are completely
valid in the general model with n additional VLdQs.
In Fig. 1 we present the effects of the different constraints in the Usd coupling. As
we can see in this figure, the most efficient constraints are provided by KL + ,
that constrains Re(Usd ), and /, that bounds Im(Usd ) to the rectangular box in the
center. However, these constraints will be difficult to improve due to the large hadronic
uncertainties. Similarly, the constraints from K are very precise on the experimental side
and the precision of this limit on Usd is determined by the hadronic parameter BK =
0.85 0.15 [23,25]. Hence, it is not expected to improve largely in the near future. On
the other hand, the decay K + + is much cleaner from the theoretical point of view
[32]. The experimental value for the branching ratio, based in the single event found so
far is
10
,
BR(K + + ) = 1.5+3.4
1.2 10
(14)
(15)
Unfortunately, it is clear from the errors that we cannot obtain a lower bound on this BR
at 80% CL, still we have included in the figure a dashed circle showing the effect of a
future lower limit which would correspond to a hypothetical value of 8 1011 and would
exclude the region to the left of the small rectangle. Moreover, it is interesting to notice
that improving the upper bound to the level of 2 1010 would provide a bound on the
same level as the combined bounds from KL + and /. In fact, these values could
be reached after the analysis of the stored data from E787 and the sensitivity of the new
2 In Ref. [11], the old value of this BR at 1 was used, therefore both a lower and an upper limit were
considered.
293
experiment E949 [34] (already approved) will reach 1011 in the next few years. Hence,
this decay will provide the most stringent and clean bounds on the Usd coupling in the near
future, although at present the bounds are still obtained from KL + and /.
In Figs. 2 and 3 we present a scatter plot of the allowed values of Usd in the general
model with n additional vectorlike down quarks, although we must emphasize that this
allowed region does not change at all even if we go to the more restricted case of a single
VLdQ. We impose all the constraints described above, with the / parameters from set I
and set II in Figs. 2 and 3, respectively. In both approaches the bound for the real part is,
1.3 105 < Re{Usd } < 4.0 106 .
(16)
This constraint is directly obtained from Eq. (8) with the limit values of Re(sd
t ) which
4 and Re(sd )
4 from Eq. (4) to O(5 ) for
)
=
4.9
10
=
2.1
10
are Re(sd
max
min
t
t
= , 0 respectively. Therefore, this implies that the additional correlations among the
parameters in the 1VLdQ model are irrelevant for the Usd bound.
The constraints on the imaginary part of this coupling depend slightly on the adopted
(1/2)
(1/2)
= 1.0 0.2, we can see that Usd is necessarily
value for B6 . For set I, with B6
positive and does not reach the origin, indicating marginally the need of new physics for
/. The allowed range is 1.0 107 < Im{Uds } < 6.5 106 . On the other hand, in
set II, we get an allowed area, 1.5 106 < Im{Uds } < 6.0 106 , including the SM
as one of the possible points in agreement with the experimental results. This bounds are
4 and Im(sd )
5
directly obtained from Eq. (10) with Im(sd
t )max = 1.4 10
t min = 6 10
294
Fig. 2. Phenomenological bounds on the FCNC coupling Usd with the / hadronic parameters in
set I.
Fig. 3. Phenomenological bounds on the FCNC coupling Usd with the / hadronic parameters
from set II.
295
assuming a 50% error, which corresponds to the Gaussian errors that we use in this
constraint. In view of these two different options to calculate /, we take a conservative
bound on Im(Usd ) including both possibilities,
1.5 106 < Im{Usd } < 6.5 106 .
(17)
It is interesting to notice, that these bounds turn out to be much more stringent than
the direct bounds on FCNC couplings usually quoted in the literature for VLdQ models
[35]. This improvement is mainly due to the inclusion of the SM contributions that
were completely neglected in the calculation of K in the bounds presented in [35], the
improvement of the experimental results in (KL + )SD , K + + , and to the
inclusion of the bound from /. In this way, our bounds basically agree with the general
bounds in [20].
3.2. B physics constraints
In this section, we study the constraints on the Ubd and Ubs couplings from B physics
FCNC. In fact, the experimental information in the B system is improving rapidly with the
new results from B factories and so, it is important to update the bounds on these FCNC
couplings. In all the following processes, the choice of set I or set II to calculate / has no
relevant effects on the Ubd and Ubs couplings. Hence, we analyze the constraints making no
distinction on the hadronic parameters used to calculate /. In first place, we concentrate
on the constraints from CP conserving processes and then we add the information from
the B 0 CP asymmetries. The main CP conserving processes constraining Ubd and Ubs are
BR(B Xd,s l + l ) and %MBd,s .
From the upper bound on BR(B Xs l + l ) 4.2 105 [36] and assuming BR(B
Xd l + l ) BR(B Xs l + l ) we have [23,37],
Y0 (xt )t + CU 2Z Ubq 2
(B Xq l + l )
2
=
(B Xc e+ e ) 2 sin4 W
Vcb 2 f (mc /mb )
1
4 104
2 sin4 W sin2 W +
4
bq
(18)
with f (z) = (1 8z2 + 8z6 z8 24z4 log z) a phase space factor due to the mass of the
charm quark. From here we get
Y0 (xt )bq
(19)
t + CU 2Z Ubq < 0.15.
3
i ) + O(5 ) 8 103 (1 0.4 e i ) with Y 1,
Replacing here bd
0
t A (1 e
it is clear that the bd
t contribution can be safely neglected and we get a bound Ubd 
2
1.7 103 . However, this is not true in the case of bs
t A 0.040 that shifts the
3
Ubs  constraint to a circle of radius 1.7 10 centered in (0.040/92.7, 0).
mixing we have [1012]
Additionally, from BB
B
M12q
2 B f2 m
G2F MW
Bq Bq Bq Bq
12 2
bq 2
S0 (xt )t
bq ,
296
bq = 1 3.3
Ubq
bq
+ 165
Ubq
bq
2
(20)
where the new parameters are defined in Ref. [23], and the mass difference %MBq =
B
2M q . The experimental values for these observables are %MB = (0.472 0.017)
12
297
Fig. 5. Constraints on the Ubs coupling from B Xs l + l in the nVLdQ model (left) and in the
1VLdQ model (right).
which is not useful to set a constraint on Ubs . Moreover, it is easy to see from Eq. (20)
that for similar values of Ubd and Ubs , the FCNC effects on %MBs will be suppressed
bs 2
2
by a factor (bd
t /t ) when compared with the effects on %MBd . This implies that
s mixing. 3 Hence,
in this model we cannot expect to observe the FCNC effects in Bs B
this coupling is only constrained by the upper bound on BR(B Xs l + l ). In Fig. 5 we
show the constraint from BR(B Xs l + l ) as a circle slightly shifted from the origin,
that implies an upper bound of Ubs  2 103 [37,41] both in the general model with
n VLdQs and in the model with a single VLdQ. Similarly, the b s branching ratio
provides constraints of the same order of magnitude [42].
Despite these strong constraints from CP conserving processes, the CP asymmetries in
B decays are still very effective to constrain the Ubd coupling [43,44]. Recently, the arrival
of the first measurements of the B J /KS CP asymmetry, aJ / , from the B factories
has caused a great excitement in the high energy physics community.
+0.32+0.09
(Belle [46]),
aJ / = 0.580.340.10
(21)
+0.41
0.790.44
(CDF [47]).
These values correspond to a world average of aJ / = 0.51 0.18, that can be
compared with the SM expectations of 0.59 aJSM
/ = sin(2) 0.82 with =
/(V V )). The errors are still too large to draw any firm conclusion. In fact,
arg(Vcd Vcb
td tb
if we take the world average at 95% CL we do not get any improvement over the constraints
from CP conserving processes. However, anticipating the improvement of the experimental
errors from B factories, we take the world average at 1 level to see the effects on the Ubd
coupling. From Eq. (20), it is straightforward to obtain aJ / as,
aJ / = sin 2 arg(bd ) .
(22)
3 This is not always true in the presence of an up vectorlike quark, as can be seen in Ref. [40].
298
Fig. 6. Allowed region for the Ubd coupling requiring a value of aJ / asymmetry to reproduce the
world average at 1 in the nVLdQ model (left) and 1VLdQ model (right).
Using the world average at 1 as the experimentally allowed range, we show in Fig. 6
the resulting region for the Ubd coupling. As we can see here, the Ubd allowed region is
sizeably modified from this constraint. The outer regions in the second and fourth quadrants
are reduced while the central region corresponding to the SM remains filled; this situation
represents an improvement over the analysis presented in Ref. [44]. Note that the results
are essentially similar for the nVLdQ and the 1VLdQ cases. As we can see here, with
the world average for the aJ / CP asymmetry there is no need of new physics in the Ubd
sector.
299
Fig. 7. Distribution of the values of aJ / for 95 000 events in the nVLdQ (left) and in the 1VLdQ
(right).
be a natural candidate for a model with clear deviations from the SM expectations in CP
asymmetries, satisfying simultaneously all other experimental constraints.
3
2
From the expression of aJ / in Eqs. (20) and (22), with bd
t  [5 10 , 1.2 10 ]
3
and Ubd  10 , it is clear that arg(bd ) can be very large. Indeed, in the general model
with an arbitrary number of VLdQ, there is no further influence on Ubd from constraints
on other elements of the mixing matrix. In fact, only the inequalities in Eq. (7) can restrict
this coupling, however, given the bounds on Di2 , they are almost inoperative in this case.
Hence, all the points inside the allowed contour in Fig. 4 are equally probable and any value
of the asymmetry is possible with a comparable probability. Still, the especial geometry
of this allowed region and the preferred orientation of tbd slightly favors the SM range
over other values. Furthermore, even in a more constrained model, as in the model with
a single VLdQ large departures from the SM range are possible. The expected values
for the asymmetry both in the n VLdQs and in the 1VLdQ model are shown in Fig. 7
as an histogram of the distribution of 95 000 events satisfying all the constraints in the
previous sections. As expected from the above discussion, in the general case all possible
values of the asymmetry have similar probability. On the other hand, in the model with a
single VLdQ, this distribution is clearly peaked on the SM range. Moderate deviations are
rather frequent and larger departures are more rare but still possible for any value of the
asymmetry.
In Figs. 8 and 9 we show, in the 1VLdQ model, the correlation of the possible values
of the asymmetry with Ubd  and Ubs , respectively. Here, we see that the range 0.55
aJ / 0.85 corresponding to the SM expected range concentrates most of the events and
can be reproduced with Ubd  = 0 and any allowed value of Ubs . However, for different
values of the asymmetry there is a clear correlation between aJ / and the minimum value
of Ubd  required to obtain this asymmetry. For instance, to obtain an asymmetry below
0.5, a Ubd  2 104 is needed. This required minimum is also true in the general
model with an arbitrary number of VLdQs. Similarly, we see in Fig. 9 that these large
values of Ubd correspond to low values of Ubs [11], however it is important to emphasize
300
Fig. 8. Correlation between the values of aJ / and Ubd  in the model with a single VLdQ.
Fig. 9. Correlation between the values of aJ / and Ubs  in the model with a single VLdQ.
301
Fig. 10. Allowed region for the Ubd coupling consistent with the BaBar result for the aJ /
asymmetry at 1 in the nVLdQ (left) and in the 1VLdQ (right).
Fig. 11. Allowed region for the Ubs coupling consistent with the BaBar result for the aJ /
asymmetry at 1 in the nVLdQ (left) and in the 1VLdQ (right).
that this correlation is only true in the minimal model, but not in a model with an arbitrary
number of VLdQs. To see this we show in Figs. 10 and 11 both in the nVLdQ and
in the 1VLdQ models an scatter plot of Re(Ubd ) versus Im(Ubd ) and Re(Ubs ) versus
Im(Ubs ) for an asymmetry corresponding to the BaBar result at 1 , 0.14 aJ / 0.54
[45]. In this Fig. 10 we see that the great majority of the allowed points are in the range
1 104 (2 104 ) Ubd  1.2 103 , in the nVLdQ (1VLdQ) model, i.e., a large,
nonvanishing Ubd coupling is required to reproduce the BaBar asymmetry. In particular,
this means that, within this model, a low CP asymmetry implies the presence of new
physics in bd transitions. This conclusion would be unchanged in the general model.
On the other hand, we see that, for these points, in the 1VLdQ model, the coupling Ubs
is always restricted to the range Ubs  3 104 ; hence, all the allowed points have
simultaneously high Ubd  and low Ubs . Indeed, it is easy to obtain in the 1VLdQ model,
302
5. Conclusions
In this work we have updated the constraints on tree level FCNC couplings in the
framework of a theory with n isosinglet vectorlike downquarks. The inclusion of the
constraints from (KL + )SD , /, K and K has allowed to improve
sizeably the bound on the Usd coupling. The precise range of allowed values for this
coupling depends strongly on the hadronic input in the calculation of /. Our summary
4 Still, it is important to emphasize that these points also require the presence of new physics in B decays.
303
Table 1
Constraints on the tree level FCNC couplings from rare
processes in the K and B systems
Re{Usd }
Im{Usd }
 Re{Ubd }
 Im{Ubd }
Ubs 
1.3 105
4.0 106
1.5 106
6.5 106
4
7 10
1.2 103
2 103
in Table 1 includes the main theoretical approaches [20,2431]. In the near future, K +
+ will be quite useful to further constrain Usd without large theoretical uncertainties.
We have calculated the constraints on Ubd and Ubs from %MBd , B Xs,d l l and the
B J /KS CP asymmetry. In Table 1, we summarize our results on these bounds
without the additional constraint from the B J /KS CP. Note that all these bounds
on Ubq are independent of the hadronic inputs in /. In addition, we have shown that
the B J /KS CP asymmetry is especially sensitive to the presence of a Ubd FCNC
coupling. In this regard, we have shown that the 1 value of the world average is already
able to exclude half of the allowed region for the Ubd coupling. However, due to the still
large experimental errors this constraint has a small effect at 95% CL. Assuming a low
value of the aJ / asymmetry below 0.52 (following BaBar range at 1 ) we have shown
that, in models with n VLdQs, treelevel FCNC in the bd sector are mandatory. The
FCNC coupling Ubd is required to be greater than 1 104 , rising to 2 104 in the
1VLdQ model. The Ubs coupling in the general model is bounded by 2 103 , although
for the 1VLdQ model the bound goes down to 3 104 .
Therefore, a clear favorable scenario for these models would be to find simultaneously
at least one order of magnitude
a low aJ / (below 0.5) and a large BR(Bd Xd l l),
close to the present
bigger than the SM expectations. Values of BR(Bd Xs l l)
experimental bounds are still possible in the nVLdQ model. Nevertheless, in the 1VLdQ
model these BR are expected to be similar to the SM values. In summary, this aJ / CP
asymmetry, together with the rare decays B Xs,d l l and K are the best options to
further constraint the FCNC tree level couplings or to discover the presence of vectorlike
quarks in the low energy spectrum as suggested by GUT theories or models of large extra
dimensions at the TeV scale.
Acknowledgements
The work of F.B. and O.V. was partially supported by the Spanish CICYT AEN99/0692.
O.V. acknowledges financial support from a Marie Curie EC grant (HPMFCT200000457).
304
References
[1] S.L. Glashow, J. Iliopoulos, L. Maiani, Phys. Rev. D 2 (1970) 1285.
[2] M.K. Gaillard, B.W. Lee, Phys. Rev. D 10 (1974) 897.
[3] J. Ellis, D.V. Nanopoulos, Phys. Lett. B 110 (1982) 44;
R. Barbieri, R. Gatto, Phys. Lett. B 110 (1982) 211;
S. Bertolini, F. Borzumati, A. Masiero, G. Ridolfi, Nucl. Phys. B 353 (1991) 591;
F. Gabbiani, E. Gabrielli, A. Masiero, L. Silvestrini, Nucl. Phys. B 477 (1996) 321, hepph/9604387;
A. Masiero, O. Vives, hepph/0104027, and references therein.
[4] M. Bando, T. Kugo, Prog. Theor. Phys. 101 (1999) 1313, hepph/9902204;
M. Bando, T. Kugo, K. Yoshioka, Prog. Theor. Phys. 104 (2000) 211, hepph/0003220;
M. Obara, G. Cho, M. Nagashima, A. Sugamoto, hepph/0105287.
[5] N. ArkaniHamed, S. Dimopoulos, G. Dvali, Phys. Lett. B 429 (1998) 263, hepph/9803315;
L. Randall, R. Sundrum, Phys. Rev. Lett. 83 (1999) 3370, hepph/9905221.
[6] For phenomenology of FCNC in models with large extra dimension, see:
A. Delgado, A. Pomarol, M. Quiros, JHEP 0001 (2000) 030, hepph/9911252;
F. del Aguila, J. Santiago, Phys. Lett. B 493 (2000) 175, hepph/0008143;
S.J. Huber, Q. Shafi, Phys. Lett. B 498 (2001) 256, hepph/0010195;
N. Rius, V. Sanz, hepph/0103086;
H. Davoudiasl, T.G. Rizzo, hepph/0104199;
S.J. Huber, Q. Shafi, hepph/0104293.
[7] G. Barenboim, F. J. Botella, O. Vives, work in progress.
[8] F. del Aguila, J. Cortes, Phys. Lett. B 156 (1985) 243;
G.C. Branco, L. Lavoura, Nucl. Phys. B 278 (1986) 738;
F. del Aguila, M.K. Chase, J. Cortes, Nucl. Phys. B 271 (1986) 61;
Y. Nir, D.J. Silverman, Phys. Rev. D 42 (1990) 1477;
D. Silverman, Phys. Rev. D 45 (1992) 1800;
G.C. Branco, T. Morozumi, P.A. Parada, M.N. Rebelo, Phys. Rev. D 48 (1993) 1167;
W. Choong, D. Silverman, Phys. Rev. D 49 (1994) 2322;
V. Barger, M.S. Berger, R.J. Phillips, Phys. Rev. D 52 (1995) 1663, hepph/9503204;
D. Silverman, Int. J. Mod. Phys. A 11 (1996) 2253, hepph/9504387;
M. Gronau, D. London, Phys. Rev. D 55 (1997) 2845, hepph/9608430;
F. del Aguila, J.A. AguilarSaavedra, G.C. Branco, Nucl. Phys. B 510 (1998) 39, hepph/9703410.
[9] L. Lavoura, J.P. Silva, Phys. Rev. D 47 (1993) 1117.
[10] G. Barenboim, F.J. Botella, G.C. Branco, O. Vives, Phys. Lett. B 422 (1998) 277, hepph/9709369.
[11] G. Barenboim, F.J. Botella, O. Vives, Phys. Rev. D 64 (2001) 015007, hepph/0012197.
[12] G. Barenboim, F.J. Botella, Phys. Lett. B 433 (1998) 385, hepph/9708209.
[13] J. Roldan, F.J. Botella, J. Vidal, Phys. Lett. B 283 (1992) 389.
[14] F.J. Botella, L. Chau, Phys. Lett. B 168 (1986) 97.
[15] L. Chau, W. Keung, Phys. Rev. Lett. 53 (1984) 1802.
[16] D.E. Groom et al., Eur. Phys. J. C 15 (2000) 1.
[17] L. Wolfenstein, Phys. Rev. Lett. 51 (1983) 1945.
[18] R. Barate et al., ALEPH Collaboration, Phys. Lett. B 465 (1999) 349;
P. Abreu et al., DELPHI Collaboration, Phys. Lett. B 439 (1998) 209;
M. Bargiotti et al., Riv. Nuovo Cimento 23 (3) (2000) 1, hepph/0001293.
[19] F. del Aguila, J.A. AguilarSaavedra, R. Miquel, Phys. Rev. Lett. 82 (1999) 1628, hepph/9808400.
[20] A.J. Buras, L. Silvestrini, Nucl. Phys. B 546 (1999) 299, hepph/9811471.
305
[21] D. Gomez Dumm, A. Pich, Phys. Rev. Lett. 80 (1998) 4633, hepph/9801298;
G. DAmbrosio, G. Isidori, J. Portoles, Phys. Lett. B 423 (1998) 385, hepph/9708326.
[22] T. Inami, C.S. Lim, Prog. Theor. Phys. 65 (1981) 297.
[23] A.J. Buras, R. Fleischer, Report No. TUMHEP27597, hepph/9704376.
[24] M. Ciuchini, E. Franco, L. Giusti, V. Lubicz, G. Martinelli, Report No. ROME991268, hepph/9910237;
M. Ciuchini, G. Martinelli, Nucl. Phys. Proc. Suppl. 99 (2001) 27, hepph/0006056;
S. Bosch, A.J. Buras, M. Gorbahn, S. Jager, M. Jamin, M.E. Lautenbacher, L. Silvestrini, Nucl.
Phys. B 565 (2000) 3, hepph/9904408.
[25] A.J. Buras, hepph/0101336.
[26] T. Hambye, G.O. Kohler, E.A. Paschos, P.H. Soldan, W.A. Bardeen, Phys. Rev. D 58 (1998)
014017, hepph/9802300;
T. Hambye, G.O. Kohler, P.H. Soldan, Eur. Phys. J. C 10 (1999) 271, hepph/9902334.
[27] S. Bertolini, J.O. Eeg, M. Fabbrichesi, E.I. Lashin, Nucl. Phys. B 514 (1998) 93, hepph/9706260;
S. Bertolini, M. Fabbrichesi, J.O. Eeg, Rev. Mod. Phys. 72 (2000) 65, hepph/9802405.
[28] E. Pallante, A. Pich, Phys. Rev. Lett. 84 (2000) 2568, hepph/9911233;
E. Pallante, A. Pich, I. Scimemi, hepph/0105011.
[29] J. Bijnens, J. Prades, JHEP 0006 (2000) 035, hepph/0005189;
J. Bijnens, J. Prades, JHEP 0001 (2000) 002, hepph/9911392;
J. Bijnens, J. Prades, Nucl. Phys. Proc. Suppl. 96 (2001) 354, hepph/0010008.
[30] S. Narison, Nucl. Phys. B 593 (2001) 3, hepph/0004247.
[31] J.F. Donoghue, E. Golowich, Phys. Lett. B 478 (2000) 172, hepph/9911309.
[32] G. Buchalla, A.J. Buras, Phys. Rev. D 54 (1996) 6782, hepph/9607447.
[33] S. Adler et al., E787 Collaboration, Phys. Rev. Lett. 79 (1997) 2204, hepex/9708031;
S. Adler et al., E787 Collaboration, Phys. Rev. Lett. 84 (2000) 3768, hepex/0002015.
[34] T.K. Komatsubara, E787 Collaboration, hepex/9905014.
[35] Y. Nir, D.J. Silverman, Phys. Rev. D 42 (1990) 1477;
Y. Grossman, Y. Nir, R. Rattazzi, hepph/9701231.
[36] S. Glenn et al., CLEO Collaboration, Phys. Rev. Lett. 80 (1998) 2289, hepex/9710003.
[37] G. Buchalla, G. Hiller, G. Isidori, Phys. Rev. D 63 (2001) 014015, hepph/0006136.
[38] A. Stocchi, hepph/0010222.
[39] A. Ali, D. London, Eur. Phys. J. C 18 (2001) 665, hepph/0012155.
[40] M. Aoki, M. Nagashima, N. Oshimo, hepph/0104063;
M. Aoki, G. Cho, M. Nagashima, N. Oshimo, hepph/0102165.
[41] M.R. Ahmady, M. Nagashima, A. Sugamoto, hepph/0105049.
[42] C.V. Chang, D. Chang, W. Keung, Phys. Rev. D 61 (2000) 053007;
M. Aoki, E. Asakawa, M. Nagashima, N. Oshimo, A. Sugamoto, Phys. Lett. B 487 (2000) 321,
hepph/0005133.
[43] G. Barenboim, G. Eyal, Y. Nir, Phys. Rev. Lett. 83 (1999) 4486, hepph/9905397.
[44] G. Eyal, Y. Nir, JHEP 9909 (1999) 013, hepph/9908296.
[45] B. Aubert et al., BaBar Collaboration, Report No. SLACPUB8540, hepex/0008048.
[46] H. Aihara, Belle Collaboration, Belle note 353, hepex/0010008.
[47] T. Affolder et al., CDF Collaboration, Phys. Rev. D 61 (2000) 072005, hepex/9909003.
[48] J.P. Silva, L. Wolfenstein, Phys. Rev. D 63 (2001) 056001, hepph/0008004;
G. Eyal, Y. Nir, G. Perez, JHEP 0008 (2000) 028, hepph/0008009;
Z. Xing, hepph/0008018;
A.J. Buras, R. Buras, Phys. Lett. B 501 (2001) 223, hepph/0008273.
[49] A. Masiero, O. Vives, Phys. Rev. Lett. 86 (2001) 26, hepph/0007320;
A. Masiero, M. Piai, O. Vives, Report No. FTUV1208, hepph/0012096.
Abstract
We present a selftuning solution of the cosmological constant problem with one extra dimension
which is curved with a warp factor. To separate out the extra dimension and to have a selftuning
solution, a three index antisymmetric tensor field is introduced with the 1/H 2 term in the Lagrangian.
The standard model fields are located at the y = 0 brane. The existence [1] of the selftuning solution
(which results without any fine tuning among parameters in the Lagrangian) is crucial to obtain
a vanishing cosmological constant in a 4D effective theory. The deSitter and antideSitter space
solutions are possible. The deSitter space solutions have horizons. Restricting to the spaces which
contain the y = 0 brane, the vanishing cosmological constant is chosen in the most probable universe.
For this interpretation to be valid, the existence of the selftuning solution is crucial in view of the
phase transitions. In this paper, we show explicitly a solution in case the brane tension shifts from one
to another value. We also discuss the case with the H 2 term which leads to onefinetuning solutions
at most. 2001 Published by Elsevier Science B.V.
PACS: 04.50.+H; 11.25.Mj; 98.80.Es
Keywords: Cosmological constant; Selftuning; Fourform field; Brane
1. Introduction
The cosmological constant problem [2] is probably the most important clue to the
physics at the Planck scale. Most attempts toward solutions of the cosmological constant
problem introduce additional ingredients [38], except the wormhole and anthropic
solutions of the problem [6,9].
The wormhole solution is based on the probabilistic interpretation that the probability
to have a universe with a vanishing cosmological constant is the largest [6]. But in
the evolving universe where the true vacuum is chosen as the universe cools down,
Email address: jekim@phyp.snu.ac.kr (J.E. Kim).
05503213/01/$ see front matter 2001 Published by Elsevier Science B.V.
PII: S 0 5 5 0  3 2 1 3 ( 0 1 ) 0 0 3 9 7  2
307
308
the RS type models, the brane(s) has a special meaning in the sense that the matter fields
can reside at the brane only. However, the effective gravitational interaction of the matter
fields is obtained after integrating out the fifth dimension y. For this effective theory to
make sense, we must require that:
(i) the metric is wellbehaved in the whole region of the bulk, and
(ii) the resulting 4D effective Planck mass is finite.
The condition (i) is to find a solution without a singularity in the region defined. In this
regard, if the warp factor vanishes at say y = ym , then ym becomes the horizon and the
universe connected to the matter brane is up to y = ym . In the RSI models, it is a disaster
if 0 < ym < 1/2 is between the two branes located at y = 0 and y = 1/2. This is because
one needs both branes for the consistency in the RSI models. But in the RSII models,
if there exists ym then one can consider the space only for ym < y < ym if ym is not
a naked singularity. Indeed, it can be shown that it is consistent to consider the space
up to this point only by calculating the effective cosmological constant by integrating
y to ym . The localized gravity condition (ii) restricts the solutions severely, since the y
integration in some solutions would give a divergent quantity for MP or MP get more
important contributions as y .
The EinsteinHilbert action with a bulk cosmological constant and a brane tension in the
RSII model does not allow a selftuning solution. Addition of the GaussBonnet term [21]
does not improve the situation, and also it does not help to regularize a naked singularity in
the selftuning model with a bulk scalar [14]. But addition of the three index antisymmetric
tensor field AMNP (and the field strength HMNP Q and H 2 HMNP Q H MNP Q ) allows a
selftuning solution [1].
In this paper, we discuss the RSII model with antisymmetric tensor field added. The
case with 1/H 2 term allowing the selftuning solution is the main motivation for this
extensive study. HMNP Q has been considered before in connection with the cosmological
constant problem [3] and the possible compactification of the seven internal space in
the 11D supergravity [22]. Even in the 5D RandallSundrum model HMNP Q is useful
to separate the 4dimensional space. The 1/H 2 term looks strange, but the consideration
of the energymomentum tensor would require a nonvanishing HMNP Q , triggering the
separation of the extra dimension from the 4dimensional space. Below the Planck scale,
we consider that the action is an effective theory. Above the Planck scale, we consider
that the quantum gravity effects may be very important, but at present the final form for
quantum gravity is not known yet.
In this paper, the selftuning solution means that it does not need a finetuning between
the parameters in the action, which is a progress toward understanding the cosmological
constant problem. The selftuning solutions are found from time independent Einstein
equations. However, the existence of the selftuning solution alone does not solve the
cosmological constant problem completely. It is because if the ansatz for a time dependent
metric allows, for example, the deSitter space solutions with the antisymmetric tensor field
added then choosing the flat space is simply choosing a boundary condition. However,
the existence of the selftuning solution and the probabilistic interpretation for the wave
309
function of the universe can provide a logical understanding of the vanishing cosmological
constant even in this case [5]. Suppose that we start from a flat universe from the
beginning a la the wormhole interpretation of the vanishing cosmological constant. But
this interpretation alone may encounter a difficulty when the phase transitions such as the
electroweak phase transition or the QCD phase transition add a nonvanishing cosmological
constant at a later epoch. However, the existence of the selftuning solution chooses the flat
space solution out of numerous possibilities when the universe goes through these phase
transitions. If there exists a selftuning solution, then the wormhole interpretation chooses
the vanishing cosmological constant even after these phase transitions.
We find that the 1/H 2 term does always allow deSitter space solutions with localized
gravity. In the RS model, however, a negative brane tension (1 < 0) does not allow
a localized gravity in the deSitter space [23], but allows only a nonlocalized gravity.
There are arguments excluding these nonlocalized gravity [24], and in the RSII model
for a negative brane tension one may exclude the deSitter space solutions. However, the
nonlocalized gravity cannot become a strong argument for the vanishing cosmological
constant. It simply means that the deSitter space solution with nonlocalized gravity
cannot materialize to our universe. At some point, we may invoke an anthropic principle
or turn to a probabilistic interpretation. Namely, as long as there exist solutions for
nonzero cosmological constants whether there results a localized gravity or not, we need
a probabilistic interpretation. For a probabilistic interpretation, the existence of the selftuning solution is crucial in choosing the flat universe.
In Section 2, we present the flat space solutions with 1/H 2 term and with H 2 term.
It is shown that 1/H 2 term allows the flat space selftuning solution but H 2 term allows
at most onefinetuning solutions. DeSitter space and antideSitter space solutions are
also commented. In Section 3, it is shown that the 1/H 2 term allows the antideSitter
and deSitter space solutions. We discuss the horizons appearing in our solutions. We also
discuss how the universe chooses the vanishing cosmological constant. In Section 4, we
present a timedependent solution such that the 4D spacetime remains flat when the brane
tension shift instantaneously to another value. Section 5 is a conclusion.
310
We also present onefinetuning solutions for H 2 term with time independent metric, and
compare with the other known tuning solutions [14,19].
2.1. A selftuning solution of the cosmological constant with 1/H 2
A selftuning solution exists for the following action,
2 4!
1
4
b + Lm (y) ,
S = d x dy g R +
2
HMNP Q H MNP Q
(1)
which will be called the KKL model [1]. Here we set the fundamental mass parameter M
as 1 and we will recover the mass M wherever it is explicitly needed. We assume a Z2
symmetry of the warp factor solution, (y) = (y). The sign of the 1/H 2 term is chosen
such that at the vacuum the propagating field AMNP has a standard kinetic energy term.
The action contains the 1/H 2 term which does not make sense if H 2 does not develop
a vacuum expectation value. Since the cosmological constant problem is at the bottom
of most cosmological application of particle dynamics, it is worthwhile to study any
solution to the cosmological constant problem. We note that this problem has led to so
many interesting but unfamiliar ideas [35,7]. Therefore, any new idea in the possible
interpretation of the cosmological constant problem is acceptable at this stage. In fact,
we found a very nice solution with the above action and hence we propose the action
(1) as the fundamental one in gravity. Being a part of gravity, we do not worry about the
renormalizability at this stage.
Flat space solution. The ansatz for the metric is taken as
ds 2 = 2 (y) dx dx + dy 2,
(2)
+3
G = g 3
,
G55 = 6
2
,
(3)
where prime denotes differentiation with respect to y. With the brane tension 1 at the
y = 0 brane and the bulk cosmological constant b , the energy momentum tensors are
TMN = gMN b g M N
1 (y)
4
1
1
P QR
+ 4 4!
g
.
H
H
+
MP QR N
MN
2
H4
H2
(4)
The specific form for H 2 HMNP Q H MNP Q in Eq. (1) makes sense only if H 2
develops a vacuum expectation value at the order of the fundamental mass scale. Because
of the gauge invariant four index HMNP Q , four spacetime is singled out from the five
dimensions [22]. The four form field is denoted as H ,
H =
$
,
g
n(y)
(5)
311
where , . . . run over the Minkowski indices 0, 1, 2, and 3. With the above ansatz, the field
equation for the fourform field is satisfied,
H MNP Q
M
(6)
g
= 0.
H4
There exists a solution for b < 0. The two relevant Einstein equations are the (55) and
() components,
2
8
,
6
(7)
= b
A
2
8
3
(8)
= b 1 (y) 3 ,
+3
A
where A is a positive constant in view of Eq. (5). It is easy to check that Eq. (8) in the
bulk is obtained from Eq. (7) for any b , 1 , and A. This property is of the specialty
of the HMNP Q field. Near B1 (the y = 0 brane), the function must be generated by
the second derivative of . The Z2 symmetry, (y) = (y), implies (d/dy)(y)0+ =
(d/dy)(y)0 . Thus,
d2
d2
d
(9)
(y) = 2 (y)
+ 2(y)
(y).
dy
dy 2
dy
y=0
This function condition at B1 leads to a boundary condition
k1 ,
y=0+
(10)
where we define ks in terms of the bulk cosmological constant and the brane tension,
b
1
k ,
(11)
k1
.
6
6
Let us find a solution for the bulk equation Eq. (7) with the boundary condition Eq. (10).
We define a in terms of A,
1
.
a=
(12)
6A
The solution of Eq. (7) consistent with the Z2 symmetry is
1/4
1/4
k
cosh 4ky + c
,
(y) =
(13)
a
where c is an integration constant to be determined by the boundary condition Eq. (10).
This solution, consistent with condition (i), is possible for any value of the brane
tension 1 . Note that c can take any sign. This solution gives a localized gravity consistent
with the above condition (ii). The boundary condition (10) determines c in terms of b
and 1 ,
1
1 k1
1
= tanh
.
c = tanh
(14)
k
6b
A schematic shape of (y) is shown in Fig. 1.
312
Fig. 1. (y) as a function of y for the flat space ansatz. It is plotted for k = 1 and a = 1.
M3
=
2ka
1
1(cosh(c))1
dx
.
(1 x 2 ) 1 12 x 2
(15)
(16)
Note that the Planck mass is given in terms of the integration constant a, or the
integration constant is expressed in terms of the fundamental mass M and the 4D Planck
mass MP,eff ,
2
M3
1
a=
(17)
.
F ,
2
MP,eff
2k
2 0
Curved space solution. The curved space solution is a timedependent solution which
will be discussed in the KKL model of Section 3.
Localization of gravity and no tachyon. The perturbed metric near the above background
is
ds 2 = 2 + h dx dx + 2h 5 dx dy + (1 + h 55 ) dy 2
1
2
( + h ) dx dx + 2h5 dx dz + (1 + h55 ) dz2 ,
K (z)
(18)
(19)
313
Then the linearized Einstein and energymomentum tensors near the above background
solution are given by
1
hMN + (M P h N)P MN P Q h P Q
2
2
3K
(M hN5 + N hM5 5 hMN )
2K
K
P
K
2
K
2
K
K
3MN
+ 2 2 h55
hP 5 3
2 2 hMN ,
K
K
K
K
K
GMN =
K
2
K
3k1
18a 2
+ 6 2 h + (z)
h55 10 h55 ,
= 3
K
K
K
K
(20)
(21)
K
2
6a 2
(22)
h
h55 ,
55
K2
K 10
K
2
T5 = 6 2 h5 ,
(23)
K
where a is given in Eq. (12) and
denotes the derivative with respect to z and M
MN N , + z2 . h MN is defined as h MN hMN MN h/2 and h is the trace
for hMN . When deriving the linearized expression for the energymomentum tensor near
the background, we used Eqs. (7) and (8).
The 5dimensional graviton has 15 degrees of freedom maximally due to its symmetric
property in 5dimensional spacetime. However, the 5dimensional graviton has at most 9
physical degrees of freedom, because it is decomposed into the tensor, vector and scalar
modes in 4 dimension spacetime, which have maximally 5, 3, 1 degrees of freedom,
respectively. Hence, we should take at least 6 gauge conditions to analyze the spectrums
including the 4dimensional massive modes from the 5dimensional graviton. After solving
the equations of motion we could, of course, require additional gauge conditions and
reduce the degrees of freedom for the massless modes.
As K(z), which is a part of metric, specifies a direction z (or y) the equation of motion
for the graviton gives a relation between h55 and other components of metric fluctuations
as shown in Eqs. (20)(23). Thus we should carefully choose the gauge conditions such that
they are consistent with the relation given by the equations for the 5dimensional graviton.
We find that the following gauge conditions are appropriate ones,
T55 = 6
M h MN = 0,
(24)
h55 = 0.
(25)
Due to the above gauge condition, the remaining degrees of freedom are 9. Under the
gauges, linearizations of the Einstein equations in the conformal coordinate Eq. (19) are
1
3K
h + h
( h5 + h5 5 h ) = 0,
2
4
2K
h = 0,
K
+6
h5 = 0,
K
(26)
(27)
(28)
314
where Eqs. (26)(28) are (), (55), (5) components of the linearized Einstein equations,
respectively.
The trace part of Eqs. (26) and (27) describe the scalar mode in the 5dimensional
graviton. Here we can confirm the consistency by checking that the trace of Eq. (26) is
the same as Eq. (27). Eq. (26) is also the equation for the tensor mode (4dimensional
graviton), and Eq. (28) is the equation for the 4dimensional vector mode. From Eq. (27),
we see the scalar mode h does not have any tachyonic state.
To solve the Eq. (28), let us separate the variables of the vector mode such as h5 (x, z) =
h 5 (x)V (z). Then Eq. (28) is
4 m2 h 5 (z) = 0,
24a 2
12k1
z2
(z) 10 V (z) = m2 V (z),
K
K
(29)
(30)
where m2 is interpreted as the 4dimensional effective mass for the KK mode. Eq. (30) is
rewritten in the y coordinate, which is defined in Eqs. (18) and (19), as
2 2
y +
y + 12k1 2 (y) + 24a 2 10 + m2 V (y) = 0,
(31)
where we used Eqs. (7) and (8), and in this coordinate h 5 (x, y) h 5 (x) V (y).
Note that under the Z2 symmetry h and h 55 have even parity but h5 has odd parity
h (y) = h (y),
(32)
h 55 (y) = h 55 (y),
(33)
h 5 (y) = h 5 (y).
(34)
(35)
We see that if the solution has odd parity, all terms except the second one in Eq. (31), are
explicitly zero at y = 0. Therefore, any stable odd parity solution of h5 is not allowed
due to the boundary conditions, even though there could exist bulk solutions. Note that
the above result was possible in our nonflat background case, because
(y = 0) = 0 and
y V (y = 0) = 0 generally in the linear derivative term in Eq. (31). Thus, the Gaussian
normal condition h 55 = h 5 = 0 [20] is a natural gauge in our case.
With Eqs. (25)(27) and h 5 = 0, we derive () equation as
K
3 5 h = 0,
(36)
K
or
(37)
15K 2 3K
2K
4K 2
15 2 3k1
39a 2
k
=
(z)
4K 2
K
4K 10
2
2
9 C
3 C
=
+
,
4 z
2 z2
315
V (z) =
(38)
where K(z) = 1/(y(z)) = exp(C). Therefore, we can see that m2 0, i.e., there is no
Tachyon state near the background, by regarding the above equation as a supersymmetric
quantum mechanics [19],
3 C
3 C
Q Q (z) z +
(39)
z +
(z) = m2 (z).
2 z
2 z
Moreover, there appears the localization of gravity on the brane as expected from the finite
4D Planck mass, due to one bound state of massless graviton, 0 (z) e3C/2 = K 3/2 , that
is, h0 = c$ eipx (c is a constant). Note that the zero mode solution 0 (z) automatically
k1
z=0+ = 0 from Eq. (37). Since V (z) 0 as
satisfies the boundary condition z + 2K
z , there is no mass gap for KK massive modes, which start with m2 = 0. The KK
massive modes just gives rise to small corrections to the Newton potential at the length
scale larger than the curvature scale 1/k as in the RS case [12].
More general case with 1/H 2n . So far we presented a model for a simple Lagrangian
with 1/H 2 term only. But we can show that more general Lagrangians containing only
negative powers of H 2 can have the selftuning solutions. Suppose that Lagrangian
contains n an /(H 2 )n with a1 > 0 and large for the 1/H 2 dominance, the last terms
of the (55) and () Einstein equations, (7) and (8), are changed to
Cn ()
n
An
(2n + 1)Cn ()
n
An
(40)
respectively, with A1 > 0. Then checking the two equations, we obtain Cn () = 8n for
the consistency. Then, the (55) equation gives
8n+2
b
.
  = 2
(41)
6
6An
n
Eq. (41) gives
0 as 0 if n 0, which guarantees the existence of the solution.
But for n < 0, there exists a naked singularity and there is no solution.
This observation is very useful when we consider the radiative corrections on the
background of our selftuning solution. Suppose that the 1/H 2 is given with possibly
small additional terms of other negative powers of H 2 . This means that our solution is not
expressible in terms of the cosh1/4 , but there is a solution. Then the radiative corrections
which are powers of H 2 can be brought approximately into the form n0 an /(H 2 +
H 2 )n where the n = 1 term is dominant. Thus, if the corrections contain only the (H 2 )n
type terms, the existence of the selftuning solution is intact.
316
M
M
S = d 4 x dy g
R
HMNP Q H MNP Q
2
2 4!
b +
L(i)
(y
y
)
.
(42)
i
m
i
The ansatze for the metric and the four form field are taken also as given in Eqs. (2) and
(5), respectively. Thus the field equation for AMNP is trivially satisfied again,
M g H MNP Q = 0.
(43)
The two relevant Einstein equations are the (55) and () components,
2
A
6
= b + 8 ,
2
A
3
+3
= b 1 (y) 2 (y yc ) 8 ,
1/4
a
sinh 4ky + c
,
for b < 0: (y) =
k
1/4
1/4
a
,
for b > 0: (y) =
sin 4ky + c
k
1/4
,
for b = 0: (y) = 4ay + c
(44)
(45)
(46)
(47)
(48)
d
d2
d2
(y).
(50)
(y)
=
(y)
+ 2 (y) (y yc )
dy 2
dy 2
dy
y=0
Then the boundary conditions for the above three cases are
1 k1
1 k2
for b < 0: c = coth
= 4kyc coth
,
k
k
k1
k2
= 4kyc cot1
,
for b > 0: c = cot1
k
k
a
1
for b = 0: c =
= a 4yc
,
k1
k2
317
(51)
(52)
(53)
2
.
6
(54)
We note that in the case of the H 2 term (not 1/H 2 ) in the action, the onefinetuning
relations between k1 and k2 appear always, e.g., the relations (53)(55), while in the case
of RSI model, the twofinetuning relations k = k1 = k2 were inevitable.
If we introduce both 1/H 2 and H 2 in the action, there does not exist a flat space selftuning solution. In this case, the derivative of the warp factor satisfies
8
A
b 2
+ 8.
= +
(55)
6
6A 6
A necessary condition for the existence of a flat space solution is
0 and 0 as
y . Certainly, Eq. (55) does not satisfy this necessary condition. On the other hand,
a necessary condition for the deSitter space regular horizon is that
= finite as 0.
Also, this condition is not met in Eq. (55) and hence there does not exist deSitter space
regular horizon.
2.3. Comparison with other tuning solutions
There appeared several selftuning solutions since early eighties [3,5,14,19]. The early
stage scenario [3,5] used field strength of three index antisymmetric tensor field H to
introduce an integration constant. The vacuum value of H = $ c introduces an
integration constant c which contributes to the cosmological constant c2 . Therefore,
there exists a value of c such that the effective cosmological constant vanishes for a
range of bare cosmological constant. Once c is determined to give the zero effective
cosmological constant, c cannot change since it is a constant. In this sense, the fourdimensional example is not a working model. Thus, the selftuning solution needed a
dynamical field to propagate. The three index antisymmetric field is not a dynamical field
in 4D spacetime.
Introduction of the extra dimension opened a new game in the selftuning solutions of
the cosmological constants [1,14,1719,26]. Here, we briefly comments on the key points
of these solutions.
The California selftuning solution [14,15] must use the RSI model set up, i.e., introduce
two branes B1 and B2, and introduce a specific form for the potential of a bulk scalar field
318
coupling to the brane matter with a desired form. But the model has a naked singularity and
to cure the problem, one has to introduce a brane at this singular point. Then, one needs
a condition at the new brane and leads to one finetuning there [15]. If one satisfies this
onefinetuning condition between parameters in the Lagrangian, there always exists a flat
space solution. Therefore, the solution is not a selftuning solution originally anticipated.
One interesting or (disastrous to some) point of this model is that it does not allow the
deSitter or antideSitter space solutions. In this sense, it is an improvement over the
original proposal [3,5] for the selftuning solutions. However, it may be difficult to obtain
a period of inflation since the deSitter space exponential expansion is not possible. One
should see whether the bulk energymomentum tensor satisfies = p = p5 /2 to have
the exponential expansion of the effective 4D space [25]. However, at present it is not
known what matter satisfies this kind of equation of state.
There appeared another interesting selftuning solution [19,27], which does not
introduce any brane. This model assumes a specific form of the 5D scalar potential
multiplied to the matter Lagrangian in 5D. Even though the metric gives a localized gravity,
the matter fields propagate in the full 5D space since the potential tends to a constant value
as y . It is not certain how we are forbidden to realize the extra dimension in this
model.
Then, there are proposals that the deSitter space solution is phenomenologically
acceptable as far as the curvature at present is sufficiently small [1618,26]. In a sense,
it also tries to accommodate the current small vacuum energy [28] with the deSitter space
solutions. (On the other hand, note that the quintessence idea is based on the solution of
the cosmological constant problem [29].) The way the proposals make the vacuum energy
small is to separate the distance between branes sufficiently large since the vacuum energy
is exponentially decreasing with the separating distance. The one finetuning at B1 is met
but the second fine tuning at B2 is not satisfied and allows the deSitter space solutions.
To have the vacuum energy decreased down to the current energy density, one needs that
the separation distance increases as t increases. Namely, the horizon point at the y axis
is required to increase. Then it is possible to relate the current vacuum energy with the
current mass energy. In Ref. [26], the relation = (5/2)m is obtained for total = 1.
The model presented in [1] allows a selftuning flat space solution, selftuning deSitter
and antideSitter space solutions. In general it is possible to introduce a period of inflation.
The transition from one flat space to another flat space can arise following the solutions of
the Einstein and field equations, as we discuss in the subsequent sections.
319
ds 2 = 2 (y) dt 2 + e2 t ij dx i dx j + dy 2
= 2 (y)g dx dx + dy 2
(56)
1
1 k1
sinh k(ym y) , ym = coth
(y) =
(57)
,
k
k
k
(4)
R
where the integration constant ym is determined from the boundary condition at the brane
and
b
1
k1
.
k ,
(58)
6
6
For a positive tension brane, we get a positive value of ym , which gives rise to an event
horizon at the finite proper distance away from the brane. Therefore, the region beyond
the horizon cannot be causally connected to the region where the brane resides. This
bulk horizon resembles the dS4 horizon on the brane. And, the dS4 solution is regular
along the bulk because the curvature tensors become finite at the bulk horizon. To say
about whether the dS4 solution could describe a dynamical compactification of the extra
dimension with the region beyond the horizon cut off, we should check its consistency
from the 4D effective cosmological constant by integrating out the fifth dimension. The
4D effective action is given by
ym
1
d x g (4)
dy 4 R (5) b 1 (y) ,
2
S4,eff
ym
2
MP,eff (4)
R 3eff ,
d 4 x g (4)
2
(59)
where g (4) and R (4) are given from g = g + h in the zero mode expansion and the
4D effective Planck mass and the 4D effective cosmological constant are determined as
follows,
1
2
MP,eff = 3 kym + sinh(2kym ) > 0
(60)
2
k
and
1
eff =
3
=
ym
2
+6
dy 4
+ b + 1 (y)
ym
2
k3
kym +
1
sinh(2kym ) > 0.
2
(61)
320
Then, when we compare the 4D Einstein equations of motion derived from the above
4D effective action with those satisfied by the dS4 brane solution, we can show that the
2
ratio eff /MP,eff
is exactly equal to the 4D cosmological constant of the brane solution.
Therefore, the dS4 brane solution can be regarded to reproduce the 4D effective deSitter
spacetime with the extra dimension dynamically compactified up to the horizon distance
in the bulk.
It is easy to observe that 1 < 0 does not allow a horizon, which implies a nonlocalized
gravity.
3.2. The KKL model
The background metric is
ds 2 = 2 (y)g dx dx + dy 2,
(62)
G = 3
(63)
3
2 g ,
+3
2
2 ,
G55 = 6
(64)
6
where > 0 for dS4 and < 0 for AdS4 background. The term arises from the time
derivatives of the metric. Then, the (55) and () components of the Einsteins equations
with 1/H 2 follows,
6
3
2
2 = b
6
2
+3
8
,
A
2 = b 1 (y) 3
3
(65)
8
.
A
(66)
321
For the antideSitter space, let us consider the asymptotic behavior of the warp factor
(y) as y . With z = 1/y, we can rewrite the (55) component as
2
d
k 2 2 a 2 10
(67)
= 4+ 4
,
dz
z
z
z4
where
k2
b
,
6
a2
1
.
6A
(68)
To get the finite 4D Planck mass from the noncompact extra dimension or have the
localized gravity, should be zero as z 0. However, the 4D cosmological constant
term should be divergent as z 0 even if the last two terms are set to be zero. Therefore,
there does not appear the localized gravity for the AdS background with nonzero effective
4D cosmological constants.
In the KKL model [1] the deSitter space metric ansatz gives
as
= k 2 + k 2 2 a 2 10,
(69)
where
k =
,
6
k=
b
.
6
(70)
The boundary condition at y = 0 relates the brane tension (k1 ) and the bulk cosmological
constant,
,
k1 = k 2 a 2 8 (0) +
(71)
(0)2
where (0) is a function of an integration constant c. The above relation (71) is valid for
both the deSitter and antideSitter space solutions.
Consider the half space y > 0. If we choose the minus sign in front of the square root of
Eq. (69), there always exist ym as in the RSII model.
Even if we choose the plus sign in front of the square root of Eq. (69) near y = 0+ , there
exist horizons. This is because there can exist a point where
vanishes for a sufficiently
large in the region where increases. For this not to be realized, should tend to an
asymptotic limit as y . If it goes to an asymptotic limit, the Planck mass becomes
infinite and there does not result a localized gravity. Therefore, we restrict to the case for
turning around and coming down, in which case there results a horizon which is determined
by the point where = 0 at y = ym . Between y = 0 and y = ym , there can (1 < 0) or
cannot (1 > 0) exist a point
= 0. Even if there exists such a point as in the
where
above example, the integral d/ k 2 + k 2 2 a 2 10 always converges, and we obtain
a localized gravity with the horizon at y = ym . Therefore, there always exist deSitter
space solutions, which have the periodic form as shown in Fig. 3. Similarly, there exist
antideSitter space solutions, which have the periodic form as shown in Fig. 2. But the
322
Fig. 2. (y) as a function of y for the antideSitter space solution. It is plotted for k 2 = 0.01, k = 1
and a = 1, where k 2 = .
Fig. 3. (y) as a function of y for the deSitter space solution. It is plotted for k 2 = 0.01, k = 1 and
a = 1, where k 2 = .
antideSitter space solution does not lead to a localized gravity. Even if we allow the antideSitter space solutions, neglecting the unphysical unlocalized gravity, the probability to
choose the antideSitter space is exponentially small [5].
The curved space solutions in the KKL model have the effective bulk cosmological
constant more negative compared to the RSII case. Therefore, the bound from the weaker
energy condition on the background matter supporting those spacetimes is completely
satisfied. On the other hand the weaker energy condition is saturated in the RSII
case [23]. The weaker energy condition reads TMN M N 0 for a null vector M , which
2 for the background metric Eq. (62). In the KKL case,
corresponds to (
/)
323
the weaker energy condition is always satisfied without the possibility of saturation,
2 4a 2 8 <
2 .
=
(72)
Existence of deSitter space solutions in the KKL model allows possible cosmological
constants in 4D space as nonnegative. Hawkings probabilistic interpretation of the
cosmological constant chooses the flat space [5]. However, we need a qualification in this
statement. Existence of nonlocalizable deSitter space solution can give a large probability.
The Euclidean action can be written schematically as
R4
+ ,
SE d 4 x MP2
(73)
2
after the y integration. The deSitter space solutions have horizons as shown in Fig. 3.
If we integrate over the whole region of the y space MP diverges. But the Planck mass
must be from the integration only up to the horizon connected to y = 0 brane which we
call the first universe. Then MP is finite in the first universe. However, if we neglect the
effect of the brane, the probability to go the second universe (the universe between the
next two horizons) is of the same order as the probability to go to the first deSitter space
universe, because the horizon points appear periodically. Summing up the probabilities to
go to either of these deSitter universes, we would obtain an infinity. Nevertheless, the
second universe does not contain the brane, and we have to exclude this possibility of
transition to the universes not containing the brane. Therefore, MP is considered to be
finite, and the probability to stay in the flat universe is maximum [5].
324
integration constant c2 corresponding to the brane cosmological constant new . This situation corresponds to a timedependent energymomentum tensor. This transition from a
flat space solution to another flat space solution is through satisfying the field equations
and is different from Wittens sudden choice of a flat space solution [3]. There can be timedependent solutions from a flat space to deSitter or antideSitter spaces, but the probability for these transitions is exponentially small compared to a flat to flat transition [5].
To study the timedependence the metric is taken as
ds 2 = n2 (t, y) dt 2 + a 2 (t, y)ij dx i dx j + b2 (t, y) dy 2.
The Einstein tensors are
3 a a
b
3 a
a
a
b
+
+ 2
+
,
G00 = g00 2
a
a a
b
n a a b
b
a
1
a b a
n a
b n
Gii = gii 2 2 +
2
2
n
a b a n a
b n
a
a
a
n
a
a
1 n
b
n
+2 +
+2
,
+ 2
2 +
b n
a
a
n
a
b n
a
3 a a n a
3 a
n
a
G55 = g55 2
+ 2
+
,
a
n a a n a
b a n
b a
a
a n
,
+
G05 = 3
a n
ba
a
(74)
(75)
where dot and prime denote differentiations with respect to t and y, respectively.
With the brane tension 1 at the y = 0 brane and the bulk cosmological constant b ,
the energymomentum tensor is
1 (y)
TMN = gMN b g M N
4
1
1
P QR
+ 4 4!
HMP QR HN
+ gMN 2 .
H4
2
H
(76)
1
H 5ij k = $ 5ij k0 0 .
g
(77)
Then, we have
H 2 = 4! f 2 h2 ,
H0NP Q H0 NP Q = 3!g00 f 2 ,
H5NP Q H5 NP Q = 3!g55 h2 ,
HiNP Q Hj NP Q = 3!gij f 2 h2 ,
H5NP Q H0 NP Q = 3!bnf h,
H5NP Q Hi NP Q = 0,
(78)
where f 2 =
2 /b2 and h2 = 2 /n2 . The TMN appearing in the Einstein equations
GMN = TMN are
6
1
4h2
(y) + 2
+
T00 = g00 b +
,
b
f h2 (f 2 h2 )2
6
1
(y) + 2
Tij = gij b +
,
b
f h2
2
4h2
T55 = g55 b + 2
,
f h2 (f 2 h2 )2
T05 =
4bnf h
.
(f 2 h2 )2
= 0
+ 5
M
H4
H4
H4
$ 0ij k5
=
0
5
= 0.
(4!)2
((
/b)2 ( /n)2 )2
((
/b)2 ( /n)2 )2
Note that is given, in the static homogeneous 4D case considered above, by
ab
(y) = 2 2A sinh 4kby + c ,
4k
325
(79)
(80)
(81)
We anticipate that a flat space solution, viz. Eq. (13), with ci = tanh1 (i / 6b ) is
given, probably by the initial condition or Hawkings choice of the cosmological constant.
During a phase transition when old changes to new , we try to show the existence of an
interpolating solution connecting two flat space solutions with c1 and c2 .
Let us consider the case that the brane tension 1 = old changes to 1 = new
instantaneously due to a phase transition by brane matter at the brane. Then the boundary
condition requires a time dependent b,
b(t) = (bnew bold ) (t t0 ) + bold,
(82)
which gives
1/4
1/4
k
(y, t) =
cosh 4kb(t)(y + y1 )
,
a
for t = t0 ,
(83)
where bold , bnew and y1 are constants. c in Eq. (13) is endowed with a timedependence as
given above. The t dependence of generates nonzero timederivatives of our metric in
the Einstein equation. We suppose that the metric dynamics gives rise to the bulk matter
fluctuation near the vacuum, T (m)M N diag(, p, p, p, p5 ),
3 b
= 2
(84)
+
,
b
2
b
b
1
+
p= 2 2 +
(85)
,
b
3
.
p5 = 2
(86)
Note that the bulk matter contribution vanishes when t = t0 , because the matter would be
t0 ).
proportional to (t t0 ), 2 (t t0 ) and (t
326
b
(87)
,
=
b
which leads to G05 = 0, then the 5D continuity equations are automatically satisfied [25],
b
+ 3 ( + p) + ( + p5 ) = 0,
(88)
b
p5
+ 3 (p5 p) + ( + p5 ) = 0.
(89)
a
(y, t) = 2 2A b(t) sinh 4kb(t)(y + y1 ) ,
(90)
4k
where A is the constant. The t dependence of generates also nonzero H ij k5 component
of the four form field in Eq. (77), that would be proportional also to the delta function,
h (t t0 ). Since the h term on the RHS of the Einstein equations appear as
hm /(f 2 h2 )n/2 with 0 m < n, the (t)function of h gives vanishing contribution at
t = 0. Thus, it turns out that T05 = 0 for any t. For t = t0 , Eq. (83) satisfies the Einstein
equations with the f 2 dependence on the RHS [1]. Near t = t0 , TMN are reduced to those
of the RS [12] due to the divergence of or h (which kills the f 2 dependence), and so the
solution becomes the RS solution [12].
We have shown above three solutions: the flat space solution with c1 = tanh1 (old /
6b ) for t < t0 , the RS solution [12] with the finetuning k = k1 , and the flat space
the field equation and the Einstein equations, old suddenly jumps to 6b and again
suddenly jumps to new . If the transition from old to new is not abrupt, there would
exist a solution smoothly connecting the two flat solutions with c1 and c2 . But the above
example shows that at the intermediate stage it would go through the RS solution [12].
The consistency of the solution is achieved by showing Eq. (87) using the divergence
(m)
of h. As the time derivatives of the metric in G55 = T55 + T55
have been identified already
(m)
with p5 of T55 , the remnant reads as
1
h2
b 1
2
= b(t)
(91)
+
,
2
2
2
6
3f h
3 (f h2 )2
from which we can see immediately that Eq. (87) is satisfied,
1
h2
b
b 1
2
= b
+
2
2
2
b
6
3f h
3 (f h2 )2
1
h2
b 1
2
+
=
t
6
3 f 2 h2 3 (f 2 h2 )2
b
,
b
(92)
where we used
1
= 0.
t 2 (t t0 )
327
(93)
Note that Eq. (93) can be shown for a specific representation of the delta function.
We should check whether the equation of motion for the four form field Eq. (80) remains
satisfied even when the constant b is changed to b(t) like Eq. (82). But we can also see
easily that even in the case, it is satisfied always regardless of t = t0 or not. When t = t0 , it
is the static case, and so the equation of motion is trivially satisfied. On the other hand, for
t t0 the equation of motion is also satisfied due to the (t t0 ) divergence of h, which is
checked easily if we use Eq. (93).
We have seen that there exists a timedependent solution when the brane tension changes
from old to new instantaneously. In this case, the universe starting from a flat universe
can go to a flat universe instantaneously. Instead of the sudden shift of the brane tension,
the transition from old to new can be smooth. We argue that in this case also, there exists
a solution since we can replace the function with a smooth function D(t; $) of t defined
in a short interval $, such that lim$0 D(t; $) = (t). The error by introducing $ is at most
$ and hence our solution is approximate, but this shows that there can exists a smooth
solution very close to our approximate solution with an error of order $.
There may exist solutions connecting deSitter space and flat space, deSitter space and
deSitter space, etc. [31]. But the final universe may be chosen probabilistically to be the
flat universe a la Hawking [5].
5. Conclusion
We considered the cosmological constant in the (4 + 1) dimension with a warp factor.
A brane, containing the matter fields, is located at the origin y = 0 of the extra dimension.
To separate the 4D space we introduced a three index antisymmetric tensor field with the
field strength H . The simple form for the H action is considered: the Lagrangian with
1/H 2 and H 2 . We found a selftuning solution, Eq. (13), with 1/H 2 . The H 2 term does
not allow a selftuning solution, but allows nonlocalizable gravity with one brane or onefinetuning solutions with two branes.
We concentrated the discussion related to the selftuning solution, Eq. (13). We also
found that there exist deSitter space and antideSitter space solutions with 1/H 2 term.
For a finite range of parameters of the bulk cosmological constant b and the brane
tension 1 , there always exists the flat space solution but it is not unique. Therefore,
when the boundary condition is changed at the electroweak or QCD phase transitions,
the cosmological constant after the phase transitions can be anything allowed by the
Einstein equations. However, the probability to choose the vanishing cosmological constant
is infinitely large compared to the others [5]. This argument is applicable in our case since
there exists the selftuning solution, Eq. (13).
This may not sound so attractive as any model not allowing deSitter or antideSitter
space solutions. However, if there exists a such model, then it may not allow inflation
328
which is probably needed in cosmology. In the selftuning model the transition from
one integration constant to another integration constant is possible through satisfying the
equations of motion.
For the selftuning solution we presented to work in cosmology, there must be a
natural explanation of the current acceleration of the expansion [28]. There appeared some
proposals [29], but these must be shown to work with the selftuning solution.
Acknowledgements
This work is supported in part by the BK21 program of Ministry of Education, Korea
Research Foundation Grant No. KRF2000015DP0072, CTP Research Fund of Seoul
National University, and by the Center for High Energy Physics(CHEP) of Kyungpook
National University.
References
[1] J.E. Kim, B. Kyae, H.M. Lee, hepth/0011118.
[2] M. Veltman, Phys. Rev. Lett. 34 (1975) 777;
For a review, see, S. Weinberg, Rev. Mod. Phys. 61 (1989) 1.
[3] E. Witten, in: R. Jackiw, H. Khuri, S. Weinberg, E. Witten (Eds.), Proc. Shelter Island II
Proceedings, Shelter Island, 1983, MIT Press, 1985, p. 369.
[4] E. Baum, Phys. Lett. B 138 (1983) 185.
[5] S. Hawking, Phys. Lett. B 134 (1984) 403.
[6] S. Coleman, Nucl. Phys. B 310 (1988) 643.
[7] A. Linde, Phys. Lett. B 200 (1988) 272.
[8] M. Duff, Phys. Lett. B 226 (1989) 36.
[9] S. Weinberg, Phys. Rev. Lett. 59 (1987) 2607;
S.M. Barr, J.F. Donoghue, D. Seckel, Phys. Rev. D 57 (1998) 5480;
T. Banks, M. Dine, L. Motl, hepth/0007206.
[10] G. Dee, J.S. Langer, Phys. Rev. Lett. 50 (1983) 383;
R.C. Brower, D.A. Kessler, J. Koplik, H. Levine, Phys. Rev. Lett. 51 (1983) 1111.
[11] L. Randall, R. Sundrum, Phys. Rev. Lett. 83 (1999) 3370, hepth/9905221.
[12] L. Randall, R. Sundrum, Phys. Rev. Lett. 83 (1999) 4690, hepth/9906064.
[13] V.A. Rubakov, M.E. Shaposhinikov, Phys. Lett. B 125 (1983) 139.
[14] N. ArkaniHamed, S. Dimopoulos, N. Kaloper, R. Sundrum, Phys. Lett. B 480 (2000) 193,
hepth/0001197;
S. Kachru, M. Schulz, E. Silverstein, Phys. Rev. D 62 (2000) 045021, hepth/0001206;
C. Csaki, J. Erlich, C. Grojean, T. Hollowood, Nucl. Phys. B 584 (2000) 359, hepth/0004133.
[15] S. Forste, Z. Lalak, S. Lavignac, H.P. Nilles, Phys. Lett. B 481 (2000) 360, hepth/0002164;
S. Forste, Z. Lalak, S. Lavignac, H.P. Nilles, JHEP 0009 (2000) 034, hepth/0006139.
[16] H.B. Kim, Phys. Lett. B 478 (2000) 285.
[17] S.H.H. Tye, I. Wasserman, hepth/0006068;
E. Flanagan, N. Jones, H. Stoica, S.H.H. Tye, I. Wasserman, hepth/0012129.
[18] J.M. Cline, H. Firouzjahi, hepph/0012090;
N. Tetradis, hepth/0012106.
[19] A. Kehagias, K. Tamvakis, hepth/0011006.
329
Abstract
We consider maximal globally hyperbolic flat (2 + 1)spacetimes with compact space S of genus
g > 1. For any spacetime M of this type, the length of time that the events have been in existence is M
defines a global time, called the cosmological time CT of M, which reveals deep intrinsic properties
of spacetime. In particular, the past/future asymptotic states of the cosmological time recover and
decouple the linear and the translational parts of the ISO(2, 1)valued holonomy of the flat spacetime.
The initial singularity can be interpreted as an isometric action of the fundamental group of S on a
suitable real tree. The initial singularity faithfully manifests itself as a lack of smoothness of the
embedding of the CT level surfaces into the spacetime M. The cosmological time determines a real
analytic curve in the Teichmller space of Riemann surfaces of genus g, which connects an interior
point (associated to the linear part of the holonomy) with a point on Thurstons natural boundary
(associated to the initial singularity). 2001 Elsevier Science B.V. All rights reserved.
PACS: 04.60.Kz
Keywords: Cosmological time; Asymptotic states; Real trees; Marked spectra
1. Introduction
We shall be mainly concerned with maximal globally hyperbolic, matterfree spacetimes
M of topological type S R, where S is a compact closed oriented surface of genus g > 1.
The (2 + 1)dimensional Einstein equation with vanishing cosmological constant actually
implies that M is (Riemann) flat.
After [9] and [26], a large amount of literature has grown up about this (2 + 1)gravity
topic, regarded as a useful toymodel for the higherdimensional case. Two main kinds
of description have been experimented. A cosmological approach points to characterize
the spacetimes in terms of some distinguished global time; for instance the constant mean
curvature CMC time has been widely studied [3,17]. A geometric timefree approach
Email addresses: benedett@dm.unipi.it (R. Benedetti), guada@df.unipi.it (E. Guadagnini).
05503213/01/$ see front matter 2001 Elsevier Science B.V. All rights reserved.
PII: S 0 5 5 0  3 2 1 3 ( 0 1 ) 0 0 3 8 6  8
331
eventually identifies each flat spacetime by means of its ISO(2, 1)valued holonomy
[16,26]. With the exception of the case with toric space (g = 1), there is not a clear
correspondence between the results obtained in these two approaches.
The aim of this paper is to show that this gap can be filled by using the canonical
Cosmological Time CT, that is the length of time that the events of M have been in
existence (see [2]). It turns out that this is a global time which reveals the fundamental
properties of spacetime. It is canonically defined by means of the very basic spacetimes
structures: its casual structure and the Lorentz distance. The cosmological time is
invariant under diffeomorphisms, therefore, the = a level surfaces Sa provide a gaugeinvariant description of space evolution in M. Both the intrinsic and extrinsic geometry
of the surfaces Sa , as well as their past/future asymptotic states, are intrinsic features of
spacetime. The asymptotic states are defined by the evolution of the observables associated
to the length of closed geodesic curves on the surfaces Sa . Remarkably, they recover and
decouple the linear and the translational parts of the holonomy. The study of the asymptotic
states also leads to understand the initial singularity (we will always assume that the space
is future expanding) and the way how the classical geometry degenerates, but does not
completely disappear. The initial singularity can be interpreted as the isometric action of
the fundamental group of S on a suitable real tree. Differently to the case of the CMC
time (for instance), the level surfaces Sa of the CT are in general only C 1 embedded into
the spacetime M. This lack of smoothness takes place on a geodesic lamination on Sa and
is a observable large scale manifestation of the intrinsic geometry of the initial singularity.
Thus the initial singularity admits two complementary descriptions: one, in terms of real
trees and, the second, in terms of geodesic laminations. The existence of a duality relation
between real trees and laminations was already known in the context of Thurston theory
of the boundary of the Teichmller space. It is remarkable that Einstein theory of (2 + 1)gravity sheds new light on this subject and puts duality in concrete form.
In [6] we have also used the cosmological time in order to study certain interesting
families of (2 + 1)spacetimes coupled to particles.
Our main purpose consists of elucidating the central role of the cosmological time and
its asymptotic states in the description of spacetimes. The cosmological time perspective
provides a new interpretation of several facts spread in the literature which are related to
Thurston work. More precisely, the present article is based on, and could be considered a
complement of, Messs fundamental paper [16].
The importance of (2 + 1)gravity has been pointed out by several authors, see for instance [8,9] and [26]; here we simply add a few comments. In general, the models which
have dynamical degrees of freedom associated with the spacetime geometry are of particular interest in physics. Indeed, gravitational interactions are supposed to be described
by general relativity in which the geometry of spacetime admits a nontrivial dynamical
evolution. A satisfactory knowledge of all the classical and quantum aspects of general relativity is still lacking; so, toy models which provide conceptual hints in this directions are
welcome. The matterfree (2 + 1)gravity model with compact space is a remarkable example of general relativistic theory because the complete classical solution [16] has been
produced. In this context, one can then find explicit answers to some open problems. In
332
our article we have explored a few general topics which are related the problem of time in
classical gravity. The resulting conceptual hints that we have obtained are:
(i) a consistent correspondence between the dynamical and the static pictures of
spacetime exists, and has been explicitly produced in the model;
(ii) one can introduce a global canonical time which corresponds to the age of the
universe, this time codifies intrinsic features of spacetime by means of the associated
asymptotic states;
(iii) the asymptotic states are characterized by spectra of observables;
(iv) the geometric structure of the initial singularity gives rise to observable effects
which can be detected at later times;
(v) the space slices of any global time, which plays the role of the age of the universe,
are not necessarily smoothly embedded into spacetime.
(q) can be interpreted as the length of time the event q has been in existence in N . For
example, if N is the standard flat Minkowski space Mn+1 , is the constant valued function, so in this case it is not very interesting. In [2] (see also [25]) one studies the properties
of a manifold N with regular cosmological time function. Recall that is regular if:
(1) (q) is finite valued for every q N ;
(2) 0 along every pastdirected inextensible causal curve.
The existence of a regular cosmological time function has strong consequences on the
structure of N and of the constant surfaces [2]. In particular when is regular, : N
(0, [ is a continuous function, which is twice differentiable almost everywhere, giving a
global time on N denoted by CT. Each level surface is a future Cauchy surface, so that
N is globally hyperbolic. For each q N there exists a futuredirected timelike unit speed
geodesic ray q : (0, (q)] N such that:
q (q) = q,
q (t) = t.
The union of the past asymptotic endpoints of these rays can be regarded as the initial
singularity of N .
The cosmological time function is not related to any specific choice of coordinates in N ;
it is gaugeinvariant and so it represents an intrinsic feature of spacetime. Thus, when
333
the cosmological time is regular, the constant level surfaces and their properties have a
direct physical meaning as they are observables.
We present now two basic examples of spacetime with regular cosmological time, which
shall be important throughout all the paper. To fix the notations, the standard Minkowski
space M2+1 is endowed with coordinates x = (x 1 , x 2 , x 0 ), so that the metric is given by
ds 2 = (dx 1 )2 + (dx 2)2 (dx 0 )2 . M2+1 is oriented and timeoriented in the usual way.
Example 1. Consider the chronological future of the origin 0 M2+1
2 2 2
I + (0) = x M2+1 : x 1 + x 2 x 0 < 0, x 0 > 0 .
Its cosmological time, : I + (0) (0, ), is a smooth submersion; the constanttime
{ = a} surfaces are the (upper) hyperboloids
2 2 2
I(a) = x M2+1 : x 1 + x 2 x 0 = a 2 , x 0 > 0 .
Hence I(a) is a complete space of constant Gaussian curvature equal to 1/a 2 , and of
constant extrinsic mean curvature 1/a. The Lorentzian length of the timelike geodesic
arc connecting any p I + (0) with 0 equals (p); 0 is the initial singularity. Note that
I(a) can be obtained from I(1) by means of a dilatation in M2+1 with constant factor a;
shortly we write I(a) = aI(1). We shall denote by SO(2, 1) the group of oriented Lorentz
transformations acting on M2+1 and by ISO(2, 1) the Poincar group. SO+ (2, 1) denotes
the subgroup of SO(2, 1) transformations which keep I + (0) and each I(a) invariant.
ISO+ (2, 1) is the corresponding subgroup of ISO(2, 1).
Example 2. Let us denote by I + (1, 3) the chronological future in M2+1 of the line {x 1 =
x 0 = 0}
2 2
I + (1, 3) = x M2+1 : x 1 x 0 < 0, x 0 > 0 .
The Lorentzian length of the timelike geodesic arc connecting any p = (x 1 , x 2 , x 0 )
I + (1, 3) with q = (0, x 2 , 0) equals the cosmological time (p). The level surfaces are
2 2
I(1, 3, a) = x M2+1 : x 1 x 0 = a 2, x 0 > 0
and have constant extrinsic mean curvature equal to (1/2a). Each surface I(1, 3, a) is
isometric to the flat plane R2 . To make this manifest, it is useful to consider the following
change of coordinates. Let 2+1 = {(u, y, ) R2+1 : > 0} be endowed with the metric
ds 2 = 2 du2 + dy 2 d 2 . Then, x 1 = sh(u), x 2 = y, x 0 = ch(u), is an isometry
between 2+1 and I + (1, 3). The level set { = a} of 2+1 goes isometrically onto
I(1, 3, a), so this is intrinsically flat. Note that the group of oriented isometries of 2+1 is
generated by the translations parallel to the planes { = a}, and the rotation of angle of
the (u, y) coordinates.
We are going to show that any maximal globally hyperbolic, matterfree (2 + 1)spacetime M, with compact space S, actually has regular cosmological time, and its initial
singularity can be accurately described.
334
3. Flat (2 + 1)spacetimes
A flat spacetime is, by definition, locally isometric to the Minkowski space M2+1 .
We assume that our maximal hyperbolic flat spacetimes are timeoriented and future
expanding, and that these orientations locally agree with the usual ones on M2+1 . The
spacetime structures on S R are regarded up to oriented isometry homotopic to the
identity.
3.1. Minkowskian suspensions
We introduce here the simplest (2 + 1)spacetimes with compact space S of genus g > 1.
Recall that the upper hyperboloid I(1) M2+1 , mentioned in the previous section, is a
classical model for the hyperbolic plane H2 (see [7] for this and other models); the Poincar
disk is another model which can be obtained from I(1) by means of the stereographic
projection shown in Fig. 1. We shall use the Poincar model in Section 4.
Take any hyperbolic surface F = H2 / homeomorphic to S. is a subgroup of
SO+ (2, 1) which acts freely and properly discontinuously on H2
= I(1). I(1) can be
identified with the universal covering of S and with the fundamental group 1 (S). can
be thought also as a group of isometries of the spacetime I + (0) and M(F ) = I + (0)/ is
the required spacetime with compact space homeomorphic to S. We call it the Minkowskian
suspension of F . This construction is wellknown; sometimes M(F ) is also called the
Lorentzian cone over F or the Lbell spacetime based on F . I + (0) can be regarded as the
universal covering of M(F ). Let us now consider the cosmological time of M(F ). The CT
of I + (0) naturally induces the CT of M(F ). Indeed, each level surface Sa of M(F ) has
I(a) as universal covering; moreover, S1 = F and Sa = aF . In this case, the CT coincides
with the CMC time and each level surface Sa smoothly embeds into M(F ). The initial
singularity trivially consists of one point.
its suspension in I + (0)
Notation. Let Y be any subset of I(1), we shall denote by Y
which is defined by Y = a(0,[ aY .
335
336
(1)
337
Remark 3.1. We have the following characterization of J0 . The interior points of this
segment make the subset of U (M0 ) (boundary of the convex set U (M0 )) of the points
with exactly two null supporting planes; the endpoints make the subset of U (M) with
more than two null supporting planes. Recall that a supporting plane at x U (M0 ) is a
plane P such that x P and U (M0 ) P = . P is null if it contains some nulllines.
The covering U (M0 ) M2+1 is flat. To get M0 , we only need to specify the action of
1 = 1 (F0 )
= Z on U (M0 ).
3.3.2. Action of the fundamental group
1 acts on A by the restriction of the action of 0 on I(1). The domain B corresponds
(via the isometry established in Section 2) to B = {(u, y, ) 2+1 ; 0 y r}, so
that the action of 1 on B transported on B is just given by the translation (u, y, )
(u + s, y, ). Finally, if is the translation (x 1 , x 2 , x 0 ) (x 1 , x 2 + r, x 0 ) on M2+1 , then
the action of 1 on C is just the conjugation of 0 by .
The CT of the covering U (M0 ) passes to the quotient M0 = U (M0 )/1 ; each level
surface Sa is only C 1 embedded into M0 , so that it is endowed with an induced
C 1 Riemannian metric. This allows anyway to define the length of curves traced on the
surface Sa and the derived lengthspace distance. Let A = A/1 , B = B/1 and C = C/1 .
Then, Sa B is a flat annulus of depth r and parallel geodesic boundary components
of length as; Sa (A C) can be isometrically embedded into aF0 , and has geodesic
boundary curves of length as. As shown in Fig. 4, Sa can be obtained by cutting F0 along
c and by inserting a annulus of depth r.
Remark 3.2. If g is an element of ISO+ (2, 1) acting on X M2+1 as g(X) = QX + w,
the transformed domain g(U (M0 )) = Q(A B C) + w is, of course, an isometric copy
of the universal covering in M2+1 . The curve = Q(0 ) is a geodesic line of I(1); is the
intersection of I(1) with a suitable hyperplane passing at the origin of M2+1 . Let us denote
by the suspension of ; then
Q(B) =
{ + v},
[0,r]
where v is the unitary (in the Minkowski norm) vector tangent to I(1), normal to , and
pointing towards Q(C ). We also denote Q(B) = B(, v, r). The shape of the CT level
surfaces in g(U (M0 )) is shown in Fig. 5. The initial singularity of g(U (M0 )) is given by
the spacelike segment J = Q(J0 ) + w.
338
t ( ) =
rvi .
i
Note that if 1 (x0 ) and 2 (x0 ) belong to the same component R, then t (1 ) = t (2 ),
whence also t (R) = t ( ) for any such that (x0 ) R, is well defined. U (M) is tiled by
+ w, (ii) B(, v, r) + w, for some translation vector w R3 .
tiles of two types: (i) R
More precisely, the tiles of the first type make the open subset of U (M)
+ t (R) .
R=
R
R
Note that each tile has its own CT; all the cosmological times fit well together and define
the CT of U (M) which passes to the quotient M.
339
Remark 3.3. The construction of M(F ) and of M(F, L) can be performed for any
hyperbolic surface F , not necessarily compact nor of finite area. Similarly, by starting from
any locally finite family of weighted geodesic lines in I(1), the simplicial deformation that
we have just described produces a globally hyperbolic spacetime structure on R2 R with
a regular cosmological time.
Remark 3.4. When F is compact, every region R (defined above) is bounded by infinitely
In fact, as F is compact, every with = 1 is of hyperbolic type [7].
many lines of L.
and for every with ( ) = , and ( ) are ultraConsequently, for any L
parallel. This means that the hyperbolic distance satisfies d(, ( )) > 0; moreover,
and ( ) have a common orthogonal geodesic line. Suppose now that a region R is
In this case, R contains a band E of infinite diameter,
bounded by finitely many lines in L.
As F is compact, H2 is tiled by tiles of the form
bounded by two halflines contained in L.
(D), where D is a fundamental domain for of finite diameter. So, one (at least) tile
= and this contradicts the fact that
(D) must be contained in E. But clearly (D) L
R is a region of H2 \L.
The same conclusion holds if F is of finite area. If F is of infinite area, we can eventually
0 just consists of
have different behaviours. For instance, in the example F0 above, L
2
one component which divides H into two regions. Other examples will be presented in
Subsection 4.1.
4. The cosmological time of (2 + 1)spacetimes
In this section we describe the main properties of the CT for an arbitrary spacetime M.
We adopt the notations of the previous sections; in particular, M is assumed to be an
expanding matterfree spacetime of topological type S R with compact surface S of
genus g > 1. The validity of the following statements can be quite easily checked for
spacetimes of simplicial type. We shall try to point out the main ideas; we postpone a
commentary on the proofs, with references to the existing literature.
Proposition 4.1. The cosmological time function, : M (0, [, is surjective and
regular, so that it defines a global time on M. It lifts to a regular cosmological time on
the covering, : U (M) (0, [. Each level surface
Sa of U (M) maps onto Sa of M and
is its universal covering. In other words, the action of 1 (S) on U (M) restricts to a free,
properly discontinuous isometric action on
Sa such that Sa =
Sa /1 (S). Each
Sa (Sa ) is a
future Cauchy surface.
4.1. Initial singularityexternal view
Let us give a description of the initial singularity of M as it appears from the exterior
point of view, that is, from the Minkowski space in which the universal covering U (M) is
placed. In Subsection 4.5 we shall show how the initial singularity can also be characterized
in terms of the observables associated with the CT asymptotic states.
340
341
= H2 .
Fig. 7. Partial tilings of F
342
343
Remark 4.11. If M is the spacetime of simplicial type which corresponds to the weighted
multicurve L on the surface F , then La is just made by the boundary components of the
flat annular components embedded into Sa , which are associated to the components of L.
The content of the last remark generalizes as follows. Recall that 1 (S) acts as on
each I(a). For every a (0, [, let us consider the map
Sa I(a)
pa :
defined as follows: pa (x) is the unique point of I(a) such that the tangent plane to
Sa
at x is parallel to the tangent plane to I(a) at pa (x). This map is welldefined, surjective
and 1 (S)equivariant. By taking the union of the pa s we get a 1 (S)equivariant map
p : U (M) I + (0), respecting the CT. This induces a map p : M M(F ) respecting the
CT.
Proposition 4.12.
(1) There exists a geodesic lamination F on F = I(1)/ , which lifts to a geodesic
and any leaf of
on I(1), such that, for every a, one has pa (
La ) = a F
La
lamination F
is isometrically mapped onto a leaf of a F . That is, the union of pa (La )s covers the
of F
.
suspension F
(2) F is the disjoint union of two sublaminations
F = L F ,
where L is the maximal multicurve sublamination of F . Note that either L or F may be
empty. Then
) isometrically into I + (0) respecting the CT;
(a) p embeds U (M)\p1 (F
1
(b) p embeds U (M)\p (L) continuously into I + (0) respecting the CT.
is the union of components of the type B(, v , r) + w, so that
(3) The set p1 (L)
1
(p ) (L) Sa is the disjoint union of flat annular components of Sa , like in the case of a
spacetime of simplicial type.
We have an immediate corollary concerning the intrinsic and extrinsic geometry of the
constant CT surfaces.
a =
a is either
Corollary 4.13. W
Sa \
La is an open dense set of
Sa . Each component of W
isometric to an open set of I(a) or is a flat band which embeds into I(1, 3, a), and projects
onto an annulus of Sa . Flat annuli do occur only if L is nonempty.
4.3. CT duality
To sum up, two geometric structures are naturally associated to the spacetime M: the
real tree T (the initial singularity) and the geodesic lamination F on F = I(1)/ which
reflects the lack of smoothness of the embedding of Sa into M. We have already noted that
for a spacetime of simplicial type these two objects are dual to each other. Here we want
to strengthen and generalize this point.
344
345
type associated to L, and let F be the = 1 level surface of this spacetime. Let us denote
by F the lamination on F which coincides with F outside the flat annuli of F and is
defined as L1 above on these annuli. If L is empty, set F = F . The measured lamination
(F , ) extends to a measured lamination (F , ) on F . The flat annular components
are foliated by closed geodesics parallel to the boundary components. These annuli are
endowed with a plain transverse measure of total mass equal to the corresponding annulus
of F with the lifted (1 equivariant) measured
depth. Take the universal covering F
by closed subsets in the very
geodesic lamination (F , ). Now define a partition of F
S1 , with respect to the lamination
L1 .
same way we have defined above the partition T of
Call again this partition T . We can give it a distance d which makes it an Rtree. If E and
E are the closure of two components of the complement of the lamination, take two points
is transverse to
x and x in these closed sets such that the geodesic segment [x, x ] of F
the leaves of the lamination. By integration, the transverse measure induces a distance on
the subset of T made by the closed sets intersecting [x, x ]. In fact, by the invariance of
the measure, this distance does not depend on the segment [x, x ]. Finally one verifies that
in this way one can actually define a distance between any two points of T and that the
resulting (T , d) is a geometric real tree.
Remark 4.17. Clearly, weighted multicurves on the surface F dually correspond to
geometric simplicial real trees; the spacetimes of simplicial type do materialize this duality.
4.4. Reconstruction of M = U (M)/
Starting from (F = I(1)/, T ) or, equivalently, from (F = I(1)/, (F , )), one can
reconstruct a cocycle t, whence M = U (M)/ . This generalizes what we have done for
a spacetime of simplicial type in Subsection 3.3.
, )
With the notations introduced at the end of the previous subsection, consider (F
on F . To recover a cocycle t do as follows: fix one base point p0 on F out from the
support of the lamination. Let p0 be its image on F . If is a loop in F based on p0 , which
which starts at
represents an element [ ] of (F , p0 ), lift it to the oriented arc in F
p0 ; up to homotopy we can assume that is transverse to the leaves of the lamination.
Let f be any continuous R3 valued function on which coincides with the unit normal
, and oriented in agreement with . Now
to the leaves of the lamination, tangent to F
we can integrate f along by using the transverse measure getting a vector t ([ ]). By
varying [ ], one gets such a cocycle t.
4.5. CT asymptotic states
The above discussion tells us that any spacetime M = U (M)/ is completely
determined by the linear part of its holonomy (or equivalently by the surface F =
I(1)/ ) and by its initial singularity i(M) = T /1 (S). The aim of this subsection is to
recover these geometric objects from the internal point of view by working inside the
spacetime. More precisely, we would like to show that F and T can be interpreted as
346
the future and past asymptotic states for the geometry of the CT level surfaces. To this
aim we shall consider the observables defined by the lengths of the curves on the CT level
surfaces. It is convenient to introduce the concept of Marked Spectrum associated with a
d) which is endowed with an action of the surfaces fundamental group
metric space (X,
1 (S). Whenever we shall refer to X, we shall actually refer to the
1 (S), so that X = X/
d, ) (see Remark 4.5).
triple (X,
Let us denote by C the set of conjugation classes of 1 (S)\{1} which coincide with the
homotopy classes of noncontractible continuous maps f : S 1 S. Each marked spectrum
is a point of the functional space (R0 )C , endowed with the natural product topology. The
Marked Spectrum sX of X (denoted also by sX
), is defined as follows: for any c = [ ] C,
1 (S), = 1,
sX (c) = inf d p, (p) .
pX
The spectrum is marked because one takes track of the map in addition to its image.
When X = F or Sa , sX (c) is just the length of a closed geodesic curve (not necessarily
simple; that is, selfcrossings could possibly occur in c) which minimizes the length among
the curves in that homotopy class. For this reason, in such a case, sX is called the Marked
= T , sT can be expressed, in dual terms,
Length Spectrum and is denoted by lX . When X
as the Marked Measure Spectrum of the corresponding measured geodesic lamination F
on F ; usually this is denoted IF . IF (c) is just the minimal transverse measure realized
by the curves in that homotopy class. When T is simplicial, that is when F is a weighted
multicurve L of F , IL (c) is easily expressed in terms of the geometric intersection number
(this also justifies the notation): assume that all the weights are equal to 1 (that is the length
of all edges of T is equal to 1); then it is easy to see that IL (c) is just the minimum
number of intersection points between L and any curve belonging to c and transverse to
the components of the lamination. For arbitrary weights one just takes multiples of the
contribution of each component of L.
Remark 4.18. Instead of the whole C, one could prefer to use the subset S C of isotopy
classes of simple closed curve in S, and take the corresponding (restricted) marked spectra.
The discussion should proceed without any substantial modification.
4.5.1. On the boundary of the Teichmller space
It is convenient, at this stage, to recall the fundamental facts about the role that the
marked spectra play in the study of the Teichmller space and in Thurstons theory of its
natural boundary. Let us denote by Tg the Teichmller space of the hyperbolic structures
on S up to isometry isotopic to the identity. It is wellknown (see [7,11,23]) that the map
l : Tg (R0 )C
defined by l(F = H2 / ) = lF , realizes a meaningful embedding of Tg onto a subset of
(R0 )C homeomorphic to the finitedimensional open ball B 6g6 . We shall identify Tg
with l(Tg ). In fact Tg is in a natural way a real analytic submanifold of (R0 )C .
347
348
349
Remark 4.26. This means, in particular, that in a far CT future the spacetime looks like the
Minkowskian suspension M(F ). In order to detect the dual effect of the initial singularity
on the embedding of Sa into M, for large value of the cosmological time one needs to
increase the accuracy in the measurement of geometric quantities. Nevertheless, this effect
is, in principle, observable for any finite value a of the CT.
Proposition 4.27. For every a (0, [, lSa /a belongs to Tg (R0 )C . Hence, the
cosmological time determines a curve M : (0, [ Tg . This is a real analytic curve which
connects F Tg with the point on the natural boundary [T ] Tg (here [.] denotes the
projective class).
Remark 4.28. Consider a spacetime of simplicial type. To prove Proposition 4.25 in
this case, one has to note that the depth of the annular regions is constant on each Sa .
When a 0, the contribution (to the length of any curve on Sa ) of the part contained in
the nonannular components becomes negligible, the length of the annuli boundaries goes
linearly to zero, so that only the transverse crossing of the annuli becomes dominant. When
a , the annuli depth goes to zero because of the rescaling by 1/a, and the length
spectrum converge to the spectrum of F . The general case follows by using the density
stated in Proposition 4.23. Concerning Proposition 4.27, in the special case of a spacetime
of simplicial type, the curve in Tg is just given by the FenchelNielsen flow obtained by
twisting the hyperbolic surface F along the closed geodesic of the multicurve (see [23]
and also [7]).
4.6. A commentary on the proofs
The identification between cocycles of a spacetime M with measured geodesic
laminations on F = I(1)/ is due to Mess [16]. In fact one can find other examples of
such a construction of cocycles from measured laminations in the contest of Thurstons
theory of bending or earthquakes (see for instance [10]).
Measured geodesic laminations emerged in the original Thurstons approach to the
natural compactification of Tg [2123]. See also [11] for the alternative approach by
using the measured foliations (see Remark 4.24 (4)). For the claim about the quadratic
differentials in Remark 4.24 (4)) see [14]. The dual approach via real trees is due to
MorganShalen [18,19]. This approach does apply to more general, higherdimensional
situations. The monography [20] contains a rather exhaustive introduction to this matter
and we mostly refer to it (and to its bibliography) for all the details. In particular one can
find in [20] a complete proof of the duality (see Proposition 4.19 and Proposition 4.20). The
delicate point is just the inversion of the map we have described above. The geometric
interpretation of the MorganShalen convergence (see Remark 4.24 (3)) is due to Paulin
and Bestvina (cf. the bibliography of [20]).
It is an amazing fact that the spacetimes materialize this subtle duality in the way we
have seen. Note also that, in the spacetime setting, the choice of the base hyperbolic surface
350
F (see Remark 4.24 (4)) is fixed by the linear part of the holonomy of M, that is by its
future asymptotic state.
Concerning Proposition 4.27, the FenchelNielsen flow generalizes to the earthquake
flow (one uses again the density 4.23) with initial data (F, F ) which has real analytic
orbits [15,24].
5. Complements
In this section we add a few comments about the flat spacetimes with compact space
of genus g = 1, and about the spacetimes with negative cosmological constant. Finally we
discuss a conjecture relating the CT and the CMC time.
5.1. Toric space (g = 1)
The case in which the surface S is a torus is a particular example of the socalled Teichmller spacetimes which we have analysed in [5]. So we simply remind
the main points. Each nonstatic spacetime determines a curve : (0, [ T1 , (a) =
(w(a), (a)), where T1 is the cotangent bundle of the Teichmller space T1 of conformal
structures on the torus up to isomorphism isotopic to the identity. Let us recall that T1 is
isometric with the Poincar disk. The cotangent vectors (a) at the point w(a) T1 is
a quadratic differential on a Riemann surface representing w(a). It is not hard to verify
that is just the complete orbit of the Teichmller flow with initial data (w(1), (1))
(see [1,5]). In particular, the projection of onto T1 is a complete geodesic connecting
two boundary points. These points can also be interpreted in terms of marked spectra. Let
us denote by H and by V the horizontal and vertical measured foliations of w(1). Then:
lima lSa /a = IH and lima0 lSa = IV .
5.2. Spacetime with negative cosmological constant
The above discussion on CT for flat spacetimes (i.e., with cosmological constant = 0)
can be adapted to the case of negative which we normalize to be = 1. We denote
by X2+1 the Universal antideSitter spacetime of dimension 2 + 1. Each spacetime is now
locally isometric to X2+1 . The role played by I + (0) in the flat case, is played now by
the diamondshaped domain D(2) (see [13] p. 132) isometric to B 2 (/2, /2) with
metric, in coordinates (y 1 , y 2 , t), ds 2 = (cos2 t)h2 dt 2 , where h2 is the usual Poincar
hyperbolic metric on the open disk B 2 .
5.2.1. AntideSitter suspensions
If F = H2 / is a hyperbolic surface of genus g > 1, then isometrically acts also
on D(2) and D(F ) = D(2)/ is the antideSitter suspension of F . Up to a translation,
the function t gives the CT and it has many qualitative properties similar to the CT of the
Minkowskian suspensions, but we have now both an initial and a final singularity, both
reduced to one point. In a sense, D(F ) can be obtained by the Minkowskian suspension
351
M(F ) by a procedure of warping and doubling; D(F ) and M(F ) have the same initial
singularity; the future asymptotic state of M(F ) becomes the level surface of the CT
on D(F ) where the expansion ends and the collapsing begins. Also the antideSitter
analogous of I + (1, 3) is easy to figure out.
5.2.2. Deforming antideSitter suspensions
We want to generalize the above warping and doubling construction. Let M =
U (M)/ as in the former flatspacetime discussion, = + t ( ). F = I(1)/ as
usually. For t (/2, 0), (0, ), set t = (/2)e . Denote h(a) the spatial metric
on Sa . On the manifold F (/2, 0) consider the metric ds 2 = cos2 (t)h( )/ 2
dt 2 , getting a spacetime D (M). Similarly, take M and D (M ), where M =
U (M )/ , = t ( ), D (M ) is obtained from D (M ) by reversing the
time and the orientation. Finally, set D(M) = D (M) D (M ), by gluing the two pieces
at t = 0. D(M) is locally antideSitter; up to a translation, t gives the CT. The asymptotic
state for t /2 (i.e., the initial singularity) is equal to the initial singularity of M.
The final singularity (t /2) coincides with the initial singularity of M . The future
asymptotic states of M and M glue at the level surface {t = 0} of the CT where the
expansion ends and the collapse begins. The orbit of D(M) in Tg is given by the union of
two earthquake rays associated to M (pointing to the future) and to M (towards the past);
note that the qualitative behaviour is similar to what we have remarked for g = 1. D(M) is
the quotient of a domain D(2)M X2+1 , which is a deformation of the diamondshaped
domain D(2). Also in this case the spacetimes with simplicial asymptotic singularities are
significant and particularly simple to be understood.
5.3. CT versus CMC
Assume again that the space S is of genus g > 1, and that the spacetimes are flat. Given
any global time on a spacetime M = U (M)/ , the asymptotic behaviour of the geometry
of the corresponding level surfaces reflects in general a property of the specific time and
not of the spacetime. On the other hand, we have seen that the asymptotic states of the
cosmological time are intrinsic features of the spacetime. In this sense, we can say that a
global time is good when it has the same asymptotic states of the CT. The CMC time,
say, is a widely studied global time. A natural question is whether is a good global
time. We conjecture that this is the case. Let us denote by Wa the { = a} level surfaces of
the CMC time.
Conjecture 5.1. (a) lima lWa = sT ; (b) lima0 lWa /a = lF .
There are some strong evidences supporting the conjecture; in particular by [3] we know
that:
(1) is a global time function with image the interval (0, +).
(2) If : (0, ) Tg is any orbit in Tg (here Tg is intended as a space of conformal
structures) then:
352
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]
W. Abikoff, The Real Analytic Theory of Teichmller Space, SpringerVerlag, Berlin, 1980.
L. Andersson, G.J. Galloway, R. Howard, Class. Quantum Grav. 15 (1998) 309322.
L. Andersson, V. Moncrief, A. Tromba, J. Geom. Phys. 23 (1997) 191205.
J.T. Beem, P.E. Ehrlich, K.L. Easley, Global Lorentzian Geometry, 2nd edn., Pure and Applied
Mathematics, Vol. 202, Dekker.
R. Benedetti, E. Guadagnini, Phys. Lett. B 441 (1998) 6068.
R. Benedetti, E. Guadagnini, Nucl. Phys. B 588 (2000) 436450.
R. Benedetti, C. Petronio, Lectures on Hyperbolic Geometry, SpringerVerlag, 1992.
S. Carlip, Phys. Rev. D 42 (1990) 2647.
S. Deser, R. Jackiw, G. t Hooft, Ann. Phys. (NY) 152 (1984) 220.
D.B.A. Epstein, A. Marden, in: D.B.A. Epstein (Ed.), Analytical and Geometric Aspects of
Hyperbolic Space, London Math. Soc. Lecture Notes Series, Vol. 111, Cambridge Univ. Press,
1987.
A. Fathi, F. Laudenbach, V. Poenaru, Travaux de Thurston sur les surfaces, Astrisque (1979)
6667.
W.M. Goldman, Inventiones Math. 93 (1988) 557607.
S. Hawking, G.F.W. Ellis, The Large Scale Structure of SpaceTime, Cambridge Univ. Press,
Cambridge.
S.P. Kerckhoff, Topology 19 (1) (1980) 23.
S.P. Kerckhoff, Comment. Math. Helvetici 60 (1985) 1730.
G. Mess, Preprint IHES/M/90/28 (1990).
V. Moncrief, J. Math. Phys. 30 (1989) 29072914.
J.W. Morgan, P. Shalen, Ann. Math. 122 (1985) 398476.
J.W. Morgan, P. Shalen, Topology 30 (1991) 143154.
J.P. Otal, Le thorme dhyperbolisation pour les varits fibres de dimension trois,
Astrisque 235 (1996).
W.P. Thurston, Bull. Amer. Math. Soc. 6 (1982) 357381.
W.P. Thurston, Bull. Am. Math. Soc. 19 (1988) 417432.
W.P. Thurston, Geometry and Topology of 3manifolds, Princeton Univ. Press, Princeton, 1982.
W.P. Thurston, in: D.B.A. Epstein (Ed.), London Math. Soc. Lecture Notes, Vol. 111,
Cambridge Univ. Press, 1987.
R. Wald, P. Yip, J. Math. Phys. 22 (1981) 26592665.
E. Witten, Nucl. Phys. B 311 (1988) 46.
Abstract
We propose a numerical technique for calculating effective actions of electromagnetic backgrounds based on the worldline formalism. As a conceptually simple example, we consider scalar
electrodynamics in three dimensions to oneloop order. Beyond the constantmagneticfield case,
serving as a benchmark test, we analyze the effective action of a stepfunctionlike magnetic field
a configuration that is inaccessible to derivative expansions. We observe magneticfield diffusion,
i.e., nonvanishing magnetic action density at space points near the magnetic step where the classical
field vanishes. 2001 Elsevier Science B.V. All rights reserved.
PACS: 12.20.m; 11.15.Ha
Keywords: Worldline; Effective action; Monte Carlo simulation
Introduction
The worldline formalism was invented by Feynman [1] simultaneously with modern
relativistic secondquantized QED, but for a long time it was used only occasionally for
actual calculations. Eventually, the observation of a close relation between the worldline
formalism and the infinitestringtension limit of string path integrals triggered further
developments of the worldline formalism for QCD [2] and QED [3], particularly for gauge
particle amplitudes. Certain computational advantages of this formalism were subsequently
recognized and led to numerous applications. Among them, the progress achieved for
QED amplitudes with allorder couplings to an external background field is particularly
remarkable [46]. The formalism works most elegantly for constant electromagnetic
background fields and can be extended to a derivative expansion in the electromagnetic
Email address: holder.gies@cern.ch (H. Gies).
1 Emmy Noether fellow.
05503213/01/$ see front matter 2001 Elsevier Science B.V. All rights reserved.
PII: S 0 5 5 0  3 2 1 3 ( 0 1 ) 0 0 3 7 7  7
354
background [7,8]. For a comprehensive review of the worldline formalism in QED and
beyond, see [9].
In the present work, we intend to demonstrate that the worldline formalism is moreover
perfectly suited for numerical computations. This is because the path integral over closed
loops in spacetime can be approximated by a finite ensemble of loops, which allows for a
simple and fast evaluation of expectation values. The latter include observables depending
on an arbitrary background field.
We illustrate this proposal by means of a simple example: the oneloop effective action
of Euclidean scalar QED in three dimensions. The computational task boils down to the
calculation of the Wilson loop expectation value using a loop ensemble.
As a first step, we verify the method in Section 2 by considering the constantmagneticfield case, which can