Nuclear Physics B 584 (2000) 3–45

www.elsevier.nl/locate/npe

Natural effective supersymmetry
Junji Hisano a , Kiichi Kurosawa b,c , Yasunori Nomura c
a Theory Group, KEK, Tsukuba, Ibaraki 305-0801, Japan
b YITP, Kyoto University, Kyoto 606-8502, Japan
c Department of Physics, University of Tokyo, Tokyo 113-0033, Japan

Received 28 February 2000; accepted 23 May 2000

Abstract
Much heavier sfermions of the first-two generations than the other superparticles provide a natural
explanation for the flavor and CP problems in the supersymmetric Standard Model (SUSY SM).
However, the heavy sfermions may drive the mass squareds for the light third generation sfermions
to be negative through two-loop renormalization group (RG) equations, breaking color and charge.
Introducing extra matters to the SUSY SM, it is possible to construct models where the sfermion
masses are RG invariant at the two-loop level in the limit of vanishing gaugino-mass and Yukawacoupling contributions. We calculate the finite corrections to the light sfermion masses at the twoloop level in the models. We find that the finite corrections to the light-squark mass squareds are
negative and can be less than (0.3–1)% of the heavy-squark mass squareds, depending on the number
and the parameters of the extra matters. We also discuss whether such models realized by the U(1)X
gauge interaction at the GUT scale can satisfy the constraints from 1mK and K naturally. When
both the left- and right-handed down-type squarks of the first-two generations have common U(1)X
charges, the supersymmetric contributions to 1mK and K are sufficiently suppressed without
spoiling naturalness, even if the flavor-violating supergravity contributions to the sfermion mass
matrices are included. When only the right-handed squarks of the first-two generations have a
common U(1)X charge, we can still satisfy the constraint from 1mK naturally, but evading the
bound from K requires the CP phase smaller than 10−2 .  2000 Elsevier Science B.V. All rights
reserved.
PACS: 14.80.Ly; 12.60.Jv; 12.15.Ff

1. Introduction
The Standard Model (SM) has enjoyed remarkable successes in describing physics
down to a scale of 10−18 m, the weak scale. However, the quadratically divergent
radiative correction to the Higgs boson mass leads to the well-known naturalness problem.
Supersymmetry (SUSY) provides a solution to this problem by extending the chiral
symmetry to the scalar partners. At present, the minimal supersymmetric standard model
0550-3213/00/$ – see front matter  2000 Elsevier Science B.V. All rights reserved.
PII: S 0 5 5 0 - 3 2 1 3 ( 0 0 ) 0 0 3 4 3 - 6

4

J. Hisano et al. / Nuclear Physics B 584 (2000) 3–45

(MSSM) is considered to be the most promising candidate beyond the SM (for reviews see,
[1,2]).
The MSSM is severely constrained by flavor physics. Introduction of SUSY-breaking
terms provides new sources of the flavor-changing neutral currents (FCNC’s) and CP
violation. For instance, the experimental value of the mass difference between KL and
KS , 1mK , leads to the upper bound on the squark contribution written as
!2
1m2˜  30 TeV 2
d
2
. 1,
(1)
sin θd˜
m
¯ d˜
m
¯ 2˜
d

in a basis where the quark mass matrices are diagonal (see, for example, [3,4]). Here, m
¯ 2˜
d

is the averaged squark mass, and 1m2˜ and sin2 θd˜ are the mass-squared difference and
d
the mixing angle between the down-type squarks of the first-two generations. In order to
satisfy this constraint, (i) 1m2˜ ' 0 (degeneracy [5,6]), (ii) sin θd˜ ' 0 (alignment [7–9]),
d
(iii) m
¯ d˜ & 30 TeV (decoupling [10–13]), or the hybrid of them is required.
In this article, we discuss the third possibility, the decoupling scenario. In this scenario,
the masses for the third generation squarks, Higgsinos, and gauginos are of the order of the
weak scale while the other squarks and sleptons are heavy enough to suppress the FCNC
and CP-violating processes. This is still natural at the one-loop level, since the squarks
and sleptons of the first-two generations are not strongly coupled with the Higgs boson.
This mass spectrum is sometimes called as effective supersymmetry or the more minimal
supersymmetric standard model. It is realized in the anomalous U(1) SUSY-breaking
models [14–24], U(1)0 models [25–27], composite models of the first-two generation
particles [28,29], or radiatively-driven models with specific boundary conditions [30–32].
However, this scenario generically suffers from a problem that the third generation
sfermions obtain the vacuum expectation value (VEV), breaking color and charge [33,
34]. At the two-loop level the heavy first-two generation sfermions contribute to the the
renormalization group (RG) equations for the third generation sfermion masses through the
gauge interactions. The RG equations for the SUSY-breaking sfermion masses m
e2r through
the SM gauge interactions are given by [35] 

dm
e2r X αA 2 A X A 2
=
8
Cr
Ts m
es
µ


s
A 
 X
X  αY  αA  X
αY
2
+2
Ys m
es +
8
Ys CsA m
e2s ,
(2)
Yr
Yr



s
s
A

0

at the two-loop level in the DR scheme [36]. Here, we have taken a limit where the
gaugino masses vanish. An index A (= 1–3) represents the SM gauge groups, and we
have adopted the SU(5)GUT normalization for the U(1)Y gauge coupling (α1 ≡ (5/3) αY ).
TrA , CrA and Yr are the Dynkin index, the quadratic Casimir coefficient and the hypercharge
for the sfermion r, respectively. We find that all sfermion mass squareds contribute to the
RG equations at the two-loop level. Thus, the contribution from the heavy sfermions may
destroy the mass spectrum by driving the light-sfermion mass squareds to be negative at

J. Hisano et al. / Nuclear Physics B 584 (2000) 3–45

5

low energies. This fact makes it difficult to construct models where such a hierarchical
sfermion mass spectrum is generated at much higher energy scale than the weak scale.
However, the hierarchical mass spectrum can be stabilized against the RG evolution if
the following relations among the SUSY-breaking mass squareds for the heavy sfermions
are imposed [37,38]:
X
TrA m
e2r = 0,
(3)
r

X
X

Yr m
e2r = 0,

(4)

Yr CrA m
e2r = 0,

(5)

r

r

for A = 1–3. These relations cannot be satisfied by the heavy MSSM sfermions alone,
since all of them have positive SUSY-breaking mass squareds. However, the relations
can be satisfied if we introduce extra fields with negative SUSY-breaking mass squareds
which transform nontrivially under the SM gauge groups, since the sum is taken over
the extra fields as well as the heavy MSSM sfermions. The extra fields should have the
supersymmetric masses to avoid the breaking of the SM gauge groups caused by their
condensations. Thus, the extra fields have to be in vector-like representations of the SM
gauge groups, and we call them extra matters. The supersymmetric masses for the extra
matters should not be much larger than the SUSY-breaking scalar masses. Otherwise, the
large radiative correction is generated below the energy scale where the extra matters
are decoupled. We refer this extension of the MSSM to the Natural Effective SUSY SM
(NESSM), hereafter.
The NESSM can be realized by assuming that the SUSY-breaking masses for the
heavy sfermions and extra matters are generated by a D-term VEV of some U(1)X gauge
interaction such that m
e2r = QX
r h−DX i. The U(1)X should be broken in order to give the
supersymmetric masses for the extra matters. In terms of the U(1)X charges, Eqs. (3)–(5)
are written as
X
TrA QX
(6)
r = 0,
r

X
X

Yr QX
r = 0,

(7)

Yr CrA QX
r = 0.

(8)

r

r

The first equation corresponds to vanishing mixed anomalies between U(1)X and the SM
gauge groups. The second and third ones mean that U(1)X has no kinetic mixing with
U(1)Y at the one- and two-loop levels. If U(1)X is anomaly-free and all fields charged under
U(1)X are embedded in the SU(5)GUT multiplets, Eqs. (6), (7) are satisfied automatically.
In this case, we have only to choose the U(1)X charges for the extra matters to satisfy
Eq. (8). Even if U(1)X is anomalous, however, we can still construct a model in which
some of the fields are decoupled at the U(1)X breaking scale and Eqs. (6)–(8) are satisfied
at low energies.

6

J. Hisano et al. / Nuclear Physics B 584 (2000) 3–45

In this article, we calculate the finite corrections to the light-sfermion mass squareds at
the two-loop level in the NESSM, assuming that all fields are embedded in the SU(5)GUT
multiplets at the GUT scale. In the NESSM, while the dangerous contributions which lead
to color and charge breaking do not exist in the RG equations for the SUSY-breaking
scalar masses, the finite corrections from the heavy sfermions and the extra matters at the
two-loop level may still drive the light-sfermion mass squareds to be negative. We find
that the finite corrections to the light-squark mass squareds depend on the number and the
parameters of the extra matters and they are less than (0.3–1)% of the heavy-squark mass
squareds.
We also discuss whether the models in which the NESSM is realized by the U(1)X gauge
interaction at the GUT scale (MGUT ' 2 × 1016 GeV) are viable or not in the light of the
experimental constraints. The anomalous U(1) SUSY-breaking model given in Refs. [37,
38] is an explicit example for such models. On this setup, the supergravity contributions
to the SUSY-breaking terms, which are generically non-universal in flavor space, are
suppressed compared with the U(1)X D-term contribution. Moreover, the breaking of the
U(1)X symmetry can naturally explain the hierarchical structure of the quark and lepton
masses by the Froggatt–Nielsen mechanism [39]. In this paper, we consider the constraints
from 1mK and K , which lead to the severest bound on the heavy-sfermion mass scale. We
find that, when both the left- and right-handed down-type squarks (only the right-handed
squarks) of the first-two generations have common U(1)X charges, the contributions to
1mK and K (1mK ) are sufficiently suppressed without spoiling naturalness even if the
flavor-violating supergravity contributions to the sfermion mass matrices are included.
This paper is organized as follows. In the next section, we show our setup and the U(1)X
charge assignments. We explain that the squark and slepton masses are inversely related
to the quark and lepton masses through their U(1)X charges. In Sections 3, we derive the
formula for the finite corrections to the light sfermion masses at the two-loop level in the
NESSM, and evaluate them numerically. In Section 4, using the constraints from 1mK
and K the lower bounds on the third generation sfermion masses at the GUT scale are
derived. We neglect the effects of the Yukawa interactions in Sections 3 and 4, since they
are model-dependent. The effects are evaluated in Section 5. We also discuss the three-loop
RG contributions to the light sfermion masses there. Section 6 is devoted to the conclusions
and discussion. In Appendix A, we review the anomalous U(1) SUSY-breaking model
given in Refs. [37,38] as one of the explicit realizations of the NESSM. In Appendix B,
we show the complete formulae for the contributions to the light sfermion masses from the
heavy ones at the two-loop level. The formulae for 1mK and K are given in Appendix C.

2. Setup
In this section, we show the setup of the NESSM on which we perform our analyses in
later sections. In the NESSM, the SUSY-breaking scalar masses for the heavy sfermions
and extra scalars are given by the U(1)X D-term. We here assume that the nonzero U(1)X
D-term is generated associated with the breaking of the U(1)X gauge symmetry caused

J. Hisano et al. / Nuclear Physics B 584 (2000) 3–45

7

by the VEV of a Φ field which has a U(1)X charge of −1 (QX
Φ = −1). In this case, the
U(1)X D-term is related to the F -term of the Φ field at a (local) minimum of the potential
as follows:
−DX =

2
gX

MX2

|−FΦ |2 ,

(9)

where MX (= gX |hΦi|) is the U(1)X gauge boson mass. Here, we have assumed that the
VEV’s and F -terms of the other fields are negligibly small for simplicity. Then, the relation

between the supergravity contribution m0 and the U(1)X D-term (mD ≡ |−DX | ) is given
by
m0 '

|−FΦ | |hΦi|
=
mD ,
Mpl
Mpl

(10)

where Mpl is the reduced Planck scale (Mpl ' 2 × 1018 GeV). This shows that the
supergravity contributions to the sfermion masses are, in general, suppressed compared
with the D-term contribution, mD , as long as |hΦi| is smaller than Mpl . Then, the U(1)X
charged sfermions have large masses of the order of the U(1)X D-term, mD , while
the neutral sfermions only receive smaller masses of order m0 from the supergravity
contributions. In order to obtain desired mass spectrum, we set m0 to be of the order of
the weak scale and |hΦi|/Mpl ' (10−1 ∼ 10−2 ).
Next, we consider the U(1)X charge assignment. We assume that all the fields of the
NESSM are embedded in SU(5)GUT multiplets at the GUT scale. It not only maintains
the successful gauge coupling unification but also guarantees vanishing U(1)Y D-term
contribution at the one-loop level (Eq. (7)). We also assume that one pair of the extra
matters are introduced in (5 + 5? ) representation of the SU(5)GUT for simplicity. The
extension to the other cases is straightforward. Then, the U(1)X charges for the 5 and
X
5? extra matters, QX
5ex and Q5? ex , are determined from those for the quarks and leptons in
the SM to satisfy Eqs. (6), (8) as
QX
5ex

= −2

3
X

QX
10i ,

(11) 

X
QX
10i + Q5? i ,

(12)

i=1

QX
5? ex = −

3
X
i=1

X
where QX
10i and Q5? i (i = 1–3) are the U(1)X charges for the quarks and leptons embedded
?
in the 10 and 5 representations of the SU(5)GUT , respectively. An index i represents the
generation.
The U(1)X charges for the SM matter multiplets are determined by the following
considerations. The naturalness argument tells us that the superparticles which are strongly
coupled with the Higgs boson are necessary to have masses around the weak scale and thus
should have zero U(1)X charges. This requires that QX
103 = 0, since the top squarks and
left-handed bottom squark are coupled with the Higgs boson through large top Yukawa
coupling. On the other hand, the superpartners of the light quarks and leptons have to

8

J. Hisano et al. / Nuclear Physics B 584 (2000) 3–45

be heavy enough to sufficiently suppress the FCNC processes, so that they should have
positive U(1)X charges. This explains the hierarchy of the quark and lepton masses
naturally by the Froggatt–Nielsen mechanism, since the Yukawa matrices are generated
through the VEV of the Φ field suppressed by suitable powers of hΦi/Mpl [39] as
W=

3
X 

(fu )ij

i,j =1

+

3
X
i,j =1

hΦi
Mpl 

(fd )ij 

QX

X
10i +Q10j

hΦi
Mpl

Ψ10i Ψ10j Hu 

QX

X
10i +Q5? j

Ψ10i Ψ5? j Hd ,

(13)

where Hu and Hd are the Higgs doublets for which we have assumed the vanishing U(1)X
charges. Thus, it is possible to reproduce the observed quark and lepton mass matrices
X
with (fu )ij , (fd )ij = O(1) if we appropriately choose the U(1)X charges QX
10i and Q5? i
[40–43].
According to the above arguments, we consider the following U(1)X charge assignments:
Model

I

II

III

IV

QX
101

1

2

2

2

QX
102

1

1

1

1

QX
103

0

0

0

0

QX
5? 1

1

1

1

2

QX
5? 2
QX
5? 3

1

1

1

1

0

1

0

1

QX
5ex

−4

−6

−6

−6

QX
5? ex

−4

−6

−5

−7

Here, the U(1)X charges for the Higgs multiplets are zero in all the models and the
charges for the extra-matter multiplets were determined by Eqs. (11), (12). Model I is
the simplest possibility realizing the decoupling scenario. In this model, the FCNC’s are
strongly suppressed due to the degeneracy of the SUSY-breaking masses between the firsttwo generation sfermions, while the hierarchy among the quark and lepton masses cannot
be explained completely. Models II–IV are motivated to explain the fermion mass hierarchy
by the Froggatt–Nielsen mechanism. In particular, in Models II, IV the second and third
X
generation doublet leptons have the same U(1)X charges, QX
5? 2 = Q5? 3 , so that the observed
1
large mixing between νµ and ντ [44] is naturally explained. In these models, the FCNC
1 Models II and IV prefer large and small angle MSW solutions to the solar neutrino problem, respectively
[45,46].

J. Hisano et al. / Nuclear Physics B 584 (2000) 3–45

9

processes are less suppressed than in Model I. Models II, III have non-degenerate SUSYbreaking masses for the left-handed down-type squarks in the first-two generations. Thus,
K 0 –K¯ 0 oscillation has dominant contribution from the off-diagonal elements in the lefthanded squark mass matrix. Model IV has non-degenerate SUSY-breaking masses for both
the left- and right-handed down-type squarks in the first-two generations, so that K 0 –K¯ 0
oscillation receives contributions from off-diagonal elements of both the left- and righthanded squark mass matrices. As a result, Model IV is more severely constrained from
K 0 –K¯ 0 mixing than Models II, III, as will be discussed later.
We now discuss the other SUSY-breaking parameters. Since the gaugino masses are
constrained by naturalness argument, we consider them to be of the order of the weak scale.
Indeed, in the anomalous U(1) SUSY-breaking model given in Appendix A, the gaugino
masses arise from the F -term of the dilaton field [48] and their sizes can be of the order
of the weak scale [49]. We treat the gaugino mass (at the GUT scale) as a free parameter
in the phenomenological analyses below. Also, the superpotential Eq. (13) generates the
SUSY-breaking trilinear scalar couplings,
X
(Au )ij = QX
10i + Q10j

(Ad )ij = QX
10i 

,
hΦi 

.
+ QX
5? j
hΦi

(14)
(15)

The (Au )ij and (Ad )ij are defined by Lsoft = (yu )ij (Au )ij ψ˜ 10i ψ˜ 10j hu + (yd )ij (Ad )ij
ψ˜ 10i ψ˜ 5? j hd . The (yu )ij and (yd )ij are the Yukawa couplings W = (yu )ij Ψ10i Ψ10j Hu +
(yd )ij Ψ10i Ψ5? j Hd , and ψ˜ and h represent the scalar components of the corresponding
supermultiplets Ψ and H , respectively. From Eq. (10), we find that Au ' Ad ' mD except
for the ones which involve only the light sfermions.
The supersymmetric masses and the holomorphic SUSY-breaking masses for the extra
matters come from the U(1)X symmetry breaking. The supersymmetric masses should be
of the same order with mD as we mentioned in the introduction. While these features
are naturally explained in the explicit model given in Refs. [37,38] (see Appendix A), we
mainly adopt a phenomenological approach in this article and take the masses as input
parameters. We parameterize the supersymmetric and holomorphic SUSY-breaking mass
parameters as
Wex = mψ Ψ3ex Ψ3? ex + m0ψ Ψ2ex Ψ2? ex ,
Lex, soft = − Fψ ψ˜ 3ex ψ˜ 3? ex + Fψ0 ψ˜ 2ex ψ˜ 2? ex 

+ h.c. .

(16)
(17)

Here, the triplet extra matters Ψ3ex and Ψ3? ex and the doublet extra matters Ψ2ex and Ψ2? ex
are embedded in the Ψ5ex and Ψ5? ex. In this article, we impose the SU(5)GUT relations on the
supersymmetric and the holomorphic SUSY-breaking masses, mψ = m0ψ and Fψ = Fψ0 , at
the GUT scale unless otherwise stated. Then, the masses for the heavy sfermions and extra
matters are completely determined by three free parameters mD , mψ and Fψ once a U(1)X
charge assignment is specified.

10

J. Hisano et al. / Nuclear Physics B 584 (2000) 3–45

3. Corrections to the light sfermions
In the NESSM, the two-loop RG contributions to the light sfermions are canceled
between the heavy sfermions and the extra-matter multiplets. However, there are still
negative finite corrections. In this section, we numerically estimate the effects of the finite
corrections in various cases, using their explicit form which is completely calculated at
the two-loop level in Appendix B. We conclude this section by showing how hierarchical
the sfermion mass spectrum can be in the NESSM without breaking color and charge,
comparing the NESSM with the original effective SUSY in which there is no extra-matter
multiplet.
Now, let us numerically estimate the finite corrections in the models discussed in
the previous section. We first consider Models I, II. In these models, the U(1)X charge
assignments are consistent with the SU(5)GUT and the extra-matter multiplets have an
invariance under the parity Ψ5ex ↔ Ψ5? ex , so that the finite corrections to the light
, take relatively simple forms. The explicit expressions for m2˜
are
sfermions, m2˜
f , finite
f , finite
given in Eqs. (B.47)–(B.49) in Appendix B. These expressions can be further simplified in
the case of one-pair extra-matter multiplets as follows:   

X αA 2 
m2ψ
1 2
2
π
CfA˜ − TrF˜ TFA
m
+
log
m2f˜, finite = 4
˜


3
m2˜
A
F 

1
+ m2ψ G(y1 , y2 )
2   

h
i
X αY
αA
(18)
Yf˜ TrF˜ YF˜ CFA˜ m2F˜ log m2F˜ ,
+4


A

where     

1
1
y2
+ y1 Li2 1 −
+ (y1 ↔ y2 ), (19)
G(y1 , y2 ) = y1 log y1 − 2y1 Li2 1 −
y1
2
y1
y1 ≡

m21
m2ψ

,

y2 ≡

m22
m2ψ

,

(20)

where F˜ denotes the heavy sfermions; m1 and m2 are the mass eigenvalues for the
extra scalars (see Appendix B). Here, we have neglected the difference between the
supersymmetric masses (and the holomorphic SUSY-breaking masses) of the triplet and
doublet extra matters for demonstrational purpose, though it will be included in later
numerical calculations.
The first term of the first line in Eq. (18) gives dominant contributions to the lightsfermion mass squareds. They are negative and could cause color and charge breaking
since mψ > mF˜ . The second term has the same form as the scalar mass squared generated
by integrating out the messenger fields in the gauge-mediated SUSY breaking mechanism
[50]. Therefore, it gives the positive contributions almost proportional to the SUSYbreaking bilinear coupling Fψ of the extra scalars. Even if Fψ = 0, however, G(y1 , y2 )

J. Hisano et al. / Nuclear Physics B 584 (2000) 3–45

11

remains positive, since m21 and m22 are smaller than m2ψ by the U(1)X D-term. The last
term which is generated through U(1)Y D-term does not give significant contributions due
to the smallness of the hypercharge gauge coupling. The renormalization-point dependence
of the term is canceled out due to Eq. (8).
In order to see the dependence of the finite corrections on mψ , mD and Fψ , we fix
the heavy-sfermion mass scale mD = 10 TeV as an overall scale and take mψ /mD and
Fψ /(mψ mD ) as remaining two dimensionless free parameters. The ratio mψ /mD must
be of order unity in the NESSM, since otherwise the heavy sfermions generate the large
radiative corrections below the scale mψ where the extra-matter multiplets decouple. Then,
Fψ /(mψ mD ) should not be much larger than one to avoid the negative mass squared for
the extra scalar. Indeed, Fψ /(mψ mD ) is of order unity in the explicit example constructed.
Fig. 1 illustrates the dependence of the finite corrections on these two parameters. As
an example, we consider the corrections to the light squarks in Models I, II and plot the
ratio of the finite corrections to the heavy sfermion mass-scale squared m2D . Here, we
only take into account the dominant two-loop contribution through the strong interaction,
(3)
that is, we drop the last term and set A = 3 and C ˜ = 4/3 in Eq. (18). The minimum
f
of the ratio is −0.36% (mψ /mD = 2.2) and −0.54% (mψ /mD = 2.7) in Models I and II,
respectively, when Fψ = 0. The difference between Models I and II in Fig. 1 comes mainly    

m2˜ /m2D = TrF˜ T A
QX˜ , which is
from the group-theoretical factor in Eq. (18), TrF˜ T A

F

F
4 in Model I and 6 in Model II. Thus, the corrections in Model II are larger than in Model I
approximately by a factor of 1.5.
The ratio given in Fig. 1 determines how hierarchical the sfermion mass spectrum can
be without breaking color and charge. Since the bare mass mq,0
˜ for the light squarks has
to be larger than these negative corrections, it gives a lower bound on mq,0
˜ . In Model I, for
example, we obtain
2
2
m2q,0
˜ > −mq,
˜ finite ' 0.4% × mD ,

(21)

Fig. 1. Finite corrections to the light squarks through the strong interaction. The percentages on the
figures represent the ratio of the finite corrections to the heavy-sfermion mass scale, m2q,
/m2D .
˜ finite
The left figure is in Model I and the right is in Model II. The left side of the solid line is excluded
since the mass squared for the extra scalar is negative and color is broken. Here, we have taken
mD = 10 TeV and evaluated all the parameters at µ = 10 TeV.

12

J. Hisano et al. / Nuclear Physics B 584 (2000) 3–45

when the supersymmetric mass for the extra matter, mψ , is not much larger than mD as
discussed in the previous section. 2 Since mq,0
˜ set the scale for the light sfermion masses,
we find that the splitting between the heavy and light sfermions cannot far exceed an
order of magnitude. In the case of the original effective SUSY models (models without
extra-matter multiplets), the RG contributions to the light sfermions, m2˜ , are written
f , log
by solving the RG equation Eq. (2) as follows: 

h
i α (µ) α
A
GUT
2

,
TrF˜ TFA
m
˜

bA


A

m2f˜, log(µ) = 4

X Cf˜
A

(22)

where bA is the coefficient for the one-loop beta function of the SM gauge coupling and
αGUT is the gauge coupling constant at the GUT scale. Here, the contributions from the
U(1)Y D-term are neglected. In Model I, for example, radiative corrections to the light
squarks through the strong interaction are estimated as
2
m2q,
˜ log (µ = 10 TeV) ' −2.3% × mD ,

(23)

2
so that the bare light-squark mass squared m2q,0
˜ should be larger than about −mq,
˜ log (µ =

10 TeV) ' 2.3% × m2D . The rough estimations in Eqs. (21), (23) show that the splitting
between the heavy and light sfermion masses in the NESSM can be more than twice larger
than in the original effective SUSY. This conclusion remains true even if full radiative
corrections are taken into account, as will be shown below.
We here comment on the effect of introducing more than one pairs of extra-matter
multiplets. If we introduce more extra-matter pairs, the extra matters could have smaller
U(1)X charges satisfying Eqs. (6), (8) and thus smaller supersymmetric masses. Thus, we
can reduce the finite corrections to the light-sfermion mass squareds by introducing more
extra matters, since it is the supersymmetric mass for the extra matters that determines
the size of the negative finite corrections. For example, the minimum size of the finite
corrections with Fψ = 0 in Model I are reduced to −0.36%, −0.30%, −0.26%, and
−0.24% by introducing one, two, three and four pairs of Ψ5ex and Ψ5? ex , respectively.
However, introducing extra matters changes the beta functions, so that the behavior of
the gauge couplings becomes worse at high energy in this case. In this article, we limit
ourselves to the case of one pair of Ψ5ex and Ψ5? ex , hereafter.
We now include the effects from the gauginos. Since the gauginos give positive RG
contributions at the one-loop level, they relax the lower bounds on the light sfermion
masses. The lower bounds on the light sfermion masses can be determined as follows. We
first set the boundary conditions at the GUT scale 2 × 1016 GeV: we, for simplicity, take the
universal gaugino mass m1/2 and give the bare mass m0 to the light sfermions at the GUT
scale. Then, we run all the soft SUSY-breaking masses using two-loop RG equations. We
2 In the region where m /m & O(10) and F /(m m ) = O(10), the finite corrections can be much smaller
ψ
D
ψ
ψ D
due to the accidental cancellation between the negative contribution from the heavy-sfermion loops and the
gauge-mediated contribution from the extra-matter multiplets. However, this does not necessarily mean that more
hierarchical superparticle mass spectrum can be realized, since the gluino mass receives the correction from the
extra-matter loops (mg˜ ' (α3 /4π )(Fψ /mψ ) ∼ 0.1mD ).

J. Hisano et al. / Nuclear Physics B 584 (2000) 3–45

13

neglect the effects of the Yukawa couplings, since they are rather model-dependent without
the assumption of universal scalar masses. The effects will be discussed in Section 5. The
heavy sfermions are decoupled at the scale mD = 10 TeV, where the full finite corrections
are added to the light sfermions. The positive contributions from the gauginos are included
from the GUT scale to the scale where the gluino decouples. After these steps, if any light
sfermions have negative mass squareds at the weak scale, the parameter region is excluded
and we need larger value of m0 at the GUT scale.
In Figs. 2 and 3, we have plotted the lower bounds on the light sfermion masses m0 as
a function of the gaugino mass m1/2 in Models I and II. The m0 and m1/2 are the GUTscale values and normalized by the heavy-sfermion mass scale mD . In each figure, the

Fig. 2. Lower bounds on the mass ratios of the light sfermions to the heavy sfermions in Model I. m0
and m1/2 represent the light sfermion and gaugino masses at the GUT scale, respectively. The regions
below the curves are excluded, since the light sfermions indicated by various lines have negative
mass squareds at the weak scale, breaking color and charge. Here, we have set mD = 10 TeV and
mψ = 2.0 × mD at the GUT scale, and neglected the Yukawa couplings.

Fig. 3. Lower bounds on the mass ratios of the light sfermions to the heavy sfermions in Model II.
We have set mD = 10 TeV and mψ = 2.5 × mD at the GUT scale.

14

J. Hisano et al. / Nuclear Physics B 584 (2000) 3–45

Fig. 4. Lower bounds on the mass ratios of the light sfermions to the heavy sfermions in Model III.
We have set mD = 10 TeV and mψ = 2.5 × mD at the GUT scale.

left graph (ESSM) is in the case of the original Effective SUSY SM (without extra-matter
multiplets) and the right one (NESSM) is of the Natural Effective SUSY SM. We have
taken the supersymmetric mass mψ for the extra-matter multiplets as mψ = 2.0 × mD in
Model I and mψ = 2.5 × mD in Model II at the GUT scale. The various curves represent
the lower bounds on the bare masses for the indicated light sfermions, below which they
have negative mass squareds at the weak scale. From Figs. 2 and 3 we find that in the
NESSM the light sfermion masses and/or the gaugino masses can take values less than
half in the ESSM without breaking color and charge. This in turn means that we can obtain
more hierarchical mass spectrum in the NESSM than in the ESSM.
Note that we have assumed that the triplet and doublet extra matters have the same
supersymmetric masses at the GUT scale. Then, at low energy the supersymmetric mass
for the triplet extra matter becomes about twice as large as that for the doublet one due to
the RG evolution. As a result, we cannot take the supersymmetric mass which minimizes
the finite corrections in Fig. 1, since then the doublet extra scalar has negative mass
squared. Thus, if we allow different supersymmetric masses for the triplet and doublet
extra matters at the GUT scale, we can further reduce the size of finite corrections. 3 In
this case, however, the running masses for the triplet extra scalars take negative values at
the high-energy scale, which means that the scalar potential has another minimum at the
nonzero values of the triplet extra scalars.
In the rest of this section, we consider Models III, IV. The finite corrections in these
models are given by Eqs. (B.44)–(B.46) in Appendix B. They take less simple forms due
to the absence of the parity between the extra-matter multiplets, Ψ5ex ↔ Ψ5? ex . In Figs. 4
and 5, we have plotted the lower bounds on m0 in Models III and IV, respectively. We have
taken the supersymmetric mass for the extra-matter multiplet as mψ = 2.5 × mD in Model
III and mψ = 2.7 × mD in Model IV at the GUT scale. All the other parameters have been
taken the same as in Models I, II.
3 The different supersymmetric masses for the triplet and doublet extra matters do not necessarily contradict
with the GUT, since we can make them split using GUT-breaking VEV such as h6(24)i.

J. Hisano et al. / Nuclear Physics B 584 (2000) 3–45

15

Fig. 5. Lower bounds on the mass ratios of the light sfermions to the heavy sfermions in Model IV.
We have set mD = 10 TeV and mψ = 2.7 × mD at the GUT scale.

The main difference of Models III, IV from Models I, II is the existence of the one-loop
U(1)Y D-term, 

h 
i
αY
(24)
Yf˜ TrR YR cos 2θ m21 log m21 − m22 log m22 ,
m2f˜, FI-1loop =

where R denotes the extra matters. This term arises because the RG evolution from the
GUT to low-energy scale splits the supersymmetric masses for the triplet and doublet extra
matters. However, the corrections Eq. (24) do not dominate the two-loop contributions,
since the hypercharge gauge coupling is much smaller
than  

the strong coupling and there
2 in Eq. (18). As a consequence,
m
is no enhancement by factors such as 4 or TrF˜ T A


Figs. 4 and 5 are roughly the same as Figs. 2 and 3. Nonetheless, we can still find the small
effects of the one-loop U(1)Y D-term. In Model III, the lower bound on the mass for u˜ 3
is tighter than that for q˜3 , since a positive one-loop U(1)Y D-term is generated. In Model
IV, the lower bound on the mass for e˜3 is much larger than that in Model II because of a
negative one-loop U(1)Y D-term.
4. Constraints from K 0 –K¯ 0 mixing
K 0 –K¯ 0 mixing gives the severest constraints on the masses for the first-two generation
sfermions in flavor-changing processes. In this section, we calculate the lower bounds on
the heavy-sfermion mass scale mD from the KL –KS mass difference 1mK and the CPviolating parameter K . Combining these bounds on mD with the bounds on the mass ratio
m0 /mD given in the previous section, we obtain the lower bounds on the light sfermion
masses m0 at the GUT scale. It turns out that no severe fine tuning is needed in the NESSM,
compared with the original Effective SUSY SM (ESSM).
We first discuss the structure of the mass matrix for the down-type squarks, since it
induces a dominant flavor violation in K 0 –K¯ 0 mixing through the gluino–squark box
diagram. We restrict our attention to the contribution from the heavy first-two generation

16

J. Hisano et al. / Nuclear Physics B 584 (2000) 3–45

squarks below. The mass matrix for the down-type squarks in the U(1)X gauge basis is
given by
!
2
0
MLL
2
,
(25)
M =
2
0 MRR
where we have set the left–right mixing mass terms zero due to the smallness of the
2
corresponding Yukawa couplings. The left- and right-handed squark mass matrices MLL
2 consist of two parts as follows:
and MRR
2
2
2
= MLL,D
+ MLL,0
,
MLL

2
2
2
MRR
= MRR,D
+ MRR,0
.

(26)

2
2
and MRR,D
come from the U(1)X D-term and are proportional to their U(1)X
MLL,D
charges,
!
!
QX
QX
0
0
101
5? 1
2
2
2
MRR,D =
(27)
mD ,
m2D .
MLL,D =
X
0 QX
0
Q
102
5? 2
2
2
and MRR,0
are contributions from supergravity, whose components are of the order
MLL,0
of the light-sfermion mass squared m20 and induce flavor violation, 4
!
!
2
2
2
2
O(m
O(m
)
O(m
)
)
O(m
)
0
0
0
0
2
2
MLL,0
=
=
,
MRR,0
.
(28)
O(m20 ) O(m20 )
O(m20 ) O(m20 )

The SUSY contribution to K 0 –K¯ 0 mixing is controlled by two parameters qualitatively,
as shown in Appendix C: the averaged squark masses m
¯ 2LL, RR and the off-diagonal
¯ 2LL and m
¯ 2RR are given by
elements δLL, RR . 5 The averaged squark masses m
q
q
2
2
X m2 ,
X
Q
m
¯
=
QX
(29)
m
¯ 2LL = QX
RR
101 102 D
5? 1 Q5? 2 mD ,
where we have neglected O(m20 ) contributions. On the other hand, δLL and δRR are the
off-diagonal elements of the squark mass matrices normalized with the averaged squark
masses, in a basis where the quark mass matrix is diagonal (see also Appendix C). In order
to change the U(1)X gauge basis to this basis, we introduce unitary matrices V L and V R
which diagonalize the down-type quark Yukawa matrix given in the U(1)X gauge basis.
We parameterize them as  

cos θL −eiαL sin θL
L
,
(30)
V =
eiαL sin θL
cos θL
and V R with the replacement L → R in V L . Then, the squark mass matrices, M2LL and
M2RR , in a basis where the quark mass matrix is diagonal are given by
4 If the U(1) charges for the first and second generations are different, the off-diagonal elements of M 2
X
LL,0
X
|QX
|QX
−QX
|
? −Q5? |
2
10
10
5
1
2 and (hΦi/Mpl )
1
2 , respectively.
and MRR,0 are suppressed by (hΦi/Mpl )
5 In the following figures, we use the exact formula for the K 0 –K¯ 0 mixing given in Appendix C.

J. Hisano et al. / Nuclear Physics B 584 (2000) 3–45
2
M2LL = V L† MLL
V L,

2
M2RR = V R† MRR
V R.

Consequently, the off-diagonal element δLL is represented as follows: 
2  
m2D
m0
X

Q
+
O
δLL = −eiαL sin θL cos θL QX
,
101
102
2
m
¯ LL
m
¯ 2LL

17

(31)

(32)

X
X
X
and δRR is obtained with the replacement L → R and QX
10i → Q5? i . If Q101 6= Q102 , δLL
is dominated by the U(1)X D-term contribution and is given definitely up to the mixing
X
angle sin θL . On the other hand, if QX
101 = Q102 , the U(1)X D-term contribution vanishes
and the second term dominates δLL . In this case, the off-diagonal element has a large model
dependence since we cannot calculate the supergravity contributions.
The contribution to K 0 –K¯ 0 mixing from the gluino–squark box diagram is calculated
in Appendix C. We have included the leading-order QCD corrections [51,34] used the bag
parameters obtained by lattice calculations [52,53]. The details are given in Appendix C.
The constraints from 1mK are summarized as follows:

m
¯ LL, RR
,
(25–35) TeV
(m
¯ LL m
¯ RR )1/2
.
(δLL δRR )1/2 <
(150–250) TeV
δLL, RR <

(33)
(34)

This shows that if δLL and δRR are of the same order, Eq. (34) gives about ten times as
severe bounds on the heavy squark masses as Eq. (33). Furthermore, if δLL and/or δRR
have CP-violating phases of order unity, the constraints from K give twelve times severer
bounds,
m
¯ LL, RR
,
(35)
(310–430) TeV
(m
¯ LL m
¯ RR )1/2
,
(36)
(δLL δRR )1/2 <
(1900–3100) TeV
as discussed in Appendix C.
Now, let us discuss Models I–IV in turn. In Model I, the averaged squark masses are
δLL, RR <

¯ 2RR = m2D .
m
¯ 2LL = m

(37)

Since the U(1)X charges for the first-two generations are the same, the D-term contribution
does not induce any flavor violation. Thus, Model I is the hybrid scenario of the decoupling
and degeneracy in a sense. Then, the flavor violation comes from the supergravity
contributions. Although we cannot calculate the off-diagonal elements in this case, we
¯ LL, RR . For simplicity, we here assume
expect δLL,RR to be of order (0.1–1)m20 /m
δLL =

m20
m2D

,

δRR =

m20
m2D

.

Then, we obtain the bound on the light sfermion mass m0 from Eq. (34) as  

m0 3
TeV,
m0 > (150–250)
mD

(38)

(39)

18

J. Hisano et al. / Nuclear Physics B 584 (2000) 3–45

if δLL and δRR do not have CP-violating phases. The ratio m0 /mD is bounded from below
in Fig. 2 in the previous section, so that the lower bounds on m0 can be estimated as
follows:
ESSM
m0 /mD > 0.18
m0 > (0.9–1.5) TeV

NESSM
m0 /mD > 0.07
m0 > (60–100) GeV

(40)

Here, the bounds on m0 /mD are those for zero gaugino mass. Note that the lower bound
on m0 depends on the third power of the ratio m0 /mD in Eq. (39). As a result, although the
bound on m0 /mD in the NESSM is only 1/2.5 of that in the ESSM, the bound on the light
sfermion mass m0 in the NESSM becomes much smaller than in the ESSM by a factor of
(1/2.5)3 ∼ 1/15.
If δLL and δRR have CP phases of order unity, the constraint from K gives twelve times
larger bounds on m0 than that from 1mK as,
ESSM
m0 /mD > 0.18
m0 > (11–18) TeV

NESSM
m0 /mD > 0.07
m0 > (0.8–1.3) TeV

(41)

Thus, it seems to require some tuning of the phases or electroweak symmetry breaking
even in the NESSM. In order to reduce the bound by a factor of 3, for example, the phase
of δLL δRR must be tuned to be 1/32 ∼ 0.1. However, including the gaugino contributions
reduces the lower bound on m0 , so that we can, in fact, realize the hierarchical spectrum
without tuning as shown in Fig. 6.

Fig. 6. Lower bounds on the bare masses for the light sfermions in Model I. The horizontal axis
represents the gluino running mass evaluated at the gluino decoupling scale, and the vertical axis the
indicated light sfermion mass at the GUT scale. The regions below the curves are excluded due to the
negative mass squareds for the corresponding sfermions. Bold lines represent the constraints from
1mK . Fine lines represent the constraints from K with O(1) CP phases.

In this case. we can estimate the lower bounds on m0 from 1mK as δLL = ESSM m0 /mD > 0. such as mψ = 2. Therefore. (43) m ¯ 2LL = 2 m2D . in the case where the U(1)X D-term contributes to the flavor violation such as in Models II–IV. This is because mD can take a smaller value as m0 decreases.2–4. the masses for the various light squarks and sleptons at the GUT scale.22 and set the off-diagonal element δLL as sin θL cos θL ∼ 0. Fig. 2. the mixing angle θL is necessary to determine δLL . the NESSM solves the CP problem in K 0 –K¯ 0 mixing without tuning of the phases or electroweak symmetry breaking. (44) √ 2 On the other hand.15. 6. In contrast. / Nuclear Physics B 584 (2000) 3–45 19 In Fig. This is peculiar to the case where the flavor violation comes from the supergravity contributions. (33):   m0 TeV. we have plotted the lower bounds on m0 . the averaged squark masses are given by √ m ¯ 2RR = m2D .J. (42) as can be seen from Eq. we here restrict our attention to the flavor violation from the U(1)X D-term contribution and give the lower bounds on the light sfermion masses using only the constraint on δLL . Note that the shape of the excluded region in Fig. so that the shape of the excluded region of m0 is similar to that of m0 /mD in the previous section. which is twice as large as the GUT-scale gaugino mass m1/2 in the previous section. Hisano et al. Then. 6 shows that there is no constraints from 1mK in the NESSM. In Model II.0 × mD . δRR is determined from the supergravity contributions and is more model-dependent than δLL . so that the U(1)X D-term contribution induces flavor violation.22. The left-handed down-type squarks of the first and second generations have different U(1)X charges. 3 in the previous section. The horizontal axis represents the running gluino mass evaluated at the gluino decoupling scale. (39). 6 is very different from that in Fig. are the same as in the previous section. we here take sin θL = 0. we turn to the other models. The only difference from the previous section is that mD is not fixed to 10 TeV but varies with the constraints from K 0 –K¯ 0 mixing. The supergravity contributions are discussed later.09 m0 > (290–400) GeV (46) . The boundary conditions at the GUT scale. Moreover. the lower bound on mD does not change with m0 . Now. including the gaugino contributions. Thus. Since the product of V L and the diagonalizing matrix for the left-handed up-type quark gives the Cabbibo angle sin θC = 0. mD > const × (m0 )2/3 . we obtain the following bound from Eq. it is natural to take sin θL ∼ sin θC .4) mD Since the bounds on the ratio m0 /mD are given in Fig.22 m0 > (700–970) GeV NESSM m0 /mD > 0. (45) m0 > (3.

(34): m0 > (2. we have plotted the lower bounds on m0 in Model II including the effect of the gaugino masses. The constraints from 1mK (Eq. Therefore. the bounds on m0 become half. 7. Lower bounds on the bare masses for the light sfermions in Model II.6–4. Hisano et al. and if we take δRR = (1/4) × m20 /m2D . we obtain the following lower bounds on m0 from Eq. (33)) are plotted. Therefore.20 J. the lower bounds on the light sfermion masses become higher by a factor of 12. we have taken sin θL = 0. we have plotted the lower bounds on m0 in Model III. the size of δRR is strongly model-dependent as we emphasized. Thus. The lower bounds on the mass ratios m0 /mD are also almost the same between Models II and III as shown in the previous section. The averaged squark masses are given by √ √ m ¯ 2RR = 2 m2D . so that the constraint is the severest among Models I–IV. 7. both δLL and δRR come from the large U(1)X D-term contribution. If we take δRR = m20 /m2D as we did in Model I. Therefore. the bounds are greatly lowered by including the effect of the gaugino masses as in Fig. However. 8. however. / Nuclear Physics B 584 (2000) 3–45 Fig. for example. In Model IV. In Fig. It shows that we can take more natural mass scale for the light sfermions in the NESSM than in the ESSM. Moreover.3) TeV in the ESSM and m0 > (430–720) GeV in the NESSM. the bounds on m0 in Fig. it seems to require a slight tuning of the electroweak symmetry breaking even in the NESSM. 8 are similar to those in Fig. If we set both angles θL and θR equal to the Cabbibo angle for simplicity. 6. the phase is necessary to be less than 10−2 to avoid extreme fine tuning of the electroweak symmetry breaking. Before turning to Model III. K 0 –K¯ 0 mixing gives the same constraints on both models. 7. If δLL has a phase of order unity.22 as in Model II. then the δLL and δRR are given by . In Fig. (47) m ¯ 2LL = 2 m2D . Since the U(1)X charge assignment for the first-two generations in Model III is the same as in Models II. the supergravity contributions are less important to restrict the parameters in Model II. we comment on the supergravity contributions.Here.

9–3.4–7. 8. 9. √ 2 Substituting these values into Eq. √ 2 δRR = sin θR cos θR ∼ 0. Lower bounds on the bare masses for the light sfermions in Model IV. / Nuclear Physics B 584 (2000) 3–45 21 Fig. m0 > (19–32) mD (48) (49) The bounds on the ratio m0 /mD are given in Fig. The constraints from 1mK (Eq.15. Hisano et al.10 m0 > (1. (34)) are plotted. 5.4) TeV NESSM m0 /mD > 0. (33)) are plotted. δLL = sin θL cos θL ∼ 0. we obtain the following bound:   m0 TeV. The constraints from 1mK (Eq. so that the lower bounds on m0 can be estimated as ESSM m0 /mD > 0. (34). Lower bounds on the bare masses for the light sfermions in Model III.J. Fig.15.23 m0 > (4.2) TeV (50) .

so that SDR scheme 0 is considered to be all-order definition of the DR scheme. The RG equations for the SUSY-breaking scalar mass squareds in this 0 scheme coincide with those in the DR scheme at the two-loop level. we discuss other RG effects on the light sfermion masses which we have not considered in the previous sections. the contributions from the heavy sfermions and extra matters in the three-loop RG equations are proportional to α3 α2 αY or αY3 . 5. Next. the RG contributions from the gauge interactions can be divided into two classes in the limit of vanishing gaugino mass in this scheme. Furthermore. We can reduce the degree of fine tuning by taking smaller values for the mixing angles sin θL and sin θR . since the terms proportional to α32 αY and α22 αY vanish due to Eqs. the three-loop RG contributions are more strongly suppressed in such models. the bottom and tau Yukawa couplings may drive the mass squareds for the doublet squark of the third generation and the right-handed stau to be negative at the one-loop level. (4). The other is the contribution from the kinetic-term mixing between U(1) symmetries. However. (5) in this case. it seems difficult to realize Model IV without fine tuning in a framework of the NESSM. the contribution from the bottom and tau Yukawa couplings. in models where the extra matters are the fundamental representation of SU(3)C and SU(2)L . we consider the effect of the bottom and tau Yukawa couplings. Then. IV the right-handed bottom squark and the left-handed slepton of the third generation obtain masses from the U(1)X D-term due to QX 5? 3 6= 0. / Nuclear Physics B 584 (2000) 3–45 In Fig 9. yukawa . we consider the correction to the light sfermions from the RG equations at the three-loop level. we must take sin θR less than 0. which exists only if there is an internal U(1) gauge symmetry. Then. and the contribution from the top Yukawa coupling. Hisano et al. When the SUSY-breaking scalar mass squareds are regarded as a D-term of an external U(1)X gauge multiplet. There are three types of contributions which potentially affect the previous analyses: the three-loop contribution through the gauge couplings. One is the contribution from mixed anomalies between the external U(1)X and the internal gauge symmetries. the RG contributions to the light sfermion masses from the heavy sfermions and extra matters only come from the U(1)Y and U(1)X mixing contribution at the three-loop level [37. In Models II.55]. First.22 J. we have plotted the lower bounds on m0 in Model IV including the effect of the gaugino mass. Other renormalization-group effects In this section. The RG contributions to the doublet squark of the third generation and the right-handed stau. Since we have imposed Eqs. m2q˜3 . in order to reduce the bounds on the masses for both the light sfermions and the gluino to 500 GeV. In consequence. the contributions are suppressed by small αY and negligible compared with the two-loop finite correction. (3)–(5) in the NESSM. We here adopt the SDR scheme defined by “the analytic continuation into superspace” [54]. are given as . Thus.38. even if there is no CP phase. yukawa and me2˜3 .03 and it causes another tuning problem. It shows that an extreme fine tuning is required in Model IV.

in the region where m0 is small enough. The size of µ is determined by the electroweak symmetry breaking. In the region where m0 is much larger than the gaugino masses. However. we have neglected it. we have plotted the lower bounds on the bare masses for top squarks. and taken At = 0 at the GUT scale for simplicity. we discuss the effect of the top Yukawa coupling. On the other hand. since At receive the positive contribution from the gaugino mass. 10 and 11. Thus. Thus. we have shown two extreme cases for the up-type Higgs mass in each choice for the sign of the µ parameter: one is the universal case mhu = m0 and the other is mhu = 0. In Figs. the RG contribution to the righthanded stau can be larger than the finite correction. 23 (51) (52) Here. There are two negative contributions to the topsquark mass squared through the top Yukawa coupling. the bound on the bare mass given by these contributions is still looser than the bounds on the light squarks from the finite correction. so that the bound is weaker than the case with µ > 0. since the effect of the top Yukawa coupling depends on the supergravity contributions. At the soft SUSY-breaking trilinear coupling associated with the top Yukawa coupling and µ the supersymmetric mass for the Higgs doublets.005% × 1 + tan2 β m2D + |Aτ |2 . the dominant effect of the top Yukawa coupling comes . the RG contribution to the doublet quark of the third generation is negligible if tan β < 10. Hisano et al. the effect of the top Yukawa coupling is much modeldependent in this region. if µ is negative. as long as tan β is sufficiently small.J. Then. For instance. and Ab and Aτ are the soft SUSY-breaking trilinear scalar couplings associated with the bottom and tau Yukawa couplings. we have assumed the universal scalar mass at the GUT scale except for the up-type Higgs mass. In the figures. IV predict the small bottom and tau Yukawa coupling constants due to QX 5? 3 6= 0 and thus require small tan β. the cancellation between two contributions from At and µ occurs. / Nuclear Physics B 584 (2000) 3–45   m2q˜3 . The left–right mixing term for top squark is given by  (53) Mt˜2 LR = mt (At + µ cot β). which gives the negative contribution in diagonalizing the topsquark mass matrix. which are model-dependent. these contributions are sufficiently suppressed. On the other hand. since we cannot predict the up-type Higgs mass at the GUT scale. if the up-type Higgs mass is much smaller than the top squark mass at the GUT scale. If µ is positive. Thus. yukawa ' −0. yukawa ' −0. the effect is negligible as shown in Figs. In the previous sections.   me2˜3 . Finally. One is the RG contribution which mainly comes from the up-type Higgs mass mhu . 10 and 11. On the other hand. the lower bound on the top squark mass is somewhat severe in this case. the At and µ are added up constructively in the left–right mixing term. the up-type Higgs mass mhu gives the dominant effect. so that we take At = 0 at the GUT scale for simplicity in the following calculations. it turns out that µ and At play an important role in calculating the negative contributions. where mt is the top quark mass. At at the weak scale is almost saturated by the radiative correction from the gluino. Here. we have assumed that tan β is not so large. The other is the contribution from the left–right mixing term. compared with the finite correction. Since Models II. However.003% × 1 + tan2 β m2D + |Ab |2 .

since . Fig. Lower bounds on the bare masses for the light sfermions in Model II are plotted by combining the constraint from 1mK (Eq. 11. / Nuclear Physics B 584 (2000) 3–45 Fig. Here.24 J. it gives the negative contribution to the top-squark mass eigenvalue of order −mt At ∼ −mt mg˜ . in order to obtain the positive top-squark mass squared at the weak scale. Since At term is generated radiatively by the gluino as At ∼ mg˜ . (36)) with the condition that mt˜1 .0.0. (33)) with the condition that m2˜ . mt˜2 > 0. we have set tan β = 3. The required increase δmg˜ of the gluino mass is determined by the following inequality: 2mg˜ δmg˜ & mt mg˜ . we have to take a larger mass for the gluino than the case without top-Yukawa contributions. The regions t1 t2 below the curves are excluded due to the negative mass-squared eigenvalues for top squarks. where t˜1 and t˜2 represent the mass eigenstates for top squarks. m2˜ > 0. 10. The regions below the curves are excluded due to the negative mass-squared eigenvalues for top squarks. from the left–right mixing term through diagonalization of the top-squark mass matrix. we have set tan β = 3. Hisano et al. Here. Thus. Lower bounds on the bare masses for the light sfermions in Model I are plotted by combining the constraint from K (Eq.

when only the right-handed squarks of the first and second generations have a common U(1)X charge. the lower bounds on the light sfermion masses become somewhat severer by including the effects of the top Yukawa coupling.and right-handed down-type squarks of the first-two generations have common U(1)X charges. we can still satisfy the constraint from 1mK naturally. 6. . we have found that we can still take the light sfermion and gluino masses as small as ∼ 400 GeV in the natural effective SUSY. we have calculated the finite corrections to the light sfermion masses from the heavy sfermions and extra matters at the two-loop level in the NESSM. Then. the supersymmetric contributions to 1mK and K are sufficiently suppressed without breaking naturalness. On the other hand. which are generically non-universal in flavor space. it gives the lower bounds on the lightsfermion bare masses similar to the ones we have derived in this article. We have also discussed whether the models in which the NESSM is realized by the U(1)X gauge interaction at the GUT scale are viable or not in the light of the experimental constraints from 1mK and K . since they do not have explicit renormalization-point dependence. The formulae we have given in Appendix B are applicable to any models which have hierarchical mass spectrum for the sfermions such as effective SUSY. the supergravity contributions to the SUSY-breaking terms. even if the flavor-violating supergravity contributions to the sfermion mass matrices are included. but evading the bound from K requires somewhat small CP phase of order 10−2 . we believe that our result gives the most conservative bound for the naturalness in the effective SUSY models. To summarize. The finite corrections increase (decrease) if the supersymmetric masses (the holomorphic SUSY-breaking masses) of the extra matters become large. 10 and 11. That is. While the sfermion mass squareds in the NESSM are RG invariant at the two-loop level in the limit of vanishing gaugino-mass and Yukawa-coupling contributions. when both the left.J. Thus. / Nuclear Physics B 584 (2000) 3–45 25 the positive contribution from the gluino is estimated as of order m2g˜ . the finite corrections may still drive the light-sfermion mass squareds to be negative. the formulae for the finite corrections can also be applied to the models where the U(1)X Dterm is generated at much lower energy than the GUT scale [25–27]. we have to take a larger mass for the gluino by the order of 100 GeV compared with the previous case as shown in Figs. assuming that all fields are embedded in the SU(5)GUT multiplets at the GUT scale. are suppressed compared with the U(1)X D-term contribution. while we have to take them larger than ∼ 1 TeV in the original effective SUSY. Conclusions and discussion In this article. Hisano et al. However. We have found that the corrections can be less than (0.3–1)% of the heavy-squark mass squareds. On this setup. In particular. We have found that.

a. . Sports and Culture of Japan. . Nf X represent flavor indices. and introduce two singlet chiral superfields Φ1 and Φ2 . The D-term potential for U(1)X is given by .38] and show that it naturally realizes the setup of the NESSM. Q5? ex < 0). and take the sign convention of U(1) charges such that ξ 2 > 0. The tree-level superpotential of the model is given by  Qa Q¯ a¯ (fφ )aa¯ Φ1 Φ2 + (fψ )aa¯ Ψ5ex Ψ5? ex . a¯ = 1. Q5? i > 0).K.19]. the SUSY-breaking model is constructed as follows. We also introduce extra-matter chiral multiplets X X X Ψ5ex and Ψ5? ex with U(1)X charges QX 5ex and Q5? ex (Q5ex . Science. / Nuclear Physics B 584 (2000) 3–45 Acknowledgements We would like to thank Hitoshi Murayama and Youichi Yamada for useful discussions. Here. X X X X Q¯ a¯ (−(Q5ex + Q5? ex )/2). and the effective superpotential of the model is exactly given by  1 Φ1 Φ2 Tr (fφ M) + Ψ5ex Ψ5? ex Tr (fψ M) Weff = Mpl  3Nc −Nf  1 Nc −Nf Λ . Then.2) + (Nc − Nf ) det M in terms of gauge-invariant composite fields M a = Qa Q¯ a¯ [57]. the dynamical superpotential is generated by nonperturbative effects of the SU(Nc ) gauge interaction. Here. The model is based on the anomalous U(1) SUSY-breaking model [18. This leads to the generation of a nonzero Fayet–Iliopoulos (FI) D-term ξ which is smaller 2 . on Priority Area 707 “Supersymmetry and Unified Theory of Elementary Particles” and by the Grant-in-Aid No. Λ is the dynamical a¯ scale of the SU(Nc ) gauge theory.10740133 (J. The anomalous U(1) gauge symmetry frequently appears in low-energy effective theories of string theory. Hisano et al. thank the Japan Society for the Promotion of Science for financial support. Anomalous U(1) model In this appendix.). (A. . We consider the SU(Nc ) SUSY gauge theory with Nf flavors Qa and Q¯ a¯ (Nc /2 < Nf < Nc ). ξ 2 ' (g 2 /192π 2 ) Tr QMpl parameterize it as ξ ≡ Mpl with  = O(0.N. Thus. We assign the U(1)X charges as Qa (−(QX 5ex + Q5? ex )/2). K. Φ1 (−1) and Φ2 (1 + Q5ex + Q5? ex ).H. respectively (Q10i . but its anomalies are canceled by a nonlinear transformation of the dilaton chiral multiplet [56]. We identify this anomalous U(1) gauge symmetry with U(1)X given in the text.26 J. This work was supported in part by the Grant-in-Aid for Scientific Research from the Ministry of Education. The matter content is anomalous under this U(1) symmetry. (6).1).1) Wtree = Mpl Then. (A. we briefly review the model given in Refs. in order to satisfy Eqs. [37. (8). We than the reduced Planck scale Mpl by a one-loop factor. Appendix A. and Y. . X we assign the U(1)X charges QX 10i and Q5? i to the quark and lepton chiral multiplets Ψ10i X X and Ψ5? i .

J. Φ2 and M at the minimum are given by solving the following equations:  X 2 2 (A.3) In this appendix.  3Nc −Nf  1 Nc −Nf  Λ (fφ )aa¯ = Φ1 Φ2 .6) M −1 aa¯ det M Mpl From Eqs. Hisano et al.6) as hΦ1 i ' hΦ2 i ' ξ. hMi ' Λ 3Nc −Nf Nc  Mpl ξ2  Nc −Nf Nc . The minimum of the potential can be obtained by making an expansion in the small parameter (Λ/Mpl )(3Nc −Nf )/Nc  1. nonperturbatively generated superpotential forces M to have VEV’s.5). First. (A. The large FI D-term.4). (A. (A. This nonvanishing D-term gives the MSSM sfermions and extra scalars the SUSY-breaking mass squareds proportional to their U(1)X charges. Now. Calculating higher order in (Λ/Mpl )(3Nc −Nf )/Nc . The point is that if fψ is larger than fφ . (A. is satisfied.4) |Φ1 |2 − 1 + QX 5ex + Q5? ex |Φ2 | = ξ . which gives supersymmetric mass terms for the singlet fields and the extra matter fields.7) Then. and the VEV of the M field is given by Eq. the VEV’s of the fields Φ1 . The dynamics of the model can be understood as follows. Then. they cannot absorb ξ completely and nonzero U(1)X D-term remains. To the leading order.5) = − 1 + QX 5ex + Q5? ex |Φ2 | (Nf − Nc )|Φ2 | + Nf |Φ1 | . we find that the VEV’s of the auxiliary fields can be written in terms of the VEV’s of the scalar fields as − F¯M † a¯ a −F¯Φ † 1 −F¯Φ † 2 . − |Φ1 |2 + 1 + QX 5ex + Q5? ex |Φ2 | + Q5ex |Ψ5ex | + Q5? ex |Ψ5? ex | 27 (A. since the singlet fields have the supersymmetric mass induced by the VEV’s of M. is absorbed by the shift of the singlet fields Φ1 and Φ2 . (13)).  2 2 2 |Φ1 | (Nf − Nc )|Φ1 | + Nf |Φ2 |   X 2 2 2 (A. fψ > fφ . we use the same letter for the chiral superfield and its scalar component. ξ . we find that both the Φ1 and Φ2 fields have nonvanishing VEV’s of order ξ . the Yukawa matrices for the quarks and leptons are generated through the VEV of the Φ1 field suppressed by suitable powers of hΦ1 i/Mpl ' ξ/Mpl =  (see Eq. avoiding phenomenologically disastrous large breaking of the SM gauge groups. let us minimize the potential explicitly. / Nuclear Physics B 584 (2000) 3–45 VD = 2  p gX X † ξ 2 − QX 5ex + Q5? ex Tr M M 2   X 2 X 2 X 2 2 . (A. We assume that this condition. only the singlet fields shift to absorb the FI D-term and the extra-matter fields do not develop VEV’s.

2  Nc − Nf 1 |Φ1 |2 + |Φ2 |2 .

.

=− Tr(fφ M).

fφ∗−1 aa¯ . Mpl   Φ1 = Tr(fφ M). Mpl (A.8) (A.10) .9) (A. ∗ ∗ Nc Nf Mpl Φ1 Φ2   Φ2 = Tr(fφ M).

28 J. / Nuclear Physics B 584 (2000) 3–45 −DX = . Hisano et al.

.

2Nf − Nc 1 |Φ1 |2 + |Φ2 |2 .

Tr(fφ M).

13) (A.14). −DX ' N ξ 2(Nc −Nf ) Mpl f (A. From Eqs.8)–(A.14) √ The dynamical scale Λ is determined so that |−DX | ' (1–10) TeV to give the heavy sfermions multi-TeV masses. we find the following useful relation: . (A. 2N −Nc ξ 4(Nc −Nf ) Mpl f  1/Nc Λ3Nc −Nf ¯ ¯ ξ.11).11) Substituting Eq.12) (A. (A.7) into Eqs. X X 2 Nc Mpl |Φ1 |2 − (1 + Q5ex + Q5? ex )|Φ2 |2 (A.2 .12)–(A. (A. −FΦ † ' −FΦ † ' N 1 2 ξ 2(Nc −Nf ) Mpl f  2/Nc Λ3Nc −Nf . we obtain −F¯M † '  Λ2(3Nc −Nf ) 1/Nc .

.

.

.

.

.

.

−FΦ1 .

2 .

−FΦ2 .

2 .

−FM .

2 .

'.

.

'.

.

15) |−DX | ' . (A.

.

.

hΦ i .

.

hMi .

. hΦ1 i .

since their sizes are estimated as 6 . 2 This relation ensures that flavor-breaking supergravity contributions are an order of magnitude smaller than the contribution from the U(1)X D-term.

.

.

.

p .

.

−FΦ1 .

.

 .

.

' ξ .

−FΦ1 .

'  |−DX | ' 102 –103 GeV.16) . (A.

M .

M .

hΦ i .

pl pl 1 √ From Eqs. Since U(1)X is broken down at very high energy scale of order ξ . The gaugino masses arise from the F -term of the dilaton field [48]. (14). Finally. the supersymmetric mass for the extra matters is actually the same order with the SUSY-breaking masses for the heavy sfermions. (A. and then it does not much affect the preceding √ analysis of the dynamics. .14). except for the ones which involve only the light sfermions (see Eqs. We have identified the U(1)X in the text with the anomalous U(1) gauge symmetry. The SUSY-breaking trilinear scalar couplings of order |−DX | are also generated by the superpotential which generates Yukawa matrices for the quarks and leptons. (15)). it does not necessarily mean that the matter content is 6 Since the gravitino mass is of the order of the weak scale.7). it is not a contradiction. (6) might seem contradicted by the fact that the anomalous U(1) gauge symmetry has mixed anomalies for all the other gauge groups including the SM ones. we also find that |−DX | ' hMi/Mpl . W = µHu Hd ) of the desired size (the weak scale) by introducing a holomorphic term K = Hu Hd in the Kähler potential [58]. Their sizes can be of the order of the weak scale [49]. Then. we comment on the anomaly. Thus. Eq. (A. However. we can induce µ-term (supersymmetric mass term for the Higgs field.

c. M ˜¯ 2 Fψ |mψ |2 + m (B. − m Then. That is. the extra scalars have the SUSY-breaking mass terms  ∗ ˜ ˜¯ 2 ψ˜¯ ∗ex ψ˜¯ ex . keeping √ Eqs. since we have adopted the SDR scheme 0 [54] which is all-order definition of the DR scheme in the text (see Section 5). ψ¯ ex ! e2 Fψ∗ |mψ |2 + m 2 eex = . 7 Appendix B. the large two-loop RG contributions are absent below the ξ scale as we have explained above. if the supertrace of the matter fields is nonzero. The superpotential of the vector-like extra matters. mA˜ receives divergent radiative correction at one loop through loops of the sfermions. First. However. Thus. so that the counterterm is needed to cancel this divergence.2) ψex − m Lex. (8) satisfied between two scales |−DX | and ξ . Finite corrections In this appendix. (A. The insertion of this counterterm gives divergent contribution to the sfermion masses at one loop. is given by Wex = mψ Ψex Ψ¯ ex .J. We use 0 the DR scheme [36] to regularize the theory. Then. Ψex and Ψ¯ ex . In this scheme. if we introduce fields Ψ5anom and Ψ5? anom of masses of order ξ with the superpotential W ∼ hΦ1 iΨ5anom Ψ5? anom . In addition. e2 ψ˜ ex (B. we calculate the contributions to the gaugino and the light sfermion masses arising from loops of the extra-matter multiplets and the heavy sfermions. the -scalar mass mA˜ is zero at the tree level and it does not appear in the relation between physical quantities. (6). However.3) (B.4) 2 can be diagonalized by the unitary matrix V as eex The mass matrix M 7 The light-sfermion mass squareds may receive negative RG and finite contributions above and at the decoupling scale of Ψ5anom and Ψ5? anom . (B. since there is no large log-factor in the contributions. . We include these contributions in the supergravity contributions when we make phenomenological analyses in the text. we explain our notations. soft = − Fψ ψ˜ ex ψ˜¯ ex + h. the mass terms for the extra scalars are written as !  2 ψ˜ ex ∗ ˜¯ ˜ e Lex = − ψ ex .1) We denote the scalar and fermion components of Ψex (Ψ¯ ex ) as ψ˜ ex and ψex (ψ˜¯ ex and ψ¯ ex ). / Nuclear Physics B 584 (2000) 3–45 29 anomalous below ξ scale. we have to carefully treat the -scalar in order to obtain the two-loop contribution to the light sfermion masses when the supertrace is nonvanishing. we can make these negative contributions smaller than the supergravity contributions by choosing their group-theoretical factors to be small. Hisano et al. respectively.17) we can match the anomalies as required by the anomalous U(1) symmetry. ψ ex Mex ˜ ∗ .

6) and define y1 ≡ m21 m2ψ .5) (B. the contribution from the heavy sfermions are obtained by taking the limit mψ → 0 and θ → 0. Note that T A is the Dynkin index of the extra matter and not the sum of the index of the extra matter and anti-matter. That is. We parameterize V as   cos θ −eiα sin θ . Diagram contributing to the gaugino masses. Hisano et al. The gaugino masses The gauginos acquire their masses through the one-loop diagram of the extra-matter multiplet shown in Fig. 12. 12.30 J. Here. [59]. y1 . . If there is a pair of vector-like extra matters. Fig. y2 . we use T A = 1/2 for a vector-like pair of fundamental extra matters.8) mg˜A = 2 2 2 4π mα − mψ m2ψ α=1. (B.9) where the trace is taken over pairs of the extra matters R. T A is the Dynkin index of the extra-matter representation.2 where A = 1–3 represents the standard-model gauge groups. V= e−iα sin θ cos θ (B. Then. and we have adopted the √ SU(5) GUT normalization for the U(1)Y gauge coupling (g1 ≡ 5/3 gY ). the gaugino masses are given by     αA y1 log y1 − y2 log y2 − y1 y2 log(y1 /y2 ) A −iα TrR TR mψ e . it is given by  2 2 X † gA mα m2α A T mψ V2α Vα1 log . (Note that mψ .1. (B. m22 = V M V . / Nuclear Physics B 584 (2000) 3–45  2 † eex diag m21 . in a normalization T A = 1/2 for a fundamental of SU(N ) and T 1 = (3/5)Y 2 for U(1)Y . sin 2θ mg˜A = 2 4π (y1 − 1)(y2 − 1) (B. y2 ≡ m22 m2ψ .) This is in agreement with the result given in Ref. Then.7) We calculate the gaugino and the light sfermion masses arising from loops of the extramatter multiplets in terms of these parameters. α and θ depend on R. B.

The other contribution is graphs involving heavy particles. 13.9). One diagram contributing to the light sfermion masses is the one-loop -scalar graph shown in Fig. the heavy sfermions do not contribute to the gaugino masses. One is the contribution directly arising from the loop diagrams: the one-loop graph involving the -scalar and the two-loop 2 . (B. (B. we obtain  i 1 h X αA 2 Λ2IR 2 A A 2 − log 2 . Λ (4π)2− IR A˜ (B. A˜ = −4 4π  µ (B.11) as δm2 A˜   2 h i 1 2gA A 2 2 2 =− TrR TR m1 + m2 − 2mψ .12) into Eq. A˜ = 2 0() 2 − 2 X 2 A µ gA Cf˜ δm2 .J. There are two types of contributions which generate the light sfermion masses. Here. and gives A˜ A˜ the light-sfermion mass squared.10). . The counterterm δm2 is determined to cancel the divergence of the -scalar mass arising from one loop of A˜ the extra-matter multiplets. respectively. It can also be seen by taking the limit mψ → 0 and θ → 0 in Eq. L = −(1/2)δm2 A˜ aµ˜ A˜ aµ˜ . B. The sfermion masses We now calculate the light sfermion masses at the two-loop level. and ΛIR is the infra-red cutoff. It contains the counterterm δm2 for the -scalar mass.12) (B. Cf˜ TrR TR mψ (y1 + y2 − 2) mf˜. / Nuclear Physics B 584 (2000) 3–45 31 Obviously. We first consider the direct contribution from the extra-matter multiplets. both of which are of order αA that through the generation of U(1)Y FI D-term at one loop and two loops. One-loop diagram contributing to the light sfermion masses which involves the -scalar A. C A˜ is the quadratic Casimir coefficient for the light f sfermion f˜.10) A in D = 4 − 2 dimensions.13) A ˜ Fig. which are of order αY and αY αA . m2f˜. (B. 13. (4π)2  Substituting Eq. (B. µ is the renormalization scale.2. Hisano et al.

[50] in the case of vanishing supertrace.13). the combination −γ + log(4π) has been absorbed by an appropriate redefinition √ of the renormalization scale.32 J. Their contribution is ultra-violet finite even in the case of nonvanishing supertrace. so that it contributes to the two-loop RG equations for the light has 1/ pole of order αA sfermion masses. (B. These graphs are identical to those considered in Ref. We will omit µ. hereafter. The result 2 . / Nuclear Physics B 584 (2000) 3–45 Here. Together with Eq. 14. we obtain the light sfermion masses induced directly by loops of the extra-matter multiplets as i 1  h X αA 2 . The remaining diagram consists of two-loop graphs involving the extra-matter multiplets given in Fig. µ → µ eγ /4π (where γ is the Euler number). Hisano et al.

2 A A 2 .

Cf˜ TrR TR mψ (y1 + y2 − 2) mf˜. (B.14) + sin2 2θ y1 Li2 1 − 2 y1 y2 R1 which is in agreement with the result given in Ref. Taking the limit mψ → 0 and θ → 0 and regarding m21 as the heavy sfermion masses m2˜ . Li2 (x) = − 0 dt log(1 − xt)/t is the dilogarithm function. 13 and 14. direct R = −4 4π  A     X αA 2  CfA˜ TrR TRA m2ψ (y1 + y2 − 2) log m2ψ + 2 +4 4π A      1 1 + (y1 log y1 + y2 log y2 ) − 2 y1 Li2 1 − + y2 Li2 1 − y1 y2      y2 y1 1 + y2 Li2 1 − . we obtain the light sfermion masses induced by loops of the heavy F sfermions as h i 1  X αA 2 . Note that Λ2IR is canceled between the diagrams of Figs. Here. [59].

direct. 2 CfA˜ TrF˜ TFA m m2f˜.

(B. .15) 4π 3 A to the light where F˜ denotes the heavy sfermions. The direct contribution m2˜ f .F˜ = −4 ˜ F˜ 4π  A      X αA 2 1 2 A A 2 2 Cf˜ TrF˜ TF˜ mF˜ log mF˜ + 2 − π +4 . direct sfermion masses is given by summing up that from the extra-matter multiplets and that from the heavy sfermions.

.

16) m2f˜. direct. direct = m2f˜. (B.

direct .R + m2f˜.

F˜ .
We next consider the light sfermion masses induced via the generation of U(1)Y FI
D-term. At the one-loop level, the light sfermion masses generated by loops of the extramatter multiplets are given by (see Fig. 15)  

 

h

i 1
αY
2
2
2

Y ˜ TrR YR cos 2θ m1 − m2
+1
mf˜, FI-1loop R = −
4π f  


h 
i
αY
(B.17)
Yf˜ TrR YR cos 2θ m21 log m21 − m22 log m22 .
+

J. Hisano et al. / Nuclear Physics B 584 (2000) 3–45

33

Fig. 14. Two-loop diagram contributing to the light sfermion masses which involves the extra-matter
multiplets Ψex and Ψ¯ ex .

Fig. 15. One-loop diagram contributing to the light sfermion masses through the generation of U(1)Y
FI D-term.

34

J. Hisano et al. / Nuclear Physics B 584 (2000) 3–45

Fig. 16. Two-loop diagram contributing to the light sfermion masses through the generation of U(1)Y
FI D-term.

This contribution vanishes if the extra-matter multiplets have an invariance under the
parity Ψex ↔ Ψ¯ ex , as can be readily seen by taking θ = ±π/4 in the expression. The heavy
sfermions also generate one-loop U(1)Y FI D-term, which gives  


h
i 1

αY
Yf˜ TrF˜ YF˜ m2F˜
+1
m2f˜, FI-1loop

F˜ = −
4π  


h
i
αY
(B.18)
Yf˜ TrF˜ YF˜ m2F˜ log m2F˜ .
+

The total contribution through the generation of U(1)Y FI D-term at one loop is

m2f˜, FI-1loop = m2f˜, FI-1loop

R +m2f˜, FI-1loop

F˜ .

(B.19)

U(1)Y FI D-term is also generated at two loops through the diagram shown in Fig. 16.
The resulting light sfermion masses from extra-matter loops are written as  

X
X



† 
2
2 2
A

gY gA Yf˜ TrR YR CR
− Vα2 V2α
Vα1 V1α
I1 m2α
mf˜, FI-2loop R =
A

+4

X
α

X
αβγ

α


Vα1 V1α 

† 
− Vα2 V2α
I2 m2α


† 

† 
− Vα2 V2β
− Vβ2 V2γ
Vα1 V1β
Vβ1 V1γ

J. Hisano et al. / Nuclear Physics B 584 (2000) 3–45

35 


† 
× Vγ 1 V1α
− Vγ 2 V2α
I3 m2α , m2β , m2γ 

X 


† 
2
2
Vα1 V1β − Vα2 V2β (δm2 )βα Ic mα , mβ
,
+

(B.20)

αβ

in the Feynman gauge. Here, functions I ’s are defined as
Z 4 Z 4 

d p
d k
1
(2p − k)2
1
2
, (B.21)
I1 mα =
(2π)4 (2π)4 (p2 − m2α )2
k2
(p − k)2 − m2α
Z 4 Z 4 

d p
1
k2 − k · p
1
d k
, (B.22)
I2 m2α =
(2π)4 (2π)4 (p2 − m2α )2
k2
(p − k)2 − m2ψ
Z 4
Z 4 

d p
d q
1
1
1
,
(B.23)
I3 m2α , m2β , m2γ =
4
2
2
2
4
2
2
(2π) p − mα p − mβ (2π) q − m2γ
Z 4 

1
1
d p
2
2
.
(B.24)
Ic mα , mβ = i
(2π)4 p2 − m2α p2 − m2β
P 2 A
CR (δm2 )αβ (ψ˜ ex )∗α
The counterterm (δm2 )αβ for the extra-scalar masses, L = − A gA
˜
(ψex )β , is determined to cancel the divergence,

(B.25)
as
(δm2 )αβ = 


1
1
2
(4π) 
×

− sin2 2θ (m21 − m22 ) − 4m2ψ sin 2θ cos 2θ (m21 − m22 ) eiα

sin 2θ cos 2θ (m21 − m22 ) e−iα sin2 2θ (m21 − m22 ) − 4m2ψ

!
. (B.26)

Substituting Eq. (B.26) into Eq. (B.20), we obtain

m2f˜, FI-2loop

R =  

h
X αY  αA  
i 1
A
2
2
Y ˜ TrR YR CR cos 2θ m1 − m2
+3
−2

4π f 

A  

X αY  αA 
1
1
Yf˜ TrR YR CRA cos 2θ − m2ψ log2 m21 + m2ψ log2 m22
+4


2
2
A  

+ m21 + m2ψ log m21 − m22 + m2ψ log m22    

m2ψ
m2ψ  

+ m21 − m2ψ Li2 1 − 2 − m22 − m2ψ Li2 1 − 2
m1
m2

36 J. Hisano et al.27) This contribution vanishes when θ = ±π/4 as it should be. / Nuclear Physics B 584 (2000) 3–45  (m2 log m21 − m22 log m22 )2 1 + YR CRA cos 2θ sin2 2θ 2 1 4 m21 − m22   − 4 m21 log m21 − m22 log m22 − m21 log2 m21 − m22 log2 m22    + m21 − m22 2 + log m21 + log m22 − log m21 log m22 . (B. The contribution from the heavy sfermions is read off by taking the limit mψ → 0 and θ → 0 as .

FI-2loop. m2f˜.

Two-loop diagram contributing to the light sfermion masses. Y where  J1 = TrR YR X † Vα1 V1α †  − Vα2 V2α α  Z 4  1 d p . there is another diagram which contributes to the light sfermion masses at the two-loop level.29)  (B. FI-2(1loop) = −gY4 Yf˜ J1 J2 + igY4 Yf˜ δm2 J2 + igY4 Yf˜ J1 δg 4 . + TrF˜ YF˜ (2π)4 p2 − m2˜ Z d 4p 1 4 2 (2π) p − m2α Y (B. 17. since they subtracted divergences not in the MS scheme.F˜ =  h i 1 X αY  αA  A 2 Y ˜ Tr ˜ Y ˜ C m +3 −2 4π 4π f F F F˜ F˜  A    X αY  αA  1 2 A 2 2 Y ˜ Tr ˜ Y ˜ C m log mF˜ + π . . 8 In addition. [34]. The diagram is shown in Fig. 8 Their result is different from ours by a finite part proportional to −γ + log(4π ). (B.30) F Fig.28) +4 4π 4π f F F F˜ F˜ 6 A reproducing the earlier result derived in Ref. 17 and its contribution is m2f˜.

FI-2(1loop) =  −  − αY 4π αY 4π 2 Yf˜ 2  2 n h h io i 1 TrR YR cos 2θ m21 − m22 + TrF˜ YF˜ m2F˜  h i h i h io n × 2 TrR YR2 + TrF˜ YF2˜ + Trf˜ Yf2˜ n h h i i Yf˜ TrR YR cos 2θ m21 − m22 + TrF˜ YF˜ m2F˜ h h io i − TrR YR cos 2θ m21 log m21 − m22 log m22 − TrF˜ YF˜ m2F˜ log m2F˜    × TrR YR2 cos2 2θ log m21 + log m22 . (B.35).31) + TrF˜ YF2˜ ˜ f f˜ (2π)4 (q 2 − m2˜ )2 (2π)4 (q 2 − m2˜ )2 F f The counterterms δm2 and δg 4 are defined by Y Y X   † †  ˜ ∗ 2 ∗ ∗ ˜ ˜ ˜ ˜ ˜ Vα1 V1β − Vα2 V2β ψex α YR ψex β + ψF˜ YF˜ ψF˜ + f Yf˜ f L = −gY δm2 Y αβ − gY4 δg 4 f˜∗ Yf˜ f˜ Y X † Vα1 V1β †  − Vα2 V2β  ∗  ∗ ˜ ˜ ˜ ˜ ψex α YR ψex β + ψF˜ YF˜ ψF˜ . (B. we obtain m2f˜.36) From Eqs. which lead to h h io 1  i 1 n 2 2 2 cos 2θ m − m m Tr Y Y + Tr . (B. (B.32) αβ They are determined by the conditions. Hisano et al.33) and (B. (B. (B.35) (B.36).34) respectively.29). δm2 = R R 1 2 F˜ F˜ F˜ Y (4π)2  h i h i h io 1  1 n 2 2 2 . / Nuclear Physics B 584 (2000) 3–45 37  X  Z 4 1 1 d q † † † †  J2 = TrR YR2 − Vγ 2 V2δ − Vδ2 V2γ Vγ 1 V1δ Vδ1 V1γ (2π)4 q 2 − m2γ q 2 − m2δ γδ  Z 4   Z 4  1 1 d q d q 2 + Tr Y . 2 Tr Y + Tr Y + Tr δg 4 = R R F˜ f˜ Yf˜ F˜ Y  (4π)2 (B.J.

the light sfermion masses induced at two loops through the generation of U(1)Y FI D-term are given by . (B.38 J. Hisano et al. / Nuclear Physics B 584 (2000) 3–45   m2 log m21 − m22 log m22 − 2YR2 sin2 2θ 1 − 1 m21 − m22 h i h i + TrF˜ YF2˜ log m2F˜ + Trf˜ Yf2˜ log m2f˜ .37) Thus.

.

38) m2f˜. FI-2loop = m2f˜. (B. FI-2loop.

R + m2f˜. FI-2loop.

FI-2loop. direct + m2f˜. direct + m2f˜. (B.40) TrR TRA m2ψ (y1 + y2 − 2) + TrF˜ TFA ˜ mF˜ = 0.43) m2f˜. However.41) TrR YR cos 2θ m21 − m22 + TrF˜ YF˜ m2F˜ = 0.19). the divergences are canceled among loops of various heavy particles under the conditions Eqs. in the notation of this appendix. i h h i  (B.42) TrR YR CRA cos 2θ m21 − m22 + TrF˜ YF˜ CFA˜ m2F˜ = 0. (B. FI-2loop. (3)–(5) discussed in the text. (B.19).16). Altogether.F˜ + m2f˜. total = m2f˜. FI-2(1loop). the light sfermion masses at the two-loop level are given by m2f˜. (B. (B. (B.16). (B. (B. m2f˜. direct =   X αA 2 CfA˜ TrR TRA m2ψ (y1 + y2 − 2) log m2ψ + (y1 log y1 + y2 log y2 ) 4 4π A      1 1 + y2 Li2 1 − − 2 y1 Li2 1 − y1 y2      y2 y1 1 +y2 Li2 1 − + sin2 2θ y1 Li2 1 − 2 y1 y2    X αA 2 1 2 2 2 CfA˜ TrF˜ TFA . The light sfermion masses generically have divergent contribution. FI-1loop + m2f˜.38). FI-1loop + m2f˜. so that they receive large negative contribution from the RG evolution. (B. These conditions are written as i h h i 2 (B. The finite case In the previous subsection.44) +4 ˜ mF˜ log mF˜ − π 4π 3 A . we have calculated the gaugino and the light sfermion 0 masses at the two-loop level in the DR scheme. we obtain the finite contribution to the light sfermion masses.39) combining Eqs.38). B. h h i i  (B.3. Using these relations in Eqs. total = m2f˜.

FI-1loop =   h i αY Yf˜ TrR YR cos 2θ m21 log m21 − m22 log m22 4π   h i αY Yf˜ TrF˜ YF˜ m2F˜ log m2F˜ . . αA (µ) and αY (µ). FI-2loop =   X αY  αA  1 1 Yf˜ TrR YR CRA cos 2θ − m2ψ log2 m21 + m2ψ log2 m22 4 4π 4π 2 2 A   + m21 + m2ψ log m21 − m22 + m2ψ log m22     m2ψ m2ψ   + m21 − m2ψ Li2 1 − 2 − m22 − m2ψ Li2 1 − 2 m1 m2  2 2 2 2 2 (m log m1 − m2 log m2 ) 1 + YR CRA cos 2θ sin2 2θ 2 1 4 m21 − m22   −4 m21 log m21 − m22 log m22 − m21 log2 m21 − m22 log2 m22    + m21 − m22 2 + log m21 + log m22 − log m21 log m22    X αY  αA  1 Yf˜ TrF˜ YF˜ CFA˜ m2F˜ log m2F˜ + π 2 +4 4π 4π 6 A  2 n h i αY Yf˜ TrR YR cos 2θ m21 log m21 − m22 log m22 + 4π h io + TrF˜ YF˜ m2F˜ log m2F˜    × TrR YR2 cos2 2θ log m21 + log m22   m21 log m21 − m22 log m22 1− m21 − m22  h i + TrF˜ YF2˜ log m2F˜ . Here. / Nuclear Physics B 584 (2000) 3–45 m2f˜. log(m /µ ).45) m2f˜.J. since the extra scalars embedded in a common SU(5)GUT multiplet could have different masses due to the difference between their supersymmetric masses caused by RG evolution below the GUT scale. + 4π 39 (B.46) It should be understood that the coupling constants αA and αY are the renormalized ones at the scale µ. FI-1loop even if we assign U(1)X charges consistent with the SU(5)GUT. we have dropped the term proportional 2 to Trf˜ Y ˜ log m ˜ which is the contribution from the RG evolution below µ. Hisano et al. −2YR2 sin2 2θ (B. Note that f f from one-loop U(1)Y FI D-term is generically nonvanishing the contribution m2˜ f . and the logarithms log m2 of various masses are 2 2 normalized  2 at the  scale µ.

F 2. FI-2loop = 4 4π 4π (B. Hisano et al. direct = 4 4π A   × TrR TRA m2ψ (y1 + y2 − 2) log m2ψ + (y1 log y1 + y2 log y2 )      1 1 + y2 Li2 1 − − 2 y1 Li2 1 − y1 y2      y2 y1 1 y1 Li2 1 − + y2 Li2 1 − + 2 y1 y2    X αA 2 1 2 A A 2 2 Cf˜ TrF˜ TF˜ mF˜ log mF˜ − π .44)–(B.47) +4 4π 3 A m2f˜. (B. M2LL and M2RR are diagonalized separately by the sfermion mixing matrices (U L )XL . FI-1loop = 0. (B. The U(1)X charge assignment is consistent with SU(5)GUT: TrF˜ [YF˜ f (m2˜ )] = 0. The mass matrix for the down-type squark is relevant to the gluino box diagram. Then.i and (U R )XR . In a basis where the down-type Yukawa matrix is diagonal. [51–53]. we calculate the constraints on the first-two generation sfermion masses from 1mK and K in effective SUSY. where the gluino is much lighter than the firsttwo generation sfermions. we not only take into account the leading QCD corrections but also make use of B parameters instead of the vacuum insertion approximation. / Nuclear Physics B 584 (2000) 3–45 Furthermore.40 J. From now we take the left–right mixing mass term M2LR = 0 because of the small Yukawa couplings. the above expression is considerably simplified when the following two conditions are satisfied: 1. since we calculate the constraints on the first-two generation sfermion masses. Constraints from 1mK and K In this appendix. j are the indices of the generation.1) Lmass = − dLi dRi 2 d˜Rj (M2† LR )ij (MRR )ij where the subscripts i. 2. Eqs. In the following calculation. the mass matrix is !  2 2 (M ) (M )  ij ij d˜Lj LL LR ∗ ˜∗ ˜ . m2f˜. In consequence.48) (B. we only consider the gluino box diagram since it gives a dominant contribution.49) A Appendix C. According to Refs.i as follows: .46) are reduced to X αA 2 CfA˜ m2f˜. (C. We here restrict the subscript i to i = 1. h i X αY  αA  Yf˜ TrF˜ YF˜ CFA˜ m2F˜ log m2F˜ . The extra-matter multiplets have an invariance under the parity Ψex ↔ Ψ¯ ex : θ = ± π 4 These conditions are satisfied in the case of Model I and Model II considered in the text.

7) (C. The operators O1. We here give a convenient expression by using the so-called mass insertion method [60. (C.4) (C.4. The above effective Lagrangian is valid for an arbitrary down-type squark mass matrix. / Nuclear Physics B 584 (2000) 3–45  U L M2LL U L† = diag m2XL . where the heavy sfermions decouple. which is defined by a geometric mean as q q m ¯ 2RR ≡ m21R m22R . The other is the off-diagonal element δLL (δRR ) of M2LL (M2RR ).2 IXL .6) (C. is written as   (C.9) become .2 IXL .J.   O5 = d¯α PL sβ d¯β PR sα . 2R .YL UYL . 9 9 (C. 61].5 and their coefficients C1.Y = − log m2X − log m2Y m2X − m2Y .7)–(C.XL UXL .YR .YL + IXL .2) where XL = 1L .2R m ¯ 2RR .   O4 = d¯α PL sα d¯β PR sβ .YR UYR . (C. The 1S = 2 effective Lagrangian at the scale mD . Hisano et al. (C.2 U1. 41 (C.9) O˜ 1 and C˜ 1 are obtained with the replacement L → R in O1 and C1 . (m21L − m22L ) and (m21R − m22R ). I˜X.Y and I˜X. (C.YL . respectively.XL UXL . since they are suppressed by m2g˜ /m2X ( 1).12) By taking a leading order of the mass differences.YR UYR .YR + IXL .2 U1.3) Leff = α32 (mD ) C1 O1 + C˜ 1 O˜ 1 + C4 O4 + C5 O5 .   11 ˜ 1 L† L† L L C1 = U1.8) (C.5) (C. which is normalized with the averaged mass as δLL ≡ (M2LL)1L .2 IXL . 9 36   1˜ 7 L† R† L R C4 = U1.11) m ¯ 2LL ≡ m21L m22L . the coefficients in Eqs.YR .  U R M2RR U R† = diag m2XR . δRR ≡ (M2RR )1R .10) where we have neglected the contributions dependent on the gluino mass mg˜ . We introduce two parameters essential for the mass insertion method. 2L and XR = 1R . One is the ¯ 2RR ) for the left. 3 3   5˜ 1 L† R† L R C5 = U1.XL UXL .Y in the above equations are IX.4.5 are defined as follows:   O1 = d¯α γµ PL sα d¯β γ µ PL sβ .YR − IXL .(right-) handed down-type squarks in the first-two averaged mass m ¯ 2LL (m generations.2 U1.Y = 0.2L m ¯ 2LL . The loop integrals IX.

6–11 are obtained by making use of the exact coefficients given in Eqs.19) where  κ1 = α3 (mb ) α3 (µhad ) 6/25 α3 (mt ) α3 (mb ) 6/23  α3 (mg˜ ) α3 (mt ) 6/21  α3 (mD ) α3 (mg˜ ) κ4 = κ1−4 . The coefficients at µhad are calculated at the one-loop level as follows [51]: C1 (µhad ) = κ1 C1 (mD ). However. C5 = − 27 m ¯ LL m ¯ RR C1 = − (C.17) 1 C4 (µhad ) = κ4 C4 (mD ) + (κ4 − κ5 )C5 (mD ). Hisano et al. −2/b3 .14) (C. C˜ 1 (µhad ) = κ1 C˜ 1 (mD ). we often use the expression in the mass insertion method in estimating the bounds on the light sfermion masses.20) (C. 1/2 κ5 = κ1 . we represent the hadronic matrix elements of the renormalized operators O(µ) at the renormalization scale µ in terms of the corresponding B parameters as follows: .18) (C. we must evolve it using RG equations to the hadronic scale µhad .21) (C.13) (C. (C.42 J. Since the above effective Lagrangian is obtained at the heavy-sfermion mass scale mD . (C. In Section 4.16) (C. Instead of using the vacuum insertion approximation. / Nuclear Physics B 584 (2000) 3–45 2 11 δLL . Figs. b3 is the coefficient of the one-loop beta function for the strong coupling between the gluino mass mg˜ and the heavy-sfermion mass scale mD . 108 m ¯ 2LL 1 δLL δRR .7)–(C. 3 C5 (µhad ) = κ5 C5 (mD ).15) C˜ 1 is obtained with the replacement L → R in C1 . (C. where hadronic matrix elements are evaluated. (C.9). C4 = 9m ¯ LL m ¯ RR 5 δLL δRR . [52] and assumed that all the light sfermions and the gauginos decouple at the gluino mass scale for simplicity. we have taken the hadronic scale µhad = 2 GeV according to Ref.22) Here.

.

.

1 .

K 0 .

O1 (µ).

K¯ 0 = K 0 .

O˜ 1 (µ).

K¯ 0 = mK fK2 B1 (µ). 3  2 .

0 1 0.

mK mK fK2 B4 (µ). K .

O4 (µ).

K¯ = 4 ms (µ) + md (µ)  2 .

0 0.

1 mK .

.

25) .24) (C. ¯ mK fK2 B5 (µ). K O5 (µ) K = 12 ms (µ) + md (µ) (C.23) (C.

(C.26) B4 (µ = 2 GeV) = 1. / Nuclear Physics B 584 (2000) 3–45 43 where mK = 497.73(10).27) B5 (µ = 2 GeV) = 0.28) We have now explained all the elements necessary for calculating the SUSY contribution to the KL –KS mass difference. . fK = 160 MeV. Hisano et al. In our calculation.03(6). we use the following values for the B parameters obtained by lattice calculations at µ = 2 GeV [62. (C. (C.J.7 MeV.60(6).63]: B1 (µ = 2 GeV) = 0. ms (2 GeV) = 125 MeV and md (2 GeV) = 7 MeV.

.

(C.SUSY = −2 Re K 0 .29) 1mK.

Leff .

.K¯ 0 . we obtain constraints from 1mK by imposing a condition that the SUSY contribution does not saturate the experimental value. Using them.

.

.

.

30) 2. (C.

K 0 .

Leff .

K¯ 0 .

31) (C. (δLL δRR )1/2 < (150–250) TeV δLL. (25–35) TeV (m ¯ LL m ¯ RR )1/2 . in the case where the δLL and δRR have no CP-violating phases. there is another constraint from K . .49 × 10−15 GeV.32) If δLL and/or δRR have CP-violating phases. The constraints can be expressed in a simple form using the mass insertion as follows: m ¯ LL. < 1mK = 3. RR . RR < (C.

0.

.

.

1 .

Im K .

Leff .

K¯ 0 .

√ 2 1mK (C. and then the above constraint Eq.3 × 10−3 . (C. < K = 2.33) The constraint is the severest when hK 0 |Leff |K¯ 0 i is pure imaginary.33) is rewritten as √ .

.

.

.

(C.34) 2.

K 0 .

Leff .

K¯ 0 .

Phys. Haber. Dine. Silvestrini. J. Nanopoulos. B 415 (1994) 293. Lett. Phys. H. M. R. E. Masiero. Gabbiani. References [1] [2] [3] [4] [5] [6] [7] [8] H. 65 (1990) 2939. L.4. R. Hagelin. J. Phys. F. S. Hall.5 × 10−3 1mK . B 477 (1996) 321. Rev.S. Lett. D. Rev. Ellis. Leigh. . Gatto. Nilles. R. Kane. 117 (1985) 75. Barbieri.E. Nucl. Randall. Phys. Nucl. B 110 (1982) 211. 110 (1984) 1. Lett. B 110 (1982) 44. G. T.L. Phys. Phys. Phys. A. L.V. Gabrielli.P. Phys. A. Kagan. Rep.J. Thus. < 2 2 K 1mK = 6. Tanaka. L. Rep. the constraints from K can be severer than those from 1mK by a factor of √ (2 2 K )−1/2 ∼ 12. Kelley. D 48 (1993) 4269.

Cohen. Pomarol.. Nakano. Zhang. Phys. Lett. T. Nomura. Lett. R. Komine. Martin. Phys. Schwarz.T.P.E. J. R. Arkani-Hamed. Lett. Pomarol. D. Sato. Phys.-J. Rattazzi. Rev. Phys.B. S.B.L. Phys. Wright. Phys. Graesser. J. Matchev. Nielsen. D 50 (1994) 2282. P. A. N.E. H.G. Phys. Nucl. Lett. M.E. Murayama. A. Tommasini. Phys. H. Rev. B 356 (1995) 45. Lett. Feng.44 [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] [39] [40] [41] [42] [43] [44] [45] [46] [47] [48] [49] [50] [51] [52] [53] [54] [55] [56] [57] [58] [59] J. Y. B 388 (1996) 588. B 473 (2000) 272. Nelson. Lett. Shafi. B 401 (1997) 38. Grojean. Proc. M. 77 (1996) 3728. J. Zhang. Z. Vaughn. Nucl.P. Nucl.B. Phys. A. Binetruy. Rattazzi. A. Eur.F. S. Dine. Q. Super-Kamiokande Collaboration. Phys. J. Dudas. Nucl. Lett. E. Rev. JHEP 9810 (1998) 008. Rev. Lett.D. Terning. Phys. Nucl. N. Y. Dimopoulos. Plümacher. L. 77 (1999) 293. D 56 (1997) 1598. M. D.L. Tavartkiladze. B 147 (1979) 277. B 445 (1999) 316. R.-J. Phys. M.A. Phys. H. N. Lett. Phys. D 55 (1997) 3177. Scimemi. D 56 (1997) 6733. M. Rev. Lett. P. D 58 (1998) 115005. Rev. Y. J. Lett. D. B 256 (1985) 557. Pokorski. J. Y. Zhang. S. Arkani-Hamed. Hisano. N. N. Y. . 84 (2000) 2572. K. Feng. E. Polonsky. Feng. Lett. R. L. Phys.E.P. Phys. J. Phys. Phys. C. Kaplan. Phys. C. Lett. Phys. Nucl. J. B 350 (1995) 49. de Carlos. P.-J.A. Riotto. S. Langacker. Rev. Phys.A. G. B 481 (1996) 85. M. Kobayashi. C. Phys. Dudas. F. Kolda. R. Mohapatra. B 546 (1999) 3. Phys. Trivedi. Savoy. Phys. Jack. Nelson. S. Vaughn. Binetruy. Giudice.G. Phys. S. A. E. P. Affleck. J. L. B 357 (1995) 573. Phys. B 454 (1999) 290. Green. Rev. J. B 369 (1996) 255. M. Giudice. Ramond. B 412 (1997) 77.A. Polonsky. Nucl. Dine. S. Phys. I. Dudas. N. G.D. D 60 (1999) 055003. Phys. Weiner. Phys. Luty.F. Phys. Yanagida. N. hep-ph/9812290. Masiero. Phys. / Nuclear Physics B 584 (2000) 3–45 Y. Arkani-Hamed. D.H. C 10 (1999) 347. B 563 (1999 ) 3. T. N. D 61 (2000) 075011. Dudas. Riotto. Giudice. Barreiro. Phys. Pokorski.R. E. Kawamura. Rev. Lett.L. Casas. Yamada. Moreno. B 431 (1998) 329. B 375 (1996) 141. hep-ph/9909281. Nomura. N. N. Dine. Y. Bagger. G. Murayama. Phys. J. M.T. Luty. Poppitz. Lett. Lepeintre. Rev. Luty. Lett. Lett. Rev. Lett. Bagger. Nir. J. Phys. Savoy. Lett. G.P. Polonsky. B 149 (1984) 117. Rev. S. Nucl. Savoy. Seiberg. Jones.T. Contino. K. Rev. G. Phys. B 402 (1997) 101.N. Hisano et al. Suppl. N.A. B 243 (1990) 250. M. Phys. Phys. Pokorski. B 477 (2000) 233. Yamaguchi. Everett. Phys. Hall. Seiberg. B 332 (1994) 100. Martin. Froggatt. M. G. B 466 (1996) 3. Kribs. Lett. E. Lett. hep-th/9404033. Martin. Wang.E. Rev. J. A. D 58 (1998) 015004. Fukuda et al. R. M. Proc.T. Kagan. Ramond. D. Phys. H. Kurosawa. Phys. B. D 50 (1994) 5481. Lett. Phys. S. B 206 (1988) 480. B 481 (2000) 67.F.M. Phys. Bagger. M. Phys. Suppl. B 445 (1998) 82. K. Arkani-Hamed. T. Binetruy.P. Phys. Phys. J. A. I. C. Kaplan. Ramond. Lett. P. Kaplan.A. Lavignac. Phys. S. B 389 (1996) 503. Lett. B 309 (1993) 337. D 55 (1997) 1138. A. A. Ciuchini et al. P. Dvali.A. Yamada. D. Phys. Agashe.A. A.. Ibanez. D 59 (1999) 015007. S. 81 (1998) 1562. M. J. I. B 473 (2000) 264. P. Phys. B 477 (1996) 353. Martin. A. 77 (1999) 3. Lett. Nelson. Terning. S.A. Arkani-Hamed. Lett. Masiero. Rev. Nucl. C. Samuel. D 55 (1997) 4262. C. R. Ross.

Lett. Phys. C. Kostelecky. Nucl.R. 45 . B. L. Nucl. 53 (1997) 181. Phys. Hall. B 267 (1986) 415. V. Hisano et al. B 453 (1999) 30. Hall. Allton et al. Phys. Proc.. S. Dugan. Sharpe. B 255 (1985) 413. Grinstein. Suppl. S.J. Phys.J. Raby. L. Nucl. / Nuclear Physics B 584 (2000) 3–45 [60] [61] [62] [63] M.A.R.

accepted 5 June 2000 Abstract Right-handed (RH) rotations do not play a role in the Standard Model. For example in GUTs large RH rotations of the quarks can be related to the observed large neutrino mixing or in particular. Germany b School of Physics and Astronomy.b.  2000 Elsevier Science B. We present a simple realistic nonSUSY SO(10) GUT with large RH and LH mixing and study the corresponding nucleon decay rates.1 a Department of Physics. Right-handed rotations 1.physik. all mixing angles are relevant for the proton decay. 0550-3213/00/$ – see front matter  2000 Elsevier Science B. and only the differences of the LH mixing angles are involved in VCKM . Carsten Merten a.physik. One can add conjectures for the structure of the fermionic mass matrices to the SM. The mass matrix entries are arbitrary parameters of the model and only the neutrinos are restricted to be massless. Proton decay. Keywords: Neutrinos. are needed. Many different conjectures are known to give the right masses of the charged fermions and VCKM (within the experimental errors) and this is clearly an indication that the mass problem is far from being solved. PII: S 0 5 5 0 .9 . The main reason for this large freedom is that right-handed (RH) rotations 2 are nonobservable in the SM.uni-wuppertal. This means that two unitary matrices.elsevier.de 1 merten@theorie. neutrino mixing and proton decay Yoav Achiman a. All rights reserved.V. that is no more true in extensions of the Standard Model.∗ . Gaußstr.uni-wuppertal. D-42097 Wuppertal. This leads to the huge freedom in the fermionic mass matrices. E-mail: achiman@theorie. Introduction What is the origin of the fermionic masses? This is one of the open questions in the Standard Model (SM). Mass matrices.nl/locate/npe Large quark rotations.Nuclear Physics B 584 (2000) 46–68 www. University of Wuppertal. Also the observed left-handed (LH) mixing matrix VCKM involves ∗ Corresponding author.de 2 A general non-hermitian matrix is diagonalized by a bi-unitary transformation. However.3 2 1 3 ( 0 0 ) 0 0 3 5 3 . 69978 Tel Aviv. one from the left and one from the right.V. GUTs. Israel Received 7 April 2000. All rights reserved. Those matrices are equal only for hermitian (or symmetric) matrices [1]. revised 31 May 2000. Tel Aviv University. 20.

Most recent models use SUSY GUTs [38]. In addition.36] because this method assumes ad hoc physics beyond the GUT and many new particles. [34] and generalized it to asymmetric matrices. Our aim was to look for the simplest possible realization of a realistic model with large mixing angles. as will be explained later. the available parameter space of low-energy SUSY shrinked recently so much that MSSM is on the verge of loosing its “naturality” [40–42]. We did not use non-renormalizable contributions to the mass matrices à la Froggatt and Nielsen [35. there is already a strong indication from the neutrino sector that large rotations are required (for the atmospheric neutrino anomaly [2– 5]. This can explain the hierarchy of the masses but not the light seesaw neutrino properties. The aim of this paper is to present a simple realistic SO(10) GUT [16–18] model with large LH as well as RH mixing angles and to use them to study the consequences for nucleon decay. C. Large rotations in the quark sector could be a natural reason for the large mixing observed in the leptonic sector in terms of neutrino oscillations. the resulting matrix elements are given there in orders of magnitude only. A predictive model for the fermionic masses must involve a family symmetry which dictates the texture of the mass matrices and protects it from getting large radiative corrections. In particular. At the same time solutions without low energy SUSY have emerged quite naturally [43] in superstring and M theory. Mixing effects were generally neglected in the conventional proton decay models [19–23] by assuming that the mixing angles are small. for the solar neutrino puzzle [6–11]). . the Grand Unified Theories (GUTs) [12–15]. Those are obtained from a product of three matrices and hence predicted up to a factor of [O(1)]3 which may be quite large [37].Y. We therefore used the famous SO(10) paper of Harvey et al. Unification of the gauge coupling constants is obtained using an intermediate breaking scale [47–55] MI ' 1011 GeV. Actually. all mixing angles are relevant. 3 However.33] in the framework of an SO(10) GUT that will add relations between the matrix elements. and also the hierarchy problem can be solved adding extra dimensions [44–46]. We know that the SM must be extended for many reasons if not alone to explain the fermionic masses and mixing angles. The considerable freedom in the mass matrices of the SM does not exist in its extensions. there is a kind of duality between the RH mixing of the quarks and the leptonic LH rotations [24– 31]. Note also that conventional GUTs are relatively simple and give much more reliable predictions for nucleon decay than SUSY theories. Achiman. In particular. We think therefore that it is worthwhile to look for a non-SUSY GUT which is consistent with all observed experimental facts. This is very useful because MI is also the right mass scale 3 We shall use it also in a forthcoming paper [39]. Merten / Nuclear Physics B 584 (2000) 46–68 47 only the differences between the mixing angles of the u-like and d-like quarks and the individual mixing can be large. We used a global U (1)F [32. In such a framework one cannot change the “weak basis” without changing the physics. Such matrices give usually large LH and/or RH mixing angles. in its most popular extension. The hope is that the fine tuning (hierarchy) problem will be solved in the more fundamental theory.

Merten / Nuclear Physics B 584 (2000) 46–68 for the RH neutrinos needed for the see-saw mechanism as well as for leptogenesis as the origin of the baryon asymmetry [56–59] and the (invisible) axion window [60–63]. 4 As will be explained in the next chapter.64. 3)126 . the GPS representation (10.e.. baryon asymmetry induced by leptogenesis and especially nucleon decay can reveal the unknown mixing angles and reduce considerably the freedom in the fermionic mass matrices. I PS . 2). GI = GPS ≡ SU(4)C ⊗ SU(2)L ⊗ SU(2)R . 2) and (15. The mass matrices and our global U (1)F details are given in Section 3. For the masses of the Higgs scalars the “extended survival hypothesis” [69] is assumed. we 4 Note that due to the quark-lepton symmetry of G the SO(10) mass relations are valid also at the scale M . Achiman. The RH neutrino masses of order MR ∼ MI will be induced via the VEV of the above mentioned (10.48 Y. i. Three solutions are found which are consistent with the observed neutrino anomalies [2–11. 2. Symmetry breaking of the SO(10) GUT We use an SO(10) GUT [16–18] which is broken down to the SM via an intermediate scale MI as follows MU MI MZ SO(10) −→ GI −→ GSM −→ SU(3)C ⊗ U (1)em . (1) where the intermediate symmetry group is the Pati–Salam one [68]. In Section 4 the numerical solutions for the mass matrices are obtained by the use of the renormalization group equations (RGEs). leptoquarks. This number is needed however to fix the RGEs used to calculate MI . 2. Our claim is that observation of other phenomena like RH currents. The observed neutrino oscillations teach us about the neutrino masses and mixing angles. C. 2.67]). The breaking at MU is done using the Higgs representations Φ210 (or Φ54 in models with D parity) while for the breaking at MI we use the SM singlet of a Φ126 . and a fit to the observed properties of the charged fermions is elaborated. 3)126 . MU and αU (MU ) as well as the values of the mass matrices at MI . This is the first evidence for physics beyond the SM. 2. The exact number of representations needed for the fermion mass matrices will be given later when we will discuss those matrices. The Higgs doublet of the SM is therefore a linear combination of the SU(2)L doublets in the representations (1. Section 6 is devoted to the conclusions.65] (except for LSND [66. 2. 1. All mixing angles for those solutions are obtained explicitly and this allows for the calculation of the nucleon decay rates in Section 5. In view of the representation content of the mass terms 16 ⊗ 16 = (10 ⊕ 126)symm ⊕ 120antisymm (2) we give the light fermions masses via the VEVs of (1. 2)10/120 and (15. The plan of the paper is as follows. In Section 2 we discuss the symmetry breaking that is dictated by the requirement of gauge unification and the Higgs representations needed to give the correct fermion masses. 2)120/126. 1.

3. (9) 12π 12π where S2 (Gj ) is the Dynkin index of the adjoint representation of the group Gj . C. The mass matrices The aim of this paper is to present models with large RH and LH mixing angles of the quarks that lead to large mixing of the leptons and to study the predictions of those .14 × 1010 GeV α1 (MI ) (46. (MU ) + = α2L 2π 1 −1 . (MU ) + = α2R 2π In general the matching conditions between gauge couplings belonging to theories symmetry groups Gj and Gk can be written as −1 −1 (MU ) = α4C (MU ) + αU (6) (7) (8) with 1 1 S2 (Gj ) = αk−1 (MI /U ) − S2 (Gk ). Merten / Nuclear Physics B 584 (2000) 46–68 49 used in the RGEs N1 = number of (1. In this case however MU (GPS ⊗ D) = 1.39)−1 α2R (MI ) (55. 1)126 = 0 and ∆R = number of (10. 3)126 = 1 for the GPS symmetry breaking. (MI ) = α1−1 (MI ) − α3−1 (MI ) + α2R 3 3 3π while at MU the conditions are (4) (5) 1 .86)−1 Quantity Value α3 (MI ) (31. αj−1 (MI /U ) − Table 1 Symmetry breaking scales and gauge couplings for N1 = 4 and N15 = 2 Quantity Value MI 6.42)−1 MU 1.31 × 1016 GeV αU (MI ) (20.08)−1 3.85)−1 α2L (MI ) (39.Y. 12π (3) −1 α2L (MI ) = α2−1 (MI ). 2. where D is the discrete D-parity which requires α2L = α2R between MU and MI . 5 2 1 −1 . a value which would lead to a too fast proton decay.16 × 1015 GeV. We checked also the RGEs for the intermediate symmetry GI = GPS ⊗ D. 2) = 4 and N15 = number of (15. Achiman.00)−1 α2 (MI ) (39. 2) = 2 as well as ∆L = number of (10. the results are given in Table 1. 3π 1 −1 .86)−1 Quantity Value α4C (MI ) (31. For the matching conditions at MI we took −1 (MI ) = α3−1 (MI ) + α4C 1 . 2. 1. Details can be found in [70].

in order to have large mixing angles one should rather use asymmetric mass matrices. To generate the mass matrices we use the method of Harvey et al. [34] as it was realized in [71] for symmetric mass matrices. Y120 =  0 0 x2 0 0 y2 0 0 −z2 0 5 This can help to solve the strong CP problem while the observed CP violation in the K-decay can come from a different origin than the CKM matrix. Practically speaking. i. Y126 =  0 0 y2  . C.74]. As we need anyhow asymmetric mass matrices we shall use the asymmetric Fritzsch texture which is also known under the name “nearest neighbour interaction” model (NNI) [75]. the fermionic part of the mass matrices has the following quantum numbers:   α1 + α1 α1 + α2 α1 + α3 (13) Mf ∼  α1 + α2 α2 + α2 α2 + α3  .. Y10 =  0 0 x2  .e. Taking all this into account the Yukawa coupling matrices (at energies µ & MI ) can have the structure       0 x1 0 0 y1 0 0 z1 0 (1) (1) (1) Y10 =  x1 0 0  . α1 + α3 α2 + α3 α3 + α3 To realize the NNI texture (10) one sees that only Higgs representations with the charges β = α1 + α2 . Hence. Y120 =  −z1 0 0  . α1 +α2 = 2α3 . Merten / Nuclear Physics B 584 (2000) 46–68 models for the nucleons decay ratios.   0 A 0 M = B 0 C . However. . Also. we still have the possibility to couple one Higgs representation to two different combinations. 0 0 x˜1 0 0 y˜1 0 0 0       0 0 0 0 0 0 0 0 0 (2) (2) (2) (14) 0 z2  . (11) Φk → exp(iβk θ )Φk . Y126 =  y1 0 0  . as it is well known the symmetric version of this texture cannot account for the large top quark mass [73. This symmetry will be chosen in such a way that the predictive Fritzsch texture [72] will be realized.50 Y. α2 + α3 and α3 + α3 can couple to the fermions. we shall use a global family symmetry U (1)F or Zn that will dictate the neutrino properties in terms of the observed masses and mixings of the charged fermions. However. (12) The invariance under U (1)F requires therefore that the βk must obey αi + αj = βk . (10) 0 D E In view of the fact that we are actually mainly interested in the predictions for the nucleon decay rates which are not sensitive to the details of CP violation we will use for simplicity real mass matrices. Achiman. For asymmetric matrices we need to make certain changes in the above scenario and add also the antisymmetric Higgs representation Φ120 . Namely. 5 The three fermion families and the different Higgs representations in 16i Φ¯ k 16j transform under the global U (1)F as follows 16j → exp(iαj θ )16j .

They also fix the matrix elements of the RH neutrino Majorana mass matrix (in terms of the VEV of (10. x1 y1 (Mu )33 = x˜1 υu(1) + y˜1 ωu(1) =  = x1 υu(1) − 3 y1ωu(1) + z1 υ˜ u(1) = (Mu )12 − 4 y1ωu(1) . Actually.  − z2 υ˜ u(2) + ω˜ u(2) . 33 y1 (27) (28) (29) (30) (31) (32) (33) (34) All those elements are obtained from 14 independent Higgs parameters which are products of VEVs and Yukawa couplings. M(Dir) ν 32    y˜1 (Dir) (1) (1) = x˜1 υu − 3y˜1ωu = (Mu )33 − 4 Mν y1 ωu(1) . 1.  (17) (18)    x˜1 y˜1 x1 υd(1) + y1 ωd(1) . Merten / Nuclear Physics B 584 (2000) 46–68 51 On top of that.  − z2 υ˜ d(2) − 3 ω˜ d(2) = (Md )32 + 4 z2 ω˜ d(2) . M(Dir) ν 23   = x2 υu(2) − z2 υ˜ u(2) − 3 ω˜ u(2) = (Mu )32 + 4z2 ω˜ u(2) .  + z2 υ˜ d(2) − 3 ω˜ d(2) = (Md )23 − 4z2 ω˜ d(2) . (25) (Mu )21 = x1 υu(1) (Mu )23 = x2 υu(2) (Mu )32 = x2 υu(2) (26) + y1 ωu(1) − z1 υ˜ u(1).  + z2 υ˜ d(2) + ω˜ d(2) .   y˜1 (1) (1) y1 ωd(1) . Those give also the right unification as discussed before. Achiman. M(Dir) ν 12  = x1 υu(1) − 3y1ωu(1) − z1 υ˜ u(1) = (Mu )21 − 4y1 ωu(1) . (Me )33 = x˜1 υd − 3 y˜ 1ωd = (Md )33 − 4 y1 (22) (23) (24) (Mu )12 = x1 υu(1) + y1 ωu(1) + z1 υ˜ u(1). C. This VEV will generate also the RH neutrinos masses MR ∼ MI . as is discussed in Appendix A we will use only one Φ126 in addition to two Φ10 and two Φ120 to generate the asymmetrical mass matrices. 2) and two in that of (15. with MR ∼ MI being a quasi-free parameter: .     x˜1 y˜1 x1 υu(1) + y1 ωu(1) . M(Dir) ν 21   = x2 υu(2) + z2 υ˜ u(2) − 3 ω˜ u(2) = (Mu )23 − 4z2 ω˜ u(2) . 2. 1.Y. In terms of the notation and discussion of Appendix A we obtained the following expressions for the mass matrix elements in terms of the VEVs and Yukawa couplings: (Md )12 = x1 υd(1) + y1 ωd(1) + z1 υ˜ d(1). 3) that breaks GPS → GSM ). (2) (2)  − z2 υ˜ d + ω˜ d . The detailed fits required four VEVs in the direction (1. (15) (Md )21 = x1 υd(1) (Md )23 = x2 υd(2) (2) (Md )32 = x2 υd (16) + y1 ωd(1) − z1 υ˜ d(1). (20) (Me )21 = x1 υd(1) (Me )23 = x2 υd(2) (Me )32 = x2 υd(2) (21) (Md )33 = x˜1 υd(1) + y˜1 ωd(1) = − 3y1ωd(1) − z1 υ˜ d(1) = (Md )21 − 4 y1ωd(1) . 3) to break the SO(10) gauge symmetry at MI .  + z2 υ˜ u(2) + ω˜ u(2) . 2). x1 y1 (19) (Me )12 = x1 υd(1) − 3y1ωd(1) + z1 υ˜ d(1) = (Md )12 − 4y1 ωd(1). 2. we need at least one Φ126 with a large VEV in (10.

given the mass matrices we have definite predictions for the proton decay. D†L Md DR = M(D) d . On the RHS of (36) we use the calculated masses of the charged fermions (see Table 2) and the (real) CKM matrix at µ = MI . Now. in the framework of the SM (with massive neutrinos) the only observable mixing matrices would be VCKM = U†L DL and U = E†L Nν . the baryon asymmetry and RH currents which can be measured in principle by future experiments. y˜1 /y1 (35) By diagonalizing the charged fermion mass matrices U†L Mu UR = M(D) u .R . C.14 MeV mτ (MI ) 1838.6 MeV mb (MI ) 1360 MeV mt (MI ) 101.2 GeV Mass Value me (MI ) 513 keV mµ (MI ) 108.3 MeV 6 There is only one neutrino mixing matrix as Mlight is always symmetric. This is done using the SO(10) relations (15–34) for the mass matrices at µ = MI = 6. Achiman. without using the neutrino sector. 7 4.R and Nν .14 × 1010 GeV. We then run the mass matrices from MI to MZ . U†L DL = VCKM . ν 7 Note however. However.38 MeV ms (MI ) 47.R . 6 Neglecting phases (as explained before) we have 12 masses and 21 mixing parameters. Table 2 Fermion masses at MI Mass Value mu (MI ) 1. diagonalize them there and compare the obtained masses and mixings with their experimentally observed values. . All those fix the proton decay rates and other GUT scale effects (like baryon asymmetry induced by leptogenesis [56–59]). Those values are obtained using the RGEs as is described in detail in [70]. only 13 of our 14 parameters are independent (2) (2) because only the combination z2 (υ˜ u + ω˜ u ) appears in the charged fermion equations. (36) and the light see-saw neutrino mass matrix T light (Maj) −1 MνR M(Dir) Mν ' −M(Dir) ν ν (37) we obtained all masses of the charged fermions and those of the three light neutrinos as well as the mixing matrices UL.4 MeV Mass Value mc (MI ) 337. Merten / Nuclear Physics B 584 (2000) 46–68  (Maj) MνR 0  = MR y1 0 y1 0 0   0 0 1   0 = y1 MR 1 0 0 0 y˜1  0 0 . This gives 18 mixing observables. Numerical solutions for the mass matrices The first step will be to calculate numerically the mass and mixing matrices in (36).52 Y. E†L Me ER = M(D) e . EL.16 MeV md (MI ) 2. DL.

For the given quantity we use the ratio y˜1 /y1 which determines the RH neutrino mass matrix (35) except for a global factor and we vary its value between 1 6 |y˜1 /y1 | 6 1000. (38) 0.05.49 6 |(U)23 | 6 0.71. or 0.Y. the neutrino sector of our model is uniquely fixed up to (2) the two parameters MR ∼ MI and z2 ω˜ u . We found two regions in the y˜1 /y1 –z2 ω˜ u parameter space which obey the atmospheric neutrino requirements together with small angle MSW [76. 2 and 3.03 6 |(U)12 | 6 0. We have in (36) 30 nonlinear equations.05 (small angle MSW). For each region we fixed a representative solution (see Table 3) and used it to calculate the corresponding neutrino properties.  2  mν2 − m2ν1 6 1000. so altogether there are 18 mixing angles. MR and z2 ω˜ u(2) which predict neutrino properties lying in the range allowed by oscillation experiments [2–11. Solution 1 (large mixing MSW) in the y˜1 /y1 –z2 ω˜ u parameter space. On the other hand each of the (real) 6 mixing matrices is parametrized by 3 angles.65]: |(U)13 | 6 0. .64.49 (large angle MSW ). 50 6 1m2atm /1m2sun ≡ m2ν3 − m2ν2 (40) (41) The value of y1 MR will be fixed at the end to give the exact absolute scale of the neutrino masses.35 6 |(U)12 | 6 0. The explicit results for the neutrinos are given in Table 4.77] explanation for the solar neutrino puzzle (models 2a. Given these three parameters the model predicts the three neutrino masses and three (2) lepton mixing angles in U. With the 13 Higgs parameters we have 31 “unknowns”. once the value of y˜1 /y1 is given. We therefore look if there are solutions of the model for reasonable values of y˜1 /y1 . (39) 0. Merten / Nuclear Physics B 584 (2000) 46–68 53 (2) Fig. C. By studying the equations in detail one can see that.b) and one with the large angle MSW (model 1) as can be seen in the Figs. 1. 1. Achiman. One of them must therefore be “given” to be able to search for solutions numerically.

Solution 2a (small mixing MSW) in the y˜1 /y1 –z2 ω˜ u parameter space. (2) Fig. . Merten / Nuclear Physics B 584 (2000) 46–68 (2) Fig. They are used to calculate the branching ratios of the nucleon decays. The corresponding mixing angles are given in Table 5. Table 3 (2) Values of y˜1 /y1 and z2 ω˜ u for the three representative solutions Solution MSW effect Model 1 Model 2a Model 2b Large mixing Small mixing Small mixing y˜1 /y1 19 24 −18 (2) z2 ω˜ u (MeV) −3450 −9175 −9925 In this way the explicit LH and RH mixing matrices are also fixed. C. 2. Achiman.54 Y. Solution 2b (small mixing MSW) in the y˜1 /y1 –z2 ω˜ u parameter space. 3.

Y. Achiman, C. Merten / Nuclear Physics B 584 (2000) 46–68

55

Table 4
Masses and mixing angles of the light neutrinos at MZ
Parameter

Value in model 1

Value in model 2a

Value in model 2b

(y1 MR /MI ) · mν1
(y1 MR /MI ) · mν2
(y1 MR /MI ) · mν3

−0.0245 eV
0.0876 eV
−2.402 eV

−8.73×10−4 eV
0.355 eV
−3.031 eV

1.16×10−3 eV
−0.467 eV
4.365 eV

(ν)

−0.487

−0.050

0.051

(ν)
θ23
(ν)
θ31

0.205

0.506

0.496

0.004

0.003

−0.003

θ12

mν2 /mν1
mν3 /mν2 

2 

mν3 − m2ν2
m2ν2 − m2ν1

−3.57
−27.43

−406.3
−8.54

815.4

71.9

−401.2
−9.35
86.4

5. Calculation of the nucleon decay rates
The partial decay rate for a given process nucleon → meson + antilepton is expressed
as follows:
!
X
X
1 2
2
2
2
2
2
2
m ρj |S| |A| |AL |
|Al Ml | + |AR |
|Ar Mr | ,
(42)
Γj =
16π nucl
r
l

where Ml and Mr are the hadronic transition matrix elements for the relevant decay
process. l and r denote the chirality of the corresponding antilepton. Al and Ar are the
relevant coefficients of the effective Lagrangian (67) given in Appendix B. A, AL and AR
are factors which result from the renormalization of the four fermion operators as follows:  

α1 (MZ ) −23/82
,
(43)
AL =
α1 (MI )  

α1 (MZ ) −11/82
,
(44)
AR =
α1 (MI )        

α4C (MI ) −5/8 α2L (MI ) −27/100 α2R (MI ) −3/20 α2 (MZ ) 27/38
A=
α4C (MU )
α2L (MU )
α2R (MU )
α2 (MI ) 
2/7  
6/23 
6/25   

α3 (MZ )
α3 (mb )
α3 (mc )
α3 (1 GeV) 2/9
×
.
(45)
α3 (MI )
α3 (MZ )
α3 (mb )
α3 (mc )
Using then [78]
α3 (1 GeV) = 0.544,

α3 (mc ) = 0.412,

α3 (mb ) = 0.226

(46)

one obtains
|AL |2 = 1.155,

|AR |2 = 1.071,

|A|2 = 23.59.

(47)

56

Y. Achiman, C. Merten / Nuclear Physics B 584 (2000) 46–68

Table 5
Mixing angles in the three different models
Parameter

Value in model 1

Value in model 2a

Value in model 2b

θL12

(u)

−0.627

−0.540

0.275

θL23

(u)

−0.055

−0.056

−0.075

θL31

(u)

0.000

0.000

0.000

(u)
θR12
(u)
θR23
(u)
θR31
(d)
θL12
(d)
θL23
(d)
θL31
(d)
θR12
(d)
θR23
(d)
θR31
(e)
θL12
(e)
θL23
(e)
θL31
(e)
θR12
(e)
θR23
(e)
θR31
(ν)
θ12
(ν)
θ23
(ν)
θ31

0.005

0.006

−0.012

0.049

0.051

0.043

0.000

0.000

0.000

−0.404

−0.317

0.498

−0.025

−0.024

−0.041

−0.018

−0.016

0.012

0.117

0.152

−0.092

0.979

1.011

0.679

0.000

0.000

0.000

−0.018

−0.015

0.015

0.733

−0.060

−0.089

0.000

−0.001

0.001

0.262

0.314

−0.301

−0.063

0.747

0.561

0.014

−0.001

0.002

−0.487

−0.050

0.051

0.205

0.506

0.496

0.004

0.003

−0.003

s (E
s (E
|S|2 = hΨNucl
r1 , rE2 , rE3 )|δ(Er1 − rE2 )|ΨNucl
r1 , rE2 , rE3 )i is the probability to find two valence
quarks of the nucleon at one point in space. We used here the value |S|2 = 0.012 GeV3
given in [79]. ρj ≡ (1 − χj2 )(1 − χj4 ) with χj = mMeson /mNucl is an SU(6) spin-flavour
symmetry breaking phase space factor.
The resulting branching ratios are given in the Tables 6 and 8. Our model also gives the
total decay rates. The predicted rates in the different models are given in the Tables 7 and 9.
Note however that these predictions have a relatively large uncertainty. This was estimated
by Langacker [80] to be
+0.5

∆τp→e+ π ◦ = 10±0.7±1.0−3.0 yrs.

(48)

Y. Achiman, C. Merten / Nuclear Physics B 584 (2000) 46–68

57

Table 6
Partial decay rates Γi /Γ of the proton for the case of vanishing fermionic mixing in comparison with
the rates obtained in our solutions
Decay channel
of the proton

Rates in %
(no mixing)

Rates in %
in model 1

Rates in %
in model 2a

Rates in %
in model 2b

p → e+ π 0
p → e+ K 0
p → e+ η
p → µ+ π 0
p → µ+ K 0
p → µ+ η
p → e+ ρ 0
p → e+ ω
p → e+ K ∗0
p → µ+ ρ 0
p → µ+ ω
p → νeC π +
p → νeC K +
Cπ+
p → νµ
C K+
p → νµ
p → νeC ρ +
p → νeC K ∗+
C ρ+
p → νµ
C K ∗+
p → νµ
p → ντC π +
p → ντC K +
p → ντC ρ +
p → ντC K ∗+

33.6

1.2

5.8

5.1
16.9



32.3


0.1
4.9


0.1



21.4
3.1
0.8
8.5
2.6
0.3
3.3
10.8
0.0
1.3
4.3
25.6
2.0
8.9
1.3
3.9
0.4
1.4
0.0
0.1
0.1
0.0
0.0

25.1
2.6
0.9
5.7
0.9
0.2
3.8
12.7
0.0
0.9
2.9
35.6
2.0
0.5
0.2
5.4
0.3
0.1
0.0
0.1
0.1
0.0
0.0

27.8
4.5
1.0
5.6
1.8
0.2
4.2
14.0
0.0
0.8
2.8
27.7
4.1
0.3
0.3
4.2
0.6
0.0
0.0
0.0
0.0
0.0
0.0

p → e+ X 0
p → µ+ X0
p → ν C X+

56.8
5.8
37.4

39.4
17.0
43.7

45.1
10.6
44.3

51.5
11.2
37.2

Table 7
Total decay rate and lifetime of the proton for the representative solutions
Quantity
Γp
τp

Value in model 1

Value in model 2a

Value in model 2b

2.54 × 10−35 yr−1
3.94 × 1034 yr

2.57 × 10−35 yr−1
3.89 × 1034 yr

2.83 × 10−35 yr−1
3.53 × 1034 yr

Taking this into account our predicted representative rates have therefore a chance to be
observed by SuperKamiokande [81,82] and ICARUS [83].

58

Y. Achiman, C. Merten / Nuclear Physics B 584 (2000) 46–68

Table 8
Partial decay rates Γi /Γ of the bound neutron for the case of vanishing fermionic mixing in
comparison with the rates obtained in our solutions
Decay channel
of the neutron

Rates in %
(no mixing)

Rates in %
in model 1

Rates in %
in model 2a

Rates in %
in model 2b

n → e+ π −
n → µ+ π −
n → e+ ρ −
n → µ+ ρ −
n → νeC π 0
n → νeC K 0
n → νeC η
Cπ0
n → νµ
C K0
n → νµ

n → νµ
n → νeC ρ 0
n → νeC ω
n → νeC K ∗0
C ρ0
n → νµ

n → νµ
C K ∗0
n → νµ
n → ντC π 0
n → ντC K 0
n → ντC η
n → ντC ρ 0
n → ντC ω
n → ντC K ∗0

62.9

9.7

15.1

0.6

1.7

2.3
7.7



0.0





40.1
15.8
6.2
2.4
12.0
6.8
0.4
4.2
0.1
0.2
1.8
6.1
0.2
0.6
2.1
0.1
0.0
0.7
0.0
0.0
0.0
0.0

46.2
10.4
7.1
1.6
16.4
4.9
0.6
0.2
1.0
0.0
2.5
8.4
0.2
0.0
0.1
0.0
0.0
0.4
0.0
0.0
0.0
0.0

49.9
10.0
7.7
1.5
12.5
8.2
0.5
0.1
0.9
0.0
1.9
6.4
0.4
0.0
0.1
0.0
0.0
0.0
0.0
0.0
0.0
0.0

n → e+ X −
n → µ+ X−
n → ν C X0

72.6

27.4

46.3
18.2
35.3

53.3
12.0
34.7

57.6
11.5
31.0

Table 9
Total decay rate and lifetime of the bound neutron for the representative solutions
Quantity
Γn
τn

Value in model 1

Value in model 2a

Value in model 2b

2.72 × 10−35 yr−1
3.68 × 1034 yr

2.80 × 10−35 yr−1
3.57 × 1034 yr

3.14 × 10−35 yr−1
3.18 × 1034 yr

Yet the essential predictions of the model are actually the branching ratios. Comparing
our branching ratios with those of the conventional SO(10) (i.e., without large mixings)
one sees clearly the suppression of the p, n → e+ X channels relative to the channels
p, n → µ+ X, ν c X. In particular p, n → µ+ π, ν c K are prominent. Note that in SUSY

Y. Achiman, C. Merten / Nuclear Physics B 584 (2000) 46–68

59

Table 10
Ratios of some partial nucleon decay rates
Ratio
Γ (p → e+ K 0 )
Γ (p → e+ π 0 )
Γ (p → µ+ π 0 )
Γ (p → µ+ K 0 )
Γ (p → ν C K + )
Γ (p → ν C π + )
Γ (p → e+ π 0 )
Γ (p → ν C π + )
Γ (n → µ+ π − )
Γ (n → e+ π − )
Γ (n → µ+ ρ − )
Γ (n → e+ ρ − )
Γ (n → ν C K 0 )
Γ (n → ν C π 0 )
Γ (n → e+ π − )
Γ (n → ν C π 0 )

No mixing

Model 1

Model 2a

Model 2b

0

0.145

0.104

0.162

0

3.27

6.33

3.11

0.003

0.098

0.064

0.157

1.040

0.618

0.693

0.993

0

0.394

0.225

0.200

0

0.387

0.225

0.195

0.113

0.469

0.347

0.722

4.16

2.48

2.78

3.96

GUTs [38] the dominant decays are into final states involving K mesons [84–86]. 8 The
special properties of our model are clearly reflected in the comparison of ratios as is done
in Table 10.

6. Conclusions
We presented an SO(10) GUT with a global U (1)F family symmetry that is consistent
with all experimental observations. The special thing about this model is that it involves
large mixing angles for the quarks, in contrast with the conventional expectations. The
large mixing in the lepton (neutrino) sector results therefore naturally.
Large rotations have considerable consequences for observables outside the Standard
Model. In particular calculations of the nucleon decay rates require the knowledge of all
mixing angles. We found rates which are obviously different from the conventional GUTs.
Our model is however only one simple example for the effects of large mixing angles. We
are studying now a SUSY GUT with all possible phases and its consequences also for other
GUT observables or RH currents. The hope is that by restricting the values of the mixing
angles, we will be able to reduce the large freedom in the present models for fermionic
masses.
8 This is also the case for non-SUSY GUTs with maximal RH mixings [87–89].

60

Y. Achiman, C. Merten / Nuclear Physics B 584 (2000) 46–68

Table 11
Higgs couplings and vacuum expectation values in SO(10) GUTs
(i)

(j )

(k)

(l)

Higgs representation

(1, 2, 2)10

(1, 2, 2)120

(15, 2, 2)120

(15, 2, 2)126

Yukawa coupling matrix

(i)
Y10
(i) (i)
υu , υd

(j )
Y120
(j ) (j )
υ˜ u , υ˜ d

(k)
Y120
(k) (k)
ω˜ u , ω˜ d

(l)
Y126
(l) (l)
ωu , ωd

Vacuum expectation values

Acknowledgements
We would like to thank M.K. Parida for discussions and especially for helping us with
threshold effects. One of us (Y.A.) is also grateful to the School of Physics and Astronomy
of the Tel Aviv University for the kind hospitality during his present visit.

Appendix A. Fermionic masses in SO(10) GUTs
Dirac masses for the charged fermions are of the form m(ΨLC )T CΨL and therefore
transform under SO(10) as follows:
16 ⊗ 16 = (10 ⊕ 126)symm ⊕ 120antisymm.

(49)

The SO(10) Higgs representations which can give masses to the fermions have the
representation content
10 −→ (1, 2, 2) ⊕ (6, 1, 1),

(50)

120 −→ (1, 2, 2) ⊕ (15, 2, 2) ⊕ (6, 3, 1) ⊕ (6, 1, 3) ⊕ (10, 1, 1) ⊕ (10, 1, 1),

(51)

126 −→ (15, 2, 2) ⊕ (10, 3, 1) ⊕ (10, 1, 3) ⊕ (6, 1, 1)

(52)

under GPS = SU(4)C ⊗ SU(2)L ⊗ SU(2)R , while for the fermions in the 16
16 −→ (4, 2, 1) ⊕ (4, 1, 2).

(53)

Hence the Dirac mass terms transform as follows
¯ 1, 2) = (15, 2, 2) ⊕ (1, 2, 2)
(4, 2, 1) ⊗ (4,

(54)

The different Yukawa couplings and the corresponding VEVs of Φ10 , Φ120 and Φ126 are
given in Table 11.
Taking now the Clebsch–Gordan coefficients into account the mass matrices get the
following contributions [90]
Md = υd Y10 + ωd Y126 + (υ˜ d + ω˜ d )Y120 ,

(55)

Me = υd Y10 − 3ωd Y126 + (υ˜ d − 3ω˜ d )Y120 ,

(56)

= υu Y10 + ωu Y126 + (υ˜ u + ω˜ u )Y120,

(57)

= υu Y10 − 3ωu Y126 + (υ˜ u − 3ω˜ u )Y120.

(58)

Mu
(Dir)

The RH neutrinos can acquire also Majorana masses in term of the (10, 1, 3) component
of Φ126 :

Y. Achiman, C. Merten / Nuclear Physics B 584 (2000) 46–68

61

¯ 1, 2) ⊗ (4,
¯ 1, 2) ⊗ (10, 1, 3) = (1, 1, 1) ⊕ · · · ,
(4,

(59)

(Maj)
MνR

(60)

= MR Y126 ∼ MI Y126 .

The corresponding VEV MI is responsible for the symmetry breaking step GPS → GSM .
The neutrinos will have therefore the 6 × 6 mass martix  

0
M(Dir)
ν 

(61)
M=
T
(Maj) .
MνR
M(Dir)
ν
Using the fact that the non-vanishing entries of the Majorana mass matrix are much larger
than those of the Dirac matrix one can approximately block diagonalize the 6 × 6 matrix
and obtain the see-saw matrix [91–93] 
T
light
(Maj) −1
(62)
MνR
M(Dir)
Mν ≈ −M(Dir)
ν
ν
as well as
heavy


light

(Maj)

≈ MνR .

(63)

is symmetric and therefore can be diagonalized using one unitary matrix Nν
light

NTν Mν

light(D)

Nν = Mν

.

(64)

Neutrino oscillations are induced via the leptonic analogue to the CKM matrix
U = E†L Nν .

(65)

Appendix B. Effective SO(10) Lagrangian for nucleon decays
The baryon number violating part of the SO(10) Lagrangian (without fermionic mixing)
is known to be [12–15]
g

β
+
+
+ d¯Rα γ µ eR
L∆B6=0 = √U X¯ µα εαβγ u¯ L γ µ uL + d¯Lα γ µ eL
2
g

β
+
C
− u¯ Lα γ µ eL
+ √U Y¯µα εαβγ u¯ L γ µ dL − d¯Rα γ µ νeR
2 

gU 0 α

β
C
C
− u¯ Rα γ µ νeR
+ √ Xµ −εαβγ d¯L γ µ dL − u¯ Lα γ µ νeL
2
gU 0 α

β
+
C
− u¯ Rα γ µ eR
+ √ Yµ εαβγ d¯L γ µ uL − d¯Lα γ µ νeL
2 

gU α


+ d¯Rα γ µ eR
+ u¯ Lα γ µ νeL + u¯ Rα γ µ νeR
+ √ X3µ d¯Lα γ µ eL
2
+ h.c.
(66)
Taking all possible fermion mixings into account one obtains [89] the effective four
fermion Lagrangian  


β +

β +
γµ dLα + A2 εαβγ u¯ L γ µ uL e¯R
γµ dRα
Leff = A1 εαβγ u¯ L γ µ uL e¯L  


β

β
α
α
¯ L γ µ uL µ¯ +
+ A3 εαβγ u¯ L γ µ uL µ¯ +
L γµ dL + A4 εαβγ u
R γµ d R

62

Y. Achiman, C. Merten / Nuclear Physics B 584 (2000) 46–68  


β +

β +
+ A5 εαβγ u¯ L γ µ uL e¯L
γµ sLα + A6 εαβγ u¯ L γ µ uL e¯R
γµ sRα  


β

β
α
α
¯ L γ µ uL µ¯ +
+ A7 εαβγ u¯ L γ µ uL µ¯ +
L γµ sL + A8 εαβγ u
R γµ sR  


β C

β C
γµ dRα + A10 εαβγ u¯ L γ µ dL ν¯µR
γµ dRα
+ A9 εαβγ u¯ L γ µ dL ν¯ eR  


β C

β C
γµ sRα + A12 εαβγ u¯ L γ µ dL ν¯ µR
γµ sRα
+ A11 εαβγ u¯ L γ µ dL ν¯eR  



β C
β C
γµ dRα + A14 εαβγ u¯ L γ µ sL ν¯ µR
γµ dRα
+ A13 εαβγ u¯ L γ µ sL ν¯ eR  



β
β
+ A15 εαβγ u¯ L γ µ dL ν¯τCR γµ dRα + A16 εαβγ u¯ L γ µ dL ν¯ τCR γµ sRα 


β
+ A17 εαβγ u¯ L γ µ sL ν¯ τCR γµ dRα
+ (terms with two s quarks)
+ (terms with c, b and t quarks)
+
C
and ν¯ e,µ,τ
+ (terms with τ¯L,R
L)

+ h.c.,

(67)

where the coefficients Ai are given as follows [70]:
e (UR )11 (UL )11 + (UR )21 (UL )21 + (UR )31 (UL )31
A1 = G  

× (ER )11 (DL )11 + (ER )21 (DL )21 + (ER )31 (DL )31 

e (UR )11 (DL )11 + (UR )21 (DL )21 + (UR )31 (DL )31
+G 

× (ER )11 (UL )11 + (ER )21 (UL )21 + (ER )31 (UL )31 , 

e (UR )11 (UL )11 + (UR )21 (UL )21 + (UR )31 (UL )31
A2 = G 

× (EL )11 (DR )11 + (EL )21 (DR )21 + (EL )31 (DR )31 

e0 (DR )11 (UL )11 + (DR )21 (UL )21 + (DR )31 (UL )31
+G 

× (EL )11 (UR )11 + (EL )21 (UR )21 + (EL )31 (UR )31 , 

e (UR )11 (UL )11 + (UR )21 (UL )21 + (UR )31 (UL )31
A3 = G 

× (ER )12 (DL )11 + (ER )22 (DL )21 + (ER )32 (DL )31 

e (UR )11 (DL )11 + (UR )21 (DL )21 + (UR )31 (DL )31
+G 

× (ER )12 (UL )11 + (ER )22 (UL )21 + (ER )32 (UL )31 , 

e (UR )11 (UL )11 + (UR )21 (UL )21 + (UR )31 (UL )31
A4 = G 

× (EL )12 (DR )11 + (EL )22 (DR )21 + (EL )32 (DR )31 

e0 (DR )11 (UL )11 + (DR )21 (UL )21 + (DR )31 (UL )31
+G 

× (EL )12 (UR )11 + (EL )22 (UR )21 + (EL )32 (UR )31 , 

e (UR )11 (UL )11 + (UR )21 (UL )21 + (UR )31 (UL )31
A5 = G 

× (ER )11 (DL )12 + (ER )21 (DL )22 + (ER )31 (DL )32 

e (UR )11 (DL )12 + (UR )21 (DL )22 + (UR )31 (DL )32
+G 

× (ER )11 (UL )11 + (ER )21 (UL )21 + (ER )31 (UL )31 , 

e (UR )11 (UL )11 + (UR )21 (UL )21 + (UR )31 (UL )31
A6 = G

Y. Achiman, C. Merten / Nuclear Physics B 584 (2000) 46–68 

× (EL )11 (DR )12 + (EL )21 (DR )22 + (EL )31 (DR )32 

e0 (DR )12 (UL )11 + (DR )22 (UL )21 + (DR )32 (UL )31
+G 

× (EL )11 (UR )11 + (EL )21 (UR )21 + (EL )31 (UR )31 , 

e (UR )11 (UL )11 + (UR )21 (UL )21 + (UR )31 (UL )31
A7 = G 

× (ER )12 (DL )12 + (ER )22 (DL )22 + (ER )32 (DL )32 

e (UR )11 (DL )12 + (UR )21 (DL )22 + (UR )31 (DL )32
+G 

× (ER )12 (UL )11 + (ER )22 (UL )21 + (ER )32 (UL )31 , 

e (UR )11 (UL )11 + (UR )21 (UL )21 + (UR )31 (UL )31
A8 = G 

× (EL )12 (DR )12 + (EL )22 (DR )22 + (EL )32 (DR )32 

e0 (DR )12 (UL )11 + (DR )22 (UL )21 + (DR )32 (UL )31
+G 

× (EL )12 (UR )11 + (EL )22 (UR )21 + (EL )32 (UR )31 , 

e (UR )11 (DL )11 + (UR )21 (DL )21 + (UR )31 (DL )31
A 9 = −G 

× (NL )11 (DR )11 + (NL )21 (DR )21 + (NL )31 (DR )31 

e0 (DR )11 (DL )11 + (DR )21 (DL )21 + (DR )31 (DL )31
−G 

× (NL )11 (UR )11 + (NL )21 (UR )21 + (NL )31 (UR )31 , 

e (UR )11 (DL )11 + (UR )21 (DL )21 + (UR )31 (DL )31
A10 = −G 

× (NL )12 (DR )11 + (NL )22 (DR )21 + (NL )32 (DR )31 

e0 (DR )11 (DL )11 + (DR )21 (DL )21 + (DR )31 (DL )31
−G 

× (NL )12 (UR )11 + (NL )22 (UR )21 + (NL )32 (UR )31 , 

e (UR )11 (DL )11 + (UR )21 (DL )21 + (UR )31 (DL )31
A11 = −G 

× (NL )11 (DR )12 + (NL )21 (DR )22 + (NL )31 (DR )32 

e0 (DR )12 (DL )11 + (DR )22 (DL )21 + (DR )32 (DL )31
−G 

× (NL )11 (UR )11 + (NL )21 (UR )21 + (NL )31 (UR )31 , 

e (UR )11 (DL )11 + (UR )21 (DL )21 + (UR )31 (DL )31
A12 = −G 

× (NL )12 (DR )12 + (NL )22 (DR )22 + (NL )32 (DR )32 

e0 (DR )12 (DL )11 + (DR )22 (DL )21 + (DR )32 (DL )31
−G 

× (NL )12 (UR )11 + (NL )22 (UR )21 + (NL )32 (UR )31 , 

e (UR )11 (DL )12 + (UR )21 (DL )22 + (UR )31 (DL )32
A13 = −G 

× (NL )11 (DR )11 + (NL )21 (DR )21 + (NL )31 (DR )31 

e0 (DR )11 (DL )12 + (DR )21 (DL )22 + (DR )31 (DL )32
−G 

× (NL )11 (UR )11 + (NL )21 (UR )21 + (NL )31 (UR )31 , 

e (UR )11 (DL )12 + (UR )21 (DL )22 + (UR )31 (DL )32
A14 = −G 

× (NL )12 (DR )11 + (NL )22 (DR )21 + (NL )32 (DR )31 

e0 (DR )11 (DL )12 + (DR )21 (DL )22 + (DR )31 (DL )32
−G 

× (NL )12 (UR )11 + (NL )22 (UR )21 + (NL )32 (UR )31 ,

63

64

Y. Achiman, C. Merten / Nuclear Physics B 584 (2000) 46–68 

e (UR )11 (DL )11 + (UR )21 (DL )21 + (UR )31 (DL )31
A15 = −G 

× (NL )13 (DR )11 + (NL )23 (DR )21 + (NL )33 (DR )31 

e0 (DR )11 (DL )11 + (DR )21 (DL )21 + (DR )31 (DL )31
−G 

× (NL )13 (UR )11 + (NL )23 (UR )21 + (NL )33 (UR )31 , 

e (UR )11 (DL )11 + (UR )21 (DL )21 + (UR )31 (DL )31
A16 = −G 

× (NL )13 (DR )12 + (NL )23 (DR )22 + (NL )33 (DR )32 

e0 (DR )12 (DL )11 + (DR )22 (DL )21 + (DR )32 (DL )31
−G 

× (NL )13 (UR )11 + (NL )23 (UR )21 + (NL )33 (UR )31 , 

e (UR )11 (DL )12 + (UR )21 (DL )22 + (UR )31 (DL )32
A17 = −G 

× (NL )13 (DR )11 + (NL )23 (DR )21 + (NL )33 (DR )31 

e0 (DR )11 (DL )12 + (DR )21 (DL )22 + (DR )31 (DL )32
−G 

× (NL )13 (UR )11 + (NL )23 (UR )21 + (NL )33 (UR )31 .
e0 = g 2 /2M 2 0 0 , where M 2 =
e = g 2 /2M 2 and G
We used here the definitions G
U
X,Y
U
X,Y
X ,Y
MX2 0 ,Y 0 ≈ MU2 is assumed.
The coefficients Ai are connected to the hadronic transition amplitudes of the elementary
processes responsible for the nucleon decays. The independent amplitudes are given in
Tables 12 and 13.
Table 12
Decay amplitudes for the elementary processes of proton decays
Decay process
+ C
p↑→ eR
u ↑ u↓

Lagrangian term 

µ
u¯ C
L γ uL 

+
e¯R
γµ dR 


Amplitude (× 30 )

Coefficient

−4

A2

+ C
p↑→ eR
u ↓ u↑
+ C
p↑→ eR
d ↑ d↓
+ C
p↑→ eR d ↓ d ↑

+
u¯ C γ µ uL e¯R
γµ dR   

L
+
µu
e
¯
u¯ C
γ
γ
d
µ
L
R 
R  

L
+
µu
e
¯
u¯ C
γ
γ
d
µ
L
R
L
R

−8

A2

−8

A2

+2

A2

+ C
p↓→ eR
u ↓ u↓

+
u¯ C γ µ uL e¯R
γµ dR   

L
+
C
µ
u¯ L γ uL e¯R γµ dR   

+
µ
u¯ C
L γ uL e¯R γµ sR   

+
µu
e
¯
u¯ C
γ
γ
s
µ
L
R
L
R   

+
C
µ
u¯ L γ uL e¯R γµ sR   

C
µ
u¯ C
L γ dL ν¯ eR γµ dR   

C γ d
µd
ν
¯
u¯ C
γ
µ
L
R
L
eR

−10

A2

−2

A2

−8

A6

+2

A6

−2

A6

+4

A9

−10

A9

+ C
p↓→ eR
d ↓ d↓
+ C
p↑→ eR
s ↑ d↓
+ C
p↑→ eR
s ↓ d↑
+ C
p↓→ eR
s ↓ d↓
C d C ↑ u↓
p↑→ νeR
C d C ↓ u↑
p↑→ νeR      

Y. Achiman, C. Merten / Nuclear Physics B 584 (2000) 46–68

65

Table 12 — continued
Decay process
C d C ↓ u↓
p↓→ νeR
C s C ↑ u↓
p↑→ νeR
C s C ↓ u↑
p↑→ νeR
C s C ↑ u↓
p↑→ νeR
C s C ↓ u↑
p↑→ νeR
C s C ↓ u↓
p↓→ νeR
C s C ↓ u↓
p↓→ νeR

Lagrangian term


Amplitude (× 30 )

Coefficient

−8

A9

+4

A11

+2

A11

0

A13

−12

A13

+4

A11

−12

A13


Amplitude (× 30 )

Coefficient

−4

A2

+10

A2

+8

A2

+8

A9

−2

A9

+4

A9

+8

A9

+2

A9

+10

A9

+4

A11

−4

A11

0

A13

+12

A13

−2

A11

+12

A13  

C
µ
u¯ C
L γ dL ν¯ eR γµ dR   

C
µ
u¯ C
L γ dL ν¯ eR γµ sR   

C γ s
µd
ν
¯
u¯ C
γ
µ
L
R 
eR  

L
C γ d
µs
ν
¯
u¯ C
γ
L
L
eR µ R   

C
µ
u¯ C
L γ sL ν¯ eR γµ dR   

C
µ
u¯ C
L γ dL ν¯ eR γµ sR   

C
µ
u¯ C
L γ sL ν¯ eR γµ dR 

Table 13
Decay amplitudes for the elementary processes of neutron decays
Decay process
+ C
n↑→ eR
u ↑ d↓
+ C
n↑→ eR
u ↓ d↑
+ C
n↓→ eR
u ↓ d↓
C uC ↑ u↓
n↑→ νeR
C uC ↓ u↑
n↑→ νeR
C d C↑ d ↓
n↑→ νeR
C d C↓ d ↑
n↑→ νeR
C uC ↓ u↓
n↓→ νeR
C d C↓ d ↓
n↓→ νeR
C sC↑ d ↓
n↑→ νeR
C sC↓ d ↑
n↑→ νeR
C sC↑ d ↓
n↑→ νeR
C sC↓ d ↑
n↑→ νeR
C sC↓ d ↓
n↓→ νeR
C sC↓ d ↓
n↓→ νeR

Lagrangian term   

µ
u¯ C
L γ uL
µ
u¯ C
L γ uL   

+
e¯R
γµ dR
+
e¯R
γµ dR   

+
µ
u¯ C
L γ uL e¯R γµ dR   

C
µ
u¯ C
L γ dL ν¯ eR γµ dR   

C γ d
µd
ν
¯
u¯ C
γ
µ
L
R 
eR  

L
C γ d
µd
ν
¯
u¯ C
γ
µ R
L
L
eR   

C
µ
u¯ C
L γ dL ν¯ eR γµ dR   

C γ d
µd
ν
¯
u¯ C
γ
µ
L
R 
eR  

L
C γ d
µd
ν
¯
u¯ C
γ
µ
L
R
L
eR   

C
µ
u¯ C
L γ dL ν¯ eR γµ sR   

C γ s
µd
ν
¯
u¯ C
γ
µ
L
R 
eR  

L
C γ d
µs
ν
¯
u¯ C
γ
L
L
eR µ R   

C γ d
µs
ν
¯
u¯ C
γ
µ
L
R
L
eR   

C
C
µ
u¯ L γ dL ν¯eR γµ sR   

C
µ
u¯ C
L γ sL ν¯ eR γµ dR

66

Y. Achiman, C. Merten / Nuclear Physics B 584 (2000) 46–68

References
[1] Y. Achiman, PRHEP-Corfu 98/026, RH Mixing Observable?, Talk at the Corfu Summ. Inst. on
Elem. Part. Phys., 1998, hep-ph/981238.
[2] M. Nahakata, SuperKamiokande Collaboration, Talk presented at TAUP 99, Paris 6–9
September 1999.
[3] Y. Fukuda et al., SuperKamiokande Collaboration, Phys. Lett. B 436 (1998) 33; Phys. Rev.
Lett. 81 (1998) 1562.
[4] M. Ambrosio et al., MACRO Collaboration, Phys. Lett. B 434 (1998) 451.
[5] M. Spurio, MACRO Collaboration, hep-ex/9908066.
[6] B.T. Cleveland et al., Homestake Collaboration, Astrophys. J. 496 (1998) 505.
[7] K.S. Hirata et al., Kamiokande Collaboration, Phys. Rev. Lett. 77 (1996) 1683.
[8] W. Hampel et al., GALLEX Collaboration, Phys. Lett. B 447 (1999) 127.
[9] J.N. Abdurashitov et al., SAGE Collaboration, Phys. Rev. Lett. 77 (1996) 4708; Phys. Rev. C 60
(1999) 055801.
[10] Y. Fukuda et al., SuperKamiokande Collaboration, Phys. Rev. Lett. 81 (1998) 1158; Phys. Rev.
Lett. 82 (1999) 2430.
[11] M.B. Smy, SuperKamiokande Collaboration, hep-ex/9903034.
[12] P. Langacker, Phys. Rep. C 72 (1981) 185.
[13] A. Masiero, in: Grand Unification with and without Supersymmetry and Cosmological
Implications, World Scientific, Singapur, 1984.
[14] G.G. Ross, Grand Unified Theories, Benjamin/Cummings, Menlo Park, 1985.
[15] R.N. Mohapatra, Unification and Supersymmetry, 2nd edn., Springer-Verlag, New York, 1992.
[16] H. Georgi, in: C.E. Carlson (Ed.), Particles and Fields, AIP Press, New York, 1975.
[17] H. Fritzsch, P. Minkowski, Ann. Phys. 93 (1975) 193.
[18] M.S. Chanowitz, J. Ellis, M.K. Gaillard, Nucl. Phys. B 128 (1977) 506.
[19] P. Langacker, Phys. Rep. C 72 (1981) 187; in: E. Kolb et al. (Eds.), Inner Space and Outer
Space, University of Chicago Press, Chicago, 1986.
[20] D.G. Lee, R.N. Mohapatra, M.K. Parida, M. Rani, Phys. Rev. D 51 (1995) 229.
[21] M.S. Chanowitz, J. Ellis, M.K. Gaillard, Nucl. Phys. B 128 (1977) 506.
[22] G. Kane, G. Karl, Phys. Rev. D 22 (1980) 2808.
[23] M.B. Gavela, A. Le Yaouanc, L. Oliver, O. Pène, J.C. Raynal, Phys. Rev. D 23 (1981) 1580.
[24] G. Altarelli, F. Feruglio, Phys. Lett. B 451 (1998) 388.
[25] Z. Berezhiani, A. Rossi, hep-ph/9811447.
[26] S. Lola, G.G. Ross, hep-ph/9902283.
[27] C.H. Albright, K.S. Babu, S.M. Barr, Phys. Rev. Lett. 81 (1998) 1167.
[28] C.H. Albright, S.M. Barr, Phys. Rev. D 58 (1998) 013002.
[29] Y. Nomura, T. Yanagida, hep-ph/9807325.
[30] M. Bando, T. Kugo, hep-ph/9902204.
[31] Y. Achiman, C. Merten, Talk presented at TAUP 99, Paris 6–9 September 1999 (to be published
in Nucl. Phys. Proc. Suppl.), hep-ph/9911314.
[32] A. Davidson, V.P. Nair, K.C. Wali, Phys. Rev. D 29 (1984) 1504; Phys. Rev. D 29 (1984) 1513.
[33] Y. Achiman, T. Greiner, Nucl. Phys. B 443 (1995) 3.
[34] J.A. Harvey, P. Ramond, D.B. Reiss, Nucl. Phys. B 199 (1982) 223.
[35] C.D. Froggatt, H.B. Nielsen, Nucl. Phys. B 147 (1979) 277.
[36] P. Ramond, hep-ph/0001006.
[37] Y. Achiman, Nucl. Phys. Proc. Suppl. 66 (1998) 400.
[38] R.N. Mohapatra, hep-ph/9911272.
[39] Y. Achiman, M. Richter, work in progress.
[40] G.F. Giudice, hep-ph/9912279.
[41] L. Giusti, A. Romanino, A. Strumia, Nucl. Phys. B 550 (1999) 3.

Y. Achiman, C. Merten / Nuclear Physics B 584 (2000) 46–68

[42]
[43]
[44]
[45]
[46]
[47]
[48]
[49]
[50]
[51]
[52]
[53]
[54]
[55]
[56]
[57]
[58]
[59]
[60]
[61]
[62]
[63]
[64]
[65]
[66]
[67]
[68]
[69]
[70]
[71]
[72]
[73]
[74]
[75]
[76]
[77]
[78]
[79]
[80]
[81]
[82]
[83]
[84]
[85]

67

A. Romanino, A. Strumia, hep-ph/9912301.
L.E. Ibañez, FTUAM-99-37, hep-ph/9911499.
M. Chaichian, A.B. Kobakidze, hep-th/9912193.
Z. Checko, A.E. Nelson, hep-ph79912186.
I. Antoniadis, hep-ph/9909212.
S. Rajpoot, Phys. Rev. D 22 (1980) 2244.
R.W. Robinett, J.L. Rosner, Phys. Rev. D 26 (1982) 2396.
Y. Tosa, G.C. Branco, R.E. Marshak, Phys. Rev. D 28 (1983) 1731.
D. Chang, R.N. Mohapatra, M.K. Parida, Phys. Rev. D 30 (1984) 1052.
J.M. Gipson, R.E. Marshak, Phys. Rev. D 31 (1985) 1705.
D. Chang, R.N. Mohapatra, J.M. Gipson, R.E. Marshak, M.K. Parida, Phys. Rev. D 31 (1985)
1718.
N.G. Deshpande, E. Keith, P.B. Pal, Phys. Rev. D 46 (1992) 2261; Phys. Rev. D 47 (1993) 2892.
V.A. Kuzmin, M.E. Shaposhnikov, Phys. Lett. B 92 (1980) 115.
F. Acampora, G. Amelino-Camelia, F. Buccella, O. Pisanti, L. Rosa, T. Tuzi, Nuovo Cimento
A 108 (1995) 375.
M. Fukugita, T. Yanagida, Phys. Lett. B 174 (1986) 45.
W. Buchmüller, M. Plümacher, Phys. Lett. B 389 (1996) 73.
M. Flanz, E.A. Paschos, U. Sarkar, J. Weiss, Phys. Lett. B 389 (1996) 693.
A. Pilaftsis, hep-ph/9812256.
R.D. Peccei, H.R. Quinn, Phys. Rev. D 16 (1977) 1791.
J.E. Kim, Phys. Rev. Lett. 43 (1979) 103.
M.A. Shifman, V.I. Vainstein, V.I. Zakharov, Nucl. Phys. B 166 (1980) 4933.
M. Dine, W. Fisher, M. Strednicki, Phys. Lett. B 104 (1981) 199.
S.M. Bilenky, C. Giunti, W. Grimus, Prog. Part. Nucl. Phys. 43 (1999) 1.
C. Bemporad et al., CHOOZ Collaboration, Nucl. Phys. Proc. Suppl. 77 (1999) 159.
C. Athanassopoulos et al., LSND Collaboration, Phys. Rev. Lett. 77 (1996) 3082; Phys. Rev.
Lett. 81 (1998) 1774.
K. Eitel et al., KARMEN Collaboration, Nucl. Phys. Proc. Suppl. 77 (1999) 212.
J. Pati, A. Salam, Phys. Rev. D 10 (1974) 275.
F. del Aguila, L.E. Ibañez, Nucl. Phys. B 177 (1981) 60.
C. Merten, PhD Thesis WUB-DIS 99-14, University of Wuppertal, 1999, (in german), hepph/0002301.
Y. Achiman, T. Greiner, Nucl. Phys. B 443 (1995) 3.
H. Fritzsch, Phys. Lett. B 70 (1977) 436; Phys. Lett. B 73 (1978) 317; Nucl. Phys. B 155 (1979)
189.
K.S. Babu, Q. Shafi, Phys. Lett. B 357 (1995) 365.
Y. Achiman, T. Greiner, Phys. Lett. B 324 (1994) 33.
G.C. Branco, L. Lavoura, F. Mota, Phys. Rev. D 39 (1989) 3443.
L. Wolfenstein, Phys. Rev. D 17 (1978) 2369; Phys. Rev. D 20 (1979) 2634.
S.P. Mikheyev, A.Yu. Smirnov, Sov. J. Nucl. Phys. 42 (1986) 913.
A.J. Buras, Les Houches Lectures, in: F. David, R. Gupta (Eds.), Probing the Standard Model
of Particle Interactions, Elsevier Science B.V., 1998, hep-ph/9806471.
J. Bolz, P. Kroll, Z. Phys. A 356 (1996) 327.
P. Langacker, in: E.W. Kolb (Ed.), Inner Space, Outer Space, Univ. Chicago Press, Chicago,
1986.
M. Shiozawa et al., SuperKamiokande Collaboration, Phys. Rev. Lett. 81 (1998) 3319.
Y. Hayato et al., SuperKamiokande Collaboration, Phys. Rev. Lett. 83 (1999).
F. Arneodo et al., ICARUS Collaboration, Nucl. Phys. Proc. Suppl. 70 (1999) 453.
V. Lucas, S. Raby, Phys. Rev. D 55 (1997) 6986.
P. Nath, R. Arnowitt, hep-ph/9708469.

68

Y. Achiman, C. Merten / Nuclear Physics B 584 (2000) 46–68

[86] J. Pati, K.S. Babu, F. Wilczek, hep-ph/9912301.
[87] Y. Achiman, J. Keymer, University of Wuppertal Preprint WUB 83-16, contr. paper #275 to Int.
Symp. on Lepton and Photon Interactions, Cornell 1983.
[88] J. Keymer, Diploma thesis WUD 84-23, University of Wuppertal, 1983, (in German).
[89] Y. Achiman, S. Bielefeld, Phys. Lett. B 412 (1997) 320.
[90] H. Georgi, D.V. Nanopoulos, Nucl. Phys. B 159 (1979) 16.
[91] M. Gell-Mann, P. Ramond, R. Slansky, in: P. van Nieuwenhuizen, D.Z. Freedman (Eds.),
Supergravity, North-Holland, Amsterdam, 1979.
[92] T. Yanagida, in: O. Sawada, A. Sugamoto (Eds.), Proceedings of the Workshop on the Unified
Theory and the Baryon Number of the Universe, KEK report 79-18, Tsukuba, 1979.
[93] R.N. Mohapatra, G. Senjanowi´c, PRL4480912.

Nuclear Physics B 584 (2000) 69–108
www.elsevier.nl/locate/npe

CFT’s from Calabi–Yau four-folds
Sergei Gukov a,∗ , Cumrun Vafa b , Edward Witten c
a Department of Physics, Princeton University, Princeton, NJ 08544, USA
b Jefferson Laboratory of Physics, Harvard University, Cambridge, MA 02138, USA
c School of Natural Sciences, Institute for Advanced Study, Olden Lane, Princeton, NJ 08540, USA

Received 25 May 2000; accepted 13 June 2000

Abstract
We consider F/M/Type IIA theory compactified to four, three, or two dimensions on a Calabi–
Yau four-fold, and study the behavior near an isolated singularity in the presence of appropriate
fluxes and branes. We analyze the vacuum and soliton structure of these models, and show that near
an isolated singularity, one often generates massless chiral superfields and a superpotential, and in
many instances in two or three dimensions one obtains nontrivial superconformal field theories. In
the case of two dimensions, we identify some of these theories with certain Kazama–Suzuki coset
models, such as the N = 2 minimal models.  2000 Elsevier Science B.V. All rights reserved.

1. Introduction
We have learned in recent years that it is fruitful to study singular limits of string
compactifications. In this paper, we consider theories with four supercharges in four,
three, and two dimensions, constructed by considering F-theory, M-theory, and type IIA
string theory on a Calabi–Yau four-fold with an isolated complex singularity. We can
connect these theories to each other by circle compactifications from four to three to
two dimensions. In addition to the choice of singularity, the description of these theories
depends on certain additional data involving the four-form flux and membrane charge in
M-theory (and related objects in the other theories).
We will analyze the vacuum structure of these theories and the domain walls connecting
the possible vacua. We argue that in many cases, the nonperturbative physics near a
singularity generates massless chiral superfields with a superpotential, leading in many
instances, especially in two dimensions or in three dimensions with large membrane
charge, to an infrared flow to a nontrivial conformal field theory. In some cases, we can
identify the theories in question with known superconformal models; for example, type IIA
∗ Corresponding author. E-mail: gukov@feynman.princeton.edu

0550-3213/00/$ – see front matter  2000 Elsevier Science B.V. All rights reserved.
PII: S 0 5 5 0 - 3 2 1 3 ( 0 0 ) 0 0 3 7 3 - 4

70

S. Gukov et al. / Nuclear Physics B 584 (2000) 69–108

at a four-fold An singularity gives an N = 2 Kazama–Suzuki model in two dimensions,
as we argue using the vacuum and soliton structure. More generally, from the ADE fourfold hypersurface singularities with appropriate fluxes, we obtain all the N = 2 Kazama–
Suzuki models [1] at level one. This is a large list of exactly solvable conformal theories,
which includes the N = 2 unitary minimal models. It is quite satisfying that strings in the
presence of singularities captures such a large class of known conformal theories in two
dimensions, and suggests that maybe even in higher dimensions, strings propagating in
singular geometries yield an equally large subspace of conformal theories. Moreover, since
the Kazama–Suzuki models are exactly solvable conformal theories in two dimensions, it
would be interesting to see to what extent its known spectrum and correlation functions can
be extracted from string theory. This would be a natural testing grounds in view of potential
application to higher dimensions where the conformal theories are less well understood.
Our result gives a relation between singularity theory, as in the Landau-Ginzburg approach to conformal theories in two dimensions [2–4], and the singularity of internal compactification geometry. This relation may well extend to non-supersymmetric examples; it
would certainly be interesting to explore this.
In Section 2, we analyze the fluxes, branes, and vacuum states near a four-fold
singularity. In Section 3, we show how to compute the spectrum of domain walls (which
can also be viewed as strings and kinks in the three and two-dimensional cases) for a special
class of singularities. In Section 4, we identify the models derived from ADE singularities
with Kazama–Suzuki models at level 1. In Section 5, we discuss some additional classes
of singularities on four-folds, and in Section 6, we discuss the reinterpretation of some of
our results in terms of branes.

2. Classification of vacua
2.1. The G-field and domain walls
For our starting point, we take M-theory on R3 × Y , where Y is a compact eightmanifold. Soon, we will specialize to the case that Y is a Calabi–Yau four-fold, so as to
achieve supersymmetry. We also will note in Section 2.5 the generalization of our remarks
to type IIA or F-theory compactification on Y .
To fully specify a vacuum on Y , one must specify not just Y but also the topological
class of the three-form potential C of M-theory, whose field strength is G = dC. Roughly
speaking, C-fields are classified topologically by a characteristic class ξ ∈ H 4 (Y ; Z). At
the level of de Rham cohomology, ξ is measured by the differential form G/2π , and we
sometimes write it informally as ξ = [G/2π]. 1
1 The assertion that ξ takes values in H 4 (Y ; Z) is a bit imprecise, since in general [5] the G-field is shifted
from standard Dirac quantization and ξ is not an element of H 4 (Y ; Z). But the difference between two C-fields is
always measured by a difference ξ − ξ 0 ∈ H 4 (Y ; Z). ξ itself takes values in a “principal homogeneous space” Λ
for the group H 4 (Y ; Z); the relation between H 4 (Y ; Z) and Λ is just analogous to the relation between H 2 (Y ; Z)
and the set of Spinc structures on Y . The shift in the quantization law of G arises precisely when the intersection

Z). we get a relation in cohomology [∂M] = G1 + G2 . Hence. Z) is not even.6] Z 1 G∧G χ − . Now. and δ(∂M) is a four-form with delta function support on ∂M. the net source of the C-field must vanish. then the right hand side of (2.1) N= 24 2 (2π)2 Y If G obeys the shifted quantization condition mentioned in the last footnote. Moreover. (2. By Poincaré duality. so that their boundaries on W are of the form on H 4 (Y . it suffices to describe a domain wall interpolating between models with the same Y and with C-fields of arbitrary characteristic classes ξ1 and ξ2 . Let us see how this comes about. this will occur only in Section 5. Gukov et al. In our examples. More generally. Because the G-field actually jumps in crossing the fivebrane. Since the left hand side of (2. In this construction. 2 To show this. then N must also jump. Z). Eq. . then ξ2 − ξ1 = [S]. Z) = H 4 (Y . With Y being compact. 2 This may also be true of models with different Y — as suggested by results of [7] in the threefold case — but this issue is much harder to explore. Hence if it equals ξ1 on one side. this model can be generalized by picking N points Pi ∈ Y and including N membranes with worldvolumes of the form R3 × Pi . Being of codimension one in space–time. / Nuclear Physics B 584 (2000) 69–108 71 Without breaking the three-dimensional Poincaré symmetries.S. with R2 a linear subspace of R3 . 2(2π) (2. then it equals ξ2 = ξ1 + [S] on the other side. We will assume that this should be understood as the average of the G-field on the two sides: G|W = (G1 + G2 )/2. representing an element of H4 (Y . The key is that there is a self-dual three-form T on the fivebrane with a relation dT = G|W − 2πδ(∂M).2) is zero in cohomology. because the fivebrane is a magnetic source of G.3) where [∂M] is the cohomology class dual to ∂M. there is a four-cycle S ⊂ Y . Z) is the class that is Poincaré dual to S. (2. models defined using the same Y but different ξ are actually different states of the same model. such that if [S] ∈ H 4 (Y. thus it can be a positive or negative integer. We are interested in the case that the membrane worldvolumes are of the form R3 × Pi .2) where G|W is the restriction of the G-field to the worldvolume W . consider a fivebrane in R3 × Y whose worldvolume is W = R2 × S.1) implies that if ξ jumps in crossing a domain wall. we include both membranes and antimembranes and let N be the difference between the number of membranes and antimembranes. this gives a relation [5. ∂M is the union of all boundaries of membrane worldvolumes that terminate on W . it is not completely obvious how to interpret the term G|W . the characteristic class ξ = [G/2π] jumps by [S] in crossing this domain wall. this fivebrane looks macroscopically like a domain wall. and we will ignore this issue until that point. H4 (Y . (2.1) is always integral [5].

For χ negative. that is.4) N1 − N2 = 2 2π 2 (2π)2 S Y Here we have used the fact that [S] = [(G2 − G1 )/2π] to convert an integral over S to one over Y . Incorporation of supersymmetry Now we wish to specialize to the case that Y is a Calabi–Yau four-fold and to look for vacua with unbroken supersymmetry. assuming that one wants unbroken supersymmetry with zero cosmological constant. The result is that G must be of type (2. We will analyze this condition in Appendix A.6) (2π)2 Y with equality only if G = 0. (2. Clearly. and integrate over S to get a cohomology relation. but for now we note that it implies that G is self-dual and hence in particular that Z G∧G > 0. the change in N in crossing the domain wall or in other words the net number of membranes whose boundary is on the fivebrane.4) is compatible with the requirement that (2. The requirements for G. it must obey K ∧ G = 0. Given that N must be positive. the relation (2. R The energy density in the bulk gets contribution from the G flux given by 8π1 2 G ∧ ∗G . The condition that the domain wall be flat and have finite tension requires that the energy density on the two sides of the domain wall should be equal far away from the domain wall. This is based on a finite energy condition. there are only finitely many choices of G that are compatible with unbroken supersymmetry. 2) and must be primitive.72 S. Gukov et al. for compact Y .7) 6 .1) should hold on both sides of the domain wall. The integration converts [∂M] into N1 − N2 . several restrictions must be imposed. So Z Z 1 G21 − G22 1 G1 + G2 =− . (2.1) implies that Z χ G∧G (2. In evaluating (2. (2. 2 12 (2π) Y This inequality together with self-duality implies that. The second basic consequence of supersymmetry is that N must be positive.5) where K is the Kähler form. there are none at all. Only membranes and not anti-membranes on R3 × Pi preserve the same supersymmetry that is preserved by the compactification on the complex four-fold Y . For this. (2. have been obtained in an elegant computation [8]. we can suppress the common R2 factor. Energetic considerations R There is another way to understand the result that Φ = N + 8π1 2 G ∧ G should not change in crossing the domain wall. / Nuclear Physics B 584 (2000) 69–108 form R2 × Pi .3).

G = ∗G so the energy density is given by N + 8π1 2 G ∧ G. is to study the behavior as Y develops a singularity. The singular hypersurface F = 0 admits the U (1) symmetry group za → eiθra za . Our main interest in the present paper. . a = 1. . z5 ). . . we begin with five complex variables za . The cancellation of the boundary terms in question in our case is again precisely the condition of constancy of Φ across the domain wall. If the model is to flow in the infrared to a superconformal field theory. G is self-dual (as R explained in Appendix A). it is convenient then to omit the part of Y that is far from the singularity and to consider a complete but not compact Calabi–Yau four-fold that is developing a singularity. In fact. Then we let X be a smooth deformation of this singular hypersurface such as F (z1 . z5 ) =  (2.8) with  a constant. however. We assume that F is such that the hypersurface F = 0 is smooth except for an isolated singularity at z1 = · · · = z5 = 0. This is also important in our applications later. Our discussion will apply directly to an isolated singularity of a non-compact variety. 5 of degree ra > 0 and a polynomial F (z1 . Hypersurface singularities are an important example and will be our focus in this paper except in Section 5.2. .e.10) (2. . . rΩ = (2. and so its constancy across a domain wall is a consequence of the finite energy of the domain wall. some of the singularities we will study — like the An singularities of a complex surface for very large n — probably cannot be embedded in a compact Calabi–Yau manifold. For practical purposes. . z5 ) = (2. / Nuclear Physics B 584 (2000) 69–108 73 and from the membranes by N . . For the supersymmetric situation we are considering. . as we will use a BPS formula for the mass of domain walls. i. . Gukov et al. For example. Due to boundary terms at infinity. such formulas are generally not valid for objects of very low codimension in space (the codimension is one in our case). or more generally X ti Ai (z1 . Under this transformation.9) i with complex parameters ti and polynomials Ai that describe relevant perturbations of the singularity F = 0. . . . .S. . . the holomorphic four-form dz2 ∧ dz3 ∧ dz4 ∧ dz5 Ω= ∂F /∂z1 has charge X ra − 1. z5 ) that is homogeneous of degree 1.11) (2.10) is an R-symmetry group if rΩ 6= 0. one of our important applications will be to quasihomogeneous hypersurface singularities. an R-symmetry must appear in the . . Behavior near a singularity So far we have considered the case of a compact smooth manifold Y . F (z1 .12) a The U (1) symmetry in (2. . In this case.. . 2.

1) no longer holds. The discussion of [9] applies to this situation. A theorem of Tian and Yau [9] asserts. and r is a “radial” coordinate near infinity which scales under za → λra za as r → λ(Σa ra −1)/4 r. with D the divisor w = 0. Flux at infinity Noncompactness of X leads to several important novelties in the specification of the model. though it has in a certain sense (using the compactification described in the next paragraph) a pole at infinity. 3 For our purposes. and identifies r with a fractional power of |w|. First of all. Gukov et al. In what sense is such an X a Calabi–Yau manifold? The holomorphic four-form Ω defined in (2. To apply the Tian–Yau theorem to the hypersurface X and deduce the statements in the last paragraph.14) and moreover (see the precise statement in Eq. we propose that it is the symmetry just identified. we do not strictly need to know that this is true. an extra term appears in (2. namely the flux Φ measured at infinity. one writes ra = ba /d with ba relatively prime integers and d a positive integer. It is very plausible that if a compact Calabi–Yau manifold Y develops an isolated hypersurface singularity that is at finite distance on the moduli space.3) of [9]) this metric is asymptotically conical. This exponent ensures that the volume form derived from (2. the R-charges normalized so that Ω has R-charge 1). . assuming (2. Note that the variety Y 0 used in the last paragraph in relation to the Tian–Yau theorem is not a Calabi–Yau manifold. then the dimensions of the corresponding operators are proportional to ri /rΩ (in other words. Then one introduces another complex variable w of degree 1/d. Rather.15) scales like |Ω|2 . but the physical applications are certainly rather natural if it is.13) a We will see the importance of this condition from several points of view. invariant under the dynamics which occurs in the “interior” of X. This flux is a constant of the motion. only that the behavior of Y near its singularity can be modeled by X. (2. and one compactifies X to the compact variety Y 0 defined by the equation F (zi ) − wd = 0 in a weighted projective space. / Nuclear Physics B 584 (2000) 69–108 superconformal algebra.11) has no zeroes or poles on X. that there is a Calabi–Yau metric on X with volume form ¯ |Ω ∧ Ω|. then the Calabi– Yau metric on Y looks locally like the conelike metric that we have just described on the hypersurface X. requiring that the dimensions be positive gives a condition rΩ > 0: X ra > 1. 4 When X is not compact. (2. If the Ai have degrees ri . 4 then we can write the conserved quantity as 3 There is no claim here that X can be globally embedded in Y . Since the ri are positive. If we absorb the constant χ/24 in the definition of Φ. χ must in any event be defined by a curvature integral and need not coincide with the topological Euler characteristic. flux can escape to infinity. (2.1). and hence (2.13).74 S. that is it looks near infinity like 2 . ds 2 = dr 2 + r 2 ds⊥ (2.15) Here ds⊥ is an “angular” metric.

16) are positive (for the same reasons as in the case of compact X). In addition to Φ. Z) i : Hcpct (2. the C-field at infinity is flat. When the 4 (X. respectively. The groups H 4 (X. Hcpct Poincaré duality in the noncompact case says that Γ and Γ ∗ are dual lattices. topologically. Z). / Nuclear Physics B 584 (2000) 69–108 Φ =N + 1 2 Z G∧G . the rough idea being that if S is a four-cycle determining an element of H4 (X. subject to the equivalence relation that β ∼ = β + d if  has compact support. For this. Local dynamics cannot change the behavior at infinity.1 are classified by H4 (X. Z) are lattices. so the restriction of ξ to ∂X is another invariant of the local dynamics. the map i is an embedding. Moreover. C-fields on 4 (X. Z) and H 4 (X. Z) X are classified topologically by ξ ∈ H 4 (X. For unbroken supersymmetry. for hypersurface singulari4 (X. but only that its restriction is a torsion class. determined by the values of N and G. and Γ intersection form on Hcpct can be regarded as a sublattice of its dual lattice Γ ∗ .17) (by “forgetting” that a class has compact support). associated with a tree diagram in which the vertices represent four-spheres and two vertices are connected by a line if and only if the corresponding four-spheres intersect. Z) and Hcpct that classify. Domain walls of the type introduced in Section 2. respectively. 5 The cohomology of X with compact support is generated by closed forms β on X with compact support. Z) is a lattice Γ with one generator for every four-sphere in the bouquet. Z) has no null vectors. Z) → H 4 (X. there is another quantity that characterizes the definition of the model — and commutes with the dynamics. which classifies the cycles S on which a fivebrane can wrap to make a domain wall. by definition. Gukov et al. 6 Such a bouquet is. . A model with given Φ has various vacuum states. However. except that the Cartan matrix may not be positive definite and thus one is not restricted to the ADE case. we have to look at precisely what Poincaré duality says in the case of a noncompact manifold X. it is reasonable to require that the flux G vanishes if restricted to ∂X. which we will call Γ and Γ ∗ . which must be specified in defining a model.16) X We can think of Φ as a constant that must be specified (in addition to giving X) in order to determine the model. The lattice Γ can actually be described rather simply. Z) (where H 4 denotes cohomology Poincaré duality says that this is the same as Hcpct cpct 5 with compact support). For finiteness of the energy.S. both terms on the right-hand side of (2. topological classes of C-fields and of changes in C-fields in crossing domain walls are in general different for non-compact X. the region near infinity in X. but perhaps topologically nontrivial. There is another way to see more explicitly how this invariant comes about. Z) (so in particular S is by definition compact). ties. 6 H4 (X. Z). the hypersurface X is homotopic to a “bouquet” of four-spheres. there is always a natural map 4 (X. 4 (X. This does not imply that the cohomology class ξ vanishes if restricted to ∂X. This makes things very simple. In fact. then the Poincaré dual class [S] is represented by a delta function δ(S) that has compact support. Thus. (2π)2 75 (2. Such a diagram has the form of a simply-laced Dynkin diagram (with vertices for four-spheres and lines for intersections of them). so there are only finitely many possible choices of N and G for fixed Φ.

∂X. If the intersection pairing on H4 (X. ∂X) reads in part i j · · · H 4 (X. . (2. By Poincaré duality. (2. Z)/i Hcpct This can be reinterpreted as follows. Z) i Hcpct Thus the value of ξ modulo changes in crossing domain walls (the right-hand side of (2. and jump in crossing a domain wall by elements of Γ ). Z) single sphere. In particular. Z) = H 4 (X. y = −u . Z). Z) = Γ ∗ /Γ. So H 4 (∂X. and each point contributes +1 to the intersection number. since yE · uE = 0. Z) → H 5 (X. X is homotopic to the four-sphere S. H 5 (X.19) i (X.21) a=1 2 2 If we assume that p  is real and write za = xa + iya . but only by something of the form i([S]) where [S] is a class with compact support. Z) −→ H 4 (∂X. (2. we see that uE is a unit vector. Z) by the group of null vectors (which can be shown to be precisely the image in H 4 (X. Gukov et al. with appropriate boundary conditions. Z) → · · · . Then S 0 intersects S at the two points u1 = · · · = u4 = 0. the G-fields. Z)). ξ cannot change by an arbitrary amount. Z) = Here H i (X.20) H 4 (∂X. ∂X. ∂X. that is due to crossing domain walls. X is the cotangent bundle of S. take values in Γ ∗ . are thus classified by  4 (X. Consider first the simple case that X is a deformation of a quadric singularity: 5 X za2 = . So the exact sequence implies that   4 (X. Z) is the same as Hcpct H3 (X. u5 = ±1. Z) = H 4 (X. and this is zero on dimensional grounds for a bouquet of four-spheres. Z) = 1 Γ /Γ = Z . The self-intersection number of S is S · S = 2.18) H 4 (X. Z) of H 3 (∂X. then we should define Γ to be the quotient of H 4 (X. everything that we have described above carries over (i embeds Γ as a finite index sublattice of Γ ∗ . ∂X. Hence S · S = S · S 0 = 2. Z) is generated by 2 [S] (whose scalar 4 (X. The exact sequence of the pair (X. product with S is 1).76 S. The dual Γ ∗ is the subgroup of H 4 (X. Z) consisting of elements whose restriction to ∂X is torsion. (2. but the dual lattice Γ = H (X. Examples We will now illustrate these perhaps slightly abstract ideas with examples that will be important later. 2 Setting uE = xE /  + yE . Z) −→ H 4 (X. The subset of X with yE = 0 is a four-sphere S. The possible values of ξ modulo changes due to the dynamics. Z). we get xE − yE =  and xE · yE = 0. y = u . / Nuclear Physics B 584 (2000) 69–108 In crossing a domain wall. ∂X.20)) can be identified with the restriction of ξ to ∂X (the left hand side). 7 The lattice Γ = Hcpct 1 ∗ 4 is generated by [S]. ∂X. With these definitions of Γ and Γ ∗ . This is the case in which the bouquet of spheres is made from just a 4 (X. y = −u . Z)/Hcpct 2 2 7 To compute this. deform S to the four-sphere S 0 defined by y = u . Z) = Γ ∗ /Γ. y = 1 2 2 1 3 4 4 3 5 0. Z) is degenerate.

(Si intersects Si+1 at the single point z1 = bi+1 . z5 ) = 0.22) a=2 Here Pn (z1 ) is a polynomial of degree n. . One approach to proving these assertions is to show that they are true for the middle-dimensional cohomology of the surface H (z1 . we can consider the corresponding surface singularity in complex dimension four: H (z1 . Y develops a singularity that looks like a quasihomogeneous hypersurface singularity F (z1 . . . and the quotient H 4 (∂X. while the dual lattice Γ ∗ = H 4 (X. z5 ) + i Here the ti are complex parameters. z3 ) is a polynomial in three complex variables that describes a deformation of an ADE surface singularity. when some complex parameters ti are varied. and are unaffected by “stabilizing” the singularity by adding two more variables with the quadratic terms z42 + z52 . 2. . (2.) Endowed with this intersection form. It can be shown that X is homotopic to the union of the Si .24) The case just considered. . z2 . zj = 0 for j > 1. . b1 < b2 < · · · < bn . if F (z1 . For i = 1. .25) F (z1 . The quotient is H 4 (∂X. In this case. . with H (z1 . with others i vanishing. Γ ∗ is the weight lattice of G. The Si generate the lattice Γ = 4 (X. Z) = Γ ∗ /Γ = Zn . .23) i=1 with real bi . the hypersurface is deformed to a smooth one which looks locally like X ti Ai (z1 . (2. / Nuclear Physics B 584 (2000) 69–108 77 A somewhat more sophisticated example is the An−1 singularity in complex dimension four: Pn (z1 ) + 5 X za2 = 0. Si does not intersect Sj if |j − i| > 1. For simplicity we take Pn (z1 ) = n Y (z1 − bi ) (2.) For any of the ADE examples. . Upon varying the complex structure of Y . where the za have degrees ra > 0 and F is of degree 1. S · S Hcpct i i i+1 = 1. and that the zj for j > 1 are all real or all imaginary depending on the value of i modulo two. Z) is the weight lattice. (2. n − 1. Γ is the root lattice of the appropriate simply-connected ADE group G.5. . Z) = Γ ∗ /Γ is isomorphic to the center of G. . we define a four-sphere Si by requiring that z1 is real with bi < z1 < bi+1 . The intersection numbers of the S are S 2 = 2. z2 . Γ is the root lattice of the Lie group An−1 = SU(n). z3 ) = 0. . Distance to singularity and hodge structure of cohomology In the present subsection. . z3 ) + z42 + z52 = 0. Z). More generally. and the Ai are perturbations of the equation. z5 ) = 0. the bouquet is associated with the Dynkin diagram of An . We suppose that.S. .3. z2 . which form a “bouquet”. z2 . we return to the case of a compact Calabi–Yau four-fold Y . (The appropriate H ’s for the other cases are written at the end of Section 2. Gukov et al. corresponds to An−1 . z3 ) = Pn (z1 ) + z22 + z32 . . .

We will now apply this kind of reasoning to address the following question. in that this is the same condition that was needed to get an R-symmetry with positive charges and to apply the Tian–Yau theorem on existence of asymptotically cone-like Calabi–Yau metrics.78 S. For analyzing this question. If the answer is yes. we use a simple scaling. what is the Hodge type of the “vanishing cohomology”. Under za → λra za .25) and restrict the integral to the region |za | < 1. we normalize the holomorphic (4. of the part of the cohomology that “disappears” at the singularity? We only have to look at middle-dimensional cohomology. / Nuclear Physics B 584 (2000) 69–108 The first question to examine is whether the singularity at ti = 0 can arise at finite distance in Calabi–Yau moduli space. Then in the limit ti → 0. that is. the large za behavior. is immaterial (as long as there is some cutoff to avoid a divergence at large za ). we can. 8 The derivation of this formula is just as in the three-fold case. replace Y by the hypersurface in (2. There would thus be vanishing cohomology of type (4. (2. 0). Small za corresponds to small λ.28) a This is a satisfying result. (2. If the answer is no.27) scales like |λ|2Σa ra −2 . If and only if there is such a singularity. The Kähler form on the parameter space is ∂2 K. then to make the integral converge as ti → 0. which depends on how the singularity is embedded in a compact variety Y . see [10] for an exposition.29) Y converges as ti → 0. there is no vanishing cohomology of type (4. For this. (2. Gukov et al. First let us ask if there is vanishing cohomology of type (4. 0)-form Ω of Y in such a way that far from za = 0 it has a limit as ti → 0. Ω scales like λΣa ra −1 and so the integral in (2. To determine the small za behavior of the integral. Then we ask if the integral Z Ω ∧ Ω¯ (2. the distance to ti = 0 will be infinite in the metric (2. Ω would be a closed four-form that is non-zero but vanishes away from the singularity. for instance. On the parameter space of a compact Calabi–Yau manifold.27) ω = dt i d t¯j Y We want to analyze a possible singularity of this integral near za = 0 in the limit that the ti go to zero. 0). whose importance will become clear: as Y becomes singular. the Kähler potential of the Weil–Peterson metric is 8 Z ¯ K = − ln Ω ∧ Ω. Hence the condition that the integral converges at small za is X ra − 1 > 0. we would have to rescale Ω so that in the limit it vanishes pointwise away from the singularity. 0).26) ∂t i ∂ t¯j where K is the Kähler potential K. . because the deformation of a hypersurface singularity has cohomology only in the middle dimension.26).

(2.31) scales like |λ|wi with ! X ra + si − 2 . The (3. z2 .30) Y and ask if it is finite as all tj → 0. Hence. then to make the integral converge. If Φa have P c = 5a=1 (1 − degree ra and F has degree one.35) where H (z1 . . then ti scales like λ1−si and (2. They are given. Ωi would represent a nonzero (3.34) The singularities that obey this condition are the ADE singularities. we need to examine the integral Z Ωi ∧ Ω¯ i . 1) cohomology class that vanishes away from the singularity. or in other words a piece of the vanishing cohomology of type (3. we would have to rescale Ωi by a function of the tj . . z5 ) = H (z1 . The condition (2. be determined by scaling. (2. To determine if Ωi is a vanishing cycle. z3 ) + z42 + z52 . 1).32) wi = 2 a Vanishing (3. then the central charge of this model is b P 2ra ) = 5 − 2 a ra . . in a suitable set of coordinates. If not. 1) cohomology arises when wi 6 0. si = 0.31) ∂ti ∂ t¯i Y Whether this integral converges can.33) a We can classify the models that obey this condition. (2.29) is the condition that the singularity is at finite distance in moduli space. . Now let us look for vanishing cohomology of type (3. 1). . is thus X ra > 2. . singularities that can arise in the dynamics of a compact Calabi–Yau four-fold have no vanishing cohomology of type (4. . z3 ) = 0 is the equation of an ADE surface singularity. (2.28) is that b c < 3. which generalizes for an n-fold singularity to b c < n − 1. where D/Dti is the covariant derivative computed using the Gauss–Manin connection.S. . . .25) scales under za → λra za as λsi . 5 and superpotential F (Φ1 . The condition that wi > 0 for all i. by F (z1 . (2.33) thus amounts to 9 b c < 1. again. The integral (2. If the function Ai in (2. and in the limit tj → 0.30) is more conveniently written as Z ∂2 ¯ Ω ∧ Ω. 0). 1) cohomology is generated by Ωi = DΩ/Dti . . so that the integral in (2. .31) diverges near z = 0. a = 1. 9 Note that in terms of b c the condition that the local singularity of the fourfold be at finite distance in moduli space (2. 1) cohomology. so that there is no vanishing (3. z2 . . Φ5 ). / Nuclear Physics B 584 (2000) 69–108 79 We have already seen that convergence of the integral in (2. Gukov et al. The most dangerous case is for Ai = 1. (2. Consider a Landau–Ginzburg model with chiral superfields Φa .

there is only fourdimensional cohomology. R). For unbroken supersymmetry in flat space–time. with extraneous degrees of freedom decoupled. 2). the complex structure modes are all non-dynamical in this sense. we will examine a singularity of a different sort for which primitiveness of G is an important constraint. the condition that G should be primitive is automatically obeyed. one expects the i (X. For hypersurface singularities of complex dimension four. P If we work on the noncompact hypersurface. The large za divergence means that.32) does not hold for all i. The remaining constraint that we have not examined yet is a severe constraint in the case of hypersurface singularities. also ensures that there is such a divergence near za = ∞. then K ∧ G is automatically zero (if not zero. For an asymptotically conical metric on a manifold X. the condition that G be a harmonic L2 form is a mild one in the following sense. we compactified X to a complete but non-Calabi– Yau variety Y 0 by adding a divisor D at infinity. R) is all of there are no null vectors in Hcpct cpct H i (X. positivity of (2. This is certainly so for the vanishing cohomology. it would be an L2 harmonic six-form). that is to consider M-theory on R3 × X. they can be specified externally as part of the definition of the problem. It must. the condition that a ra + si > 2. instead. and the “primitive” cohomology in this situation is the cohomology that vanishes when restricted to D. so we expect L2 harmonic forms of degree four only. whose support is far from D. they vary quantum mechanically in the theory at the singularity. where K is the Kähler form. G must be a harmonic L2 form of type (2. in M-theory on R3 × X. This is the condition that G should be of type (2. 10 In Section 5. R) in space of L2 harmonic forms of degree i to be isomorphic to the image of Hcpct H i (X. In the ADE examples. / Nuclear Physics B 584 (2000) 69–108 Application to hypersurface We have developed this discussion for the case of a compact Calabi–Yau manifold Y that develops a hypersurface singularity. Gukov et al. R) in H i (X. D is an ample divisor. for singularities of this type. They correspond.80 S. be integral and “primitive”. the image of H i (X. so one expects that all of the four-dimensional cohomology is realized by L2 harmonic forms. where X is a hypersurface that develops the given singularity. This is the natural framework for studying M-theory near a singularity. Only those modes for which wi > 0 can be specified externally and represent coupling constants. to coupling constants of the theory near the singularity. If G is an L2 harmonic four-form on a manifold whose L2 harmonic forms are all four-forms. What about the requirement that G be primitive? Primitiveness means that K ∧ G = 0. For hypersurface singularities with asymptotically conical metrics of the type predicted by the Tian–Yau theorem. In Section 2. 2) [8]. For other examples. Thus. and therefore some of the complex structure deformations of X are dynamical. Now let us consider the Hodge type of the G-field in the hypersurface case. R). but it is more in the spirit of the present paper to decompactify Y and focus on the hypersurface itself. as well.2. Let us therefore now explain the significance of the above results for this case. the modes that deform the singularity of X have divergent kinetic energy and are not dynamical. Assuming 4 (X). . For 10 A different explanation of this is as follows. which R ensures that there is not a divergence of |Ωi |2 near za = 0.

) Here is another way to see the distinguished nature of the ADE singularities. The Kähler structure is parametrized classically by h1. Z). 2) is likely to play an important role in the dynamics of these models. in the supergravity approximation to M-theory.37) . (For fourfold examples where moduli are dynamically frozen see [11]. Variations of the Calabi–Yau metric of Y arise either from variations of the complex structure or variations of the Kähler structure. there is vanishing cohomology of types (3. 1) and (1. 2). Interpretation of constraints on G Since the constraints on G found in [8] have played an important role in this discussion. the vanishing cohomology is all of type (2. and (1. to avoid having to deal with the dynamical complex structure modes and the Hodge structure of the singularity. the expectation values of the Ti and Ki are arbitrary.4 = G1. which we promote to chiral superfields Ti . 2). 2. this R is not so. P so the L2 harmonic forms have this property. 2) is equivalent to positive definiteness of the intersection form on H 4 (X. (2. with a ra < 2.q is the dimension of the Hodge group H p. 2) and integral. Gukov et al. Since some of the complex structure modes are dynamical whenever there is vanishing (3. so again we see that the ADE singularities are the ones with vanishing cohomology that is entirely of type (2. We first suppose that G is zero. (Instantons can lift this degeneracy [12]. having the primitive cohomology be entirely of type (2. For other singularities.36) and the Kähler structure of X so that G ∧ K = 0. After picking an integral four-form G (which must be such that X G ∧ G > 0 or the equations we are about to write will have no solutions). and negative definite on the primitive cohomology of types (3. requiring that it be of type (2. Under such conditions. as we have seen above.3 = 0. we must adjust the complex structure of X so that G0. the intersection form on H 4 (X. For an intersection form specified by a bouquet of spheres to be positive definite is a condition that singles out the ADE Dynkin diagrams. we will pause here to attempt to gain a better understanding of these constraints. it is generically very hard to find a non-zero four-form that is of type (2. 2) will generally put restrictions on the complex structure of X. 2). If hp. Hence. (2.1 .S. The variations of the complex structure are parametrized classically by complex parameters ti .q (Y ). 1) cohomology. in particular. then the number of Ti is h3. Compactification of the C-field on Y gives rise to h1.1 chiral superfields that we may call Ki . Z) is positive definite on the primitive cohomology of type (2. If G = 0.1 real parameters ki . / Nuclear Physics B 584 (2000) 69–108 81 ADE singularities. 1). (2. Once an integral four-form G is picked. In this paper. we will study in detail only the ADE singularities.) For non-zero G. We consider compactification of M-theory on a compact four-fold Y . 3). 2). the restriction on complex structure that is entailed in making G be of type (2. As we explain in Appendix A. 3).1 U (1) gauge fields ai on R3 whose duals are scalars φi that combine with the ki to make h1.4.

38). We want to describe an effective action for the Ti and Kj that accounts for (2.q denotes the (p.37).38). (2. we look for superpotential interactions. Thus. then G2 − G1 = 2π[S].37).41) W X Here K is a complexified Kähler class whose real part is the ordinary Kähler class K. Also. . supersymmetric compactifications on R3 × Y were considered. we propose to let Ω be a holomorphic four-form on Y . (2.36). With W as in (2.36).36). a function of the Ti but a section of a line bundle over the moduli space M of complex structures on Y (on which the Ti are coordinates). (2. whose real part is (2.38) W (Ti ) = 2π Y This object is not. Let L be the line bundle over M whose fiber is the space of holomorphic four-forms on Y . one can readily guess the interaction responsible for (2.37): Z e (Ki ) = K ∧ K ∧ G.36). and this means [13] that W should be a section of L. So (2. q) part of G.42) should be the change in W in crossing the domain wall.37).3 = 0. we want a superpotential e (Ki ) that will account for (2. we may conjecture a generalization of the discussion of [8]. The tension of a domain wall obtained by wrapping a brane on a four-cycle S is the absolute value of R S Ω.) do not lift vacuum degeneracies. and a superpotential W To obtain (2. The Kähler form of M can be written Z ¯ ¯ Ω ∧ Ω. / Nuclear Physics B 584 (2000) 69–108 In (2. the linear dependence on Ω in (2. Since supersymmetric actions of the general kind d 4 θ (. so this integral is Z 1 Ω ∧ (G2 − G1 ). The e = 0 is K ∧ G = 0. Thus. the condition for unbroken supersymmetry in flat space is W = dW = 0.39) ω = −∂ ∂ln Y in other words it is ∂ ∂¯ ln |Ω|2 for Ω any section of L.40) 2π X In a theory with four supercharges.4 = 0. since the objects dΩ/dti generate H 3. In a similar spirit. the tension of a supersymmetric domain wall is the absolute value of the change in the superpotential W . More generally. In that paper. Gukov et al. If G changes from G1 to G2 in crossing the domain wall.38) is the right behavior of a superpotential. a supersymmetric compactification might have a (negative) cosmological constant in the . So we have found the supersymmetric interaction that accounts for (2. since it is proportional to the choice of Ω. (2. condition d W and imposes no further condition.82 S.1(Y ). W (Ti ) that will account for (2. . the condition dW = 0 is that G1. strictly speaking.36) is to consider supersymmetric domain walls. Gp. W e = 0 is a consequence of this. At this point. and take Z 1 Ω ∧ G. a statement that is clearly compatible with (2. In supergravity with four supercharges.36) R and (2. Another way to justify (2. the condition W = 0 is that G0.

Ti . Kj ) in section 3 by studying the soliton structure. it is interesting to compactify one of the directions in R3 on a circle and consider type IIA on R2 × Y . this is a very natural thing to do. such description of the theory would involve a superpotential function W e b with respect that the function W (Ti . This is demonstrated in Appendix B. Indeed.39). in view of mirror symmetry and other T-dualities. For fixed choices of Ti and Kj . since the superfields Ti and Kj are massless if G = 0. corresponding to the different choices of G. Ti . while other superfields are massive. Physical interpretation We will now discuss the physical interpretation of the superpotentials that we have computed. The condition for such an AdS compactification with unbroken supersymmetry is that W 6= 0 but DW/Dti = 0 (here. We will interpret the theory near the singularity as a theory of dynamical chiral fields Φα such that the critical points of the superpotential as a function of Φα are given by the possible choices of G-field. This is under investigation [15]. near a hypersurface singularity. The above analysis carries over immediately for supersymmetric vacua with a nonzero value of the Ramond– Ramond four-form G. In view of the form of (2. Kj ) = W (Ti ) + W (Kj ) is obtained by extremizing W to the Φα . in type IIA string theory. leading to a supersymmetric compactification on AdS3 × Y . / Nuclear Physics B 584 (2000) 69–108 83 noncompact dimensions. Gukov et al. the extremization with respect to Φα has different solutions. the mirror of G0. the relation of the change in the superpotential in crossing a domain wall to (2. and the mirror of G1. function W 2. most vacua will actually lead to AdS3 compactifications in the case of compact Y . with Y a Calabi–Yau four-fold. However. but there are close analogs for type IIA on R2 × Y and (if Y is . In M-theory on a compact Calabi–Yau four-fold Y .S. smooth Calabi–Yau four-fold. Going back to supersymmetric compactifications to R3 . there are different vacuum states in the theory at the singularity that are specified by different choices of the G-field.5. we must be careful and use the appropriate covariant derivative D/Dti ). a more complete b (Φα .4 = 0. but for suitable examples we will identify the superpotential b (Φα . We have computed the superpotential as a function of the superfields Ti and Kj with all other degrees of freedom integrated out. Thus. Analogs for type IIA and F-theory We have formulated the discussion so far in terms of M-theory on R3 × Y . and not just the four-form.3 would be the RR two-form. Hence. since W 6= 0. Going to a non-compact manifold has the effect of decoupling gravity and lets us avoid inducing a cosmological constant. For Y a large.3 = 0 but dropping the condition G0.4 would be the RR zero-form (responsible [14] for the massive deformation of type IIA supergravity). Kj ). However. one may guess that this should be done by keeping the condition G1. It is very difficult to see the superfields Φα explicitly. one naturally thinks that one should construct a more general effective superpotential to incorporate the possibility of turning on a full set of Ramond–Ramond fields.42) shows that W cannot vanish in all vacua. we have argued that as one approaches a singularity.

is impossible to obey if g and p are non-zero and integral. G is an element of the primitive cohomology of Y that is odd under the involution that acts as −1 on the elliptic fibers and trivially on the base. we have the following structure. p. Let Y be a four-fold that is elliptically fibered over a base B. as we have discussed briefly in Section 2. one would want to incorporate the Ramond–Ramond zeroform and two-form. H − → (cτ¯ + d)−1 H − .) If G is primitive. (Hi is a three-form on B with values in the one-dimensional cohomology of the fibers. Hence. Self-duality of G gives a relation between g and p which. One obvious change is that domain walls are now constructed from four-branes (with worldvolume R × S ⊂ R2 × Y ). 2. Gukov et al. then it is in particular self-dual (see Appendix A). In terms of a type IIB description. H R denote the NS and Ramond three-form field strengths. Let H NS . Then a four-form G on Y has at the level of cohomology an expansion X Hi ∧ θ i . In going to F-theory. We view τ as varying over B with monodromies around the loci of seven-branes by SL(2. in type IIA. the space-filling membranes that contribute to the formula (2. which are interpreted physically as the Neveu–Schwarz and Ramond–Ramond three-form field strengths of type IIB superstrings.84 S. Let θ i . Also. and 3 on B. Z) transformations τ→ aτ + b . cτ + d Under such transformations H + → (cτ + d)−1 H + . / Nuclear Physics B 584 (2000) 69–108 elliptically fibered) for F-theory on R4 × Y . Let B denote the base of F-theory “visible” to type IIB. alongside the Ramond–Ramond four-form.16) for the flux at infinity are replaced by space-filling fundamental strings. . while g and p are ordinary four. Also we need to discuss the F-theory analog of the G-field. Another obvious change is that. in the limit that the area of the fibers of Y → B is very small. i = 1. (2. be a basis of integral harmonic one-forms on the fibers. in the type IIA case. Consider H + = H R − τ H NS . the surviving part of G in the F-theory limit is contained in the Hi .and two-forms on B. the space-filling membranes that contribute to the flux Φ at infinity are replaced by space-filling threebranes. With g = p = 0. and let χ be an integral two-form generating the two-dimensional cohomology of the fibers. g and p must be integral. The analysis of [8] carries over to type IIA. H − = H R − τ¯ H NS . Our analysis of the vacuum structure also carries over readily to this case. 2.4. and Hi are respectively forms of degree 4. For G to be integral. with G now understood as the Ramond– Ramond four-form field.42) G=g+p∧χ + i where g.

and and one can regard L as a fairly generic deformation of a hypersurface singularity ∆ = 0. Hence. recall that one can interpret F-theory on R4 × Y in terms of type IIB on R4 × B with (p. where ∆ = 4f 3 (z1 . Then we identify H + = H. . / Nuclear Physics B 584 (2000) 69–108 85 A supersymmetric configuration in this context is obtained by choosing an integral (1. Actually. Φ =N + 2 4π τ2 B where N denotes the number of D3-branes. If we pick coordinates z1 . z2 . Gukov et al. In this case a given model is specified by fixing Z 1 1 H ∧ H¯ . 11 Singularities with f and g both zero are composite 7-branes of various types (the order of vanishing of discriminant would be bigger than 1). q) type of the sevenbrane is fixed. then the elliptic fibration over C3 can be described very explicitly by a Weierstrass equation for additional complex variables x. z2 . there is no monodromy around which the type of brane can change. Thus. q)-sevenbranes on a certain locus L ⊂ B. the full structure of (p. We are interested in a singular point of L at which 4f 3 + 27g 2 = 0 but f and g are not both zero. z3 ). and in particular simply-connected. This is a rather complicated structure in general. Moreover we require that H ∧ k = 0 where k denotes the Kähler class of B. z3 on C3 . For such cases the simplifications described in the text do not arise and the full structure of (p. y: y 2 = x 3 + f (z1 . (2.2 ⊗ L where L is a line bundle over B whose first chern class is c1 (L) = −12c1 (B). One reason for this is that near an isolated singularity. the detailed construction of ∆ in terms of f and g is irrelevant. q) sevenbranes is relevant.S. The (p. To describe the domain walls. Domain walls across which the Hi jump are described by a five-brane wrapped on R3 × V ⊂ R4 × B with V a three-cycle in B. z2 . q) sevenbranes also simplifies in this situation. The deformation of an isolated surface singularity is topologically a bouquet of two-spheres. z2 . H is a section of Ω 1. given by ∆ = 0. where L ⊂ C3 is developing a singularity. (2. 11 Near such a singular point. q) type of the five-brane varies as V wraps around the discriminant locus in B. the (p.44) A singular behavior of the elliptic fibration Y just corresponds in this language to a singularity of the hypersurface L ⊂ C3 . Alternatively. H − = H¯ . well defined modulo transformation by (cτ + d)−1 around the 7-branes.43) The fibers degenerate over a singular locus L which is the discriminant of the cubic. z2 . it simplifies considerably. one can replace B by C3 . but to study the local behavior near a singularity. the F-theory analog of a Calabi–Yau fourfold singularity is a more elementary-sounding problem: the study of a D7-brane in R10 = R4 × C3 whose worldvolume is R4 × L. 2) form on the base. z3 ). z3 ) + 27g 2 (z1 . H . and one can think of it (for example) as a D7-brane. z3 )x + g(z1 .

E6 . H H H H An−1 . for reasons that we have already explained. A D5-brane can end on a D7-brane. and as the notation suggests.47) . for both the surface L and the fourfold Y . whose first Chern class is an element of H 2 (L. In crossing such a domain wall. which is at x = y = 0). To obtain the ADE singularities. z2 . The D7brane supports a U (1) gauge field. Thus. where V is a three-manifold in C3 whose boundary lies in L. let us explain how they arise in the F-theory context. Poincaré cohomology class [∂V ].86 S. = z13 + z1 z23 + z32 . (2. The equation for the discriminant locus L ⊂ C3 reduces to H = 0 (plus higher order terms that are irrelevant near the singularity). the equation becomes y 2 = x 3 + 3ax 2 + H (z1 . z2 . = z13 + z24 + z32 .45) Here a is an arbitrary non-zero constant. 12 If there are null vectors in Γ . So the singularity of the elliptic four-fold is just the “stabilization” of the singularity L. and in crossing a domain wall it can jump by an element of a sublattice Γ . its boundary couples magnetically to the gauge field on the D7-brane. Z) → H 2 (L. the first Chern class jumps by the 2 (L. (2. let us describe the vacuum states and domain walls in this context. This group is a lattice Γ ∗ . / Nuclear Physics B 584 (2000) 69–108 Now. whose rank is the number of two-spheres in the bouquet. An example of an elliptic four-fold fibration Y with an isolated singularity is given by the following Weierstrass equation:  y 2 = x 3 − 3a 2x + H (z1 . z3 ) + 2a 3 . and 12 the natural map 2 (L. Z) gives an embedding of Γ in Γ ∗ . n 2 2 = z1 + z1 z2 + z3 . there is a slightly more elaborate story as mentioned in Section 2.2. Z). If we shift x to x + a.46) and this makes it obvious that the singularity of the elliptic fibration is obtained by “stabilizing” the surface singularity H = 0 by adding the quadratic terms 3a 2x 2 − y 2 (the x 3 term is irrelevant near the singularity. Γ is endowed with an integral quadratic form (the intersection pairing). E7 . Γ ∗ is the dual lattice of Γ with respect to this pairing. E8 . Gukov et al. Z) = Hcpct duality for noncompact manifolds asserts that Γ and Γ ∗ are dual. = z13 + z25 + z32 . Dn+1 . Thus we have a familiar i : Hcpct situation: the vacuum is determined by a point in a lattice Γ ∗ . and H is a quasihomogeneous polynomial describing a singularity in three variables at z1 = z2 = z3 = 0. z3 ). (2. Hence the domain walls across which the first Chern class jumps are built from fivebranes of topology R3 × V . Z). we need only select the appropriate H : H = z1n + z22 + z32 . which is an element of Γ = H2 (L. The ADE singularities will furnish important examples in the present paper.

that is C∞ and Φ. For a special case in which the role of C∞ has been analyzed (for Q = RP7 ) see [19]. This is the case of M-theory at a quasihomogeneous four-fold singularity (for the present discussion this need not be a hypersurface singularity) with a large value of the conserved quantity Φ that was introduced in Section 2. Gukov et al. so that after deforming to a smooth X. such that the total membrane charge (including the contribution of the C-field) is Φ. one natural question is whether a nontrivial conformal field theory arises in the infrared.S. or F-theory on a singular geometry. an affirmative answer to this question is strongly suggested by recent literature. There are at least massless modes associated with the motion of the membranes on X. Note that the vacua with N 6= 0 do not have a mass gap even after deforming to smooth X. one must set Φ to the smallest possible value for a given value of C∞ . This system is believed [16–18] to be described in the infrared by a conformal field theory that is dual to M-theory on AdS4 × Q. Φ must equal the minimum of Z 1 G∧G .48) Φ =N + 2 (2π)2 X We suppose that the four-fold X is a cone over a seven-manifold Q. / Nuclear Physics B 584 (2000) 69–108 87 2. (2. or. as it is usually called in two dimensions. the “massive” models just described. the soliton structure. the G-field apparently “disappears” at the singularity.2: Z G∧G 1 . and not the detailed way of decomposing Φ in terms of N and G as in (2.6. N will be zero for all vacua. for type IIA on a Calabi–Yau threefold near the . with a specified (flat) C-field at infinity that we call C∞ .48). To get a theory that after deformation of the parameters flows in the infrared to massive vacua only. the massive vacua are in correspondence with the choices of G that achieve the minimum. and with a very large number of membranes near the singularity. with a constant curvature (but topologically trivial) C-field on AdS4 that depends on Φ. and the decomposition of Φ into membrane and G-field terms is lost. deforming X to a smooth hypersurface. Conformal field theory: first results Given type IIA. Such corrections would come from appropriate instantons. In one situation. and a flat but topologically nontrivial C-field on Q that is equal to C∞ .49) 2 (2π)2 X with G running over the elements of the coset of Γ ∗ /Γ determined by C∞ . We recall that C∞ determines a coset in Γ ∗ /Γ . (2. When X develops a singularity. For example. To get a massive theory. we need to know that there are no quantum corrections to the classical geometry (which we will use to find the solitons). The analysis will be made by analyzing the domain wall structure. M-theory on R3 × X with X such a smooth hypersurface has vacua corresponding to all choices of N and G obeying (2. The AdS4 dual of this CFT depends only on what one can measure on Q. Our goal in the next two sections will be to analyze. M-theory.48). To justify the analysis. Consider M-theory on R3 × X. That decomposition arises if one makes a deformation of the theory. for the ADE singularities.

In our case. The basic strategy for identifying a supersymmetric theory based on its BPS soliton structure is the classification approach of [23] to N = 2 supersymmetric theories in two dimensions. we will propose a description in terms of an effective superpotential for a certain set of chiral superfields that generate the same soliton structure. respectively. as we have explained above. Likewise. i. in the F-theory. are absent in the examples we are considering. Such effects. we will look for supersymmetric domain walls using the classical geometry near the singularity. the Euclidean D2-brane instantons wrapped around the S3 in the conifold smooth out the singular classical geometry [20–22]. In fact. so there are no instantons. to analyze the small Φ theories. whose compact geometry consists of a bouquet of two-spheres. This is in fact the F-theory version of the description of the corrections to conifold geometry in type IIA compactification (and reduces to it upon compactification on T2 ). which in particular requires that the boundary of the instanton be a nontrivial compact cycle in L. / Nuclear Physics B 584 (2000) 69–108 conifold singularity. This is to be contrasted with the seemingly similar problem of F-theory on a Calabi–Yau threefold. which would require nontrivial three or six-cycles on X.or M5-branes. and in a few cases in three dimensions. their boundaries would be five. we are really studying. L is a complex curve. Gukov et al. a sevenbrane on L ⊂ C3 . For the instanton to affect the quantum moduli space it has to be BPS.88 S. instanton one-branes can and do modify the classical geometry. wrapped Euclidean M2. q) type).50) .and one-dimensional.. For example. Viewing them as instantons. For F-theory on a four-fold. if the singularity of the surface L is not isolated. Thus. the local geometry of the deformed singularity has nontrivial four-cycles only. In that case. Since the C3 has no nontrivial cycles. Thus there is no room for instantons. Consider a theory with N = 2 supersymmetry in two dimensions with k vacua. (2. L. In this case. has only two-cycles. so their tensions go to zero. Thus the classical singularity survives quantum corrections. So if L has no nontrivial compact fiveor one-dimensional cycles. For M-theory or type IIA near a four-fold hypersurface singularity. The domain walls are constructed from branes whose volumes vanish as the hypersurface X becomes singular. however. a similar statement holds. This description will make clear that one should expect flow to a nontrivial IR conformal field theory in the two-dimensional cases. and consider the integral k × k matrix S given by S = 1 − A. then it would generically also have nontrivial one-cycles and would thus receive corrections. a superpotential is generated by wrapped Euclidean fivebranes [12]. with nontrivial one-cycles. then there are no instantons. the relevant instantons will have to end on L. in M-theory compactifications on suitable Calabi–Yau four-folds. for the massive models derived from ADE singularities.e. In type IIB string theory the only possible candidate instantons which could end on a sevenbrane are fivebranes and onebranes (of appropriate (p. Thus one can reasonably hope to get a description in terms of an effective theory that contains only light degrees of freedom and generates these domain walls. and quantum corrections do not modify the singular classical geometry. in order to have finite action.

our analysis will identify the nonperturbative massless fields and superpotential near the singularity also in three and four dimensions.6. the superpotential we find may eventually be important in analyzing four dimensional CFT’s that arise from string theory. the description of the large charge behavior appears to be less simple. For the type IIA and F-theory compactifications. the relation above between the spectrum of the solitons and the charges of chiral fields is still a very powerful connection and in particular fixes the central charge of the corresponding conformal theory. −Pm (z1 ) + z22 + · · · + zn+1 . our task now is to analyze the soliton structure for certain hypersurface singularities. Geometry of domain walls As explained in Section 2. Above two dimensions. It was argued that this massive deformation comes from a CFT in the UV limit with central charge b c and k chiral fields with R-charges qi which satisfy    c) Eigenvalues S −t S = exp 2πi(qi − 12b (even the integral part of qi can be determined from Aij [23]). i. Instead of specializing to four-folds. we will consider the Ak singularities which were introduced in Section 2. it is believed to flow to a trivial IR theory for all n. but it is less common for a theory with a given superpotential to flow to a nontrivial IR conformal field theory. Also. Gukov et al. In fact.S. the solitons completely characterize the conformal theory. For instance. while in three dimensions this is expected only for n = 3 [25]. the domain wall or soliton analysis still identifies an effective superpotential. In any event.e. and in case of deformations of minimal models. Aij = Trij −solitons(−1)F F. so with some modification of our construction. The soliton analysis will give detailed information about the behavior for small membrane charge. and in four dimensions. which is the opposite limit from the AdS description discussed above that governs the large charge behavior at least for the M-theory compactifications. In other words. it proves insightful to consider the more general problem of identifying BPS states of wrapped n-branes in a Calabi–Yau n-fold near an isolated singularity. This is a strong restriction.2. a Φ n superpotential can become relevant as a perturbation to certain fixed points [26]. a theory with a single chiral superfield Φ and superpotential W = Φ n is believed to flow to a nontrivial CFT in two dimensions for all n > 2. To study the behavior near an Ak singularity. / Nuclear Physics B 584 (2000) 69–108 89 where 1 represents the identity matrix and A is a strictly upper triangular matrix whose Aij entry for i < j is the number of nearly massless BPS solitons interpolating between the ith sector and the j th sector weighted with the index (−1)F F [24]. 3. any theory which upon mass deformation has the same solitonic structure as that for a massive deformation of a minimal model is equivalent to it! For non-minimal models. even in four dimensions. we consider a local model for a Calabi–Yau n-fold given by 2 = 0..

is minimized and is given by Z V =α Ω C for some choice of phase α. Moreover. / Nuclear Physics B 584 (2000) 69–108 where Pm (z1 ) is a polynomial of degree m = k + 1 in z1 . A supersymmetric cycle is a Lagrangian submanifold (that is. Or stated equivalently. C which is the volume of cycle C. i. For a minimal supersymmetric cycle the quantity Z V = |Ω|. minimal supersymmetric cycles. The most singular geometry arises when Pm (z1 ) = z1m . the Kähler form vanishes on it). we get a singular geometry. For a particular choice of Pm . The intersection form on the compact homology is symmetric if n is even and antisymmetric if n is odd. Gukov et al.2.e. For a generic polynomial Pm (z1 ).90 S. we explained how to construct these spheres in Section 2.. on a minimal supersymmetric n-cycle the holomorphic n-form Ω of Calabi–Yau is real (with a suitably chosen overall phase) and gives the volume of the n-brane. When Pm has two equal roots. the geometry is not singular and the compact homology of this manifold has a basis made of m − 1 spheres of real dimension n intersecting each other according to the Am−1 Dynkin diagram. We would like to consider minimal wrapped n-branes. the condition is that . in this geometry.

Z .

Z .

.

.

.

|Ω| = .

Ω .

..

.

Suppose C is a supersymmetric minimal cycle. =q zn+1 z22 + · · · + zn2 − Pm (z1 ) We would like to minimize the volume form given by |Ω|. Note that the radius of this sphere is |Pm (z1 )| scaling that . The holomorphic n-form Ω. To count the minimal supersymmetric cycles. we follow the strategy in [27] and decompose the geometry to the “fiber and the base” as follows. the P manifold (being defined by j >1 zj2 = Pm (z1 )) has for its only nontrivial cycle a sphere 1/2 . Consider the image of C on z1 . C C which is the condition for minimizing the volume of C among the Lagrangian submanifolds in a given homology class. is given by Ω= idz1 · · · dzn dz1 · · · dzn . up to an overall complex scale factor. because for a fixed z1 . from which one can deduce by Sn−1 z1 . This is a one-dimensional subspace.

.

Z .

dz2 ∧ · · · ∧ dzn−1 .

.

.

(3.1) .

= |Pm (z1 )|(n−2)/2 .

.

.

The inverse image of C over a point in the z1 plane must. be a minimal cycle. zn Sn−1 z1 up to an irrelevant multiplicative constant. hence . and so must be (n − 1)-dimensional. if not empty.

The minimization of |Ω| will be done in two steps: We first consider Lagrangian submanifolds Cf (z1 ) for a fixed z1 which R minimize the Cf (z1 ) |Ω| and next consider the minimization of the volume interval over an interval I in z1 . / Nuclear Physics B 584 (2000) 69–108 91 the image of C in the z1 plane must be one-dimensional.S. In this way we get using (3. Gukov et al.1) .

Z Z .

.

Z Z .

n−2 .

.

|Ω| = .

Ω .

(3. = |Pm (z1 )| 2 dz1 .2) .

.

i.. I Cf (z1 )×I S n−1 (z1 ) I We now minimize the volume of the supersymmetric n-cycle with respect to the choice of the one-dimensional line segment I representing the image of the supersymmetric cycle on the z1 plane. . it gives rise to a well defined (meromorphic) one-form on a hyperelliptic cover of the z1 plane. Moreover. Indeed the topology of the cycle is an Sn which can be viewed as an Sn−1 sphere fibered over the interval. precisely if the line segment terminates at zeroes of Pm . In fact. three. (3. BPS domain walls in theories with four supercharges in dimensions two. the end-points of the segment in the z1 plane correspond to critical points in W . Strictly speaking we have constructed the supersymmetric cycle by assuming that the condition that the cycle Cf be Lagrangian is the same as being Lagrangian relative to the Kähler form induced on the fiber. (3. For example if the Kähler form has a piece of the form k = · · · + fi dz1 ∧ dzi∗ + · · · the condition of Lagrangian gets modified. where at the boundaries of the interval the radius of Sn−1 vanishes. like when the polynomial Pm has real coefficient one can use a Z2 antiholomorphic involution to argue that the cycles we constructed are both Lagrangian and supersymmetric. This is not necessarily true. and these are the only allowed boundaries. Let us define a function W with n−2 dW = Pm 2 dz1 . dW = 0 (for n = 2 the endpoints are defined by the condition that Pm (z1 ) = 0).3) corresponds to a well defined function of z1 . the condition for minimal volume is that the image of the curve in the W plane be a straight line along the direction specified by α −1 . One can allow the line segment to end at some special points on z1 where Pm (z1 ) = 0. These conditions are identical [28] for finding solitons in an N = 2 Landau–Ginzburg theory in two dimension (or more generally. If n is odd. In the more general case we proceed as follows: consider a generic Y Pm (z1 ) = (z1 − ai ).e. or four) with superpotential given by W ! If n is even. i Consider a one parameter family of Calabi–Yau metrics where ai (t) = tai .2) is minimized if along the segment on z1 plane the condition |Pm (z)| = αPm (z) is satisfied for some z-independent phase α. the D-brane worldvolume is closed and smooth. In special cases.3) In terms of W . The expression (3. branched over the zeroes of Pm (z1 ).

which are of course our main focus in the present paper. we see that in this case W = z1 . Gukov et al. Let us consider some special cases. There are m roots for Pm (z). From (3. we expect m(m − 1)/2 solitons (up to the choice of orientation) to complete the adjoint representation of U (1)m−1 to the SU(m) gauge multiplet. as was anticipated. We have already shown that the conditions for finding the solitons in this geometry are the same as those in an LG theory with the superpotential W . n = 3. Note that this gives m(m − 1)/2 solitons up to the choice of the orientation. However to construct the Kähler metric as a function of t we note that it can be mapped to the previous metric by defining e z1 = tz1 .3).92 S. Here dW can be viewed as a meromorphic one-form on a hyperelliptic Riemann surface over z1 branched over the roots of Pm (z1 ). the analogy becomes more precise: compactification on the four-fold leads to a theory in two dimensions with N = 2. the above geometry is the complex deformation of the Am−1 singularity.3) shows that W is a polynomial of degree m + 1 in the z1 plane. In this case W is defined by 1 dW = Pm (z1 ) 2 dz1 . the definition of W in (3. The geometry of these solitons for this class of conformal theories would correspond to straight lines on the Jacobian of this surface defined by the integrals dW and is presently under study [31]. The cases for a K3 surface and for a Calabi–Yau threefold have already been considered in [27] (see also [29. the type IIA on a deformed A1 singularity leads to an N = 2 theory with the same W for its . if we use our four-fold in type IIA superstring theory. Solitons for CY3 For the case of Calabi–Yau threefolds. and it is natural to identify the corresponding W with the superpotential of an N = 2 Landau– Ginzburg theory. In this way as t → ∞ the mixed terms in the Kähler form are dominated by the terms purely in the fiber direction (for n > 2 which is the case of main interest). and the solitons correspond to straight lines between the roots. however. Therefore in this limit the condition of Lagrangian submanifold in the fiber that we have used becomes accurate. and will be reviewed below. In this case. For any choice of the polynomial Pm (z1 ). We will indeed argue that for a certain choice of the membrane charge. / Nuclear Physics B 584 (2000) 69–108 Note that the BPS states will be the same for all t. Solitons for K3 In the case n = 2. for i 6= 1. Solitons for CY4 For the case of four-folds. because the condition of the BPS charges getting aligned does not change as we change t (the BPS charges only receive an overall rescaling). Thus we use the z variables but rescale the Kähler form accordingly. e zi = t n/2 zi .30]).

We denote that representation as Rk . where k = 1 and ξ is a weight of the fundamental representation of SU(m). In general the pre-image of a straight line connecting the images of critical points in the W plane will not connect the critical points in the z1 space. For k = 0. The number of choices of ξ is the dimension of Rk or m!/k!(m − k)!. there is only one vacuum (ξ = 0). as discussed in Section 2. . we have not attempted to check this correspondence explicitly. ξ is restricted to a fixed coset in Γ ∗ /Γ .6. note that even though we have m critical points. we have m vacua. It would be interesting to compare the formula for W that we have deduced from the soliton structure to the analysis of Section 2. . Let us consider the next simplest case. In this case. because both capture the mass of the BPS soliton. To find the degeneracy of the solitons between these vacua. m − 1. As explained in Section 2. or Zm . This is the number of vacuum states of the k t h model. though. we must fix the value C∞ of the C-field at infinity and also the flux Φ = N + 12 ξ 2 . For k = 0. we use the analysis of Section 3.2. This can be made very explicit in the case of the Am−1 singularity. Before we discuss these subtleties. so that N = 0 for all vacua. For some choices of Pm (z1 ) we do have exactly the same number of solitons as in the K3 case. ξ must be a weight of the k-fold antisymmetric tensor product of the fundamental representation of SU(m). (4. where Γ and ∗ Γ are the root and weight lattices of the Lie group SU(m). in order to specify the problem fully. Gukov et al. it is no longer true in general that we have m(m − 1)/2 solitons. For the theory to have a mass gap.4. / Nuclear Physics B 584 (2000) 69–108 93 superpotential. if Pm is such that the hypersurface X is smooth.1) where ξ = [G/2π] is the characteristic class of the C-field. . it is known that the number of BPS states jumps [24]. Although this is guaranteed to work. Identifications with Kazama–Suzuki models Let us consider in more detail type IIA strings propagating on a smooth hypersurface X obtained by deforming the Am−1 singularity: −Pm (z1 ) + z22 + z32 + z42 + z52 = 0. we find closely related Kazama–Suzuki coset models at level 1. For example. In fact as we change the polynomial Pm (z1 ).S. C∞ takes values in Γ ∗ /Γ . we set Φ equal to the minimum value of 12 ξ 2 (for all ξ in the given coset). then to minimize ξ 2 . For more general choices of the membrane charge. unlike the K3 case the image in the z1 plane is not a straight line. if C∞ = k. As we explained in Section 2. 4. it has been shown [28] that for Pm (z1 ) = z1m − µn for any constant µ. The coset in which ξ takes values is determined by C∞ . which is isomorphic to the center of SU(m). and the theory is trivial and massive. We found that the solitons . . there is one soliton for each pair of m critical points z1 = ωµ with ωm = 1.

. We now consider the other choices of C∞ . which we will call the k-fold symmetric combination of the k = 1 model. This model can also be viewed as an N = 2 Kazama–Suzuki coset model at level one. and so can be expressed as a polynomial in the elementary symmetric functions X zi1 · · · zil . we still get a description in terms of a chiral field with the same superpotential.2) The LG model with superpotential (4. . . but in most instances (the exception being Φ 3 in three dimensions). . xk ) = W (z1 ) + · · · + W (zk ). . a theory in three or four dimensions with a Φ m+1 superpotential is believed to flow to a free theory in the infrared. the two-dimensional theory with superpotential W is the Am minimal model [3. . xk . / Nuclear Physics B 584 (2000) 69–108 are exactly the same as those for an N = 2 LG theory with a chiral field Φ and a superpotential W obeying dW/dΦ = Pm (Φ). . we claim that the deformed LG superpotential is given by W (x1 . zk ) = z1m+1 + z2m+1 + · · · + zkm+1 . Consider the function of k variables W (z1 .94 S. of the form SU(m) .4]. In this case the soliton data are enough to determine the theory [23] as discussed at the end of Section 2. We argue that it has exactly the same solitonic spectrum as a deformation of the following LG theory. with ∂W = Pm . . SU(m − 1) × U (1) For the M-theory or F-theory near an Am−1 four-fold singularity. the condition for the existence of a soliton in the LG theory is exactly the same as the condition for a BPS wrapped fourbrane in type IIA near the Am−1 singularity. SU(m − k) × SU(k) × U (1) For the deformed singularity. Gukov et al. It is invariant under permutations of the zi . . and as we found in Section 3. . xk ) = W (z1 ) + · · · + W (zk ). . (4. In fact we can identify the m vacua with the m critical points of Pm . It is not too difficult to show [23. so that ξ is a weight of the k-fold antisymmetric product of the fundamental representation of SU(m) with some k > 1. .33] that the set of vacua of a LG theory made of a k-fold symmetric combination . Let us see why this LG superpotential has exactly the same solitonic spectrum that we get for type IIA at an Am−1 four-fold singularity with C∞ = k mod m.2) has been conjectured in [32] to be equivalent to the following Kazama–Suzuki coset model at level 1: SU(m) . . where again what we mean by this expression is that the superpotential is W (z1 ) + · · · + W (zk ) regarded as a polynomial in the elementary symmetric functions x1 . . . . So the type IIA theory near an Am−1 four-fold singularity is governed by the Am N = 2 minimal model. xl = i1 <···<il The superpotential we consider is thus W (x1 .

/ Nuclear Physics B 584 (2000) 69–108 95 of a given LG theory (in the sense introduced above) can be identified with the k-fold antisymmetric tensor product of the space of vacua of the original LG theory. getting W (x1 . In other words the solitons are in 1–1 correspondence with those roots of SU(m) that appear as solitons for the one-variable LG superpotential given by W . . The analysis of those cycles in Section 3 depends only on the geometry of the hypersurface and makes no reference to C∞ . . z2 ) = z12 + z22 in terms of the symmetric functions x1 = z1 + z2 . 13 As we already discussed. So if we label the vacua by |i1 . . as expected. z2 . x2 ) = x12 − 2x2 . This is the same result that we get from type IIA near the Am−1 singularity with C∞ = k mod m. z3 ) + z42 + z52 = 0 with H being an Am−1 singularity. As far as the allowed solitons. it is clear that soliton states are just the products of soliton states in the individual one-variable theories. and that irreducible solitons (which cannot break up into several widely separated mutually BPS solitons) are solitons in just one of the variables zi . all the roots appear with multiplicity 1. Then we write W (z1 . and Γ ∗ is the weight lattice of G. In the case W = zm+1 − az. Thus we can identify the vacua of the k-fold symmetric combination of the k = 1 model with the weights of the k-fold anti-symmetric tensor representation Rk . x2 = z1 z2 . with is denoting a vacuum in the sth one-particle theory. it suffices to consider the case that k = 2 and that each individual model has only one vacuum. to make possible a deformation to a massive theory. . then the allowed solitons only change one vacuum index at a time. Just as in the SU(m) case. To illustrate why. they can be constructed in the decoupled theory with superpotential W (z1 . . . Whichever roots appear act in the natural way on the weights of the representation Rk . Other ADE singularities So far we have mainly considered the local singularity to be H (z1 .1. we need to pick Φ so that ξ ranges over the weights of the smallest representation with a given 13 The main point that must be shown is that the vacua in the different factors must be distinct. This function has no critical points. 4. . The quotient Γ ∗ /Γ is isomorphic to the center of G. ik i. C∞ takes values in ∗ Γ /Γ . and its vacua correspond to the fundamental weights of the SU(m) lattice. So we start with m = 2: W (z) = z2 . zk ) = W (z1 ) + · · · + W (zk ) before re-expressing this in terms of the symmetric functions xi . This structure for the solitons of the deformed Kazama–Suzuki model was suggested in [34] where it was argued to correspond (with a specific choice of Kähler potential) to an integrable model. for this model. the k = 1 model has a one-variable superpotential W (z). So the solitons of a LG theory that is constructed as a k-fold symmetric combination of a one-variable theory are in natural correspondence with the solitons of the one-variable theory. Here we would like to generalize this to the case where H determines a D or E type singularity.S. the solitons are constructed by finding supersymmetric four-cycles. In this description. on the LG side. . so the combined model has no supersymmetric vacua. where Γ is the root lattice of the appropriate simply-connected ADE group G. . Hence the solitons for any k are in a natural sense the same as the solitons of the k = 1 model. Indeed. The general structure is quite like what we have seen for Am−1 . Gukov et al.

For D2n . Apart from the Grassmannians SU(m)/SU(k) × SU(m − k) × U (1) that we have already encountered.. for fundamental rep. there are two inequivalent choices. So E6 and E7 lead to one massive theory each. for spinor rep. / Nuclear Physics B 584 (2000) 69–108 nontrivial action of the center of G. . so we cannot use it to get a conformal theory with a massive deformation. . there is only the fundamental 27-dimensional representation (and its conjugate. . (If we pick the trivial action of the center. we will get the trivial representation and a massive free theory.. In the Dn case. SO(10) × U (1) E7 . So there are essentially two choices of C∞ leading to massive theories based on the Dn singularity. the dimension of the chiral ring is equal to the dimension of the corresponding representation of G. but they differ by an outer automorphism of D2n and give equivalent theories. it is known that for a level one G/H KS model. For the E6 theory. It is natural to conjecture that. 2. For the Dn case. 3. . the structure of the solitons for a special (integrable) deformation of the KS model at level 1 was studied in [34] and it was conjectured that the solitons exist precisely for each allowed single root acting on the corresponding weight diagram. E8 is simplyconnected with trivial center. This in turn is equal to the dimension of the cohomology of G/H and it was conjectured in [32] that the chiral ring is isomorphic to the cohomology ring. and for the E6 and E7 cases one has E6 . the relevant choices are the vector and spinor representations..). these Hermitian symmetric spaces are as follows. which gives an equivalent theory).96 S.) The appropriate representations are the representations with Dynkin label 1. U (n) for fundamental rep. given by SO(2n) . Gukov et al. Such a Hermitian symmetric space determines a series of N = 2 Kazama–Suzuki models (at level 1. The distinguished representations that we have described are in one-to-one correspondence with nodes of index 1 on the ADE Dynkin diagram and thence with Hermitian symmetric spaces G/H (where H is obtained by omitting the given node from the Dynkin diagram). SO(2n − 2) × SO(2) SO(2n) . Though we have not analyzed the BPS spectrum of the D4-branes in this case to find the multiplicity of the solitons for each 14 The higher level KS models do not generally admit an LG description. 14 Moreover. As a first check. as we have found for Am−1 . E6 × U (1) for fundamental rep. the massive models obtained from type IIA at an ADE singularity are the massive deformations of the corresponding level 1 Kazama–Suzuki (or KS) models. for E7 there is the fundamental 56dimensional representation. which in turn was shown to arise from the ring of an LG theory. there are two different spinor representations. Thus the G/H theories at level 1 were identified with specific LG models.

Gukov et al. and SU(2)R rotates the three complex structures on X. but will briefly analyze two cases in the present section. transforming as (3. To isolate the behavior near the singularity. This hyper-Kähler manifold is conveniently obtained by considering a U (1) gauge theory with eight supercharges. Hyper-Kähler singularities First we will consider what one might call hyper-Kähler singularities — singularities near which Y admits a hyper-Kähler structure. 2. In this description. but it may have a physical interpretation in terms of a membrane probe of the singularity. with the Ai transforming as the 3. it is natural to expect that at least for specific deformations. X is a cone over a seven-manifold Q described by X β Aiα A¯ i = δαβ .2) i 15 We will present this gauge theory as a formal device. 15 There is an SU(3) global symmetry group. There is also an SU(2) R-symmetry group. The bosonic parts of the Ai can be regarded as a complex field Aiα . broken if dE 6= 0 to SU(3) × U (1)R . (5. the cotangent bundle of complex projective two-space. then X has an SU(3) × SU(2)R symmetry. The moduli space X of zero energy states of the classical gauge theory is the space of solutions of (5.1) i with σE the traceless 2 × 2 Pauli matrices. just as in the Am−1 case. A Calabi–Yau four-fold can. however.1) divided by the action of the gauge theory. 3. and three hypermultiplets Ai . 5. develop singularities of many different types. i = 1. . 2) under SU(3) × SU(2)R . (5.1. and it is possible to add an SU(2) triplet of Fayet–Iliopoulos terms dE to the D-flatness equations. Other types of singularity The only four-fold singularities that we have so far considered in any detail are hypersurface singularities. 2. We cannot offer any sort of overview of the possibilities. it is manifest that if dE = 0. with X = T ∗ CP2 . 5. α = 1. In this case we would reproduce the solitonic structure anticipated in [34]. though Y may not be globally a hyperKähler manifold. The SU(3) preserves the hyper-Kähler structure. we replace Y by an asymptotically conical hyper-Kähler manifold X that is developing a singularity. An important fact here is that M-theory compactification on R3 × Y with Y hyper-Kähler has N = 3 supersymmetry in three dimensions. We will focus on a very concrete example. / Nuclear Physics B 584 (2000) 69–108 97 root. It is quite satisfying that we apparently get all the Hermitian symmetric space KS models at level 1 in such a uniform way from considering singularities of CY four-folds. If dE = 0. of charge 1. because the space of covariantly constant spinors on a hyper-Kähler eight-manifold is three-dimensional. the solitons are given by the root lattice of the corresponding group with multiplicity 1.S. The D-flatness condition is X β Aiα A¯ i = dE · σEαβ . A manifestly SU(2)-invariant way to exhibit the D-flatness equations is as follows.

5) with W = [CP2 ]. B i = Ai1 and Ci = A¯ i 2 . e−2iθ .98 S. Indeed. and with charges 1 and −1 under the U (1) gauge group. with dE splitting as a real component dR and complex component dC . Since X is contractible to CP2 .4) can be interpreted to mean that Ci lies in the cotangent space to CP2 . The cohomology with compact support is. That is because the center of SU(2)R is equivalent to a U (1) gauge transformation. eiθ . where the U (1) acts by right multiplication by  (5. The non-zero Betti numbers are h0 = h2 = h4 = 1. the self-intersection number W · W is equal to the Euler characteristic of W . The quantities B 2 by the gauge group. H 4 (X. in three space– time dimensions with N supercharges. With dC = 0. and SU(2)R acts by right multiplication by SU(3) elements that commute with (5. R) = H 4 (CP2 . The R-symmetry group that acts faithfully on X is actually SO(3)R = SU(2)R /Z2 . after dividing dR > 0. the dual of this. one can break up the bosonic part of the Ai into pairs of complex fields B i . Though turning on dE breaks the SO(3)R symmetry of X to SO(2).3) diag eiθ . / Nuclear Physics B 584 (2000) 69–108 It is fairly easy to see that this manifold is a copy of SU(3)/U (1). SU(3) acts on SU(3)/U (1) by left multiplication. In M-theory on R3 × X.3). The Euler characteristic of CP2 is 3. regarded as the zero section the B of T ∗ W . so in our example W ·W =3 (5. To get more insight. For any manifold W . at the point determined by ei . R) → H k (X. The appearance of an SO(3)R symmetry at dE = 0 is a hint that M-theory on R3 × X flows to a superconformal field theory in the infrared as dE → 0. the natural map k (X. the three space–time supersymmetries transform as a vector of SO(3)R . Gukov et al. (5. let us now analyze the possible G-fields on the smooth manifold X with dE 6= 0.) This description breaks SO(3)R to SO(2)R = U (1)R . Ci . just on dimensional grounds. X is equivalent to T ∗ CP2 . (Compared to the previous description. the second equation in (5. By an SO(3)R rotation. For k = 4. R). R) is zero except for k = 4. it preserves the hyperKähler structure and all of the supersymmetry of M-theory on R3 × X.4) i One must also divide by the actionpof U (1). define a point in CP . Now let us explain why for dE 6= 0. i X j B i Ci = dC . the superconformal algebra contains an SO(N )R R-symmetry group. Thus X is isomorphic to T ∗ CP2 . by Poincaré duality. transforming as 3 and 3¯ of an SU(3) symmetry group. so the non-zero Betti numbers with compact support are h4cpct = h6cpct = h8cpct . In terms of a description that makes manifest only half the supersymmetry of the gauge theory. one has H 4 (X. The D-flatness equations of the U (1) gauge theory are in this description X X |B i |2 − |Cj |2 = dR . one can set dC = 0 and P P ei = B i / dR + j |Cj |2 obey i |B ei |2 = 1 and. Hence. Even if dE 6= 0. R) is generated i : Hcpct cpct . X is asymptotic to a cone over Q at big distances.

(5. Z)/Hcpct 3 restriction of the C-field to ∂X = Q are determined by the value of k modulo three. Φ changes by an integer.8) is that if k is shifted by an integer multiple of 3 (the 3 is needed so as to leave fixed the restriction of G to Q). then α ∧ K.) In our situation. We do not know a Landau–Ginzburg or other semiclassical description for this N = 3 model with two vacua (but see below). with all vacua having N = 0. the membrane flux at infinity is Z (k + 12 )2 G∧G 1 . (There is a slightly more general formulation if X is not a complex manifold. for the following reason.7) and the fact that α · α = 3. preserves all of the supersymmetries. 16 so we need Z 1 G ∈Z+ . turning on a nonzero G-field. is hence (1 − c1 + c2 )(1 + c1 + c2 ) = 1 − c12 + 2c2 . (5. For sufficiently large Φ. proportional to α. The total Chern class of T ∗ CP2 .8) = N + Φ =N + 2 6 (2π)2 X In evaluating the integral. so the dual lattice H 4 (X. In fact. so in the present example we expect precisely one L2 harmonic form α. then α generates 4 Hcpct (X.7) 2π 3 4 (X. restricted R to CP ⊂ T CP . α is necessarily primitive with respect to all of the complex structures. Z) is generated by α/3. so c2 (T ∗ CP2 ) = −c12 + 2c2 . According to the discussion in Section 2. For an asymptotically conical manifold.S. If we normalize the four-form α to represent the class [CP2 ]. / Nuclear Physics B 584 (2000) 69–108 99 by the class [W ] = [CP2 ]. . since if K is any of the Kähler forms. This arises for Φ = 3/8. the standard conjectures would suggest 16 Let the total Chern class of the tangent bundle of CP2 be 1 + c + c . The different possibilities for the One also has H 4 (Q. with k ∈ Z. the claim follows. this model (at dE = 0) is expected to flow to a nontrivial superconformal field theory in the infrared. Z) = Z . α · α = 3. and the nonzero intersection number (5. we require    α G = k + 12 . we used (5. A check on (5.6) 2π 2 CP2 and in particular G cannot be zero. Z) = H 4 (X. the integral of c2 (X)/2 over CP2 is a half-integer. with k = 1 and k = −2.5) implies that i 6= 0. would be an L2 harmonic six-form. There is precisely one case of a model having more than one vacuum. the general flux quantization law for G is not that G/2π has integral periods but that the periods of G/2π coincide with the periods of c2 (X)/2 mod integers. if not zero. in dimension four. Z) (or rather its image in real cohomology). on a space–time X. we must turn on such a G-field. Hence.8) is obeyed. In the presence of N membranes and a G-field. According to [5]. The total Chern class of the cotangent 1 2 2 ∗ 2 bundle of CP2 is then 1 − c1 + c2 . Hence. Gukov et al. Indeed. one expects the space of L2 harmonic forms to coincide with the image of i. which is odd. A supersymmetric vacuum is then found by finding a nonnegative N and an integer k in the given mod 3 coset such that (5. (5. Also. Since CP2 c12 = 9. a model is specified by fixing the value of Φ and also by fixing the value of k modulo 3.

100 S. one finds that there is one blow-up mode and no complex structure deformation. and consider the Z4 symmetry za → iza . determined by the C-field. In the spirit of [17. the “zero section” of L. We expect that instead of T ∗ CP2 . Gukov et al. we would like to find an effective action for N membranes probing the R3 × T ∗ CP2 singularity (in the limit that CP2 is “blown down”) that will give a gauge theory dual of M-theory on AdS4 × Q. with the C-field on Q being determined by the value of k modulo three. For a study of their dynamics in the abelian case. z4 . in which it is replaced by the total space W of a line bundle L = O(−4) over CP3 .37]) suggests that the sum over multiple covers of CP3 will produce a pole at V = 0. asymptotic in closed form to the flat metric on C4 /Z4 . (Gauge theories with Chern–Simons interactions are essentially the only known classical field theories in three space–time dimensions without gravity with N = 3 supersymmetry. However. / Nuclear Physics B 584 (2000) 69–108 that the SCFT in question is dual to M-theory on AdS3 × Q. If this is so. fivebrane wrapping on CP3 will produce a superpotential [12].2. one could in a similar way analyze T ∗ F . analogy with other multiple cover formulas (such as the formula for multiple covers by fundamental strings [36. since the Hodge numbers hi. . It is natural to wonder whether.) Such a model might have roughly the right properties. with complex coordinates z1 . Blowup of orbifold singularity The other kind of singularity that we will briefly examine is a simple orbifold singularity. We begin with C4 . in fact. leading to a blow-down of the exceptional divisor CP3 ⊂ W . CP3 is embedded in W as an exceptional divisor. interaction. see [35]. though the multiple cover formula for multiple fivebrane wrapping in M-theory is not known. proportional roughly to e−V with V the volume of CP3 . . 18]. Consider an N = 4 supersymmetric gauge theory in three dimensions with gauge group S(U (N) × U (N)) and hypermultiplets ¯ Break N = 4 to N = 3 with some Chern–Simons consisting of three copies of (N. 5. If one analyzes this type of orbifold in type IIA string theory. N). One can also consider a collection of intersecting CP2 ’s (with a suitable normal bundle over it) and carry out a similar analysis. where F is a two-dimensional Fano surface. this gauge theory can be interpreted physically as the long wavelength theory of a membrane probe of the R3 ×T ∗ CP2 solution of M-theory. though we will not do so here. .0 (CP3 ) are zero for i > 0. One might at first think that one could approach the C4 /Z4 orbifold singularity in M-theory by a motion in Kähler moduli space. The quotient C4 /Z4 is a Calabi–Yau orbifold. because of the SU(4) symmetry of W . Moreover. Thus. it is actually possible to describe this metric by quadrature. . The blow-up corresponds to a very simple resolution of the singularity. such a description might be roughly as follows. We have no good way at present to determine if the model flows to a nontrivial SCFT also for small Φ. there will not be interesting long distance physics associated with the behavior of M-theory near a . More generally. W admits a Calabi–Yau metric. Physical interpretation of gauge theory? So far the U (1) gauge theory with three charged hypermultiplets has been considered just as a mathematical device.

W is contractible to CP3 . Let us raise the general question of this sort. Given this. given the asymptotically conical nature of the Calabi–Yau metric on W . then since the G-field must vanish in cohomology on a fivebrane worldvolume (because of the existence of a field T on the fivebrane with dT = G). 4. there is only one option for how the SU(2) group R that acts on the cohomology of a Kähler manifold (see Appendix A) can act on the L2 harmonic forms on W : they transform with spin 1. in the following interesting way. one certainly cannot expect to study M-theory on C4 /Z4 while ignoring the superpotential. R) → H k (W . We already explained in Section 2. though there is an L2 harmonic four-form on W . in the presence of the G-field the superpotential would be absent. using the fact that c1 (L)3 |W 6= 0. it follows that M-theory or type IIA at a four-fold singularity can be described by type IIA with a singular sixbrane R3 × L. with worldvolume Rn × Xp+1−n embedded in R10 or R11 depending on whether we are dealing with string theory or M-theory. 4. however. where L ⊂ C3 is a singular complex surface. The answer to the question of whether a supersymmetric G-field is possible turns out. Brane perspective We will conclude this paper by pointing out a reinterpretation of the problem in terms of singularities of branes. and the question of the behavior near the C4 /Z4 singularity would be restored. so its nonzero Betti numbers are h0 = h2 = h4 = h6 = 1. shows that this is so. / Nuclear Physics B 584 (2000) 69–108 101 C4 /Z4 singularity. Hence. with one class each in dimension 2. one has the dual Betti numbers h2cpct = h4cpct = h6cpct = h8cpct = 1. For cohomology with compact support. it is not primitive. At any rate. These give models for studying type II strings at a Calabi–Yau threefold singularity capturing Argyres–Douglas points of N = 2 conformal theories [38] and are presently under study [31]. Suppose we have a p-brane of some kind.S. there are no corrections to the classical Rn × L geometry. say of the form y 2 = x n locally (for n > 2).6 that F-theory on a four-fold singularity can be reinterpreted as type IIB with a D7-brane that has a worldvolume R4 ×L. 6. and one cannot turn on a G-field without breaking supersymmetry. it is important that in the cases that we have looked at. we expect the space of L2 harmonic forms on W to be three-dimensional. Is it possible to include a G-field on W while preserving supersymmetry? If so. Analogous phenomena have been noted in the past in the context of N = 2 conformal theories with NS or M5-branes worldvolumes with singular geometry R4 × Σ where Σ is a Riemann surface which develops a singularity. First of all. By successive circle compactifications. This suggests that k (W . to be “no”. As we discussed in Section 2. Consequently. R) may be nonzero for k = 2. and 6. We assume that this geometry preserves some number of supersymmetries in Rn . Is this singular geometry smoothed out in the . A topological the map i : Hcpct analysis. or type IIB with a singular fivebrane R2 × L. Gukov et al. and 6. Let us assume X develops a singularity.6.

(A.1) And the raising operator J+ is the adjoint operation of contraction with K: Gi1 i2 . the four supercharges are are the ∂ and ∂¯ operators and their adjoints. then they are all automatically primitive. So if q-branes can end on this particular p-brane. Appendix A. the SU(2) arises as an R-symmetry group. The space of . this SU(2) action arises as follows.2) Conceptually. For M5-branes. Since an (n − p)-form has J3 eigenvalue (n − p)/2. such as the G-field on a Calabi–Yau four-fold.) A diagonal generator J3 of SU(2) multiplies an (n − p)-form by (n − p)/2. The middle-dimensional cohomology of W is two-dimensional and can be decomposed as follows. R) consists precisely of the R-invariants. and there is an SU(2) symmetry that comes from rotations of the three extra dimensions. For a noncompact Kähler manifold X. For n − p > 0. (A. or 0 = J+ G = K ∧ G. with p = n. so we denote it as R. For a middle-dimensional L2 harmonic form G. and K ∧ G = 0 says that G has support disjoint from H . The lowering operator J− acts by wedge product with the Kähler form K: G → K ∧ G. pp. R) to consist of the harmonic forms that transform with spin precisely (n − p)/2. (See [39].in . For a middle-dimensional L2 harmonic form that is not R-invariant can be raised and lowered to make L2 harmonic forms of other dimensions. this definition means that the primitive part of H n (X. if all of the L2 harmonic forms are in the middle dimension. Gukov et al.102 S. the condition is the absence of compact two-cycles in X. or the space of L2 harmonic forms on any Kähler manifold.. Primitive forms The de Rham cohomology of a compact Kähler manifold X. we declare the primitive part of H n−p (X.. where n is the complex dimension of X. For a middledimensional form. From this point of view. since D(p − 2)branes and fundamental one-branes can end on them. then K is cohomologous to a form supported on H . / Nuclear Physics B 584 (2000) 69–108 quantum theory? A necessary condition for that is the existence of instantons which end on X. it clearly transforms under R with spin at least |n − p|/2..in → K i1 i2 Gi1 i2 . which is the condition on G given in [8]. and dimensionally reduce it to 0 + 1 dimensions. An important illustrative case is that of a complex surface W . the condition is the absence of compact q-cycles in X. This gives a supersymmetric system in which the Hilbert space is the space of differential forms on X. for a mathematical introduction. the condition is the absence of topologically nontrivial compact one-cycles and p − 2-cycles in the geometry of X. 122–126.. primitiveness is equivalent to either 0 = J− G. admits an SU(2) action which is as follows. Begin with the supersymmetric nonlinear sigma model in four dimensions with target space X. If there is a projective embedding such that K is the class of a hyperplane section H . So for Dp-branes in type IIA or IIB string theory.

the middle-dimensional primitive cohomology is of type (2. it suffices to consider the case that X = W1 × W2 . Since the αi are anti-self-dual. This type of argument can clearly be generalized to other dimensions. We can similarly analyze the primitive (3. Indeed. (3. In Section 2. . those that transform as 15 have R = 1. 1) and are of the form α = ai j¯ dzi ∧ d z¯ j¯ where ai j¯ is traceless. we learn that the primitive (3. with the Wi complex surfaces. 1) cohomology of a four-fold is anti-self-dual. 2)-form G = α1 ∧ α2 . In sum. the primitive (2. they again all have the same eigenvalue of ∗. with α a primitive (1. q) form by p − q. by considering the case G = α ∧ β. αi is a primitive (1. From this. The (2. 2) forms and the multiples of the Kähler class K. while those that transform as 6 have R = 1. The (1. the (4. while the (3. 1)-forms that transform as 10 of SU(4) have R = 0 and so are primitive. Tracelessness of a means that α is annihilated by the raising and lowering operators and so is primitive. p) forms for p = 0. 1) and (1. G is self-dual. since they are closed under the action of R. the (2. the Hodge ∗ operator acts on the primitive (p. 0) cohomology of a four-fold transforms trivially under SU(4) and is primitive. Four-fold compactifications to AdS3 The purpose of this appendix is to extend the analysis in [8] of supersymmetric compactification to R3 × Y with a G-field to the case of AdS3 × Y . Since ¯ ¯ h¯ ¯ d z¯ i ∧ d z¯ j transforms as the 6. From this. and to consider on X the primitive (2. Finally. with 15 the adjoint representation. we focus on the SU(4) action. we learn that the (4. . To determine the sign. 1)-forms transform irreducibly under SU(4). By setting G = β1 ∧ β2 with βi a (2. 2. 1)-form on Wi . The (2. we . 1) cohomology. In particular. At any point P ∈ X. it follows that (2. 2. 2)-form hij dzi ∧ dzj or e ij transform as 6 ⊗ 6 = 1 ⊕ 15 ⊕ 20. Since the primitive (3. 0)-forms transform as 6 under SU(4). 2) forms are clearly primitive (the lowering operator would map them to (3. it follows that they all transform with the same eigenvalue under the Hodge ∗ operator. We look first at the (p. 0) and (0. or of type (1. Since the generator of the center of U (4) simply multiplies a (p. 2) and self-dual. / Nuclear Physics B 584 (2000) 69–108 103 self-dual forms at a given point is three-dimensional. 0)-form on another surface W2 . 3)-forms) but the Kähler class K is not (as K ∧ K 6= 0). The anti-self dual two-forms are of type (1. The (0. 1)-forms transform as 6 ⊕ 10. 0) forms transform in the trivial representation 1 of SU(4). 1) and anti-self-dual. 2)-forms transform in an irreducible representation of SU(4).4. 0) form on Wi for i = 1. 0) or (0. 1)-form on a surface W1 and β a primitive (2. Gukov et al. Closer to our needs in this paper is the case of a complex four-fold X. 2)-forms a (2. Appendix B. 0) cohomology on a four-fold is self-dual.S. 1)¯ forms ai j¯ dzi ∧ dzj transform as 4 ⊗ 4 = 1 ⊕ 15. 4 − p) cohomology of a four-fold as (−1)p . 2. So for a complex surface. 2)-forms that transform as 20 of SU(4) have R = 0. 0) and (0. . and those that transform as 1 have R = 2. . 0) or (0. 1. the holonomy group U (4) acts on the differential forms at P . the self-dual forms are the (2. where for i = 1.

and accordingly here we will find that it determines the AdS3 cosmological constant. 2 Let us take the following ansatz for the metric:  2 = ∆−1 ds32 (x µ ) + ds82 (x m ) ds11 (B.104 S. Below we show that these two functions are related by the supersymmetry conditions that we are going to formulate in a moment. 2. N . Here we introduced the warp factor ∆(x m ) and a scalar function f (x m ).4). . the form of which will be fixed by the field equation and supersymmetry conditions. Since ψM vanishes in the background. . lower case letters a. . (B. n. . and Greek letters µ. both of which depend only on the coordinates on X. The field equation for G that follows from the action (B. stand for the three-dimensional Lorentzian indices 0. .1) S11 = 2 2 2 6 The eight-form anomaly polynomial can be expressed in terms of the Riemann tensor [40]:    1 1 1 2 2 4 trR trR + − . we only have to check that the variations of the gravitino vanish for some Majorana spinor η: 1 δψM ≡ ∇M η − ΓM N ∂N (log ∆)η 4  1 3/2 P QRS ∆ Γ (B.2) I8 = (2π)4 768 192 In these units the five-brane tension T6 = 1/(2π)3 . Gukov et al. run from 0 to 10 and denote eleven-dimensional indices. . Now we examine conditions for the configuration described above to be supersymmetric. ν. . Finally. are real indices tangent to X.1) looks like: 1 d ∗ G = − G ∧ G + (2π)4 I8 . / Nuclear Physics B 584 (2000) 69–108 argued that the (4. corrected by the σ -model anomaly on the five-brane worldvolume. 1. . We follow the notations of [8] where capital letters M.4) and for the three-dimensional components of G: Gµνρm = µνρ ∂m f (x m ).5) This compactification leads to the maximally symmetric three-dimensional space–time. We also allow for arbitrary internal components Gmnpq . has the following form:  Z Z  √ 1 1 1 1 d 11 x −gR − G ∧ ∗G + C ∧ G ∧ G − (2π)4 C ∧ I8 . as well as the gravitino ψM .3) (B. The bosonic part of the eleven-dimensional effective action. . . from the beginning of the alphabet denote holomorphic indices tangent to X. . − 288 where the first two terms come from the covariant derivative in the eleven-dimensional metric (B. b. We assume that the other components of G. . (B. . (B. 0) part of G is the superpotential. . vanish. .6) ΓM P QRS − 8δM GP QRS η = 0. m.

we obtain the additional equation  1  ∂m ∆−3/2  mnpqrst u Grst u.10). (B. .12) − 288 6 From the decomposition (B.15) ∂m ∆−3/2 Gmnpq = 4! . Without loss of generality we can take: γ9 ξ = ξ. we obtain the solution to (B. . (B.Mn = Γ[M1 .9) and the equations (B. Γm = 1 ⊗ γm . (B. as follows from the field equation (B.6) takes the form:  1 δψµ ≡ ∇µ η − ∂n (log ∆) γµ ⊗ γ9 γ n η 4   1 1 3/2 ∆ γµ ⊗ γ9 γ mnpq Gmnpq η + ∆3/2 (∂m f ) γµ ⊗ γ m η = 0..11) determines whether it is space-filling membranes or space-filling antimembranes that can be included without breaking supersymmetry. If the sign is changed. (B. threeHere we have written G dimensional components of the four-form field strength: Gµνρm = µνρ ∂m ∆−3/2 (B.11) Here γ9 is the eight-dimensional chirality operator that anti-commutes with all the γm ’s and satisfies γ92 = 1.3). . The sign in (B.10) ∇µ  = γµ . They satisfy: {ΓM . ΓN } = 2gMN .14) into (B.12): f = ∆−3/2 .13) / for the total contraction Gmnpq γ mnpq . Substituting (B. . Gukov et al.8) for the antisymmetrized product of gamma-matrices. 144Λ∆−3/2ξ = Gξ. (B. 2 and ξ is a commuting eight-dimensional complex spinor of definite chirality.. the corresponding supersymmetric vacuum (with the same cosmological constant Λ) can be obtained from that with γ9 ξ = +ξ by changing the sign of the function f and the chirality of . The µ-component of the supersymmetry condition (B. where the eleven-dimensional gamma-matrices Γ M are hermitian for M = 1. we make the 11 = 3 + 8 split: Γµ = γµ ⊗ γ9 . 10 and anti-hermitian for M = 0. . / (B. . / Nuclear Physics B 584 (2000) 69–108 105 Following [8]. In particular. ΓMn ] (B.9) where  is an anti-commuting Killing spinor in three dimensions: Λ (B.14) have the form similar to the membrane solution with the “effective” membrane charge density 12 G2 − (2π)4 I8 .7) We use the standard notation ΓM1 . We decompose the supersymmetry parameter as: η =  ⊗ ξ + ∗ ⊗ ξ ∗.11).3) for the internal components.S. (B.

To avoid cluttering we omit tilde from the notation and. if we choose m to be an anti-holomorphic index and use (B.19). e (B. = 0. 1). e γ a and e ea ξ = 0. and eme ξ− ∇ use it to define an almost complex structure: ξ. By means of the rescaling transformations: δψm ≡ ∇m ξ − gmn . we find that the (1. {γ a .16) can be written in the compact form: Λ 3/4 1 ∆ e γme γ npqe ξ + ∆−3/4Gmnpq e ξ + c. Now we return to the supersymmetry condition (B.6) and consider its m-component: Λ 1 γm ξ + ∆3/2γ npq Gmnpq ξ 2 24 3 1 (B.. (B. gmn → gmn = ∆3/2e ξ ξ → ξ = ∆−1/4e the equation (B.16) + ∂m (log ∆)ξ − ∂n (log ∆)γm n ξ + c. 3) piece of the field G must be zero: Ga¯ b¯cd ¯ = 0. γ b = 0. [8]).g. . γ = 2g a b . we omit it in the present discussion.17) 2 24 Then. / Nuclear Physics B 584 (2000) 69–108 which uniquely determines G given its cohomology class. it is convenient to think of γ a¯ and γ a as creation and annihilation operators that satisfy the algebra:   a b¯ ¯ ¯ γ . Furthermore. finally. Because the metric is of type (1.18) which is actually integrable. Gukov et al. 4 8 where we used the explicit form of the solution (B. obtain: 12Λ∆3/2γ a γm ξ − Gmnpq γ a γ npq ξ + c. For example.c. following [8]. we multiply the differential equation (B.106 S.19) To obtain the algebraic constraints on the field G. 17 Hence X is a complex manifold. we choose e ξ to be a covariantly constant spinor of unit norm. γa¯ act on the Fock “vacuum” ξ as annihilation operators: Namely.20) Components of this equation with different gamma-matrix structure must vanish separately. = 0.17) by γ a which kills the first term in that equation.13) and standard properties of gammamatrices. γ b } = γ a¯ .c.c. 17 Since the proof is very standard (see. (B. because Je is covariantly constant: ep Jemn = 0 ∇ the four-fold X is a Kähler manifold with ga b¯ Jea b¯ = ie being a Kähler form. = 0. ξ †e γmne Jemn = ie (B. and is identically obeyed if G is self-dual (as we will find) and closed. γ γ a¯ ξ = 0. e.

taking the trace over the holomorphic index ‘a’ in the main supersymmetry condition (B. Kazama. must be primitive: cd ¯ = 0. A 4 (1989) 325. Algebraic geometry and effective Lagrangians. Phys. The work of S. Sethi. [6] S. E. N. B 451 (1995) 109. M. Zamolodchikov. Warner. Inv. Strominger.2) piece of G. H. [8] K. Gukov. C. Witten.4. Candela. T. Morrison. Math. Phys. [7] B.2) . D. E. [3] C.G. Hitchin. Gukov et al. we get the relation: 96Λ∆3/2ξ = Gabcd γ abcd ξ ∗ . Lett. Nucl. [4] E. M-theory on eight manifolds.P. 22 (1997) 1. [13] J. Greene. was supported in part by NSF grant PHY-9513835. Tian. A. Quantization of Newton’s constant in certain supergravity theories. [16] A. Pantev. Yau. Phys. / Nuclear Physics B 584 (2000) 69–108 107 And G(2. Phys. Kehagias. Yau. B 355 (1991) 455. to appear. Bagger. B 115 (1982) 202. 106 (1991) 27. Moduli space of Calabi–Yau manifolds. Constraints on low-dimensional string compactifications. X. Nucl. Katz. Non-perturbative superpotentials in string theory. Nucl. Seiberg. [5] E. Phys. Lerche. On flux quantization in M-theory and the effective action. [15] S. [14] J. Black hole condensation and the unification of string vacua. MacPherson. Acknowledgements We have benefited from discussions with M. Phys. D. Nucl. which is natural given the discussion in Section 2.V. B 218 (1989) 51. [12] E.20). Lett.S. Witten. JHEP 9711 (1997) 004.-T.21) Finally. References [1] Y. Phys. Nucl. Becker. B 477 (1996) 155.0 . [2] A. Fayet–Iliopoulos potentials from four-folds. 46 (1987) 160. J. (B. Phys. Shapere. Witten. Catastrophes and the classification of conformal theories. Becker. Lett. Suzuki. B 480 (1996) 213. and S. de la Ossa. Phys. Bershadsky. A. B 474 (1996) 343. was supported in part by grant RFBR No 98-02-16575 and Russian President’s grant No 96-15-96939. Strominger. Morrison. Vafa. [10] P. the (2. Witten. New vacua for type II string theory. [9] G. New N = 2 superconformal field theories and superstring compactification. Geom. .C. Phys. [11] W. was supported in part by NSF grant PHY-98-02709 and that of E. Vafa. JETP Lett. S. Phys. Martinec. B 388 (1996) 736. Lett.B. Gab ¯ cd ¯ J (B. A. N. The work of C. hep-th/9805131. R. Complete Kähler manifolds with zero Ricci curvature II. N.-T. and shows that Λ is proportional to G4. New type IIB vacua and their F-theory interpretation.W.22) This extends the corresponding formula (2. B 217 (1989) 431. Polchinski.45) in [8] to compactifications with non-zero cosmological constant. S. Mod. Lett.

Becker. Vafa. K. B 411 (1997) 274. hep-th/9810201. Phys. M. On classification of N = 2 supersymmetric theories. Cecotti. membranes and nonperturbative string theory. S. Rev. P. 77 (1996) 3296. Morrison. P. Lett. Phys. C. [29] J. Strassler.P. W. [24] S. Kapustin. 151 (1993) 245. Phys. [19] S. Nucl. Argyres. M. Shenker. Commun. [35] A. Phys. Green. A new supersymmetric index. Phys. Polytopes and solitons in integrable N = 2 supersymmetric Landau– Ginzburge theories. Vafa. S. Lett.J. Nucl. Fusion residues. 158 (1993) 569. Lett. Douglas.S. work in progress. B 358 (1991) 571. Math. B 498 (1997) 101.P. [23] S. Intriligator. Phys. Witten. Nucl. Sethi. Lett. Aharony. Hanany. A. Nucl. [25] O. B 477 (1996) 746. Lerche. Cecotti. Commun. Phys. Nucl. Candelas. Rabin. J. Plesser. C. A 6 (1991) 3543. A relation between N = 8 gauge theories in three dimensions.Vafa. hep-th/9809162. 243 (1990) 257. Superconformal field theory on three-branes at a Calabi–Yau singularity. [38] P. N. M. Phys. Warner. / Nuclear Physics B 584 (2000) 69–108 [17] I. [39] P. N. Vafa. Wiley Interscience. Morrison. Parkes. E. X. Gukov et al. [40] L. Seiberg. . 386 (1992) 405. JHEP 9811 (1998) 003. P. Vafa.P. Nucl. [28] P. Intriligator. [32] W. Topological field theory and rational curves. Warner. C. Phys. Seiberg. Warner. A pair of Calabi–Yau manifolds as an exactly soluble superconformal theory. Aspects of N = 2 supersymmetric gauge theories in three-dimensions. R. Phys. Intriligator. D. [34] W. Phys. [22] N. 1978. B 448 (1995) 93. Alvarez-Gaumé. Nucl. Lett. [31] A. Strominger. B 536 (1998) 199. Nucl. Witten. Five-brane. Mathur. Phys. New phenomena in SU(3) supersymmetric gauge theory. E. Fendley. [33] K. B 499 (1997) 67. hep-th/9902033. De la Ossa. B 456 (1995) 130. Lerche. Lerche. B 324 (1989) 427. K. Fendley.108 S. Phys. Vafa. N.P.C. Math. Phys. Harris.M. A comment on duality in N = 1 supersymmetric non-abelian gauge theories. Nucl. C. Strassler. N. Nonspherical horizons. Shapere. JHEP 9904 (1999) 021. Becker. B 351 (1995) 230. Principles of Algebraic Geometry. [30] J. B 234 (1983) 269. A. [37] P.D. Phys. Integrable deformations and scattering matrices for the N = 2 supersymmetric discrete series. [18] D. Phys. B 359 (1991) 21. Self-dual strings and N = 2 supersymmetric field theory. Phys. C. M. L. [21] H. [26] D. Phys. Mayr. [27] A. N. Schulze. Summing up D-instantons. N. BPS geodesics in N = 2 supersymmetric Yang–Mills theory. Hypermultiplet moduli space and string compactification to threedimensions. B 388 (1996) 521. Klebanov. Aspinwall.R. Warner. [20] K. Griffiths. Nucl.C. Chiral rings in N = 2 superconformal theories. Nucl. Mod. C. Klemm. Lett. [36] P. Vafa. On mirror symmetry in three-dimensional Abelian gauge theories. Warner. Kutasov. Ooguri. Phys. C.

a. University of Maryland at College Park.2]. USA d School of Natural Sciences. 4) dilaton supergravities are found and their coupling to matter multiplets is discussed. respectively.d a Department of Physics. Princeton.V. James Gates Jr. All rights reserved. the study of Calabi–Yau four-fold compactifications has become of particular importance for physical ∗ Corresponding author. such as quantum cohomology and mirror symmetry [1. Olden Lane. USA Received 25 May 2000. At the same time substantial progress has been made in understanding the mathematical aspects of Calabi–Yau three-folds. accepted 13 June 2000 Abstract We consider two-dimensional supergravity theories with four supercharges constructed from compactification of Type II string theory on a generic Calabi–Yau four-fold. Since then. Edward Witten b.  2000 Elsevier Science B.6 . PII: S 0 5 5 0 . NJ 08540. E-mail: gukov@feynman.elsevier.edu 0550-3213/00/$ – see front matter  2000 Elsevier Science B. 2) and N = (0. NJ 08544. Princeton.V.edu 1 gatess@wam.umd.Nuclear Physics B 584 (2000) 109–148 www. In type IIA and type IIB cases. I found a way to make it work. MD 20742-4111. new superspace formulations of N = (2.princeton.∗ .3 2 1 3 ( 0 0 ) 0 0 3 7 4 . All rights reserved. CA 91125. It was not until the discovery of F-theory [3] that it was realized that N = 1 four-dimensional heterotic string vacua can be equivalently described as F-theory compactifications on elliptically fibered Calabi–Yau four-folds. Stanislaw Ulam 1. Pasadena. compactification of heterotic string theory on Calabi–Yau manifolds was the primary candidate for constructing realistic models in four dimensions with N = 1 supersymmetry. USA b California Institute of Technology. College Park. Princeton University.nl/locate/npe Two two-dimensional supergravity theories from Calabi–Yau four-folds S. CIT-USC Center For Theoretical Physics. This was also a strong motivation to study type II superstrings on Calabi– Yau three-folds which share many common properties with the corresponding heterotic compactifications.c. Sergei Gukov b.1 . USA c Joseph Henry Laboratories. Institute for Advanced Study. Introduction For a long time.

to describe type IIA compactifications on Calabi–Yau four-folds we need a generalization of this theory that includes interaction with twisted chiral multiplets and the possibility to turn on the superpotential and as well the twisted superpotential. 2) dilaton supergravity interacting with some number of chiral and twisted chiral multiplets. is not new and presented for the sake of completeness. acting on the matter fields. The construction in Section 4 is based on the Goldstone mechanism in the superspace formulation of non-minimal gauged N = (2. In a special case. 4) dilaton supergravity.110 S.J. as well as Section 3. It is invariant under the “mirror transformation” which. 2) (respectively. in Sections 4 and 5 we present superspace construction of general N = (2. In the Appendix A we present a straightforward but technical worldsheet calculation of string amplitudes .. so we shall not discuss non-perturbative phenomena in this paper. when the area of the elliptic fiber shrinks to zero. see. Most of this section. M-theory compactification on a Calabi–Yau four-fold is well described by F-theory compactification on the same Calabi–Yau manifold. et al. Thus. N = (0. Another characteristic feature of this supergravity theory is that the supergravity multiplet contains a real dilaton field. Surprisingly. Instead. 2) superstring theories. In the low-energy limit the theory is described by supergravity coupled to matter. 2) dilaton supergravity coupled to matter. e. On general grounds. In Section 6 we perform the Kaluza–Klein reduction of type IIB string theory on a Calabi–Yau four-fold and describe the component action of the resulting N = (0. The generalization posits that this mapping also applies to the effective action. a compactification of type IIA (IIB) string theory on a Calabi– Yau four-fold leads to a N = (2. type IIA string theory is related to M-theory via compactification on an extra circle (of small radius). / Nuclear Physics B 584 (2000) 109–148 applications. 2) dilaton supergravity to matter multiplets is the subject of Section 5. from the Kaluza–Klein reduction of type IIA string theory on a Calabi–Yau four-fold in Section 3 we find that a suitable low-energy theory is N = (2. 4)) effective field theory in two dimensions. it turns out that manifestly supersymmetric formulations of such theories has not been given previously. On the other hand. One of the most striking outcomes in the study of compactifications on Calabi–Yau three-folds is the great success in understanding non-perturbative phenomena in N = 2 field theories in four dimensions. we consider classical supergravity theories interacting with two-dimensional non-linear sigma-models that can be constructed from Calabi–Yau four-folds. The superspace formulation of the new N = (0. For example. a component action of this N = 2 dilaton supergravity was constructed in [6]. On the other hand. where we discuss local integration in superspace. Compactification of F-theory on an elliptically fibered Calabi–Yau fourfold is closely related to the corresponding compactifications of type IIA and M-theory. Coupling of the new N = (2.g. However. [4] for introduction and references. exchanges chiral multiplets and twisted chiral multiplets. We thus generalize the proposal of [5] where it was suggested that the “kinematic structure” of the mirror transformation has its origin in a mapping between chiral and twisted chiral multiplets when these superfields are regarded as the fundamental degrees of N = (2. Namely. 2) supergravity. understanding of Calabi–Yau four-fold compactifications still is quite far from that stage. when all matter multiplets are chiral and massless. Gates Jr. 4) dilaton supergravity is presented in Section 7.

. correspondingly.q) ∈ H p. are used for the two-dimensional spinor indices “+” and “−”. 2) dilaton supergravity. n.1 = h1.1 = h1. . β. . . . . The non-vanishing cohomology groups have the forms ωi following dimensions [8]: h1. an arbitrary variation of the metric of the Calabi–Yau four-fold X that respects SU(4) holonomy looks like: ¯ δga b¯ dza d z¯ b + δgab dza dzb + c.2 = h2. b.1) For the Euler number of X we have: χ = 8 + h1. run from 0 to 9 and denote ten-dimensional Lorentz indices.S. and in the Appendix B we list extra derivative constraints arising from de-gauging N = (2. X is classified by the Hodge numbers hp. q)(p. Finally. The (1. .J. Sometimes we also use capital latin letters A. h2.q which count the number of harmonic (p. Latin letters m. in Appendix D we repeat the derivation [7] of the chiral density projection formula in N = (2. A Calabi–Yau space X is a compact Kähler manifold with complex dimension four and SU(4) holonomy group. Greek letters α. . .4 = h4. real and holomorphic indices tangent to X. 2) non-minimal supergravity.and (3. Namely. . (2.1 − h2. .2 = h3. .1 + h3. . .0 = h0. . Appendix C contains components of the covariant derivative in N = (2. . 1)-forms are related to the deformation parameters of the Kähler form and the complex structure of X.3) . Capital letters M. 2. 2) dilaton supergravity needed in Section 5.1 + 2h3.2) 6 We denote by Ω a covariantly constant (4.c. represent. h0. As a topological space. . therefore. i = 1.1 δgab = h X j =1 1.1 . . where 3.2 = 2(22 + 2h1. We begin in the next section with a summary of notations and definitions used throughout the paper. respectively. h3. . 0)-form. and a.1 ). hp.. It follows that X is a Ricci-flat manifold and. Calabi–Yau four-folds: some conventions and definitions We study compactification of type II string theory on M(2) × X where M(2) is a maximally symmetric homogeneous two-dimensional space–time and X is a Calabi– Yau four-fold. / Nuclear Physics B 584 (2000) 109–148 111 corresponding to the target space metric of the effective two-dimensional theory.1 j φ j wab . .1 = h3.q .4 = 1. B. and light-cone indices “ ” and “ ”. 1). h2. respectively.3 . Gates Jr. and µ.3 . . et al. . .q (X). it can be used as a background for type II string compactification. (2. to denote both spinor and vector indices. . . We use the following notations for the space–time indices.0 = h4. N. . ν. .1 − h2.3 . (2. iδga b¯ = h X i=1 s i ωai b¯ . .

1) ∧ ωl . Mc (X).1) (1. Notice. et al.4) (2. respectively. j 4V X Z 1 √ ¯ d 8 z g ωak b¯ ωl a b .4) and (2. V = d 8z g = 4! X X Z 1 8 √ d z g wi ab w¯ jab Gφi φ¯ ¯ = ¯ . with the Kähler potential. in the ‘large volume limit’ we may describe the low-energy effective theory studying compactification of type IIA supergravity on the Calabi–Yau space X.5) (2. When the volume of X is large compared to the string scale.9) K(φi . MK (X). φ¯ i¯ ) = − ln X 3. Compactification of type IIA string theory on Calabi–Yau four-folds In this section we describe the effective two-dimensional theory constructed from compactification of type IIA string theory on a Calabi–Yau four-fold X.1) ∧ ωk (1. The moduli space of a Calabi–Yau space is locally a product of the moduli space of complex deformations.2) (2.3(X): j (1. .5) will become clear in the next section where we identify expectation values of the fields si and φj with the Kähler and complex structure moduli.10]: Z Z 1 √ K ∧ K ∧ K ∧ K. we write: 1. the metric Gφi φ¯ ¯ defined above is the Weil–Petersson j metric on the moduli space of complex structure of X.112 S.1) hsi i ωi (2. ab H 1. / Nuclear Physics B 584 (2000) 109–148 ¯ from the forms w we can construct elements in By the appropriate contraction with Ω. type IIA supergravity is a good low-energy approximation to type IIA string theory.1) ∧ ωj (1. let us start this section recalling some facts about type IIA supergravity itself. Gates Jr.6) X Z dij kl = (1.1) ωi (1. In particular. Yimn¯ = ωi (2.1 b¯ K = iga b¯ dz ∧ d z¯ = a h X (1. and the (complexified) moduli space of Kähler structure. Therefore. (2.8) i=1 for the Kähler form on X. [9]: ! Z Ω ∧ Ω¯ .7) X The notations (2. cf. With this motivation.J. Gσk σ¯ l = 2V X Z (1.1) ∧ ωm ∧ ωn¯ . (2.3) ωj ¯ ¯ = Ω¯ a¯ b¯ c¯d¯ g dd wdf d z¯ a¯ d z¯ b d z¯ c¯ dzf . j In what follows we will use some integrals over the Calabi–Yau space X [9.

dilatonindependent way under gauge transformations. The action of type IIA supergravity is invariant under 16 left and 16 right supersymmetry transformations. such that the left supersymmetries are chiral while the right supersymmetries are anti-chiral with respect to the ten-dimensional chirality operator Γ11 . a vector field AM . the fundamental strings filling twodimensional space–time do not break supersymmetry further. the dilaton ϕ.1). where the eight-form I8 is proportional to the Euler density of X. incorporation of fermionic zero-modes completes the resulting spectrum into appropriate N = (2.1 (1.1) zk ωk + c.c. / Nuclear Physics B 584 (2000) 109–148 113 The bosonic field content of type IIA supergravity contains the metric gMN . Via dimensional reduction type IIA dilaton ϕ becomes a real scalar field in the two-dimensional theory. Gates Jr. G = dC.2) N= 24 2(2π)2 X To cancel the tadpole for the B-field one has to introduce N fundamental strings filling two-dimensional non-compact space. 2) supergravity theory coupled to matter. With the appropriate choice of orientation.3).1) 2 12 4 48 where we introduced the gauge-invariant field strengths: F = dA. B.1) Ajµ ωj + h X k=1 (2. To find the spectrum of the effective low-energy theory we perform Kaluza–Klein reduction of type IIA supergravity to two dimensions. The dots in the Lagrangian (3. By supersymmetry.1) stand for higher order terms among which we find the Chern–Simons term B ∧ G ∧ G and the anomaly term B ∧ I8 . (3.3) i=1 1.1 complex scalars φi and h1. and tensor fields BMN and CMNP . and C transform in a natural.S.1 C= h X j =1 2. The bosonic part of the Lagrangian (in string frame) looks like:     √ 1 1 1 1 L(10) = −g e−2ϕ R (10) + 4(∇ϕ)2 − H 2 − F 2 − G0 2 + · · · . With this choice of normalization the fields A. (3. The ten-dimensional metric gMN decomposes into the twodimensional metric gµν . H = dB. et al.1 B= h X r i ωi(1. The antisymmetric tensor fields BMN and CMNP can be expanded into harmonic modes as follows: 1. (3.4) .1 real scalars sj defined in (2. After integration over a compact eight-manifold X these topological terms produce a global anomaly [11.12]: Z 1 χ − G ∧ G. Below we describe the decomposition of type IIA bosonic fields in harmonics of X. 2) supermultiplets.J. h3. Since X admits a nowhere vanishing complex spinor of definite chirality. compactification of type IIA string theory on X is described by N = (2. G0 = G + A ∧ H. (3.

2) chiral matter was studied some time ago [6]. where it was found that the most general N = (2. (2. After a further compactification on a circle they naturally remain as chiral superfields in two dimensions. When (2. but now a T -duality becomes possible and zk can be alternatively described if one prefers as twisted chiral superfields.1 chiral multiplets : 1. ϕ.4) yields h2. 1)-modes take value in a torus.114 S. Taking into account the vector field Aµ from the Ramond–Ramond sector of type IIA theory. To summarize. so these fields are massless.1 h (twisted) chiral multiplets : gµν .1) = 0 and H ∧∗ ωk = 0 there is no superpotential for background fluxes satisfy G ∧ ωk the corresponding harmonics zk . 2) non-linear sigma-model is based on a target space with two non-commuting complex structures J± .3) and (3. if we start in eleven dimensions. The target space of this sigma-model is parametrized by some number of chiral and twisted chiral multiplets. This kind of dilaton supergravity coupled to N = (2. A T-duality transformation on the torus then converts them from ordinary chiral superfields to twisted chiral superfields. Ajµ .15] and/or twisted . Using the formulas (2.1 complex scalar modes in three dimensions.4) in the Lagrangian (3. This agrees with the result of [13]. To find the effective action for the light fields we have to substitute (2.5) − Gσi σ¯ j ∂µ σ i ∂µ σ¯ j − Yimn¯ σ i Dµ zm D µ z¯ n¯ + · · · . i σ¯ j . [6. the reduction of the C-field (3.10]:   ¯ L(2) = e−2ϕ V R (2) + 4(∇ϕ)2 − Gφi φ¯ ¯ ∂µ φ i ∂ µ φ¯ j j   1   1 (3. ¯ z¯ k . et al. Furthermore. 1)-forms on the complex structure [10]. Gates Jr. 2) dilaton supergravity coupled to a non-linear sigma-model. cf. However. Aµ . Vector fields Aµ and Aiµ do not have propagating degrees of freedom in two dimensions and play the role of auxiliary fields in the supergravity multiplet and twisted chiral multiplets.1 h twisted chiral multiplets : 2. respectively. background fluxes of Ramond–Ramond field strengths induce effective superpotential [14. φ¯ ¯ . It is natural to choose zk to be scalar components of chiral superfields.J. 2 4 where the covariant derivative Dµ acting on zm contains a connection corresponding to the holomorphic dependence of the basis of (2. Indeed. / Nuclear Physics B 584 (2000) 109–148 It is convenient to combine real fields s i and r i into complex scalars σ i .1) and integrate over the internal space X. (3.5) and (2.3).6) we obtain the following effective action for the bosonic modes. while ker(J+ + J− ) is parametrized by twisted chiral superfields. for our purposes we need to generalize the component construction of [6] to include twisted chiral multiplets.4). we end up with the following list of N = (2. 2) supermultiplets: a gravitational multiplet : h3. so that the space ker(J+ − J− ) is parametrized by chiral superfields. These modes are scalar components of chiral superfields since there is no notion of “twisted chiral superfields” in three dimensions. zk . φi . The complex scalar fields zk which come from (2.1) (2. we find that in the large volume limit compactification of type IIA string theory on a Calabi–Yau four-fold X leads to N = (2. σj.

its Hodge dual is given Indeed. z¯ k¯ .6) S = d x d 2 θ d 2 θ¯ E −1 exp(−K). we have to incorporate these terms in the construction as well.5) as: K = Kc (φi . zk .J.11) ω =− ω ∧K∧K+ 2 3 ( X K ∧ K ∧ K ∧ K) Therefore. σ¯ j ) = − ln X ω(2) is a harmonic 2-form on a Calabi–Yau four-fold X. For the sake of simplicity. z¯ k¯ ) + Λ2 (φi . the Kähler potential [19]. φ¯ i¯ ) + KK (σj . σ¯ j . (3. 2) superspace is parametrized by .18]. σj . let us assume for a moment that h2. et al.5) with up to two derivatives or four fermions are determined by a single real function K(φi . θ¯ − ). anti-commuting coordinates θ α = (θ + . θ − ) and their complex conjugates θ¯ α˙ = (θ¯ u . We postpone the discussion of the superspace measure E till the next sections where superspace formulation will be discussed in detail. In addition to the usual space–time coordinates x µ . It is invariant under the generalized Kähler transformation: K → K + Λ1 (φi . (3.8) it follows that locally we can write the Kähler potential that gives the effective action (3. 2) superspace where the supersymmetry becomes manifest [17. σj . if by the following neat formula: R 1 (2) 2 ( X ω(2) ∧ K ∧ K ∧ K) ∗ (2) R K ∧ K ∧ K.10) KK (σj . zk ) + Λ¯ 1 (φ¯ i¯ . Hence. Kc (φi . one can verify that the metric Gσi σ¯ j can be obtained from the Kähler potential: ! Z K∧K∧K∧K .9) on the moduli space of the complex structure: ! Z Ω ∧ Ω¯ .8) From the condition (3. N = (2.5) in the following form: . Now we simply assume that the suitable measure exists. The most elegant and convenient way to do this is in N = (2. similar to the action of matter coupled N = 1 supergravity in four dimensions: Z Z 2 (3.9) where Kc is the Kähler potential (2. σj . (3. σ¯ j . z¯ k¯ ) + Λ¯ 2 (φ¯ i¯ . (3. ϕ). The main advantage of the superspace formulation is that due to the extended supersymmetry. φ¯ i¯ ) = − ln X Similar to the case of Calabi–Yau three-folds [9].5) can be written in a compact form. / Nuclear Physics B 584 (2000) 109–148 115 chiral superpotential [14–16] in the two-dimensional theory. all the term in the action (3.S. Then. we can write (2. zk ). Gates Jr. σ¯ j . Then the metric is block diagonal: Gφi σ¯ j = ∂ 2K = 0. we expect that the action of the matter fields (3.7) The target space metric is given by the second derivative of the Kähler potential. φ¯ i¯ .1 = 0. σ¯ j ). ∂φi ∂ σ¯ j (3.

Even in the absence of the superpotentials. torsionless. respectively. to the leading order the metric on the target space is Kähler. we will include a superpotential and twisted chiral superpotential that violate the R-symmetries.1) ∧ ωl(1.J.13) W (φi ) = 2π X and for the twisted chiral superpotential: Z e (σj ) = 1 eK ∧ F . (3. σ¯ j ) .116 S.1) (1.12) where the upper (lower) row is assigned a U (1)A charge +1 (−1) while the right (left) column is assigned a U (1)V charge +1 (−1). . higher derivative interactions among the massless fields obtained by integrating out massive string states would be expected to violate the R-symmetry.1) ∧ K ∧ K 4V Gσk σ¯ l = − X 1 − 72V 2 Z ωk(1.14) X e (σj ) are holomorphic functions The superpotential W (φi ) and the twisted superpotential W of the fields φi and σj . The explicit expression for the chiral superpotentials generated by the most general P Ramond–Ramond flux F = (RR fields) in terms of the Calabi–Yau moduli was derived in [15]: Z 1 Ω ∧ G. Even though we will not explicitly include massive string modes in the present paper. We will denote their linear combinations as U (1)A and U (1)V . The classical action (3. / Nuclear Physics B 584 (2000) 109–148 Z 1 (1.10): Gσk σ¯ l = − 1 ∂ 2 KK (σj . and equal to the metric on the moduli space of the Calabi–Yau space X. Q¯ −.1) ωk ∧∗ ωl 2V X Z 1 = ωk(1. Mc (X) × MK (X).5) is invariant under two U (1) R-symmetries.1) ∧ K ∧ K ∧ K . Taking into account the superpotential terms.1) ∧ K ∧ K ∧ K X ! Z ! ωl(1. The action of these symmetries on the supercharges can be represented as: Q− Q+ Q¯ u . 2 ∂σk ∂ σ¯ l Hence. the action of the matter fields reads as: . it is easy to see that the above metric indeed follows from the Kähler potential (3. These R-symmetries are not symmetries of the string theory — there is no way to assign R-transformations to massive string modes to preserve them. (3. W 2π (3. X Using the explicit expression (2. et al.8) for the Kähler form K. Gates Jr.

A simple way to see this is to assume.22]. 3 Of course.17) where Kc and KK are given by the tree-level formulas (2. In particular.15]. (3. respectively. that there exists a supersymmetric vacuum corresponding to a non-zero H -flux and consider a BPS soliton connecting such a vacuum to the vacuum with zero H -flux. σ¯ j ) DW e. = + W Dσi ∂σi ∂σi (3. In type IIA string theory this soliton would correspond to an N S5-brane wrapped over a Poincaré dual supersymmetric 5-cycle. These models can have a variety of T-duality symmetries..c. here we assume that X is elliptically fibered.20].18) eK G φi φj¯ (Dφi W ) Dφ¯ ¯ W¯ − 3|W |2 . on the contrary. is independent of this assumption. Dσi (3. this theory is dual to compactification of type IIA string theory on X.J. Of particular interest are mirror symmetries [21. After a further compactification on a torus T 2 . one can also show that any H -field flux breaks all the supersymmetry. / Nuclear Physics B 584 (2000) 109–148 Z S(2) = Z Z 117 Z d 2 x d 2 θ d 2 θ¯ E −1 e−K + d 2 x d 2 θ E −1 W (φi ) Z Z . a symmetry that maps 2 Investigating the supersymmetry conditions as in [14.10). a non-zero H -flux R lifts all the supersymmetric vacua. there is a contradiction since Calabi–Yau 4-folds do not have supersymmetric 5-cycles. A mirror symmetry is.16) we should use the appropriate covariant derivatives: ∂Kc (φi . 2 In the two-dimensional theory this effect corresponds to generation of a superpotential that lifts (part of) supersymmetric vacua.15. It is natural to interpret this in terms of the effective superpotential W ∼ C ∧∗ H for the scalar fields zk . Gates Jr. However. of course. Therefore.g. Dφi ∂φi ∂φi e ∂W e ∂KK (σj . then there are no such vacua at all.17). 2) supersymmetry. From the formulas (2. . The result. see. A simple way to see that one has to use the covariant derivatives instead of ordinary ones is to consider first compactification of F-theory on the same Calabi–Yau space 3 X.15) Generic values of Ramond–Ramond fluxes completely break the N = (2.S. j where the covariant derivative Dφi W = DW/Dφi is defined in (3.2). e.2) and the quantization condition of the G-flux [12] it follows that there is a finite number of choices for [G] ∈ H 4 (X) corresponding to supersymmetric vacua. [15].9) and (3. In the component action of the effective N = 1 four-dimensional theory there is a scalar potential:   ¯ (3. e (σj ) + c. if the vacuum values of the fields φi and σj satisfy: DW =0 Dφi and e DW = 0. if h2. φ¯ i¯ ) ∂W DW = + W. However. From the superspace construction in Section 5 it follows that in the equations (3. et al.17) also appear in the component action of the two-dimensional theory in question. however.16) then type IIA compactification on the corresponding Calabi–Yau manifold is supersymmetric [14.1 > 8 + h1.1 . It is clear that after the dimensional reduction of the four-dimensional component action the covariant derivatives (3.1 + h3. + d 2 x dθ + dθ − Ee−1 W (3.

θ − ↔ θ − that also exchanges chiral multiplets and twisted chiral multiplets and changes the superspace measure in a way consistent with other definitions of mirror symmetry. e (σj ). Gates Jr.20). tree-level amplitudes in type IIA string theory must agree with the corresponding amplitudes in the effective two-dimensional theory. It can be interpreted in terms of the supergeometrical coordinate transformation . By definition. the latter implies that under the mirror symmetry we have σi ↔ φj which is consistent with our interpretation of (vevs of) these fields as the Kähler and the complex structure moduli of the Calabi–Yau space X. 2) gravity theories have at least one gauged U (1) R-symmetry (see [23] for a general presentation). W (φi ) ↔ W (3. This operation and the conformal field theories associated with X and X corresponds [5] to a transformation which exchanges chiral multiplets and twisted chiral multiplets. 2) dilaton supergravity that includes chiral and twisted chiral multiplets on equal footing does not seem to exist in the literature. Recall that via a space–time T-duality transformation. 2) dilaton supergravity without gauged symmetry. A superspace formulation of such a supergravity theory is presented in the next section. In the Appendix A we illustrate this by a worldsheet calculation which independently proves that the target space metric is block-diagonal.118 S. we are interested in a theory where the supergravity multiplet contains a real dilaton field. φi ↔ σj .q (X) (3. Hence mirror symmetry just exchanges these two descriptions. e hp. In other words.8).q (X) = h4−p. It may seem that twisted chiral fields zk violate the invariance under (3. cf.19) e are equivalent.J. those fields can be described by either chiral or twisted chiral superfields. / Nuclear Physics B 584 (2000) 109–148 type IIA string theory on a four-fold X to type IIA string theory on the mirror variety e such that: X. 4. Although a superspace model of N = (2. Theories where the entire U (1)A ⊗ U (1)V symmetry group is gauged are called non-minimal (or reducible). Therefore. et al. 2) dilaton supergravity itself. This last property is a distinguishing feature of the new formulation since all the known N = (2. 2) supergravity where the dilaton is a complex field was constructed in [24].15) describes dynamics of the light modes in type IIA string theory on X in the large volume limit. Superspace formulation of N = (2.20) The mirror map has no effect on the N = (2. In particular. so that in the absence of matter fields it must be mirror-symmetric. A superspace formulation of N = (2. 2) theories also interchanging: U (1)A ↔ U (1)V . the mirror symmetry relates different quantum N = (2. 2) dilaton supergravity In this section we present a superspace construction of N = (2. as opposed to . (3. the low-energy effective action (3.

& ↓ U (1)V .S. it has to be a combination of consistent truncation of the non-minimal N = (2. in accordance with the results of Section 3 where we studied compactification of type IIA string theory on Calabi–Yau four-folds. et al.20).1) U (1)A & . For example. In other words. the so-called Wess–Zumino gauge. Since the broken U (1)A ⊗ U (1)V phase of the theory has exactly the same field content as the gauge symmetric phase plus the field content of the Goldstone multiplets minus the parts that go into the longitudinal components of the U (1)A ⊗ U (1)V gauge fields. minimal theories with either of the U (1) R-symmetries gauged can be obtained by truncation of the non-minimal U (1)A ⊗ U (1)V gauged supergravity [23]. we construct N = (2. It is very well known how to obtain one supergravity theory with a smaller holonomy group from a supergravity theory with a larger holonomy group. it would be interesting to study black hole solutions in this dilaton supergravity. This process has been for a long time called “degauging” (see [17]. there are no mirror-symmetric N = (2. Notice. extra component fields not present in the gauge symmetric phase begin to appear. For example.7). The basic idea of de-gauging is to break the gauge symmetry introducing extra matter field in a Goldstone-like mechanism. however. On the other hand. (4.J. More precisely.b. The reason is that the massive system no longer possess the U (1) gauge invariance. For example. we find that in total Goldstone multiplets should have three real scalars. 2) supergravity multiplet contains a real dilaton field ϕ. An advantage of this approach is that both the original and the resulting theories are manifestly invariant under the mirror symmetry (3. the vertical arrow corresponds to de-gauging U (1)A ⊗ U (1)V symmetry. 1 where the theory with the trivial holonomy group is the new N = (2. a massive gauge multiplet contains some extra component fields which could be eliminated in the massless multiplet. Gates Jr. 2) dilaton supergravity via de-gauging U (1)A ⊗ U (1)V non-minimal gauged supergravity. The relation between different supergravity theories can be schematically represented in the form of the following diagram: U (1)A ⊗ U (1)V . let us count the number of real Goldstone scalars. 2) multiplets with such . On the other hand. so that the total number of degrees of freedom increases. To see this. It contains a U (1) gauge vector field.3. consider an abelian vector multiplet in a four-dimensional N = 1 gauge theory. 2) supergravity to a minimal one plus a de-gauging of the latter. Goldstone supermultiplets must appear. 2) dilaton supergravity we are going to construct. Following these steps. And their component fields come from that part of the vector multiplet that was ignored in the symmetric phase. cf. However. that various arrows in this diagram have different meaning. All the other fields can be set to zero by a supersymmetric choice of gauge. Section 5. gaugino and an auxiliary field. We also find that the new N = (2. This toy model teaches us that when a symmetry is broken in superspace. [25]. We hope that apart from this obvious application there may also be many other aspects of the new supergravity to explore. / Nuclear Physics B 584 (2000) 109–148 119 minimal theories where only U (1)A or U (1)V factor is gauged.

(4. Left-right symmetric N = 2 superspace is parametrized by the bosonic coordinates x µ = (x . i . Therefore. 2) supergravity that leads to only one massless scalar (the dilaton). Φi | = φi . To start the construction. unfortunately. 2 D− Σ¯ j | = ζ− . one real scalar from this multiplet would become the longitudinal component of the gauge vector field. Σj | = ζ−. ]Σj | = Bj . In order to de-gauge the resulting minimal gauged supergravity we would have to introduce extra chiral (or twisted chiral) Goldstone superfield. with all the θ coordinates put to zero. at the first step we would have to make a consistent truncation that would eliminate one of the gauge fields. . In any case. [Du . In what follows we promote the complex scalar fields φi and σj defined in the previous section to the chiral and twisted chiral superfields Φi and Σj . x ).2) for a chiral superfield Φ. D−. ]Φ¯ i | = A¯ i . D−. D− Φi | = ψ− i i [D+ . Assuming that the latter route (which is not.120 S. i [Du . i [D+ . respectively. D− ]Σ¯ j | = B¯ j . Φ = D+ Φ¯ = D− Φ¯ = 0 (4. Similarly.J. D−. we find the components of a twisted chiral multiplet Σj : Σj | = σj . and: Du Σ = D− Σ = D+ Σ¯ = D−. j Du Σ¯ j | = ζu . The simplest constraints are linear in derivatives and look like: Du Φ = D−. . mirror-symmetric) is equivalent to the vertical arrow on the above diagram. 2 j (4. Du and D−. if we were following another route via a minimal gauged supergravity. with all other (anti-)commutators vanishing. i . et al. ¯ D− Φi | = ψ− . j D+ Σj | = ζ+ . (θ + . two anti-commuting complex spinor coordinates θ α = . j D−. θ¯ − ). The spinor derivatives D+ . satisfy {D+ . D+ Φi | = ψ+ i Du Φ¯ i | = ψu . in agreement with what we expect. Components of the chiral superfield Φi can be obtained using the projection method: Φ¯ i | = φ¯ i . } = ∂ . we conclude that the vertical arrow should be a more economical de-gauging. i. Gates Jr. D− . for example. Σ¯ = 0 (4. Du } = ∂ and {D− . and the other would become a dilaton. we expect that there is an economical de-gauging of the non-minimal N = (2. Indeed. D− ]Φi | = Ai . Σ¯ j | = σ¯ j . Similarly. let us arrange component fields found in the previous section into superfields. Φi | denotes the leading component of the superfield Φi .4) 2 2 where.3) for a twisted chiral superfield Σ. First we give the definitions in flat superspace and then extend them to curved superspace. Irreducible matter superfields are defined by imposing some constraints on general complex superfields. θ − ) and their complex conjugates θ¯ α˙ = (θ¯ u . / Nuclear Physics B 584 (2000) 109–148 a field content.5) . we will regard the compact fields zk as the scalar components of (twisted) chiral superfields Zk .

etc.S. ∇± ] = − ∇± . (∇ R) − R R + (∇+ ∇− F ) − F F X [∇ . ∇+ } = 0. ∇− } = 0. . [X . R| = H. On the other hand. and denote by Λα the Lorentz spin-connection. / Nuclear Physics B 584 (2000) 109–148 121 To define chiral and twisted chiral superfields in curved N = (2. Y 0 . 2 2 i i 0 0 [Y . we end up with U (1)V minimal theory. we obtain U (1)A minimal gauged supergravity theory.26]:   1 1 ¯ 1 ¯ 2 . ∇− } = − R(X − i Y¯ 0 ). 1 ¯ ¯ 1 ¯ (4.8). 2) dilaton supergravity we start with non-minimal gauged supergravity and then de-gauge U (1)A ⊗ U (1)V symmetry. then the covariant derivative in this theory has the form: ∇α = Eα B DB + Λα X + Aα Y + A0α Y 0 . . (4. ∇± ] = ∓ ∇± . } = i∇ . {∇− .7) [Y . {∇− . ∇−. [Y. if we set R = 0. F )Y + (∇ 2 R)Y 0 + · · · + c. . ] = ± ∇±. Gates Jr. et al. ∇u } = i∇ . ∇ ] = 2 2 2 i i (4.11) . 2 2 The constraints which define non-minimal N = 2 U (1)A ⊗ U (1)V supergravity are given by: {∇+ . ] = ± ∇±.c. If we introduce superfields Aα and A0α representing U (1)V and U (1)A gauge connections. X . If we set F = 0 in the constraints (4. ∇±. ∇± ] = + ∇± . Y. The easiest way to see this is to compute the commutator [23. the spinor derivatives Dα must be appropriately replaced by the covariant derivatives ∇α . the covariant derivative ∇α by ∇ so that: bα + Aα Y + A0α Y 0 ∇α = ∇ (4. We denote the leading components of the superfields F and R as follows: F | = G. In our construction of N = (2. 2 2 where the dots stand for the covariant derivative terms like (∇− F )∇u . 2 2 i i . ∇± ] = ± ∇± . ∇±. 2) supergravity theory without gauged symmetry.. 2) superspace.6) where Eα B is the supervielbein.9) + (∇+ ∇−.8) {∇+ . ∇−. {∇+ . and U (1)A symmetry generators. act on the covariant derivative ∇α in the following way [23]: 1 1 [X . The Lorentz generators. (4. } = − F¯ (X¯ − i Y).10) In order to obtain N = (2. (4. U (1)V symmetry generators. we replace bα which includes only derivatives and Lorentz generator. [Y. {∇+ .J. 2 2 where the chiral superfield R and the twisted chiral superfield F are related to the twodimensional curvature R (2) and the abelian field strengths of the graviphoton gauge fields.

so it can describe twodimensional N = (2. b+ .12) which define N = (2.12) on the covariant derivative ∇ The new supergravity theory does not have gauged symmetry and contains a real dilaton field ϕ. b+ . b− . } = − F¯ X + (λ−. 2) dilaton supergravity can be written in the following simple form: b+ } = iη+ ∇ b− . . e η− ≡ λ− + e λ− then the conditions (4.8). the fields λα and e new supergravity multiplet.J. 2) supergravity without gauged symmetry. ∇ λ−. b+ . Gates Jr. b+ . + e b+ − (λ+ − e b−. . To the λα can also be identified with the dilatino field of the leading order in θ α . ∇ {∇ 2 2 Note that these equations. . ∇ η− ∇ η+ ∇ {∇ 2 2 2 b−. b− . )∇ λ+ )∇ {∇ 2 2 2 i i bu } = i ∇ b + (λu + e b+ − (λ+ + e bu . } = i ∇ b + (λ−. ∇ b− } = iη− ∇ b+ . and using (4. } = − 1 F¯ X + i e b+ − i e b−. ∇ b− } = i(λ− − e b− . . b+ . ∇ {∇ {∇ 1 i i b− } = − RX ¯ + e b+ + e b− . as we will show in a moment. ∇ λu )∇ λ+ )∇ {∇ 2 2 i i b−.12) {∇ 2 2 λα have appeared.20). as well as the conditions (4. et al. b− . ¯ + i (λ− + e b+ . are manifestly invariant under the mirror symmetry transformation (3. Substituting (4. To summarize. They have U (1)V gauge symmetry is broken. In the theory we are constructing the U (1)A ⊗ λα are dynamical.12). ∇ λ− )∇ λ+ )∇ {∇ 2 2 2 i 1 i b−. − e b− − (λ− − e b−. The new covariant derivative ∇ the U (1)V gauge connection nor the U (1)A gauge connection. . there exists a unique mirror-symmetric two-dimensional N = (2. b+ . e η+ ≡ λ+ − e η− ≡ λ− − e λ− .122 S. They are components of the non-minimal where the new fields λα and e gauged supergravity multiplet that could be eliminated by a gauge transformation in the U (1)A ⊗ U (1)V gauge-symmetric phase.7). If we introduce four linear independent spinors (along with their conjugates): λ+ . ∇ λ−. we derive the form of the commutator bα operators: algebra for the ∇ b+ } = i(λ+ + e b+ . ∇ η−. ∇ λ+ )∇ {∇ λ− )∇ {∇ b− } = − 1 RX b+ + i (λ+ − e b− . } = i ∇ b + i η−. b− . so that the fields λα and e to be identified with the spinorial derivatives of a new matter Goldstone multiplet. ∇ {∇ 2 2 b−. dilaton supergravity defined by the set of constraints (4. / Nuclear Physics B 584 (2000) 109–148 bα contains neither and the remaining derivatives are real. ∇ η+ ∇ {∇ 2 2 2 bu } = i ∇ b + i ηu ∇ b+ − i η+ ∇ bu . η+ ≡ λ+ + e λ+ . 2) bα . ∇ b− − i η− ∇ b−. )∇ λ− )∇ (4.11) into (4. b+ .

5 However. but rather what remains after the Goldstone mechanism takes place. One might notice that a real superfield V contains one massless vector field.11) into (4. Since local integration measures of the new N = (2. λ− = −e λ− = i(∇ b−. Vanishing of these terms leads to a set of constraints which is equivalent to the set of constraints obtained from the Jacobi identities (4. we take the following ansatz for the spinors λα : b+ V ).13) impose no further constraints on V . (4. λ−. λ+ = i(∇ λ+ = e bu V ). 5 It is also worthwhile to stress here that massless superfield V is not a Goldstone multiplet itself. We outline the result in Appendix B. ∇ bγ ) } = 0. ∇α = 0. Instead of deriving derivative constraints on the spinor fields λα and e from the Jacobi identities (4.14) so that the Jacobi identities (4. ) stands for the graded antisymmetrization λα directly symbol. This means that (4. } is the graded commutator.8) one also finds terms proportional to gauge symmetry generators Y and Y 0 . ∇ {∇ in terms of an unconstraint real superfield V : V = V¯ (4. and only define the other superfields (like F and R) in terms of the derivatives of V . An advantage of the first solution is that the covariant derivative ∇ 2 b commutes with itself. [∇ (4.S. 2) dilaton supergravity theory imposing constraints (4. }. in the second case we find especially simple form of the anti-commutators b− } and {∇ b+ . b[α . For the same reason two-dimensional superfield V does not have an irreducible transverse component. and [ .16) 4 In ordinary field theories.13) ∇ where [ .12) bα . 4 For theories described by covariant derivatives ∇ Bianchi identities are simply Jacobi identities:  bβ .15) Below we present more evidence for this identification. While substituting (4. the constraints on some fields.J. ∇ b−. like in the usual gauged supergravity theories [23]. in agreement with the result of the previous section. the fields satisfy Bianchi identities because they are expressed in terms of the potentials. On the other hand. = −i(∇ (4. the Bianchi identities in this theory may lead to further on the covariant derivatives ∇ bα . There are two simple solutions to the Jacobi identities corresponding to either ηα or bα antie ηα put to zero. λu = −i(∇ λu = e b− V ). V ). / Nuclear Physics B 584 (2000) 109–148 123 Since we define new N = (2. A nice property of this solution is that V is manifestly mirror-symmetric. Gates Jr. they are identities and impose no extra constraints. Namely. unlike a similar four-dimensional superfield.13).13) are satisfied. 2) dilaton supergravity can be nicely derived from the corresponding expressions of the U (1)A ⊗ U (1)V theory only for the solution corresponding to e ηα = 0. In both cases the remaining spinor superfields can be expressed b+ . in what follows we discuss in detail only this case. massless vector fields in two dimensions do not have propagating degrees of freedom. et al. = −e λ−. It is natural to identify the dilaton field with the leading scalar component of V : ϕ = V |. .13) we use an equivalent approach which is much easier.

} = − F¯ X . b−. F ) − 1 λ− R + 1 λ−. ) ∇ b−. F X + R ∇ b− + F ∇ b−. . λ− ) + (∇ b− λ−. V + 4(∇ λ = i∇ λ = −e Further commutators of the covariant derivatives with vector indices follow from the consistency of the Bianchi identities (4. F¯ = 4∇ (4. ∇ b− } = − RX b−. ) ∇ b− . ∇ {∇ (4.17) where the superfields R¯ and F¯ can be obtained from the Bianchi identities. . V )∇ b− + (∇ b− V )∇ b−. ∇ b+ V . ∇ (4. . Gates Jr.17). ∇ {∇ {∇ 2 2 bu } = i ∇ b + (∇ bu V )∇ b+ + (∇ b+ V )∇ bu . b−.18) We also find the following expressions for the gauge connection: b V − 2∇ b+ ∇ b+ V )(∇ bu V ). + (∇ b−. 5. 2) supergravity coupled to chiral superfields Φi and Zk . ∇ (∇ [∇ 2 2 2 2 2 2   bu R) b+ F ) + 1 λu R¯ − 1 λ+ F X + i R¯ ∇ b+ . bu V + 4(∇ λ = i∇ λ =e b V − 2∇ b− ∇ b− V )(∇ b−. b+ . b ] = − i (∇ bu + i F ∇ ¯ + i (∇ b− . ∇ [∇ 2 2 2 2 2 2 where we used (4.15): . ∇ {∇ {∇ 1 1 ¯ . + (∇ b−. V ). R) b− F¯ ) + 1 λ−. b+ . b+ . Substituting (4. bu . b+ . ¯ + i (∇ b+ . − i F¯ ∇ b− . the most general action of N = (2. R¯ − 1 λ− F¯ X − i R¯ ∇ b ] = − i (∇ b−. . ∇ [∇  b ] = 2iλ+ λu + (∇ bu λ+ ) + (∇ b+ λu ) ∇ bu . ∇ [∇  b ] = − 2iλ− λ−.12). and twisted chiral superfields Σj looks like (3. et al. b ]= bu . } = i ∇ b + (∇ b−. b− ∇ R¯ = 4∇ b−.13):  b ] = − 2iλ+ λu + (∇ bu λ+ ) + (∇ b+ λu ) ∇ b+ . ∇ [∇ 2 2 2 2 2 2   i i i i b− R) + (∇ b−. ∇ {∇ b−. Lagrangians for matter multiplets coupled to N = 2 dilaton supergravity In order to couple matter fields to new N = (2. λ− ) + (∇ b− λ−. / Nuclear Physics B 584 (2000) 109–148 bα V and e which implies ηα = 2i ∇ ηα = 0.16) into (4. b− . 2) dilaton supergravity we have to repeat the analysis of [23]. ∇ [∇ 2 2 2 2 2 2   b+ R) + i (∇ bu F¯ ) − 1 λ+ R + 1 λu F¯ X − i R ∇ b+ − i F¯ ∇ b ] = i (∇ bu . b− . b− . b−. . ∇ [∇  b ] = 2iλ− λ−. It is worthwhile to stress here that one would obtain a different result de-gauging the corresponding commutators in the U (1)A ⊗ U (1)V non-minimal supergravity [27]. Up to terms with two derivatives or four fermions. b+ .J. ∇ b− } = −2(∇ b− V )∇ b− .19) [∇   b−. we find the following supergravity algebra: b+ } = −2(∇ b+ V )∇ b+ .124 S. Solving the set of constraints in Appendix B we get: b+ V .

2) and (5. . (5. the superspace measures E −1 and E −1 are related as follows: Z Z d 2 x d 4 θ E −1 L = d 2 x d 2 θ E −1 ∇¯ 2 L|. 2) supergravity theories such formulas were derived by Grisaru and Wehlau [7]. In the case of the U (1)V ⊗ U (1)A gauged supergravity the symmetry between chiral and twisted chiral fields is restored by the contribution of the anticommutator term {∇+ . sometimes it is called a chiral density projector because in the non-minimal N = (2. For gauged N = (2. ∇+ − iψ − ∇+ + iψ u ∇−˙  ¯ + ψ − ψ u + ψ u ψ − ∇u ∇− L|.2) is a D-type superinvariant. 2 4 −1 d x d θ E L = d 2 x e−1 ∇ 2 + iψ − ∇+ − iψ u ∇−  .1) In order to obtain the component action corresponding to (5. Thus. the derivation of this formula goes through as in [7] for the case of U (1)A theory. the twisted chiral density projection formula (5.2) Here ψµα is the gravitino field and L is an arbitrary scalar function of superfields. Although (5.2) and using the commutation relations (4. + − 12 H¯ − ψ − ψ u + ψ − ψ u Lc |. et al. the same projection formula is also valid in the non-minimal U (1)A ⊗ U (1)V supergravity theory [23]. Z¯k¯ ) Z Z Z Z . we can obtain the component projection formula for any chiral superspace integral: Z Z  . Therefore. 2) dilaton supergravity can bα .S. } ∼ F¯ |.1).11) and (4.16) into (5. More explicitly. In fact.7)–(4.4) By mirror symmetry the twisted chiral density projection formula in the U (1)V gauged supergravity theory has the following form: Z Z  d 2 x d 4 θ E −1 L = d 2 x e−1 ∇−. 2) supergravity ∇¯ 2 L is a chiral superfield (for a general L). e (Σj ) + c. Gates Jr. Φ¯ i¯ . Σj . one needs the appropriate projection formulas.c.5) can be derived using the methods of [7] or [28].8) we obtain the following density projection formula: . The projection formulas in the two-dimensional N = (2. + d 2 x d 2 θ E −1 W (Φi ) + d 2 x dθ + dθ − Ee−1 W (5. Σ¯ j .3) In particular. ∇−. replacing ∇¯ 2 L by an arbitrary covariantly chiral Lagrangian Lc .J. be obtained from (5. In the case of the minimal U (1)A theory the local density projection formula has the following form: Z Z  . . Zk . (5.5) replacing ∇α by a new covariant derivative ∇ substituting (4. (5. + − 12 G (5. + − 12 H¯ − ψ − ψ u + ψ − ψ u ∇¯ 2 L|. d 2 x d 2 θ E −1 Lc = d 2 x e−1 ∇ 2 + iψ − ∇+ − iψ u ∇−  . / Nuclear Physics B 584 (2000) 109–148 Z S= 125 Z d 2 x d 2 θ d 2 θ¯ E −1 L(Φi .5) As explained in [23].

2) dilaton supergravity.7) lead to the same result: Z Z    b+ V ) ∇ b− V ) b− − (∇ b+ − (∇ d 2 x d 4 θ E −1 L = d 2 x e−1 ∇  .J. the twisted chiral density projection formula (5. − (∇ d x d θ E L = d 2 x e−1 ∇   b+ V ) + iψ u ∇ b−.5) yields: Z Z    2 4 −1 b−.126 S. .6) and (5. in order to reproduce the right exponential dependence on the dilaton field in (3. and also can be verified explicitly using the commutation relations (4. b− + i ψ − − λ− ∇ b+ ∇ b+ − i ψ u − λ+ ∇ b− = d 2 x e−1 ∇   .7) + − 12 G We note that for a given superspace Lagrangian L. bu ∇ (5. / Nuclear Physics B 584 (2000) 109–148 Z 2 4 d xd θE −1 Z    b+ V ) ∇ b− − (∇ b− V ) b+ − (∇ d 2 x e−1 ∇  . both projection formulas (5. Gates Jr. As expected. V ) ∇ b+ − (∇ b+ V ) − iψ − ∇ b+ V ) b−. . . − (∇ b+ − (∇ = d 2 x e−1 ∇   − u u − b b ¯ b−˙ V ) + − 1 G b−.6) is the right projector. . Now we are ready to derive component actions for various superspace Lagrangians L. Obviously. Let us start with a simple example corresponding to pure dilaton supergravity. 2) supergravity theory [7]. − (∇ − iψ − ∇  ¯ + ψ −ψ u + ψ uψ − ∇ b− L|.8) Sgrav = d 2 x e−1 exp(−2ϕ) R (2) + 4(∂µ ϕ)(∂ µ ϕ) .17) in the new N = (2. − (∇ ∇u ∇− L|. Substituting this in the projection formula (5. V ) b+ − (∇ b−. b−.7)) and seting all the fermions to zero. in Appendix D we repeat the calculation of Grisaru and Wehlau [7] in the new N = (2. (5. we obtain the following action for bosonic fields: Z   (5.6) (or (5.6). b+ V ) b+ − (∇ + iψ − ∇  b 2 .6) bu ∇ + − 14 H¯ − ψ − ψ u + ψ − − λ− ψ u − λ+ ∇ L= In order to convince even hard boiled sceptics that (5. V ) ∇ b+ V ) b+ − (∇ b−. 2) dilaton supergravity. b− V ) + − 1 H¯ − ψ − ψ u + ψ − ψ u ∇ b− − (∇ ¯ L| − iψ u ∇ 2 Z     b−.5). et al. L. the result is equivalent to (5. b− V ) + − 1 H¯ − ψ − ψ u + ψ − ψ u ∇ b− − (∇ ¯ L| − iψ u ∇ 2 Z    . we have to take the function L in the form: Lgrav = exp(−2V ). + iψ u ∇ 2 +ψ ψ +ψ ψ This follows from the corresponding property of the local density projectors in gauged N = (2. b+ V ) b+ − (∇ + iψ − ∇  b 2 . Similarly.

11) The subscripts on L denote derivatives with respect to the scalar fields. Σj .5) describing compactification of type IIA string theory on a Calabi–Yau four-fold X.8) for dilaton and graviton fields agrees 6 with the first two terms in the effective action (3. et al.8) has exactly the same form as the action of N = 0 dilaton gravity studied long time ago.J. where L is a function of all matter superfields but V . see. Now we consider coupling of N = (2. (5. V ) L|. e.11) one may find helpful some formulas from Appendix D where we discuss in detail the component action of a free chiral superfield. (log L)φi φ¯j = (∂ 2 /∂φi ∂ φ¯ j ) log L.6). Φ¯ i¯ . A careful reader may notice that (5. it is convenient to introduce the Kähler potential K: L = exp(−K) 6 We will account for the extra volume factor V in a moment. [25].S. Σ¯ j ) of chiral superfields Φi and twisted chiral superfields Σj we get the action of the bosonic fields (the vielbein determinant e−1 is suppressed): h   L = e−2ϕ L R (2) + 4∂µ ϕ − 12 log L ∂ µ ϕ − 12 log L   + 12 (log L)φi φ¯j (∂µ φi ) ∂ µ φ¯ j − 12 (log L)σi σ¯ j (∂µ σi ) ∂ µ σ¯ j i + 12  µν (log L)φi σ¯ j (∂µ φi )(∂ν σ¯ j ) + (log L)σj φ¯i (∂µ φ¯ i )(∂ν σj ) . Clearly. + − 12 H¯ − ψ − ψ u + ψ − ψ u   bu V ) ∇ b−. It is convenient to absorb exp(−2V ) in the definition of the supervielbein determinant E0−1 = E −1 exp(−2V ). / Nuclear Physics B 584 (2000) 109–148 127 Here we used (4. so that L = 1 corresponds to pure supergravity action (5.12) . we find the modified projection formula: Z d 2 x d 4 θ E0−1 L Z    b+ − 3(∇ b+ V ) ∇ b− − 3(∇ b− V ) = d 2 x e−1 exp(−2ϕ) ∇   ..9) d 2 x d 4 θ E −1 exp(−2V )L. bu − 2(∇ (5. . Commuting exp(−2V ) to the left in (5. Once again. Gates Jr. we are interested in superspace form of the effective action (3. b+ V ) − iψ u ∇ b− V ) b+ − 3(∇ b− − 3(∇ + iψ − ∇  .11) has the structure reminiscent of N = 1 supergravity in four dimensions.g. In particular. 2) dilaton supergravity to matter fields. e. Deriving (5.16) and the formulas for the other components of the real superfield V derived in Appendix C. like in N = 1 four-dimensional theory. Moreover.. we note that bosonic action (5. (5. − 2(∇ b−. In particular.8). the action (5.g.10) × ∇ Applying this projection formula to an arbitrary function L(Φi .5) of type IIA theory on a Calabi–Yau four-fold. to reproduce the exponential dependence on ϕ = V | we take the superspace action in the form: Z (5.

9) takes the form: Z Z d 2 x d 4 θ E0−1 e−K = d 2 x d 4 θ E −1 e−2V e−K . we constructed superspace Lagrangians describing N = (2.1). it is similar to the superspace action of N = 1 supergravity in four dimensions.9)–(3.7): K → K + Λ1 (φi . (5.10). the superspace action is expected to be a function of 2V + K. Therefore. σ¯ j ) + Λ2 (φi . This form of the superspace action might be expected for a number of reasons. unlike (5.2) and (5.14) are manifestly invariant under the generalized Kähler transformations (5.13) gives the effective action (3. we provide a superspace formulation of the effective field theories constructed from compactification of type IIA string theory on Calabi–Yau four-folds.15). (5. σj ) + Λ¯ 1 (φ¯ i¯ . Namely. σj ). the chiral density projection formula must look like: Z Z Z  2  . both the superspace action (5.1). the chiral and twisted chiral density projectors in N = (2. so that the linear combination e ϕ = ϕ + 12 K remains invariant.10). σ¯ j ) + Λ¯ 2 (φ¯ i¯ . where K is the total Kähler potential given by (3. .13) Performing the superspace integration.16) d 2 x d 2 θ E −1 W = d 2 x e−1 ∇ 2 . However.5) if K is the total Kähler potential (3. one finds Lagrangian for the bosonic fields: h  ϕ ϕ )(∂ µ e ϕ ) − 12 Kφi φ¯j (∂µ φi ) ∂ µ φ¯ j R (2) + 4(∂µ e L = e−2e  + 12 Kσi σ¯ j (∂µ σi ) ∂ µ σ¯ j − 12  µν Kφi σ¯ j (∂µ φi )(∂ν σ¯ j ) i (5. Gates Jr. 2) dilaton supergravity do not simply follow from the full superspace projector (5.5) of type IIA theory on a Calabi–Yau four-fold.15) Under this Kähler transformation the original dilaton field ϕ is shifted in the opposite way. type IIA supergravity on a Calabi–Yau four-fold has a breathing mode corresponding to rescaling of the volume V → c2 V and simultaneous shift of the dilaton ϕ → ϕ + log c. 2) sigma-models with torsion coupled to dilaton supergravity. by dimensional arguments and from an examination of the index structure of the possible terms. These terms are superinvariants obtained by integration only over a half of the superspace. cf. [6]. b + · · · − 1 H¯ − ψ − ψ u + ψ − ψ u W | (5. Therefore. (3.128 S. Now we are in position to identify the function K that would reproduce the effective action (3.13) and the corresponding component action (5. (5. as well as more general N = (2. et al. / Nuclear Physics B 584 (2000) 109–148 so that the superspace action (5. the superspace action (5. where we introduced a new dilaton field e ϕ = ϕ + 12 K invariant under generalized Kähler transformations (3. 2) dilaton supergravity coupled to matter superfields and found projection formulas that allow one to rewrite integrals over the entire superspace in terms of component fields. Since the Kähler metric is invariant under (5. Incorporation of superpotential terms is more subtle. First of all. To summarize. Moreover.15) as well. cf.5).14) + Kσj φ¯i (∂µ φ¯ i )(∂ν σj ) .9). cf. Unfortunately.J.

11).S. Gates Jr.16) and (5. Moreover. mψ − ∼ W (5. are independent on degauging.16) and (5.17) may not be correct. ∼ W. 7 By the similar reasoning.17) lead to the expected structure of the scalar potential (3. 2) supergravity that was needed in [14. Extending the computation of (5. ∇ (5.15] in order to find the superpotentials (3.17) = d 2 x e−1 ∇ 2 + ψ ψ + ψ ψ W |. e.5): Z Z e d 2 x d 2 θ Ee−1 W Z   − u u − e b+ + · · · − 1 G ¯ b−. the terms quadratic in the gravitino are rather general and. Although normalization and coefficients in (5. from (5.1) we get the action of the auxiliary fields: .18). So. let us demonstrate that (5. the twisted chiral density projection formula must look like (5.18) It is this property of N = (2. we infer that one effect of the superpotential is to produce a mass term for the gravitino fields [23]: mψ −.13) and (3. / Nuclear Physics B 584 (2000) 109–148 129 where the dots stand for term containing λα or terms linear in covariant derivatives.14) induced by Ramond–Ramond fluxes. in particular.J. et al.

2 .

.

2 .

Laux ∼ L .

12 H − i A¯ i (log L)φi .

+ .

12 G − i B¯ j (log L)σj .

19) −i W 2 Integrating out the auxiliary fields and using (5. Namely.17). 4) supergravity theory will be presented in the next section. + (log L)φi φ¯j Ai A¯ j   − (log L)σi σ¯ j Bi B¯ j − i(Wφi − (log L)φi )Ai − W 12 H¯ + iAi (log L)φi   ¯ + iBi (log L)σi + c. 6. . find that the supergravity multiplet includes a real scalar dilaton field instead of an SU(2) gauge field [29. A manifestly supersymmetric formulation of such N = (0. Compactification of type IIB string theory on Calabi–Yau four-folds Compactification of type IIB string theory on a Calabi–Yau four-fold X leads to a chiral N = (0. e 1G eσi − (log L)σi Bi − W (5. all left-moving modes are 7 These terms will not affect the action of bosonic fields.30]. since we deal with N = (0. In the low-energy limit this theory is described by N = (0.12). 31]. In this section we perform a Kaluza–Klein reduction on a Calabi–Yau fourfold X and.18):    ¯  − |W |2 − |W e e Dσ¯ j W e |2 Dσi W Laux ∼ eK Kφ−1φ¯ (Dφi W ) Dφ¯j W¯ − Kσ−1 i σ¯ j i j with the covariant derivatives (3.c. In this section we show that Kaluza–Klein harmonics combine into N = (0. 4) supersymmetric effective field theory in two non-compact dimensions. 4) supergravity coupled to scalar superfields. in particular. we find the expected scalar potential (3. there is a significant difference between right-movers and left-movers. 4) supersymmetry. 4) scalar superfields [30. Furthermore.

As in type IIA theory. the axion l. the dilaton ϕ.3).7) + H · M · H + F +··· . Finally. from the reduction of gMN we find h1. cλ + d (6.1 complex scalars φj and the two-dimensional metric gµν . We will also see that the difference between the zero-point energy of the left-movers and the right-movers is proportional to the Euler number of X [32]. the field strength FMNP QR is a singlet under the SL(2. In the large volume limit the bosonic spectrum of light modes in type IIB string theory includes the metric gMN .   1 |λ|2 − Re λ . c and d satisfy ad − bc = 1.5) where the integer numbers a. Non-perturbative type IIB string theory is invariant under SL(2. while H transforms as a “vector”. Z) duality group. Z) duality group.1) (6.2) (6.4) FMNP QR = ∂[M DNP QR] + B[MN 4 Then. in order to find the zero-mode spectrum we have to expand type IIB supergravity fields in harmonic (p. it is convenient to define the following quantities: λ = l + ie−ϕ . In order to see the action of this group on the supergravity fields.3) and 3 NS RR ∂P BQR] .130 S.6) Although this equation can not be derived from any action. M= 1 Im λ − Re λ  NS  ∂[M BNP ] . . h3. SL(2. q)-forms on the space X. Together with the fermionic superpartners all these fields fit tensor fields BMN MN into type IIB supergravity multiplet. Namely. it is the Kaluza– Klein mode corresponding to the volume of the Calabi–Yau four-fold: V = dij kl s i s j s k s l . for a moment we ignore this subtlety and write bosonic “type IIB supergravity Lagrangian” simply as:  √ 1 1 Tr(∂M · ∂M) L(10) = −g − R (10) + 4 16  5 2 3 T (6.1 real scalars si . It turns out that one of the scalars si comes into the two-dimensional supergravity multiplet. The metric modes are exactly the same as in (2. 16 24 where the dots stand for higher derivative terms. et al.J. Z) acts on a complex scalar λ in the usual way: λ→ aλ + b . the 4-form tensor DMNP Q and two RR and B NS . / Nuclear Physics B 584 (2000) 109–148 supersymmetry singlets. Gates Jr. Namely. b. HMNP =  RR ∂[M BNP ] (6. The five-form field strength F is self-dual: F = ∗F. (6. (6.

g. (6. ω2 ) = X ω1 ∧ ω2 .1 − 2h2. instead of D one has P to expand the field strength F ∼ i ∂µ ui · ωi(4) and impose the self-duality condition (4) (6.2) = 45 + 3h1. where we denote by B+ (X) (respectively B− (X)) the space of (anti-)self-dual 4-forms on X. Let us call the corresponding dimensions b± = dim B± (X). [8]): Z  χ 1 7p2 − p12 = + 32 τ (Q) = b+ − b− = 45 3 X R of the quadratic form Q(ω1 . we also have b4 = b+ + b− = 2 + 2h3. we have to distinguish carefully self-dual and anti-self-dual harmonics of F . Gates Jr.2) = h1. while the (4.6).9) and b− = −1 + h1.2 . 4)-forms on a Calabi–Yau four-fold are self-dual [14].1) ri ωi .J. / Nuclear Physics B 584 (2000) 109–148 131 where dij kl are the intersection numbers of X given by (2.8) i=1 we get pairs of real scalars ri and ti . Then b+ and b− are related by the Hirzebruch signature 8 (see.1 + h2.1 − 2h2. R) = B+ (X) ⊕ B− (X). 1 2 2 4 8 . All these modes are both right-moving and left-moving. from (6.11) (2.1 − 1. 3) are anti-self-dual. et al.1 (6. All the forms of Hodge type (3.1 in number. R) is self-dual or anti-self-dual the scalar field ui is left-moving or right-moving.10) Actually. Expanding the doublet of tensor fields as: 1. Hence.1 − 1 scalars sˇi = V − 4 si satisfying the condition: 1 dij kl sˇi sˇ j sˇk sˇ l = 1. e.. To this end we recall some topological properties of Calabi–Yau four-folds.1 (6. (6. Therefore. 0). 1) or (1. we can be a little bit more precise. B RR = i=1 h X (1. Namely.1 B NS = h X 1. Expansion of the self-dual field D is a bit subtle. Depending on whether the form ωi ∈ H 4 (X.1) and (2. (6.1 + 2h3.1 + 4h3.and (0. p = − 1 tr R 4 + 1 (tr R 2 )2 . There is a decomposition of the space of the middle-dimensional forms on X: H 4 (X.1) ti ωi .S.1 . h1.1 (1.2) we find: b+ = 47 + 3h1.10) we find that: (2. On the other hand.7).9) and (6. the Kähler deformations of the metric yield h1.12) b+ and b− 8 The explicit form of the Pontryagin classes is given by: p = − 1 tr R 2 .1 + 4h3. using (2. With this mode excluded. respectively. Therefore.

sˇj .17) . Furthermore. the former are singlets with respect to four left supercharges satisfying:  i j Q+ . n− − n+ = 2 ind(D Using (6. Note that the Kaluza–Klein modes of the self-dual field F associated with (2. vj . therefore. tj . So. is not determined by supersymmetry.J. (6.1) + (∂µ q)Ω + c. however. et al. do not lead to new propagating degrees of freedom. 4) supersymmetry. while pk and q are complex.15) and the explicit expression for the Rarita–Schwinger index on a Calabi–Yau four-fold X:  Z  1 37 2 p1 − 31p2 = −4h1. rj ..15) for the total number of the right-moving fermions. 4) supersymmetry. ui and q must be left-moving. in type IIB compactification on on a Calabi–Yau four-fold X not all the fermionic modes can be determined by N = (0.1 .2 3.1 + 4h1. their number (minus the number of rightmoving fermions) is given by the Rarita–Schwinger index: / 3/2 ).1 (6. Since the fermions in question come from the type IIB gravitinos. 1)-harmonic forms give 2-form field strengths of massless vector fields in two dimensions and. 2) theory constructed from compactification of type IIA string theory.13) F= where the scalar fields ui and vj are real.1 / 3/2) = (6.1 i=1 j =1 k=1 b+ b− h X X X (+) (∂µ ui )ωi + (∂µ vj )ωj(−) + (∂µ pk )ωk(3. it has to be computed separately.14) ind(D) 1440 4 1 X in accordance with N = (0. Unlike N = (2. As we explained above. φ¯ i¯ . In the right sector we still can use supersymmetry arguments to conclude that two-dimensional supergravity multiplet / fermions and 2 ind(D) / Rarita–Schwinger fields that come from the contains 2 ind(D) corresponding spin. all the right-moving scalars found above (φi . p¯i¯ .16) ind(D 180 4 X we obtain: n− = 4h2.1 + 4h2. 9 Note that n is divisible by 4. Gates Jr. Simple counting gives: n+ = 4h3.132 S.c. 9 There are also left-moving fermions which are supersymmetry singlets. The number of left-moving fermions. vj and pk must be right-moving.21 and spin. Q+ = δ ij P+ . pi . + (6.1 − 4h3. On a Calabi–Yau fourfold the Dirac index is given by:  Z  1 7 2 p − p2 = 2 / = (6. ϕ and l) are accompanied by right-moving fermions. In particular. / Nuclear Physics B 584 (2000) 109–148 Now we expand the self-dual field strength F as: 2.2 2.23 fields in type IIB supergravity.

ϕ. c and d real numbers obeying ad − bc = 1. (2. ti ) transforms as a vector under the general SL(2. 4) supergravity. cf. In order to find the low-energy effective action one has to substitute (2. [30]. 4) theory is SL(2. 4) supergravity theories [29. l. However.33]. with a. including the supergravity itself.c. In compactification of type IIB string theory on a Calabi– Yau four-fold we find that the gravitational multiplet includes a real scalar V instead of SU(2) gauge field. Note.13) into (6. In the next section we describe the superspace formulation of this N = (0. only the scalar multiplets Σi transform nontrivially under this symmetry. are SL(2.1 scalar multiplets Σj : gµν .5) we get the following two-dimensional action for bosonic fields:   √ ¯ L(2) = −g V − 14 R (2) + 12 (∂+ λ)(∂− λ) + 12 Gφi φ¯ ¯ ∂+ φ i ∂− φ¯ j j  ¯ + 12 Gφi φ¯ ¯ ∂0 pi ∂+ p¯ j j          1 + 2 Gσi σ¯ j ∂+ s i ∂− s j + ∂+ r i ∂− r j + ∂+ t i ∂− t j + ∂0 v i ∂+ v j    1 + 12 Qij ∂0 ui ∂− uj + (∂0 q)(∂− q) + c. T ∗ Mc (X) × T ∗ MK (X).7). while all the other fields.18) This Lagrangian describes non-linear sigma-model interacting with N = (0.S. Gates Jr. Apart from ϕ. vj . R) transformation. V. R)-singlets. et al. Integrating over the internal space by means of the formulas (2. 4) supergravity using the Goldstone approach. .3). N = (0. In other words. 4) supersymmetric sigma-models. all the matter multiplets include four real scalar fields in accordance with the general classification of scalar superfields [29–31. Since the moduli space itself is a Kähler space. φ¯ i . The complex scalar λ = l + ie−ϕ transforms as (6. φi .1 scalar multiplets Φi : h1. sˇj . this result agrees with the general analysis of N = (0.4).J. The target space of the left-moving fields is the cotangent bundle to the moduli space of the Calabi–Yau manifold X. . The doublet of real scalar fields (ri .8) and (6. 2 (6. rj . tj .30]. R) symmetry of classical type IIB supergravity. the content of the gravitational multiplet is different from what was usually studied in N = (0. / Nuclear Physics B 584 (2000) 109–148 133 Now we are ready to assemble the supermultiplets. ri and ti . According to [33]. R) duality transformation. Another interesting feature that we expect to see in this N = (0. 4) supersymmetric sigma-model is based on a target space which has three covariantly constant (with respect to ∇+ ) complex structures which obey the quaternionic algebra: Jr Js = −δrs + frs t Jt . (6. l.5) under SL(2. Combining the left-moving bosonic modes with the fermion fields we get the following supermultiplets: a gravitational multiplet : h3. Performing a reduction of the ten-dimensional supersymmetry conditions one can easily check that these fields indeed represent bosonic components of the supermultiplets as stated. b. pi . all the scalar fields listed above are singlets with respect to this symmetry. p¯i .

19) P+ − P− = 24 in the type IIB vacuum corresponding to compactification on the same Calabi–Yau fourfold X. ∇u i .J.2) in type IIB compactification on a Calabi–Yau four-fold. for the sake of simplicity. 4) dilaton supergravity In this section we construct N = (0. Therefore. due to N = (0.1 + 6h3. cf. et al. for example. First let us remind that gauged N = (0.1) on the covariant derivatives ∇A ≡ (∇+i . (7. Superspace formulation of N = (0.19). ∇ } = − 12 Σ +i ∇+i + Σ¯ + i ∇u i + RX + iFi j Yj i (7. we assumed that there are no background fluxes. Gates Jr.134 S.e.2) . one can easily obtain the formula (6. In the left sector we have 48 + 6h1. Below we present a superspace construction of this theory obtained via de-gauging SU(2) symmetry. One can interpret (6. ∇u j } = i2δi ∇ . / Nuclear Physics B 584 (2000) 109–148 Finally. in type IIB superstring compactification on Calabi–Yau four-folds. Hence. Here. rj . (6. As we demonstrated in the previous section such a theory has a number of distinct features which are absent in the existing superspace formulations of two-dimensional N = (0.1 fermionic modes. in type IIA vacuum we had to include N = χ/24 fundamental strings filling two-dimensional space–time to cancel the tadpole. 4) supersymmetry. ∇ ): ∇A = EA B DB + ΛA X + iAAi j Yj i . ui .1 − ¯ ϕ and l along 2h2. 4) dilaton supergravity that arise. ∇ . ∇ } = 0. in the right sector bosonic and fermionic contributions cancel each other. the total vacuum momentum in the left sector is non-zero and is given by the following formula:  1 48 + 6h1.   [∇ .2).30]. In order to see this. q. 4) supergravity is defined in superspace by the following set of constraints [34]: [∇+i .1 + 6h3. we remark that from T-duality with type IIA string theory on a Calabi–Yau fourfold we expect an anomaly similar to (3. with n− = 4h1. (6. P+ = 0.17). j [∇+i . 4) supergravities. φ¯i¯ . ∇ } = −i Σ¯ + i X − Σ¯ + j Yi j . ∇+j } = 0. q.. i.1 . Namely. 7.   [∇+i .19) as the difference in the zero-point energy of the left-moving and the right-moving Kaluza–Klein modes [32].1 − 6h2. sˇj . unlike the standard formulations with gauged SU(2) R-symmetry [29.20) P− = 24 Using P+ = 0 and the explicit expression (2. Under a T-duality in one of the space–time directions the winding modes of these strings transform into χ/24 momentum modes: χ (6. tj . Recall that due to the global anomaly (3.2) for the Euler number. we note that a free boson on a circle has vacuum energy −1/24 and a periodic fermion has vacuum energy +1/24. new supergravity does not have a gauged symmetry and the supergravity multiplet contains a real dilaton field V.1 bosonic modes corresponding to the fields φi . [∇+i .

3) to derive the form of the commutator algebra for the bA operators. b+i . We next use (7. For this purpose.  k  k Yj . ∇u i = −δji ∇+˙ k + 12 δjk ∇u i . et al. ∇+i Σ +j = 12 δi R + iFi j . we have completed half of the de-gauging process. We write [ . ∇ } = ∇ . The leading component of A+i j k is a component of the gauged supergravity multiplet that could be eliminated by a gauge transformation in the SU(2) gauge-symmetric phase. The first step in obtaining two-dimensional N = (0. The second half consists of specifying the spinorial SU(2) connections in terms of some components of another (matter) multiplet that is consistent with the two-dimensional N = (0. DA denotes the flat space fermi and bose derivatives DA ≡ (D¯ +i . The superfield AAi j is SU(2) gauge connection. ∂ . 4) supergravity with gauged SU(2) symmetry. 4) supergravity theory that does not contain a gauged SU(2) is to note that the covariant derivative in (7. In the case of rigid supersymmetry this multiplet is described by: . Finally. D+ i . / Nuclear Physics B 584 (2000) 109–148 135 Here EA B is the supervielbein. 2 [X .1) lead to a set of Bianchi identities that are solved if: ∇u i Σ¯ +j = 0. b+k − iΣ + i X . 2 [X .  k Yj . ∇ A straightforward set of calculations leads to:   b+j } = −i A+i j k + A+j i k ∇ b+k . ∇ b } = iA i k ∇ b+k . j ∇+i Fj k = −2δik ∇ Σ¯ + j + δjk ∇ Σ¯ + i . ∇ [∇ (7. ∇+i = δik ∇+j − 12 δjk ∇+i . Yj .31].S. ∇ } = −∇ .2) can be split as: bA + iAAk l Yl k . ∇+i R = i2∇ Σ¯ + i . and X and Yji are the Lorentz and SU(2) symmetry generators.  1 X . At this stage.J. it can not describe two-dimensional N = (0. Similarly. ∇ = 0. } for the graded (anti-)commutator. ∇A = ∇ (7.4) ∇   i 1 +i + b }=− Σ ∇ b+i + Σ¯ i ∇ bu + RX . b+i . while ΛA stands for the Lorentz spin-connection. Gates Jr. respectively. 4) supergravity theory. b . for the action of SU(2) gauge symmetry generators we have:  k Yj .∇ [∇ 2 where the connection superfields now explicitly appear on the right-hand side of the equations. The Lorentz generators act on ∇A as follows: 1 [X . ∇ ∇ i  b = iA i k ∇ b+i . ∇ [∇  bu j = i2δ j ∇ b + i A¯ + j i k ∇ b+k + iA+i k j ∇ bu k . ∇u i = ∇u i . b+i . we introduce the second of the four distinct N = (0. ∂ ).3) bA does not contain the SU(2) connection nor Since the superspace covariant derivative ∇ the generator. ∇+i } = ∇+i . The constraints (7. 4) scalar multiplets (SM-II) that were discussed in [30. ∇ = 0.

Gates Jr. ∇ [∇    b+j }. SU(2)-doublet. ∇ b+k + [∇ b+i }. it can be set to zero in the Wess–Zumino gauge. A+ij k contains SU(2) representations of spin. b+i . ∇ b } = iA i k ∇ b+k .∇ b+i + Σ¯ + i ∇ bu i + RX [∇ 2 we can calculate the Bianchi identities:    b+j }. On the other hand. V¯ = V.136 S.21 and spin. D+i λ− j = 0. Du i λ− j = δji ∂ V + i∂ ϕj i . The scalar and spinor field of the matter multiplet become the dilaton and dilatino. D+i ϕj k = 2δik λ− j − δjk λ− i . b+i . ∇ b+i . ∇ [∇     k j j bu j } = i2δ ∇ b + i 2λ¯ −k δ − λ¯ −j δik ∇ b+i . In general. We note that the chirality of the dilatino is opposite to that of the gravitino. we must impose the following constraints:   A+ij k = 2λ− j δik − λ− i δjk . ∇ [∇ [∇   b } = − 1 Σ +i ∇ b . ∇ b+i = 0. The triplet of spin-0 fields ϕj i can be “eaten” by the triplet of spin-1 fields in the minimal N = (0. / Nuclear Physics B 584 (2000) 109–148 D+i V = iλ− i . + δji ∇ etc. when the SU(2) symmetry is broken. ∇ b+k + i 2λ− k δ j − λ− i δ j ∇ bu . (7. b+k . ∇ b+i λ¯ −j = iλ− i λ¯ −j − iA ∇ i j b V.5) . ∇ b+j + [∇ b+k }. We are thus led to conjecture that the new form of two-dimensional N = (0. In the locally supersymmetric theory one has to replace Dα by ∇α . [∇ i i i k b } = iA i k ∇ b+i .J. In order to gain a control over the component field content of the theory. ϕ i i = 0. gravitino. b+k − iΣ + i X .4).4) as its starting point. ∇ b+j . 4) supergravity with a component spectrum given by: eµ ν ψµ +i Aµi j λ− i V graviton. vector auxiliary fields. However. ∇ bu k + [∇ bu k }. et al.   A¯+ i j k = − 2λ¯ − j δik − λ¯ − i δjk . b+j } = −i[λ− i ∇ b+j + λ− j ∇ b+i ]. b+i . the above constraints eliminate the pure spin.23 . 4) supergravity multiplet and thus become their longitudinal components via the usual Goldstone mechanism. ∇ b+j + [∇ bu k }. dilatino field and real dilaton field may be constructed with (7. ∇ [∇ These will be satisfied if: b+i λ− j = −λ− i λ− j . At lowest order in θ α . the field A+ij k is a component field that is absent in the SU(2) gauge-symmetric phase of the theory.23 representation of SU(2). This eliminates the local SU(2) symmetry. SU(2)-triplet. ∇ b+i = 0. ∇ b+j . ∇ b+i . With this result substituted into (7. SU(2)-doublet. part of this field becomes dynamical. ϕ¯j i = ϕi j .

The explicit expressions for the density projectors can be obtained by a straightforward but tedious computation substituting (7. The work of E. (7. a current J and two weight 3/2 supercurrents G± with J -charge Q = ±1.J. Schwarz for useful discussions. Similar to the three-fold case. Recall that in a Kaluza–Klein reduction two-dimensional chiral superfields come from harmonic (3. 1)forms on X. 1)-forms. Gates Jr.3) into (7. Martin Roˇcek and John H.e. et al.. N = (2. is supported in part by the Caltech Discovery Fund. d 2 x d 2 θ E −1 L| = i d 2 x 12 e−1 C ij ∇ Z Z   j bu i + i4eψ +i ∇ bu L|. 4) dilaton supergravity theory. For a general superspace Lagrangian L. i.8).G. 2) superconformal algebra consists of two N = 2 supervirasoro algebras — one left-moving and one right-moving — each generated by an energy–momentum tensor T . The research of S. 4) gauged supergravity can be obtained by means of the following projection formula: Z Z   1 bu i ∇ bu j L| d 2 x d 2 θ E −1 12 Cij ∇ d 2 x d 2 θ d 2 θ¯ E −1 L = 2 Z   1 b+i ∇ b+j L|. 4) dilaton supergravity. 12).W. we identify these fields with marginal operators in (c. that the Zamolodchikov metric on the moduli space of Calabi–Yau four-folds is block diagonal (3. S. is supported in part by NSF Grant PHY-9513835 and the Caltech Discovery Fund.G. the component action of N = (0.S.J. Acknowledgements We are grateful to Marc Grisaru. so that twodimensional space–time is flat. Namely. is supported by the NSF grant No PHY-98-02551. Let us further assume that there are no background Ramond–Ramond fluxes. Appendix A. 3) that is responsible for the twodimensional space–time. / Nuclear Physics B 584 (2000) 109–148 137 Let us now briefly comment on density projectors in the new N = (0. grant RFBR No 98-02-16575 and Russian President’s grant No 96-15-96939. It is the first part that will be interesting to us.6) d 2 x d 2 θ¯ E¯−1 12 C ij ∇ + 2 where the corresponding chiral and anti-chiral density projector formulas look like: Z Z   b+i + i4eψ¯ + i ∇ b+j L|. d 2 x d 2 θ¯ E¯−1 L| = i d 2 x 12 e−1 Cij ∇ Similar formulas also hold in the new two-dimensional N = (0. N = (2. while twisted chiral superfields correspond to harmonic (1.6). c) . From the worldsheet viewpoint. we are going to show that two-point correlation function of vertex operators corresponding to chiral and twisted chiral superfields is zero. 2) superconformal field theory (SCFT) to free conformal theory with central charge c = (3. this compactification corresponds to adjoining c = (12. Worldsheet calculation of type IIA string amplitudes Consider compactification of type IIA string theory on a Calabi–Yau four-fold X. c) and (a.

w) 1 Φ(±1.1) and Φ(−1. ¯ = (0.138 S. z¯ ) = + reg. w) ∂z w−z However. w) (A.1) (z.1) (z. z¯ 0 ) has no singularity as w → z0 . = 2πi (A. 1) and (3.3) ¯ and In the last equality we used the fact (A. Consistency of the de-gauging procedure requires that all terms in (4. respectively.−1) i j¯ metric Gσi σ¯ j for the twisted chiral multiplets.1) (1. Appendix B. 2) non-minimal supergravity In this appendix we collect some more technical formulas that arise in the construction of N = (2. z¯ 0 ) I dw − σ¯ j φi 2G (w. respectively.−1) One might think that Gφi σj would be non-zero since the corresponding OPE has a singular part:   ∂ Ψ(±1.1) (z. w) σ¯ j (z0 .−1) (z0 .2) Now we are ready to demonstrate (3. 2G± (w.: derivative ∇ .1) . 2G∓ (w. / Nuclear Physics B 584 (2000) 109–148 multiplets. w−z (A. repeating the above arguments in the right sector one can easily see that Gφi σj is φ¯ ¯ j i · Φ¯ (−1. z¯ ) + reg.−1) (z0 .1) and Φ(−1.J. so that the commutator algebra of the new covariant bα has the form (4. z¯ ) · Φ(1. OPE is singular in both left and right sectors only for hΦ(1. and have conformal weight 1/2. In fact. z¯ ) · Φ(1.1) (z. et al.g. z¯ ) ¯ · Φ(±1. Another operator product expansion that will be useful to us is the following: ¯ · Φ(±1. They can be obtained in the operator product expansion of the supercurrents with operators Ψ : ¯ · Ψ(±1. 1) moduli: Gφi σ¯ j |z − z0 |4 φi σ¯ j = Φ(1. e. z¯ 0 ) = 0. z¯ ) = reg.1) where “reg” stands for the regular part.12).1) (z. w) ¯ · Ψ(1.1) are not the lowest multiplets must satisfy 2h = Q and 2h¯ = Q.8). 0).1) (z. Therefore. 2) dilaton supergravity via de-gauging U (1)A ⊗ U (1)V non-minimal supergravity.1) (z. They are neutral. Let us call this operators Φ(1.−1) i also zero. components in the corresponding multiplets.8) with gauge symmetry generators Y and Y 0 vanish. we conclude that the target space is locally a product of the manifold Mc (X) parametrized by the chiral fields φi and the manifold MK (X) spanned by the twisted chiral fields σj . Consider a matrix element of the target space metric Gφi σ¯ j that mixes (1. z¯ ) = 2G∓ (w.1) σ¯ j σi · Φ¯ i which correspond to the metric Gφ φ¯ for chiral multiplets and the and hΦ φ (−1. Q) ¯ so Φ(1. Extra derivative constraints arising from de-gauging N = (2. Gates Jr. Φ(1. The lowest components of (anti-)chiral (Q.2) that the operator product of G− (w.

+ ∇ λ− ) − 2e λ−. e λ+ ) = 0. λ = −i(∇ By virtue of the above relations. the Jacobi identities (4. cf. b−e b−. λ = −i(∇ b− λ−. Appendix C. ). + e λ−.S. ∇ b+ } = {∇+ . we obtain the following commutation relation: b+ } = i(λ+ + e b+ b+ . λ+ ) − F¯ = 0. (∇ Similar calculations lead to the following set of constraints: b+ λ+ − ie λ+ λ+ = 0. / Nuclear Physics B 584 (2000) 109–148 139 b+ . only half of the spinor fields are independent. et al. ∇ b e ∇− λ− + i λ− λ− = 0. e e λ−.13) are automatically satisfied. − λ−. λ+ Y + e {∇ λ+ Y 0 } λ+ )Y 0 + iλ+ ∇+ + ie λ+ ∇+ = −2(∇+ λ+ )Y − 2(∇+e b+ + λ+ Y + e λ+ )Y 0 = i(λ+ + e λ+ )(∇ λ+ Y 0 ) − 2(∇+ λ+ )Y − 2(∇+e     b+ + − 2(∇+ λ+ ) + ie = i(λ+ + e λ+ )∇ λ+ λ+ Y + − 2(∇+e λ+ ) + iλ+e λ+ Y 0   b+ + − 2(∇ b+ λ+ ) + ie = i(λ+ + e λ+ )∇ λ+ λ+ + ie λ+ λ+ Y   b+e λ+ ) + iλ+e λ+ + iλ+e λ+ Y 0 + − 2(∇     b+ + − 2(∇ b+ λ+ ) + 2ie b+e λ+ ) + 2iλ+e λ+ Y 0 . ∇+ } − 2{∇+ . b+e λ+ − iλ+e λ+ = 0. 2) dilaton supergravity evaluated at θ = 0 are. e b+e λ−. Gates Jr. + e λ− λ−. λ+ + 2iλ+ λ−. (∇ b+e λ+ ) − iλ+e λ+ = 0. ∇ i b−e b+e λ+e λ− + ∇ λ+ + 2ie λ− − i(λ+e λ− + λ−e λ+ ) − R¯ = 0. + ∇ b−. λu + ∇ λ+ ) + 2e λu + (λ+e λu + e λ = −i(∇ b b e e λ = −i(∇+ λu + ∇u λ+ ) + 2λ+ λu + (λ+ λu + λ+ λu ). ∇ 2 i b−. 2) dilaton supergravity The expressions for the covariant derivatives in N = (2. ∇ b−. + ∇ λ+ λ−. Furthermore. λ− ) + 2λ− λ−. = i(λ+ + e λ+ )∇ λ+ λ+ Y + − 2(∇ Therefore.J. + (λ−e λ−. + ∇ λ+ + i(λ+e λ−. In Section 4 we discuss two ηα are put to zero. ). ∇ 2 b+e bue e λ+e λ+ λu ). [17]: . − i(e b+ λ−. Components of covariant derivatives in N = (2. b− λ+ + i(e b+ λ− + ∇ λ+ λ− − e λ− λ+ ) = 0. ∇ b e e ∇− λ− − iλ− λ− = 0. + e λ−e λ− λ−. In both cases the natural solutions to these constraints: when either ηα or e remaining spinor superfields can be expressed in terms of an unconstrained real superfield. ∇ λ+ )∇ {∇ plus two constraints (Jacobi identities): b+ λ+ ) − ie λ+ λ+ = 0. − (λ−e λ−.

/ Nuclear Physics B 584 (2000) 109–148 bα | = ∂α . ψµα˙ . b+ ∇ ∇ 2 2 2 2 2   . 2   2  . ∇ bα | + ψµα˙ ∇ bα˙ | bµ | = Dµ + ψµα ∇ ∇ = Dµ + ψµα ∂α + ψµα˙ ∂α˙ . ∇ bβ | = 1 {∇ bα ∇ ∇ 2 (C. . in addition to the ordinary connection. i i i i i + − u − b−.2) bµ is: while the θ α component of ∇ bα . i i + bu | = D + ψ + λu ∂+ + i ψ − ∂− + i ψ u − λ+ ∂u + i ψ − ∂−. b− | = − 1 GX . b− | = − 1 H¯ X .17) we obtain the following results: b+ | = iλ+ ∂+ . From (C.3) . b−. et al. extra terms that are bilinear in the gravitini ψµα . | = −iλ−. ∂− + ψ ∂u + ψ − λ− ∂−. b− ∇ ∇ 2 2 2 2 2 and from (C.  ¯ X b |=D ∇ b+ | + i ψ u D | − 1 ψ − H¯ + ψ − G b+ ∇ ∇ 2 4   i + iψ + λ+ + ψ u ψ + + λu 2  i b b − 2iλ+ λu − (∇u λ+ ) − (∇+ λu ) ∂+ + ψ u ψ − ∂− 2  i u u i u −.1) where Dµ is the fully covariant gravitational derivative with the Lorentz connection ωµ = Λµ | that includes.J. The bβ is defined by [17]: θ α component of ∇ bβ }| bα . Gates Jr. (C. bu ∇ ∇ b− ∇ b− | = iλ− ∂− . ∇ b−. ∇ bµ ]| + ∇ bµ ∇ bµ | = [∇ bα | bα ∇ ∇ bµ ]| + Dµ ∇ bα | + ψµβ ∇ bβ ∇ bα | + ψµβ˙ ∇ b ˙∇ b bα .3) we derive the series of identities that appears below: . | = − 1 H X . ∇ ∇ b+ ∇ ¯ . b−.140 S. µ µ µ (C. Dµ is defined to be: 10 Dµ = eµ + ωµ X . Specifically. | = D + ψ ∂+ + ψ + λ−. We also need expressions for the higher θ components of the covariant derivatives. | = − 1 GX b+ ∇ ∇ ∇ 4 4 bu ∇ b−. b+ ∇ ∇ bu | = −iλu ∂u . bu ∇ ∇ ∇ 4 4   . i i b ¯ 1 1 + i b − b b b λu + ψ H ∇− ∇ | = D ∇− | + ψ D | − (∇u R) + (∇+ F ) + 2 2 2 2 2 10 In the notations of [23] this would correspond to ω = γ = ϕ . + ψ ψ − λ+ ∂u + ψ ψ ∂− .2) and (4. ∇ = [∇ β α |. . ∂−. .

− ¯ H + ψ ψ ∂u + ψ − ψ − − λ− ∂−. R) b− F¯ ) + 1 λ−. i − u ¯ ∂u + i ψ − ψ − − ψ − λ− − 2ψ − λ−. G bu | + ψ D | + (∇ |=D ∇ 2 2 2 2 2 1 . i b−. − 2iλ− λ−. + ψ ψ ∂u + ψ − ψ − − λ− ∂−. et al. i G + ψ − ψ + ∂+ + 2ψ − λ− + ψ − (ψ − + λ−. 2 2 hi i + i 1 1 b− R) + (∇ b−. 2 1 2 i + 2 i + 2 − bu ∇ b ∇ b b−. ¯ b+ R) + (∇ bu F¯ ) − λ+ H + λu G b | = D ∇− | + ψ D | + (∇ 2 2 2 2 2  1 +¯ 1 u ¯ − ψ G− ψ H X 4 4   i i + ψ − ψ + − H ∂+ + ψ − ψ − + λ−. + ψ ψ −G 2 2 . F ) − λ− H + λ−. i i + u u + ψ ψ − λ+ − iψ λu ∂u + ψ + ψ − ∂−. 2   . i − 2ψ u λu ∂u + G + ψ + ψ − ∂−. . ∂− 2 2   . . ) ∂− 2  .S. i 1 − ψ −G − ψ −H X 4 4    i i i + + + ψ ψ + λu ∂+ + H + ψ + ψ − ∂− + ψ + ψ u − λ+ 2 2 2  . ∇ ∇ b b+ ∇ ∇ b b− ∇ ∇ b bu ∇ ∇ 141 . i 1 u bu | + ψ D | − ψ − G + ψ − H X + i ψ + ψ + + λu ∂+ |=D ∇ 2 4 2  i + − bu λ+ ) + (∇ b+ λu ) + ψ ψ ∂− + 2iλ+ λu + (∇ 2   . i −. H¯ − 1 λ− G b+ | + ψ − D | − i (∇ ¯ + i (∇ ¯ |=D ∇ 2 2 2 2 2  1 + ¯ 1 u + ψ H+ ψ G X 4 4  . . − (∇ 2  . Gates Jr. + ¯ + ψ − ψ − + λ−. . ) ∂− + iψ λ− + ψ ψ + λ−. − − b−. . b− | + i ψ − D | − 1 ψ + H¯ + ψ u G X + i ψ − ψ + ∂+ |=D ∇ 2 4 2    i −. . ∂−. . i . λ− ) − (∇ b− λ−. i i −. ∂− + i ψ − ψ u ∂u + ψ ψ ∂+ + 2ψ − λ− − G 2 2 2   . . . u . 2 2  1 i − i i 1 . / Nuclear Physics B 584 (2000) 109–148    1 λ+ − ψ u G X 2   .J. . u i . i + ψ − ψ − − λ− − H¯ ∂−. 2  .

i b + (∇− R) + (∇− F ) − λ− H + λ− G − ψ G − ψ H X 2 2 2 2 4 4    i i i + ψ + ψ + + λu ∂+ + H + ψ + ψ − ∂− + ψ + ψ u − λ+ 2 2 2 . . we get: bµ | into [∇ bα ∇ Substituting the above expressions for the components of ∇ b ]| b . −.∇ b ]|. + ψ ψ ∂u + ψ ψ − λ− ∂− 2 2  bu | + i ψ + D | + ψu D ∇ 2   i b. ∇ b |] + ψ α ∇ bα ∇ b | + ψ α˙ ∇ bα˙ ∇ b | − ψα∇ bα ∇ b | − ψ α˙ ∇ bα˙ ∇ b | b . ∇ b |=D ∇ b−. b ∇ ∇ so that: b ]| = [∇ b |. D ] + D[ ψ α ] ∂α + D[ ψ α˙ ] ∂α˙   i + i b i b. i + ψ − ψ − − λ− − iψ − λ−. 1 − 1 −. .19). 1 1 . . − ψα∇ b . 1 + b + ψ D ∇u | + ψ D | + (∇ − R) + (∇ − F ) − λ− H 2 2 2 2  1 − 1 −.∇ [∇   = [D . ) + ψ − ψ u ∂u + 2iλ− λ−. + λ− G − ψ G − ψ H X 2 4 4   i i + ψ + ψ + + λu ∂+ + H + ψ + ψ − ∂− 2 2    . 2 where we also used the commutation relations (4. . u i −. ∂− 2 h i b−. ∂−.J. i 1 b− | + ψ − D | − ψ + H¯ + ψ u G X + i ψ − ψ + ∂+ + ψ− D ∇ 2 4 2    i −. Gates Jr.∇ [∇  bα ∇ b | + ψ α˙ ∇ bα˙ ∇ b | = [D . D ] + D[ ψ α ] ∂α + D[ (ψ α˙ ] ∂α˙ ) + ψ α ∇ bα ∇ b | − ψ α˙ ∇ bα˙ ∇ b |. the ∇ b |∇ b | + ψα∇ b |=∇ bα ∇ b | + ψ α˙ ∇ bα˙ ∇ b |.142 S. + ψ ψ − λ+ − 2ψ λu ∂u + G + ψ ψ ∂− 2 2   . . b is given by: b component of ∇ Furthermore. λ− ) + (∇ b− λ−. | + i ψ − D | − 1 ψ + G ∇ 2 4 2  i − − + ψ ψ + λ−. − − . . b b + iψ λ− + ψ ψ + λ− − 2iλ− λ− − (∇− λ− ) − (∇− λ− ) ∂− 2   i −. + (∇ 2 i  . et al. . / Nuclear Physics B 584 (2000) 109–148  ¯ + ψ u H X + i ψ − ψ + ∂+ b−. i + u i u + − . 1 .

∂− + H¯ + ψ − ψ u ∂u 2 2 2  i −. i − −. 2 (C. i − 2ψ u λu ∂u + G + ψ + ψ − ∂−.J. | + i ψ − D | − 1 ψ + G ¯ + ψ u H X + i ψ − ψ + ∂+ + ψ− D ∇ 2 4 2  i − − i + ψ ψ + λ−. ∂−. i i i + G + ψ − ψ + ∂+ + 2ψ − λ− + ψ − ψ − + λ−.∇ [∇ . | + i ψ − D | − ψ− D ∇ 2   1 i 1 i ¯ − 1 ψ +G ¯ − 1 ψ u H¯ X b+ R) + (∇ bu F¯ ) − λ+ H + λu G + (∇ 2 2 2 2 4 4    i − − i i − + ¯ ∂u + ψ ψ − H ∂+ + ψ ψ + λ−. . λ− ) + (∇ b− λ−. / Nuclear Physics B 584 (2000) 109–148 143   . b−. i b− | + ψ − D | − ψ− D ∇ 2       i b 1 1 1 1 i b ¯ λu + ψ + H − λ+ − ψ u G X − (∇u R) + (∇+ F ) + 2 2 2 2 2 2    . Gates Jr. ) + ψ − ψ − − λ− − iψ − λ−. i i + ψ + ψ u − λ+ − iψ u λu ∂u + ψ + ψ − ∂−. + ψ ψ − λ− ∂−. b−. ∂− + ψ − ψ u − G 2 2 2   . + 2iλ− λ−. i u − i i + ψ ψ ∂− + ψ u ψ u − λ+ ∂u + ψ u ψ − ∂−. i b−. 2 2 2  . . −. 2   . + (∇ 2  . ∂−. 2   . et al.4) Now one can compare this huge formula with the leading component of the commutator b ] computed directly from the Bianchi identities (4. 2 2  .S. − − + ψ ψ − ψ λ− − 2ψ λ−.  i 1 b+ | + ψ u D | − ψ − H¯ + ψ − G ¯ X − ψ+ D ∇ 2 4    i bu λ+ ) − (∇ b+ λu ) ∂+ + iψ + λ+ + ψ u ψ + + λu − 2iλ+ λu − (∇ 2   . .13): b . ∂− + ψ − ψ u ∂u 2 2    . bu | + i ψ u D | − 1 ψ − G + ψ − H X − ψu D ∇ 2 4  i + − i + + + ψ ψ + λu ∂+ + ψ ψ ∂− 2 2 bu λ+ ) + (∇ b+ λu ) + 2iλ+ λu + (∇    . .

J. 2 2 2 In the same fashion we obtain the relations defining the other components of the superfields R and F : 1 b ¯ i 1 b i λu F¯ + λ+ R − (∇ u F ) + (∇+ R) 2 2 2 2   i + − + − − = D[ ψ ] + ψ H + ψ ψ + ψ iψ − λ− 2    i −. ¯ i λ− R − λ− F¯ − (∇ ( ∇ R) − F ) ∇u + − − 2 2 2 2   i i 1 b ¯ 1 b. i i + ψ − ψ − ψ + + ψ u ψ + ψ + + λu + ψ − ψ − ψ + 2  2 2   i u + + + bu λ+ ) − (∇ b+ λu ) − ψ iψ λ+ + ψ ψ + λu − 2iλ+ λu − (∇ 2   i  i . . ¯ i λ− R − λ− F¯ − (∇ −F ) − R) − (∇ 2 2 2 2   i . λu F¯ + λ+ R − (∇ ( ∇ + R) F ) + ∇ + u 2 2 2 2 (C. − ψ u ψ + ψ − 2 2 2  i −. i i − ψ + ψ u ψ − − ψ − 2ψ − λ− + ψ − ψ − + λ−. 2  i . . ¯ R − λ− F¯ .4) and (C. et al. + ψ ψ + λ− − 2iλ− λ− − (∇− λ− ) − (∇− λ− ) + ψ u H + ψ + ψ − 2 2  i −. ¯ 1 b ¯ b i . F + (∇ − R) + (∇ −F ) 2 2 2 2  i = D[ ψ +] + ψ + ψ + ψ + + λu 2  i . F + (∇ ( ∇ + F ) R) + ∇ − − 2 2 2 2   i 1 b i 1 b ¯ b−. for example. − − + ψ ψ ψ + λ−. / Nuclear Physics B 584 (2000) 109–148  b . . S = (∇ − F ) + λ− − R) + (∇ 2 2 2 2 b+ | = ∂+ in (C. i = D[ ψ u] + ψ + ψ + ψ u − λ+ − 2ψ u λu + ψ − ψ − ψ u 2 2 . ¯ 1 i b. − − − ψ ψ ψ + λ−.5) where we denoted: i b ¯ 1 . λ− R − λ−. ¯ 1 b ¯ i . i − − ψ G + ψ − ψ + − ψ u ψ + ψ + + λu − ψ − ψ − ψ + − H . we find: By comparing the coefficients of ∇ i 1 b ¯ i 1 b. . Gates Jr.5). 2 1 b. b+ λ− R − λ−. − i b b .144 S.∇ b ] = λu S − λ+ S¯ + i(∇ b+ S) bu S) + i(∇ ¯ X [∇   1 b. . . λu R¯ + λ+ F − (∇ ( ∇ R) + + F ) ∇ + u 2 2 2 2   i 1 b ¯ i 1 b b− .

− ψ ψ ψ − ψ − λ− − 2ψ − λ−. Namely. 2) dilaton supergravity. i i = D[ ψ −] + ψ + G + ψ + ψ − + ψ − ψ − ψ − − λ− + ψ u G + ψ + ψ − 2 2  2   . i −. i + u −˙ − ψ ψ ψ − ψ ψ ψ − λ− − ψ ψ ψ 2 2 2  . − ψ − (∇ 2 2 2 2 4 4 Appendix D. 1 1 − ψ − G − ψ − H − ψ − ψ + H¯ + ψ u G 4 4 4  i 1 i b− R) + (∇ b−. . 2 b+ S) bu S) + i(∇ ¯ λu S − λ+ S¯ + i(∇  i b. 1 1 . . Gates Jr. −. + ψ − ψ − ψ u − ψ + ψ u ψ u − λ+ 2 2  . D ]X + ψ + (∇ − R) + (∇ − F ) − λ− H + λ− 2 2 2 2   1 . . i b G = [D . i b+ R) + i (∇ bu F¯ ) − 1 λ+ H + 1 λu G ¯ − 1 ψ +G ¯ − 1 ψ u H¯ . + λ− G − ψ G − ψ H 2 4 4  1 . + ¯ ¯ − ψ ψ G + ψ u H + ψ + ψ − H¯ + ψ − G 4  4      1 1 + 1 u i b ¯ 1 i b ( ∇ λ ψ λ ψ F ) + + − H − G R) + + ψ − (∇ + + u u 2 2 2 2 2 2  . . i − − − − b b .S. i − −. − −. 2) dilaton supergravity Here we repeat the derivation [7] of the local density projection formula in N = (2. we start by writing the most general expression for the chiral projector with the right dimension and index structure: . + ψ 2iλ− λ− + (∇− λ− ) + (∇− λ− ) + ψ ψ − λ− − iψ λ− 2  i u + −. Derivation of the projection formula in N = (2.  1 −. u i − ¯ − bu λ+ ) − ψ H + ψ ψ − ψ u 2iλ+ λu + (∇ 2    . . . et al. b+ λu ) + i ψ + ψ u − λ+ − iψ u λu − i ψ − ψ − ψ u − G +(∇ 2 2 i 1 1 i ¯ + (∇ bu R) b+ F ) λu R¯ + λ+ F − (∇ 2 2 2 2  i . 1 .J. ¯ . . 1 + ψ u ψ −G + ψ −H 4   . . / Nuclear Physics B 584 (2000) 109–148 145   i + ψ u ψ + ψ u − λ+ − 2ψ u λ+˙ 2  i i . . . F ) − λ− H + ψ u (∇ 2 2 2  1 − 1 −.

as it should be. Specifically. 2) dilaton supergravity takes the following form:  Z Z   . we evaluate (D. Φ| b− ∇ ¯ ∼ ψ D φ¯ + ψ − λ− A. the right-hand side must be invariant as well. 1 ¯ − − + − H − ψ ψ + ψ − λ− ψ − λ+ ∇¯ L|. . one can easily compute the relevant components:  bu ∇ b−.3) 4 One can easily check that this expression is equivalent to the chiral density projector (5.2). . Substituting these into (D.2) Clearly. X+ = i ψ − − λ− . Φ| ¯ b+ ∇ ¯ ∼ −ψ − D φ¯ + ψ u − λ+ A. 4 With these expressions for Xα and Y the result does not depend on whether we use the chiral projector (D. et al. Gates Jr.4).1) where the coefficients Xα and Y are to be determined.1) or its complex conjugate. ∇  b− ∇ bu ∇ b−. the chiral projection formula in N = (2. of course. It suffices to consider only bosonic terms. (D.6) obtained by de-gauging the corresponding projector in the non-minimal N = (2. b− ∇ ¯ 2 Φ¯ Φ| + X− ∇ ¯ 2 Φ¯ ∇ + X− ∇ (D.146 S.   . 2 4 −1 2 −1 b b b+ − i ψ u − λ+ ∇ b− ∇+ ∇− + i ψ − − λ− ∇ d xd θE L= d xe    u  b2 . . i + i ψ − − λ− ψ − D φ¯ − H¯ A¯ + iψ u ψ − A¯ 4   . So. ¯ − i ψ − − λ− ψ u − λ+ A. this condition completely determines the unknown coefficients Xα and Y . Φ| ¯ ∼ D D φ¯ + i ψ u ψ + − λ+ ψ + D φ¯ b+ ∇ ∇  . As we will see in a moment. / Nuclear Physics B 584 (2000) 109–148 Z 2 4 d xd θE −1 Z L= b b+ ∇ b− + X+ ∇ b+ + X− ∇ b− + Y ]∇ ¯ 2 L|. u .1) for ¯ of a free chiral multiplet: the kinetic action L = ΦΦ Z Z h    2 4 −1 ¯ b b b2 ∇ b+ ∇ b− Φ | ¯ 2 Φ¯ Φ| + ∇ ¯ 2 Φ¯ ∇ d x d θ E ΦΦ = d 2 x e−1 ∇    2  b b ¯ 2 Φ¯ ∇ ¯ 2 Φ¯ ∇ b+ Φ | + ∇ b Φ | b− ∇ − ∇    b b b+ ∇ b+ Φ | ¯ 2 Φ¯ Φ| + X+ ∇ ¯ 2 Φ¯ ∇ + X+ ∇      b b b− Φ | + Y ∇ b2 Φ Φ| . 1 Y = − H¯ − ψ − ψ u + ψ − − λ− ψ u − λ+ . Using the formulas in Appendix C along with the definition (4.  X− = −i ψ u − λ+ . Following [7]. + − ¯ bu ∇ b−. the left-hand side of this formula is invariant under complex conjugation.J. d 2 x e−1 [∇ (D. we find that the resulting component action is invariant under complex conjugation only if we put:  . ∇  . 2) supergravity.

2) supergravity. Nojiri. Nucl. B 480 (1996) 213. S. Haack.. Rev. M. Warner. Nishino. International Press. Gates Jr. Becker.J. Giddings. invariant actions. Grisaru. Gates. On flux quantization in M-theory and the effective action. S. Rocek. B 492 (1997) 623. 0) heterotic string with Wess–Zumino term. Strominger. Quant.J. Harvey. Yau. superpotentials and calibrations. P. M. E. Klemm. Witten. Lectures at the 1987 ICTP summer Workshop in High Energy Physics and Cosmology. S. A. Phys.J. Mod. Phys. S.C. Grisaru. S. Dixon. Hull. M. Gates Jr. Grisaru. S. A. Phys. Lian. Vafa. D 58 (1998) 084026. Off-shell formulation of N = 2 non-linear sigma-models.T. Phys. M. S.E. B 518 (1998) 515. N. Nucl.-T. S. S. Nucl. Gates Jr. Vafa. T. J. 7 (1990) 1033. Phys. X. Odintsov. hep-th/9705131. Grav. C. M-theory on eight-manifolds. Siegel. Duality in heterotic vacua with four supercharges. A. Yau (Ed. E. Nucl.S.). C. Troost. B 286 (1992) 78. Twisted multiplets and new supersymmetric nonlinear sigma models. Wehlau. Nucl. Vector multiplets and the phases of N = 2 theories in 2D through the looking glass.E.E.D. Phys. E. M. Becker. A 1 (1986) 191. Gates. 4 (1987) 11. G. Prepotentials for (2. Phys. B 248 (1984) 157. Callan. Papadopoulos. Superspace.-T. Evidence for F-Theory. Class.J. D = 2 supergeometry. hep-th/9912181. Quantum cosmology in the models of 2d and 4d dilatonic supergravity with WZ matter. Klemm. M. A 10 (1995) 753. B 477 (1996) 155. Candelas. J. Rabinovici.J. A. Gates. 22 (1997) 1. Rocek. Benjamin-Cummings. S. Phys. CFT’s from Calabi–Yau four-folds. Chiral rings in N = 2 superconformal theories. Grav. de la Ossa. Lerche.T. Howe. W. Rev. Mirror Symmetry II. Class. S. C. D 45 (1992) 1005. J. B.T. Phys. S. Mod. JHEP 9711 (1997) 004. B. Lett. Sevrin. K. B 324 (1989) 427. E. M. S. S. Lett. Phys.. P. hep-th/9911011. Geom. Lett. Evanescent black holes. Nucl. Roan. B 541 (1999) 615. Phys.-S.M. Nucl. Grisaru. M. / Nuclear Physics B 584 (2000) 109–148 147 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] S. J. Superspace measures. S. Phys. Phys. Lett. A study of general 2D. Yau (Eds. 2) supergravity. Kadoyoshi.-T. B 460 (1996) 579. B 584 (2000) 69. Nucl. N = 2. Wehlau. M. de Wit.S.and F-theory compactifications. Solitons. C. M.T. Witten. Phys. Nucl. Gates. Essays on Mirror Manifolds. Greene. Witten. 1997. B. Phys. et al. 1983. International Press. Sethi.T. N = 2 matter coupled to supergravity in superspace. Calabi–Yau fourfolds for M. Int. 0) superstring and σ -models. B 496 (1996) 403.P. Vafa. B 457 (1995) 219. Fayet–Iliopoulos potentials from four-folds. Gukov. Lerche. Phys. C. preceding paper in this issue. and component projection formulae for (2. Moduli space of Calabi–Yau manifolds. Phys. Two-loop finiteness of D = 2 supergravity. 1992. Phys. Supergravity with N = 2 in two dimensions.J. B 352 (1995) 43. Phys. . Wehlau. Vafa. E.). H. Alnowaiser. Sezgin. W. J. Nucl. Nucl. B 355 (1991) 455. B 355 (1995) 117. Phys. Ectoplasm has no topology. Alternative N = (4.J. The (4. Grisaru. H. Gukov. M. W. C. A. E. Nicolai. L. On the geometry behind N = 2 supersymmetric effective actions in four dimensions. Nucl. Quant. Bergshoeff. Constraints on low-dimensional string compactifications. M. Louis.

Phys. [34] R. Phys. et al. S. Townsend. Dasgupta.J. A note on low-dimensional string compactifications.M. B 345 (1995) 233. Lett. Manifest (4. Rana. B 389 (1996) 264. Phys. B 398 (1997) 285. Gates Jr. Lett. Gates Jr.148 S. S... Hull. Potentials for (p. [33] C.J. / Nuclear Physics B 584 (2000) 109–148 [31] S. Phys. Mukhi. 0) and (1. sigma models and the ADHM instaton construction. Rana.J. G. 0) supersymmetry.K. 1) supersymmetric sigma models with torsion. [32] K. Lett. Lett. 0) supergravity-scalar multiplet systems for a cognoscente. L. B 316 (1993) 291. . A canticle on (4. Papadopoulos. P. L. Gates Jr. Dhanawittayapol.

but from a practical viewpoint a reduction to a smaller subset of the fields can be more manageable and explicit.1 . All rights reserved. All rights reserved. Introduction In this paper. 0550-3213/00/$ – see front matter  2000 Elsevier Science B. it is the pseudoscalars that are the most difficult to handle. In this paper we address this problem in two cases.nl/locate/npe Geometry of the embedding of supergravity scalar manifolds in D = 11 and D = 10 M.V. Pope b.elsevier.7 . C. TX 77843. University of Pennsylvania. The consistency of the full SO(8) reduction on S 7 was proven in [1]. The N = 2 truncation includes a total of six scalar fields. comprising three dilaton/axion pairs.Nuclear Physics B 584 (2000) 149–170 www. to give gauged supergravities in lower dimensions. 2 Research supported in part by DOE grant DE-FG03-95ER40917. PII: S 0 5 5 0 . for the purposes of lifting lowerdimensional solutions back to the higher dimension. In terms of the original 1 Research supported in part by DOE grant DE-FG02-95ER40893. arising from D = 11 via compactification on S 7 . College Station.2 a Department of Physics and Astronomy.3 2 1 3 ( 0 0 ) 0 0 2 1 5 . We briefly discuss the use of these embedding ansätze for the lifting of solutions back to the higher dimension. although at a somewhat implicit level. Lü a. 1. A proof of the consistency of the full gauged SO(8) reduction on S 7 from D = 11 was given many years ago. Philadelphia. The major complexity of the spherical reduction Ansätze comes from the spin-0 fields. and of these.1 . where there are three pairs of dilatons and axions in the scalar sector.  2000 Elsevier Science B.N.V. Cvetiˇc a. PA 19104. The first example is the embedding of the scalars in the U (1)4 maximal abelian truncation of SO(8) gauged N = 8 maximal supergravity in D = 4. The other example involves the truncation of SO(6) gauged supergravity in D = 5 to a subsector containing a scalar and a pseudoscalar field. USA Received 15 March 2000. USA b Center for Theoretical Physics. accepted 4 April 2000 Abstract Several recent papers have made considerable progress in proving the existence of remarkable consistent Kaluza–Klein sphere reductions of D = 10 and D = 11 supergravities. with a potential that admits a second supersymmetric vacuum aside from the maximally-supersymmetric one. Texas A&M University. One arises in a truncation of SO(8) gauged supergravity in four dimensions to U (1)4 . H. we discuss two examples where non-trivial subsets of the scalar sectors of gauged supergravities are obtained by spherical reduction from a higher dimension.

150 M. alternatively. Such solutions can be constructed from the purely dilatonic ones by means of SL(2. where there are 35 scalars in the 35v . Again. discussed in [3]. / Nuclear Physics B 584 (2000) 149–170 SO(8) representations of the full theory. and two singlets corresponding to the original dilaton and axion of the type IIB theory. including the antisymmetric tensor field strengths. Cvetiˇc et al. and 35 pseudoscalars in the 35c of SO(8). and we shall not complete this part of the analysis in this paper. R) transformations. 20 pseudoscalars in the 10 and 10. subject to . one with maximal N = 8. In [2]. SO(6) symmetry. and it is by making use of this expression that we are able to obtain the D = 4 results in this paper. the complexities of the antisymmetric tensor embedding have prevented us from obtaining a full nonlinear result in that sector. Although an explicit interpolating solution is not known it is in principle describable within the truncation we are making. if an “existence proof” for the consistency of the reduction ansatz has independently been constructed. one can only be sure that the reduction is fully consistent with all the equations of motion of the higher-dimensional theory if one has the complete ansatz for all the higher-dimensional fields. the three dilatons come from the 35v . we shall concentrate on elucidating the geometrical structure of the embedding in D = 11 or type IIB supergravity. a further simplifying truncation was performed. In principle it allows one to read off the ansatz for the 4-form field strength. we shall concentrate on the ansatz for the Kaluza–Klein reduction of the metric tensor. and it is this that we use in order to obtain an explicit expression for the metric embedding for our 2-scalar truncation. and the three axions come from the 35c . in which a complete proof of the consistency of the S 7 reduction is exhibited. the inclusion of the three axions as well as the three dilatons leads to a considerably more complicated structure in the reduction ansatz. comprising 20 scalars in the 200 of SO(6). Thus the status of our D = 5 embedding is that. SU(1) × U (1) symmetry. The truncation we shall consider retains two spin-0 fields. This particular truncation is of interest because it is large enough to include the fields that participate in two distinct supersymmetric vacua of the D = 5 gauged theory [4]. comprising one scalar from the 200 . we can appeal to the results of [1]. On the other hand. the general ansatz for the metric tensor is rather explicit. and one pseudoscalar from the 10 and 10. The reduction ansatz becomes considerably more complicated when axions are included. In both of our examples. (Or. In the present example. In this case there are 42 spin-0 fields in total. In the case of our D = 4 example. A conjecture for the metric reduction ansatz appears in [4]. as was already seen in the case of the single dilaton/axion pair of the N = 4 gauged SO(4) truncation. which is closely analogous to the known construction in D = 4 given in [1]. Specifically. Strictly speaking.) Obtaining the ansatz for the antisymmetric tensor fields is notoriously difficult. arising from type IIB via compactification on S 5 . since no proof for the consistency of the full S 5 reduction to SO(6) gauged N = 8 maximal supergravity currently exists. These results can be used in order to study the eleven-dimensional geometrical structure of general domainwall solutions in D = 4 supported both by the three dilatonic scalars and also the three accompanying axions. The second example is a truncation of the SO(6) gauged N = 8 maximal supergravity in D = 5. although only an implicit procedure for its construction is presented. in which the three axions were set to zero. In D = 5 the situation is less clear. and the other with N = 2.

it was shown in [1] (having been previously proposed in [12]) that the ansatz for the inverse of the internal S 7 compactifying metric is gˆ mn (x.13]. and restore it later. y) dy m dy n .11]. (4) d sˆ11 where ˆ mn(x. Cvetiˇc et al. (1) V= v k`I J uk` KL which transforms under local SU (8) and rigid E7 [10. and gmn (y) is gmn (x. 3 and subject to the assumption that the conjecture for the metric ansatz in [4] is correct.12]  2 = ∆ˆ −1 ds42 + gmn (x. our result can then be used to study the geometry of the RG flow describing the transition between the two supersymmetric extrema of the associated scalar potential. are the 28 Killing vectors on the and det(g (x.7] and massive type IIA supergravity on S 4 [8] have been constructed. In particular. N = 2 U (1)4 gauged supergravity in D = 4 from D = 11 2. 4 For now. y) = K mI J 1 2    K mI J K nKL + K nI J K mKL uij I J + vij I J uij KL + v ij KL . and three associated pseudoscalars σi . so that it becomes the round S 7 metric. y). Additionally. We shall also set the gauge coupling constant g equal to 1 for now. we introduce three scalars λi . whose kinetic Lagrangian is  1X (∂λi )2 + sinh2 λi (∂σi )2 . y)) mn . y) = ∆g We use the parameterisation of the uij I J and vij KL matrices described in [13]. y) ≡ ∆ˆ −1 g mn (x. (3) ∆ˆ 2 = det(gmn (y)) where gmn (x. The eleven-dimensional metric ansatz will be given by [1. y). y) dy m dy n = ∆ˆ −1 ds42 + gˆ mn (x. The three dilaton/axion pairs in D = 4 The 35v + 35c of spin-0 fields in SO(8) gauged supergravity in D = 4 are described in terms of a 56-vielbein V. then our explicit results for the 2-scalar metric ansatz is valid. we shall leave out the Kaluza–Klein gauge fields from the construction of the metric. the truncation to three dilaton/axion pairs is naturally accompanied by the four U (1) gauge fields of the maximal abelian U (1)4 subgroup of SO(8). These gauge fields are easily incorporated in the Kaluza–Klein ansatz.1. / Nuclear Physics B 584 (2000) 149–170 151 the assumption of an ultimate consistency of the S 5 reduction scheme. . Recently. 4 where gˆmn (x. and we shall add them in at the end of the derivation.M. (5) L=− 2 i 3 Further evidence for the consistency of the S 5 reduction was obtained in [5]. the complete consistent reductions of D = 11 supergravity on S 4 [6. y) is the inverse of g mn (x. 2. with the block-diagonal form   IJ vij KL uij . In principle. y) with the scalar fields all set to zero. As discussed in [2. (2) round S 7 . where certain scalar plus gravity truncations in Kaluza–Klein sphere reductions were proved to be consistent. more evidence for the consistency of the S 5 reduction was presented in [9].y) is the inverse of gˆ mn (x. In terms of the quantities uij I J and vij KL .

χi ). L=− 2 (8) i After some algebra. (7) sin σi sinh λi = χi e . we introduce the “standard” dilaton/axion pairs (ϕi . we find that uij I J and vij KL are given by KL 1 Pij QKL 4 uij = c1 (Pa1 a2 Qa1 a2 + Pa3 a4 Qa3 a4 ) + c2 (Pa1 a3 Qa1 a3 + Pa2 a4 Qa2 a4 ) + c3 (Pa1 a4 Qa1 a4 + Pa2 a3 Qa2 a3 ) + P12 c1 c2 c3 Q12 + c1 s2 s3 ei(σ2+σ3 ) Q34 + c2 s1 s3 ei(σ1+σ3 ) Q56 + c3 s1 s2 ei(σ1 +σ2 ) Q78  + P34 c1 c2 c3 Q34 + c1 s2 s3 e−i(σ2+σ3 ) Q12 + c2 s1 s3 ei(σ1 −σ3 ) Q78  + c3 s1 s2 ei(σ1−σ2 ) Q56 + P56 c1 c2 c3 Q56 + c1 s2 s3 ei(σ2−σ3 ) Q78 + c2 s1 s3 e−i(σ1 +σ3 ) Q12 + c3 s1 s2 ei(−σ1+σ2 ) Q34  + P78 c1 c2 c3 Q78 + c1 s2 s3 ei(−σ2+σ3 ) Q56  + c2 s1 s3 ei(−σ1 +σ3 ) Q34 + c3 s1 s2 e−i(σ1+σ2 ) Q12 . we make the following definitions: ci ≡ cosh λi . 1 4 vij KL Pij QKL  = −s1 eiσ1 εa1 b1 εa2 b2 Pa1 a2 Qb1 b2 + e−iσ1 εa3 b3 εa4 b4 Pa3 a4 Qb3 b4  − s2 eiσ2 εa1 b1 εa3 b3 Pa1 a3 Qb1 b3 + e−iσ2 εa2 b2 εa4 b4 Pa2 a4 Qb2 b4  − s3 eiσ3 εa1 b1 εa4 b4 Pa1 a4 Qb1 b4 + e−iσ3 εa2 b2 εa3 b3 Pa2 a3 Qb2 b3 + P12 s1 s2 s3 ei(σ1+σ2 +σ3 ) Q12 + s1 c2 c3 eiσ1 Q34  + s2 c1 c3 eiσ2 Q56 + s3 c1 c2 eiσ3 Q78 + P34 s1 s2 s3 ei(σ1−σ2 −σ3 ) Q34 + s1 c2 c3 eiσ1 Q12  + s2 c1 c3 e−iσ2 Q78 + s3 c1 c2 e−iσ3 Q56 + P56 s1 s2 s3 ei(−σ1+σ2 −σ3 ) Q56 + s1 c2 c3 e−iσ1 Q78  + s2 c1 c3 eiσ2 Q12 + s3 c1 c2 e−iσ3 Q34 + P78 s1 s2 s3 ei(−σ1−σ2 +σ3 ) Q78 + s1 c2 c3 e−iσ1 Q56 (9) . / Nuclear Physics B 584 (2000) 149–170 To shorten the subsequent formulae. cos σi sinh λi = sinh ϕi − 12 χi2 eϕi . related to (λi . ϕi In terms of these fields. the scalar kinetic terms are  1X (∂ϕi )2 + e2ϕi (∂χi )2 . si ≡ sinh λi . (6) Also. Cvetiˇc et al. σi ) by cosh λi = cosh ϕi + 12 χi2 eϕi .152 M. for future convenience.

exactly analogous to writing the downstairs metric as ds 2 = gmn dy m dy n . Y bi ≡ χi Yi2 . For example.2. − 2b1b2 Y32 K12 K78 − Y (14) In order to proceed further. the inverse of the 2-sphere metric ds 2 = dθ 2 + sin2 θ dφ 2 is written as ∂s2 = ∂θ2 + 1 sin2 θ ∂φ2 . It is also advantageous to redefine the SO(8) basis relative to the one we have used so far.M. Yi2 − Y (12) sin σi sinh λi = bi . which amounts to a triality rotation under which Kij → 1 k` 2 (Γij ) Kk` . 6) and a4 ranges over (7. such as a1 . range over the values (1. it follows that the inverse metric (14) for the system with 3 dilatons and 3 axions is a direct sum of a 4 × 4 part involving the ∂φi basis vectors. Some useful results on this topic are collected in Appendix B. Cvetiˇc et al. 153 (10) Here. 2. (13) . The derivatives do not act on any other objects m n s here. in order to provide a compact way of summarising all the components of the uij I J and vij KL matrices. (11) Yi ≡ e(1/2)ϕi . 2). / Nuclear Physics B 584 (2000) 149–170  + s2 c1 c3 e−iσ2 Q34 + s3 c1 c2 eiσ3 Q12 . it is useful to look at the geometry of the 7-sphere in some detail. The action of transformation. we substitute these results into the ansatz (2) for the inverse S 7 metric. After doing this. 4). it is just a convenient way of writing all the components of g mn in one formula. It is advantageous to introduce a new parameterisation for the dilaton/axion pairs. we have introduced P and Q simply as arbitrary antisymmetric tensors. as follows:  ei ≡ 1 + χi2 Yi4 1/2 Y −1 . The index notation is as follows. Yi2 + Y  ei2 . The metric ansatz for the three dilaton/axion pairs From the results in Appendix B. similarly a2 ranges over (3. is given explicitly in Appendix A. 8). 5 The notation for writing the inverse metric is ∂ 2 ≡ g mn ∂ ∂ . Next. i and so cosh λi = cos σi sinh λi = 1 2 1 2  ei2 . a3 ranges over (5. we find that the inverse internal metric (2) takes the form 5 ∂ˆs2 ≡ gˆ mn ∂m ∂n   2 2 2 2 2 2 2 2 e12 K57 = Y12 K13 + K14 + K23 + K24 + K58 + K67 + K68 +Y   2 2 2 2 2 2 2 2 e22 K37 + K16 + K25 + K26 + K38 + K47 + K48 +Y + Y22 K15   2 2 2 2 2 2 2 2 e32 K35 + K18 + K27 + K28 + K36 + K45 + K46 +Y + Y32 K17 2 2 2 2 e12 Y22 Y e12 Y e22 Y e32 K34 e32 K56 e22 Y32 K78 + Y12 Y +Y +Y + Y12 Y22 Y32 K12  e12 K56 K78 − 2b2b3 Y12 K12 K34 − Y  e22 K34 K78 − 2b1b3 Y22 K12 K56 − Y  e32 K34 K56 . Indices with a “1” subscript.

Z4 = µ24 + Y12 Y22 µ21 + Y12 Y (20) The function Ξ is given by e22 Y e32 µ42 + Y e32 µ43 + Y e22 Y32 µ44 e12 Y22 Y e12 Y Ξ = Y12 Y22 Y32 µ41 + Y12 Y   e22 + Y32 Y e32 Y12 µ21 µ22 + Y e12 µ23 µ24 + Y22 Y   e22 µ22 µ24 e12 + Y32 Y e32 Y22 µ21 µ23 + Y + Y12 Y   e32 µ22 µ23 .154 M. e12 Y e32 µ23 + Y22 µ21 + Y Q3 = Y22 Y e22 µ22 + Y e12 µ23 . we find ∂ˆs23 = Y12 (µ1 ∂µ2 − µ2 ∂µ1 )2 + Y22 (µ1 ∂µ3 − µ3 ∂µ1 )2 e12 (µ3 ∂µ4 − µ4 ∂µ3 )2 + Y32 (µ1 ∂µ4 − µ4 ∂µ1 )2 + Y e22 (µ2 ∂µ4 − µ4 ∂µ2 )2 + Y e32 (µ2 ∂µ3 − µ3 ∂µ2 )2 . − 2b1b3 Y22 ∂φ1 ∂φ3 − Y (16) where Q1 = Y12 Y22 Y32 µ21 + Y12 µ22 + Y22 µ23 + Y32 µ24 . +Y (18) where µi µi = 1. Z 2 = µ2 + Y 2 Y 2 µ2 + Y 2 2 3 1 1 2 3 1 3 4 e22 µ22 + Y32 Y e22 µ24 . Z3 = µ23 + Y12 Y32 µ21 + Y12 Y e32 µ22 + Y22 Y e32 µ23 . Because of the block-diagonal structure. Cvetiˇc et al. For the 4 × 4 part. Z1 = µ21 + Y e2 Y 2 µ2 + Y e2 Y 2 µ2 . we can invert the two parts separately. e12 + Y22 Y e22 Y32 µ21 µ24 + Y + Y12 Y (21) . we straightforwardly invert the inverse metric to obtain   1 X 2 2 µi Zi dφi2 + 2b2 b3 µ21 µ22 dφ1 dφ2 − µ23 µ24 dφ3 dφ4 d sˆ4 = Ξ i  + 2b1 b3 µ21 µ23 dφ1 dφ3 − µ22 µ24 dφ2 dφ4   2 2 2 2 (19) + 2b1 b2 µ1 µ4 dφ1 dφ4 − µ2 µ3 dφ2 dφ3 . where e22 Y e12 Y e12 Y e32 µ22 + Y e32 µ23 + Y e22 µ24 . / Nuclear Physics B 584 (2000) 149–170 and a 3 × 3 part involving the ∂µi basis vectors (which are constrained by the fact that µi µi = 1): ∂ˆs2 = ∂ˆs24 + ∂ˆs23 . e12 Y e22 µ24 + Y32 µ21 + Y Q4 = Y32 Y (17) For the 3 × 3 part. (15) For the 4 × 4 inverse metric. e22 Y e32 µ22 + Y12 µ21 + Y Q2 = Y12 Y e32 µ22 + Y e12 µ24 . e32 µ23 + Y e22 µ24 . we find X  2 2 e2 µ−2 ∂ˆs24 = i Qi ∂φi − 2b2 b3 Y1 ∂φ1 ∂φ2 − Y1 ∂φ3 ∂φ4 i   e22 ∂φ2 ∂φ4 − 2b1 b2 Y32 ∂φ1 ∂φ4 − Y e32 ∂φ2 ∂φ3 .

where it is understood that the µi coordinates are expressed in terms of (ξ.M. 2 µ2 = c sin 12 θ. the inverse metric is then expressed in terms of the three unconstrained basis vectors (∂ξ . d sˆ72 = d sˆ42 + d sˆ32 :    1 X Zi dµ2i + µ2i dφi2 + 2b2 b3 µ21 µ22 dφ1 dφ2 − µ23 µ24 dφ3 dφ4 d sˆ72 = Ξ i  + 2b1 b3 µ21 µ23 dφ1 dφ3 − µ22 µ24 dφ2 dφ4  + 2b1 b2 µ21 µ24 dφ1 dφ4 − µ22 µ23 dφ2 dφ3  + 12 b12 (µ1 dµ1 + µ2 dµ2 )2 + (µ3 dµ3 + µ4 dµ4 )2  + 12 b22 (µ1 dµ1 + µ3 dµ3 )2 + (µ2 dµ2 + µ4 dµ4 )2   1 2 2 2 (25) + 2 b3 (µ1 dµ1 + µ4 dµ4 ) + (µ2 dµ2 + µ3 dµ3 ) . we first note that the determinant of (25). To do this. We can now work out the eleven-dimensional metric ansatz. ∂θ˜ ). ∂ξ = −sc−1 (µ1 ∂µ1 + µ2 ∂µ2 ) + cs −1 (µ3 ∂µ3 + µ4 ∂µ4 ). (26) = det(gˆmn ) = Ξ2 16Ξ 16Ξ 3 . Since we know the inverse metric in the form (18). by introducing three angular coordinates as follows: µ1 = c cos 12 θ. given by (4). µ4 = s sin 1 θ˜ . ∂θ . we obtain the result for the downstairs 7-metric. it is helpful first to solve the constraint µi µi = 1 explicitly. s = sin ξ . Having done so. (23) Substituting into (18). expressed in terms of the four ∂µi basis vectors formed from the the constrained µi . It then follows that ∂θ = 12 (µ1 ∂µ2 − µ2 ∂µ1 ). ∂θ˜ = 12 (µ3 ∂µ4 − µ4 ∂µ3 ). µ3 = s cos 1 θ˜ . 2 (22) where c = cos ξ . / Nuclear Physics B 584 (2000) 149–170 155 There remains the problem of inverting the 3 × 3 part ∂ˆs23 of the inverse metric. the downstairs metric can then be re-expressed elegantly in terms of the redundant set of four dµi differentials. Finally. θ. θ˜ ) using (22). and hence it can be straightforwardly inverted. Cvetiˇc et al. in the form   1 X 2 Zi dµ2i + 12 b12 (µ1 dµ1 + µ2 dµ2 )2 + (µ3 dµ3 + µ4 dµ4 )2 d sˆ3 = Ξ i  + 12 b22 (µ1 dµ1 + µ3 dµ3 )2 + (µ2 dµ2 + µ4 dµ4 )2   2 2 1 2 (24) + 2 b3 (µ1 dµ1 + µ4 dµ4 ) + (µ2 dµ2 + µ3 dµ3 ) . is  2 2 2 2  2 2  µ1 µ 2 µ 3 µ 4 µ21 µ22 µ23 µ24 s 2 c2 s c . adding this to the 4 × 4 metric d sˆ42 given in (19).

Y2 Y3 2 Y1 Y3 3 Y1 Y2 4 In the absence of axions. we may note that our result (28) is consistent with previouslyobtained special cases.156 M. In particular. it is natural to define Y1 Y2 Y3 . (27) and hence from (4) that the ansatz for the eleven-dimensional metric takes the following rather explicit form: 2 = Ξ 1/3 ds42 + Ξ 1/3 d sˆ72 d sˆ11 = Ξ 1/3 ds42 + g −2 Ξ −2/3 X   × Zi dµ2i + µ2i dφi2 + 2b2 b3 µ21 µ22 dφ1 dφ2 − µ23 µ24 dφ3 dφ4 i + 2b1b3 µ21 µ23 dφ1 dφ3 − µ22 µ24 dφ2 dφ4 + 2b1b2 µ21 µ24 dφ1 dφ4 − µ22 µ23 dφ2 dφ3   + 12 b12 (µ1 dµ1 + µ2 dµ2 )2 + (µ3 dµ3 + µ4 dµ4 )2 + 12 b22 (µ1 dµ1 + µ3 dµ3 )2 + (µ2 dµ2 + µ4 dµ4 )2 + 12 b32 (µ1 dµ1 + µ4 dµ4 )2 + (µ2 dµ2 + µ3 dµ3 )2     . X1 = Y1 Y2 Y3 . Zi = ∆Xi−1 . Denoting their potentials by Ai(1) . the first factor is the determinant of 4 × 4 block involving the φi coordinates. it is a simple matter to incorporate also the associated U (1)4 gauge fields that naturally accompany this truncation of the maximal supergravity. . 4. X2 = Y2 Y3 Y1 Y3 Y1 Y2 implying that we shall have X Xi µ2i . (30) where Y1 2 Y2 2 Y3 2 µ + µ + µ . 2. Cvetiˇc et al. ∆= ∆ = Y1 Y2 Y3 µ21 + (31) (32) (33) i It can be seen that the metric ansatz (28) therefore indeed reduces to the one given in [2] if the axions are set to zero. / Nuclear Physics B 584 (2000) 149–170 where in the first expression. From (3). (28) Note that we have reinstated the gauge-coupling constant g in this expression. for i = 1. X4 = . θ. and the second factor is from the 3 × 3 block involving the (ξ. we simply replace each occurrence of dφi in (28) by dφi → dφi − gAi(1) . then the function Ξ reduces to Ξ = ∆2 . it follows that ∆ˆ = Ξ −1/3 . (29) Finally in this section. Having obtained the Kaluza–Klein metric ansatz for the three dilaton/axion pairs. θ˜ ) coordinates. 3. X3 = . if we set the three axions χi to zero.

In spherical Kaluza–Klein reductions it is always much more difficult to obtain the ansatz for antisymmetric tensors than for the metric. the three dilaton/axion pairs.) The contributions to the 4-form ansatz can be organised into different sectors. and the present case is no exception. In [2. and the associated U (1)4 gauge fields. and LKin and LCS are the kinetic terms and i = dAi(1) . 2. The ansatz for the 4-form field strength b(4) of In principle. The scalar potential the Chern–Simons terms for the four U (1) gauge fields F(2) is given by V = −4g 2 3 X  ei2 . one can only obtain limited guidance from those results that are presented in [1].M.1. Since these are instructive and useful in their own right. Yi2 + Y (35) i=1 The kinetic terms for the gauge fields are h e22 Y e32 ∗F 1 ∧ F 1 + Y e12 Y22 Y32 ∗F 2 ∧ F 2 e12 Y LKin = − 12 |W |−2 P0 Y (2) (2) (2) (2) 4 4 e22 Y32 ∗F 3 ∧ F 3 + Y12 Y22 Y e32 ∗F(2) + Y12 Y ∧ F(2) (2) (2)  e12 ∗F 1 ∧ F 2 − Y12 ∗F 3 ∧ F 4 + 2P1 b2 b3 Y (2) (2) (2) (2)  2 1 3 2 2 4 e + 2P2 b1 b3 Y2 ∗F(2) ∧ F(2) − Y2 ∗F(2) ∧ F(2) i e32 ∗F 1 ∧ F 4 − Y32 ∗F 2 ∧ F 3 .11]. where the complete and explicit ansatz for the S 7 reduction to the bosonic sector of N = 4. W ≡ P0 − 2ib1 b2 b3 . this technique is still useful here. and up to a point. we should like to obtain also the ansatz for the 4-form field strength F eleven-dimensional supergravity. We begin with a summary of the four-dimensional theory comprising gravity. it has been possible to determine the field-strength ansatz by brute-force methods. D = 4 Lagrangian The complete Lagrangian for four-dimensional N = 8 SO(8)-gauged supergravity was obtained in [10.3. SO(4) gauged supergravity in D = 4 were obtained. we find that the fourdimensional bosonic Lagrangian for this N = 2 truncation is given by  1X ∗dϕi ∧ dϕi + e2ϕi ∗dχi ∧ dχi − V ∗1 + LKin + LCS . / Nuclear Physics B 584 (2000) 149–170 157 2. Adapting these results to the notation of this paper.13]. 2 3 L4 = R∗1 − (34) i=1 where V is the potential for the scalar fields. (This method was used successfully in [3]. and in all except one of these we have obtained complete results. In other truncations. + 2P3 b1 b2 Y (2) (2) (2) (2) where P0 ≡ 1 + b12 + b22 + b32. Unfortunately.  (36) . simpler than the case in hand. the truncation to the N = 2 U (1)4 -gauged subsector was discussed.3. it seems to be worthwhile to present those results that we have obtained here. Cvetiˇc et al.

+Y   e12 Y22 Y32 P0 ∗F 2 + 2b1 b2 b3 F 2 R2 = Y (2) (2)   1  2 1 e + Y1 P1 b2 b3 ∗F(2) + b1 P0 + 2b22 b32 F(2)   4  4 + b2 P0 + 2b12 b32 F(2) − Y22 P2 b1 b3 ∗F(2)   3  3 + b3 P0 + 2b12 b22 F(2) − Y32 P3 b1 b2 ∗F(2) .2.   2 e2 2 3 3 R3 = Y1 Y2 Y3 P0 ∗F(2) + 2b1 b2 b3 F(2)   4  4 + b1 P0 + 2b22 b32 F(2) − Y12 P1 b2 b3 ∗F(2)   1 1 e22 P2 b1 b3 ∗F(2) + b2 P0 + 2b12 b32 F(2) ] +Y   2 2 2 2 2 − Y3 P3 b1 b2 ∗F(2) + b3 (P0 + 2b1 b2 )F(2) . P3 ≡ 1 + b12 + b22 − b32. in which there is one scalar and one axion [3]. + b3 P0 + 2b12 b22 Y From (34). 4. 3. we find that the equations of motion for the gauge fields are  d |W |−2 Ri = 0.3. Cvetiˇc et al. 2. / Nuclear Physics B 584 (2000) 149–170 P1 ≡ 1 − b12 + b22 + b32. +Y (2) (2) (37)  (38) (39) (40) b(4) 2. The ansatz for F b(4) was obtained for the In previous papers the ansatz for the 4-form field strength F 4 U (1) truncation in absence of the three axions [2]. for i = 1. the Chern–Simons terms for the gauge fields are h 1 2 e2 e2 1 e2 2 2 2 e2 Y LCS = −|W |−2 b1 b2 b3 Y 1 2 Y3 F(2) ∧ F(2) + Y1 Y2 Y3 F(2) ∧ F(2) 4 4 e22 Y32 F 3 ∧ F 3 + Y12 Y22 Y e32 F(2) + Y12 Y ∧ F(2) (2) (2)   e12 F 1 ∧ F 2 − Y12 F 3 ∧ F 4 + b1 P0 + 2b22 b32 Y (2) (2) (2) (2)   1 3 2 4 e22 F(2) ∧ F(2) − Y22 F(2) ∧ F(2) + b2 P0 + 2b12 b32 Y  2 1 i 4 2 3 e3 F(2) ∧ F(2) − Y32 F(2) ∧ F(2) . where   e12 Y e22 Y e32 P0 ∗F 1 + 2b1 b2 b3 F 1 R1 = Y (2) (2)   2  2 e12 P1 b2 b3 ∗F(2) + b1 P0 + 2b22 b32 F(2) +Y    e22 P2 b1 b3 ∗F 3 + b2 P0 + 2b12 b32 F 3 +Y (2) (2)   4  4 e32 P3 b1 b2 ∗F(2) + b3 P0 + 2b12 b22 F(2) .   4 4 e32 P0 ∗F(2) + 2b1 b2 b3 F(2) R4 = Y12 Y22 Y   3  3 + b1 P0 + 2b22 b32 F(2) − Y12 P1 b2 b3 ∗F(2)   2 2 + b2 P0 + 2b12 b32 F(2) ] − Y22 P2 b1 b3 ∗F(2)    e32 P3 b1 b2 ∗F 1 + b3 P0 + 2b12 b22 F 1 . Based on those results.158 M. Finally. and for the N = 4 gauged SO(4) truncation. it can be . P2 ≡ 1 + b12 − b22 + b32 .

which in turn means that one cannot write an ansatz that can be substituted directly into the eleven-dimensional action. / Nuclear Physics B 584 (2000) 149–170 159 b(4) as the sum of three terms. and we used this in writing down our results in (41) and (43). An implication from this is that one cannot write the ansatz directly on the potential Aˆ (3) . the scalar equations of motion in question here are those of the three dilatons ϕi . each with its own seen to be natural to write the ansatz for F characteristic contribution to the whole. b(4) = 0. at least within the usual secondb(4) in the S 7 order formulation of eleven-dimensional supergravity. One must work at the level of the equations of motion. (To be precise. to write an ansatz for F b reduction that identically satisfies d F(4) = 0. Thus we are led to the following construction for the 4-form field strength: b0 + F b00 b(4) = −2gU ε(4) + F F (4) (4)   1 2Y1−1 ∗dY1 − χ1 Y14 ∗dχ1 ∧ d µ21 + µ22 2g   1 2Y2−1 ∗dY2 − χ2 Y24 ∗dχ2 ∧ d µ21 + µ23 + 2g   1 2Y3−1 ∗dY3 − χ3 Y34 ∗dχ3 ∧ d µ21 + µ24 . This will be written (We shall justify these expressions below. It will be the determination of Aˆ 0 that in terms of a potential Aˆ 0 . This feature was already b(4) does not identically satisfy d F It will be noted that F seen in the truncations in [2] and [3]. as F (3) (4) (3) (3) presents the greatest difficulty. in combination with certain non-linear admixtures of b(4) from Aˆ 0 . It is not possible. Cvetiˇc et al. The term be given by X  2 i b00 = − 1 |W |−2 dµ ∧ dφ − gA ∧ Ri . since it gives a contribution F (4) (4) b(4) = 0 implies the four-dimensional equations of To see how the Bianchi identity d F motion for the scalars and the gauge fields. whose the three axion equations of motion. e22 µ22 + µ24 + Y32 µ21 + µ24 + Y +Y and ε(4) denotes the volume form on the four-dimensional spacetime. + 2g + (41) where    e12 µ23 + µ24 + Y22 µ21 + µ23 U = Y12 µ21 + µ22 + Y    e32 µ22 + µ23 . does not enter into the discussion of the b0 = 0.) Of course the contribution to F (3) precise form we have not been able to determine. F i i (4) (1) 2g 2 (42) b00 F (4) will (43) i b0 .M. b0 that identically satisfies d F Bianchi identity. First. there will be terms .) The remaining term is F (4) b0 = d Aˆ 0 . we note from the structure of (41) and (43) that after acting with d we shall have two distinct classes of terms. The point is that the Bianchi identity will be satisfied by virtue of the D = 4 equations of motion for the scalar fields and the U (1) gauge fields being satisfied. b(4) = 0 provides b(4) must satisfy the Bianchi identity d F In fact the requirement that F us with very important clues as to the correct form of the reduction ansatz.

The b00 . Secondly. We need e1 . only consider Y1 and Y . of the dilaton equation plus an admixture of the axion equation. and the second quantity in square brackets is the axion equation of motion. Of course they are all proportional to ε(4) . It means that |W |. leads to an enormous simplification when we vary LKin and last equation in (49). etc. where the first quantity in square brackets is the dilaton equation of motion. arising (with similar independent expressions involving the (Y2 . we get the combinations of the form  (45) d 2Y1−1 ∗dY1 − χ1 Y14 ∗dχ1 .160 M. where ω(3) is a 3-form dµ2i ∧ (dφi − 12 gAi(1)) ∧ ω(3) coming from the action of d on F (4) living in the four-dimensional spacetime. Cvetiˇc et al.) The requirement of the vanishing of these terms will imply the scalar equations of motion. all of the scalar quantities with i = 2 or i = 3 labels are invariant under this transformation. and all the bi are invariant. This is a combination of the ϕ1 and the χ1 equations of motion. there will be terms of the form b00 . (44) i This can immediately be seen to imply precisely the equations of motion for the four U (1) gauge fields. δ δ − χ1 . δb (48) (49) (Of course since we are focusing on the scalars with the index i = 1 at the moment. the Pa . and give terms of this type come only from d F (4) X   2 i dµi ∧ dφi − gA(1) ∧ d |W |−2 Ri = 0. Clearly. The kinetic terms of these scalar equations come from the action of d on the final three lines in (41). is an especially simple one to compare with the scalar equations of motion coming from the four-dimensional Lagrangian (34). and these just vary by the very simple rules given in (49). / Nuclear Physics B 584 (2000) 149–170 of the form dµ2i ∧ ω(4) . i. It means that we are looking at the combination that comes from the following variation of the D = 4 Lagrangian: ˆ 4 ≡ δL4 − χ1 L4 . This particular combination.. χ2 ) and (Y3 . (ω(4) will comprise terms of the form ε(4) . since it is the simpler one. and of the form d∗dYi . δL δϕ1 δχ1 (47) If we define a symbol δˆ to denote this specific combination of field variations. δˆ ≡ δϕ1 δχ1 then we find the great simplification that ˆ 12 = Y12 .e. δˆY ˆ 1 = 0. given in (39). Let us consider the second type of contribution first. The vanishing of these terms will imply the fourdimensional equations of motion for the gauge fields. δY e12 = −Y e12 . δb LCS given by (36) and (38). In fact it is     (46) d∗dϕ1 + e2ϕ1 ∗dχ1 ∧ dχ1 − χ1 d e2ϕ1 ∗dχ1 .) The ˆ 1 = 0. It remains to check that the terms of the form dµ2i ∧ ω(4) coming from the Bianchi identity vanish by virtue of the four-dimensional scalar equations of motion. χ3 ) pairs). where ω(4) is a 4-form living entirely in the four-dimensional spacetime.

4. Note that the contributions to the scalar equations of motion from the potential V given in (35) arise from the action of the exterior derivative on the term −2gU ε(4) in (41). The scalar kinetic terms in the Lagrangian (34) are invariant under global SL(2. on the other hand. and the direction cosines µi . dr 2 . Cvetiˇc et al. and can be examined in isolation from the more complicated contributions from the four-dimensional gauge fields. This part of the calculation can be seen quite easily. to all orders in scalar fields and gauge field strengths.M. (H1 H4 )1/2 (H2 H4 )1/2 . A0(3) could in principle be determined by substituting the expressions for uij KL and vij KL appearing in (9) and (10) into them. To the extent that such expressions are implicit in the work of [1]. It is given by [14] ds4 = (gr)4 (H1 H2 H3 H4 )−1/2 dx µ dxµ + (H1 H2 H3 H4 )1/2 eϕi = Y12 = fi . r f2 = (52) This solution can be oxidised back to D = 11 [14]. Domain wall solutions and their oxidation The four-dimensional U (1)4 Lagrangian (34) supports a four-charge AdS black hole solution [13]. At leading order. (H1 H2 )1/2 f3 = (H2 H3 )1/2 . where it acquires the interpretation of being a continuous ellipsoidal distribution [14–19] of M2-branes. The scalar potential in (34). it becomes a relatively straightforward matter to verify that the b(4) vanish precisely terms of the form dµ2i ∧ ω(4) that arise in the Bianchi identity for F as a consequence of the scalar equations of motion following from (34). g2 r 2 (51) where f1 = (H3 H4 )1/2 . (H1 H3 )1/2 `2 Hi = 1 + i2 . for example that it is of the general from X   j  hij µ2i dµ2j − µ2j dµ2i ∧ dφi − gAi(1) ∧ dφj − gA(1) . these terms will give rise to the linearised ansatz for the axions χi . 2. If explicit expressions for the complete ansatz for the N = 8 SO(8) gauged supergravity embedding were available. supported by the scalar fields only. In the extremal limit. We know some aspects of The contribution F (4) (3) it structure. (50) Aˆ 0(3) = i6=j where the functions hij depend on the scalars ϕi and χi . a procedure in principle exists for reading off A0(3) . χi = 0. corresponding to the usual fractional-linear group action on each of the axion/dilaton pairs. / Nuclear Physics B 584 (2000) 149–170 161 With these observations. b0 = Aˆ 0 in (41) remains undetermined. is invariant only under the SO(2)3 subgroup transformations . R)3 transformations.3]. the four U (1) gauge fields decouple and the solution becomes AdS domain wall. of the type that has proved successful in previous (simpler) cases [2. It is not clear that attempting such a substitution would be simpler than a brute-force direct attack on the problem.

the following conjecture for the inverse S 5 metric has been proposed [4]: n e eKLcd Ω ac Ω bd . The 2-scalar D = 5 embedding in type IIB In this section. in a distributed M2-brane solution one has F(4) = d 3 x ∧ dH −1 . The metric reduction ansatz The set of 42 spin-0 fields in the complete SO(6) gauged N = 8 supergravity in D = 5 [20] are described by a 27-bein V. In the early stages of the derivation. 3. ∆ˆ 2 = det(gmn (x. (57) . in terms of an SL(6. they can be oxidised back to D = 11. Applying these global transformations to the original domain walls we obtain new solutions. The eleven-dimensional metric is given by substituting the solution into (28). Cvetiˇc et al. y) dy m dy n d sˆ10  = ∆ˆ −2/3 ds52 + gˆmn (x. Thus for a distributed M2-brane the field strength F(4) always carries three worldvolume indices.1. VI α ab ). we retain all four of the scalar fields of the truncation discussed in [4]. (55) Having obtained the SO(2)3 rotated domain-wall solutions. and its embedding in the type IIB theory via an S 5 reduction. R) basis. y) ≡ ∆ˆ −2/3 g mn (x. with Yi2 = eϕi =  1 2 fi cos2 λi + sin2 λi . where H is the harmonic function in the transverse space. y) = 2KImJ KKL (56) e is the inverse of the vielbein V. 3. we consider the embedding of the 2-scalar truncation of D = 5 gauged supergravity discussed in the introduction. VI J ab V gˆ mn (x. for which the components of the vielbein are decomposed as (V I J ab . y) dy m dy n . To see this we note from (50) that with non-vanishing axions the field strength F(4) will involve components lying purely in the internal S 7 . and gmn (y) is where V the undeformed round S 5 metric where the scalar fields are set to zero.162 M. fi χi = 1 2 (fi − 1) . 2 fi cos2 λi + sin2 λi (54) ei are hence given by The Y ei2 = Y f 2 + 14 (f − 1)2 sin2 (2λi ) fi (fi2 cos2 λi + sin2 λi ) . / Nuclear Physics B 584 (2000) 149–170 τi → τi0 = cos λi τ + sin λi . The ten-dimensional metric ansatz will then be 2 = ∆ˆ −2/3 ds52 + gmn (x. y))/ det(gmn (y)). which transforms under local USp(8) and global E6 . − sin λi τ + cos λi (53) where τi ≡= χi + ie−ϕi . In terms of this decomposition. The truncation to four spin-0 fields is described in [4]. These solutions with non-vanishing χi no longer simply describe distributed M2-branes. R) × SL(2. By contrast.

3. ϕ1 . where X (60) (61) At this stage. r = 12 (ϕ2 − ϕ1 ). where we now drop the “1” subscript on ϕ1 . It is easily verified from (60) and (61) that this is a consistent truncation. / Nuclear Physics B 584 (2000) 149–170 163 The process of making the 4-scalar truncation in the vielbein V has been described in detail in [21]. Since the final result is considerably simpler than the intermediate stages we shall. r. θ = 2φ. It is convenient also at this stage to perform a labelling of indices on the Killing vectors Kij in (58). and y2 = 12 ϕ. 4) are cycled: 2 → 3. We now adopt a description of the round 5-sphere that is precisely analogous to the one that we introduced in Appendix B for S 7 .M. Substituting this into the metric ansatz (56) is a mechanical exercise that is most conveniently implemented by computer. ϕ2 . present the final answer. The potential (61) reduces to   (62) V = cosh2 ϕ X2 (2 − cosh2 ϕ) + 2X−1 − 12 X−4 sinh2 ϕ . (59) Note that the D = 5 scalar Lagrangian for this truncation is L = −2 3 X (∂ϕi )2 − sinh2 (ϕ1 − ϕ2 )(∂θ )2 − V . θ ) are related to the quantities (ρ. Cvetiˇc et al. Thus we shall have r = − 12 ϕ. φ) appearing in [21] by ρ = X−1/2 . We find that the inverse 5-sphere metric ∂ˆs25 ≡ gˆ mn ∂m ∂n is given by   2 2 2 2 + K25 + K35 + K45 ∂ˆs25 = X−1 cosh 2y2 (cosh 2r − sin θ sinh 2r) K15  2 2 2 2 + K26 + K36 + K46 + cosh 2y2 (cosh 2r + sin θ sinh 2r) K16  + 2 cos θ sinh 2r sinh 2y2 (K26 K35 − K25 K36 + K16 K45 − K15 K46 )   2   2 2 2 + K34 + K23 + K14 + X2 14 3 − cos θ + 2 cos2 θ cosh 4r K12  2 2 + K24 + 2 cos2 θ sinh2 2rK12 K34 + cosh2 2y2 K13  − 2 sinh2 2y2 K13 K24 2 . y2 = 12 (ϕ1 + ϕ2 ). we shall end up with three . 3 → 4 and 4 → 2. This corresponds to setting θ = 0 and ϕ2 = 0 [21]. i=1 √ − = e 6ϕ3 /2 . without further ado. This time. + X−4 K56 (58) The scalars (X. under which the index values (2. and the scalar potential V takes the form [4]    V = g 2 X2 1 − cos2 θ sinh2 ϕ1 − sinh2 ϕ2   + X−1 cosh 2ϕ1 + cosh 2ϕ2   1 −4 X 2 + 2 sin2 θ − 2 sin2 θ cosh 2(ϕ1 − ϕ2 ) + 16  − cosh 4ϕ1 − cosh 4ϕ2 . we impose the further truncation to the 2-scalar subsector that we really want to consider. y2 .

/ Nuclear Physics B 584 (2000) 149–170 “direction cosines” µi . (70) we find that the 5-dimensional internal metric d sˆ52 ≡ gˆmn (x. and cyclically. One can define left-invariant 1-forms σi . (68) S3. (67) ∆ It is helpful at this stage to reparameterise the direction cosines µi . φ.   −2 2  2 −2 −2 2 ∂ˆs23 = ∆ cosh2 ϕ X µ−2 1 ∂φ1 + µ2 ∂φ2 + X µ3 ∂φ3 2 − sinh2 ϕ X(∂φ1 + ∂φ2 ) − X−2 ∂φ3 . (65) Note that the 2 × 2 inverse metric ∂ˆs22 is just equal to the metric for the single-scalar truncation when ϕ = 0. and make redefinitions of the azimuthal angles (φ1 . multiplied by a factor of cosh2 ϕ. The 3 × 3 inverse metric is equal to cosh2 ϕ times the ϕ = 0 metric. subject to the condition µi µi = 1. + 2 ∆ (66) Note that the determinant of d sˆ32 is given by (µ1 µ2 µ3 )2 /(∆3 cosh4 ϕ). (69) These satisfy dσ1 = −σ2 ∧ σ3 . σ3 = dψ + cos ϑ dφ. the inversion gives the metric   sech2 ϕ −1 X dµ21 + dµ22 + X2 dµ23 . φ2 = 12 (ψ − φ). y) dy m dy n = d sˆ22 + d sˆ32 becomes . we arrive at the following expression for the inverse 5-sphere metric ∂ˆ52 : ∂ˆs25 = ∂ˆs22 + ∂ˆs23 . (63) where the 2 × 2 and 3 × 3 blocks are given by   ∂ˆs22 = cosh2 ϕ X−1 (µ1 ∂µ3 − µ3 ∂µ1 )2 + (µ2 ∂µ3 − µ3 ∂µ2 )2  + X2 (µ1 ∂µ2 − µ2 ∂µ1 )2 . Defining also c ≡ cos ξ. µ2 = cos ξ sin 12 ϑ. After manipulations similar to those in Section 2. with the correction term appearing in its second line. φ1 = 12 (ψ + φ). In fact (ϑ. s ≡ sin ξ. ψ) are just the Euler angles on as σ1 + iσ2 = e−iψ (dϑ + i sin ϑ dφ). φ2 ) as follows: d sˆ22 = µ1 = cos ξ cos 12 ϑ. Cvetiˇc et al. and three azimuthal angles φi . The inverse of the 3 × 3 block ∂ˆs23 is straightforward to calculate. (64) and ∆ ≡ (µ21 + µ22 )X + µ23 X−2 . µ3 = sin ξ. and we find d sˆ32 =   sech2 ϕ −1 2 2 X µ1 dφ1 + µ22 dφ22 + X2 µ23 dφ32 ∆ 2 tanh2 ϕ 2 µ1 dφ1 + µ22 dφ2 − µ23 dφ3 .164 M. For the 2 × 2 block.

by completing the square. we arrive at the conjectured ten-dimensional metric ansatz for this two-scalar truncation:  X∆1/2 2 c2 X−1 dξ + σ12 + σ22 1/2 cosh ϕ 4∆ cosh ϕ  2 2 1/2 2 s Ω c2 sinh2 ϕ c X∆ cosh ϕ 2 σ3 + 3/2 σ3 .2. axion and 3-form field strengths of the type IIB theory vanish when ϕ = 0. Finally. as indicated by the cross-term between the interval dφ3 on S 1 . substituting our result for the internal hatted metric d sˆ52 into (57). but with a slightly more complicated structure. We know that when ϕ is taken to be zero. Since it is a pseudoscalar. the remaining scalar field X enters in the ansatz for the self-dual 5-form. this reduces to the metric ansatz encountered in the N = 4 gauged SU(2) × U (1) supergravity embedding obtained in [22]. 3 3 2Ω ∆2 cosh2 ϕ X dξ 2 + (73) where Ω ≡ X3 c2 + s 2 cosh2 ϕ. we obtain the result d sˆ52 = c2 X−1  c2 X 2 2 2 σ σ + σ + 1 2 4Ω 3 cosh2 ϕ 4∆ cosh2 ϕ  2 s2Ω c2 sinh2 ϕ σ dφ + − .M. In that case. the scalar X parameterises deformations of S 5 corresponding to inhomogeneities of codimension 1 of the foliation by S 3 × S 1 . there is a sort of “twist” in the S 3 × S 1 product structure of the homogeneous foliating surfaces. in this twoscalar reduction. the ansatz must reduce to one encompassed by the results in [22]. When the pseudoscalar ϕ is included too. and the 1-form σ3 on S 3 . whilst the dilaton. it is advantageous to rewrite the metric (71) as the sum of squares of just five quantities. the scalar field X parameterises inhomogeneous deformations of S 5 viewed as a foliation of S 3 × S 1 surfaces. After doing this. the field ϕ enters at the . / Nuclear Physics B 584 (2000) 149–170 d sˆ52 =   sech2 ϕ  X∆ dξ 2 + 14 X−1 c2 σ12 + σ22 + σ32 + X2 s 2 dφ32 ∆ 2 tanh2 ϕ 2 c σ3 − 2s 2 dφ3 . dφ3 − + 4Ω 2Ω ∆ cosh ϕ 2 = ∆1/2 cosh ϕ ds52 + d sˆ10 (75) 3. In addition. Cvetiˇc et al. With the pseudoscalar ϕ non-vanishing. In that case. (72) In the absence of the pseudoscalar field ϕ. + 2 4∆ 165 (71) where ∆ = Xc2 + X−2 s 2 . In particular. (74) This expression reduces to the one found in [22] if ϕ = 0. the inhomogeneities remain of codimension 1. The field-strength ansatz There does not seem to be any straightforward way to determine the ansatz for the Kaluza–Klein reduction other fields of the ten-dimensional type IIB theory.

(2) The relevant bosonic equations of motion of the type IIB theory are    1 b1 2 1 b1 2 b2 bMN = 1 H R 96 MN + 4 F(3) MN − 12 F(3) gˆ MN    b2 2 − 1 F b2 2 + 14 F (3) MN 12 (3) gˆ MN . It turns out that the required 2form harmonics are given by the construction Yab = η¯ − Γab η+ . / Nuclear Physics B 584 (2000) 149–170 linearised level in the ansatz for the NS–NS and R–R 2-form potentials Aˆ (2) ≡ Aˆ NS (2) and RR ˆ A [23]. For this to be consistent with the type IIB equations of motion. The ansatz for the self-dual 5-form H b includes a Freunnd–Rubin term G(5) = 4ε(5) (we have set the gauge coupling g = 1 here). which therefore requires λ = 1 or λ = 3. b(5) ∧ F b(3) . namely m2 = −3. (77) We are assuming here that the dilaton and axion of the type IIB theory vanish in the reduction. the linearised ansatz for pseudoscalars ϕ will be of the form Aˆ (2) = ϕY(2) .166 M. There are ten such harmonics on S 5 . b(3) ∧ Fb ∗ˆ F b(3) ∧ F b(3) = ∗ˆ Fb ¯ (3) ∧ Fb ¯ (3) . Solving for the Killing spinors. (82) where η− and η+ are any two Killing spinors of the minus and plus kinds. H where we have introduced the notation that ˆ RR Aˆ (2) ≡ Aˆ NS (2) + iA(2) . b(3) = −iH d ∗ˆ F b b b(5) = − i F ¯ dH 2 (3) ∧ F (3) . ∗ˆ F (78) We shall restrict our discussion from now on to the linearised level. We want the mass for the 10 and 10 members of the massless multiplet. we find that one of the ten . (79) where Y(2) is a complex 2-form spherical harmonic satisfying d∗Y(2) = iλY(2) (80) b(5) + ∗ˆ G b(5) b(5) ≡ G on the unit round 5-sphere. it is necessary that ¯ (3) = 0. In the notation that we are using here. Substituting into the type IIB equations of motion. which can be written in terms of the Killing spinors. respectively. There are Killing spinors η± satisfying Da η± = ± 2i Γa η± . one finds that the pseudoscalar ϕ satisfies the linearised equation of motion   (81) d∗dϕ + λ(λ − 4)ϕε(5) ∧ ∗Y(2) = 0. the required harmonics are those with λ = 3 (there are none with λ = 1). Cvetiˇc et al. In fact. (76) b(5) = ∗ˆ H b(5). A 2-form harmonic with eigenvalue λ gives a pseudoscalar ϕ with m2 = λ(λ − 4). and substituting into (82).

as discussed in [21]. which has an interpretation as the RG-flow equations on the strongly coupled field theory side. will play a significant rôle in the construction of the reduction ansatz at the full non-linear order. 3. 3 (84) (The fully-supersymmetric stationary point is at X = 1. Conclusion In this paper. both of which involve pseudoscalar as well as scalar fields. one of which is a scalar. Oxidation of five-dimensional solutions Given the conjectured metric reduction ansatz. sinh ϕ = √1 . or a closely related construction. In principle. One may expect that this harmonic. with a truncation from N = 8 to the N = 2 theory with U (1)4 gauge fields. but we have not yet completed this investigation. ϕ = 0. where the problem was studied in the absence of the three axionic scalars. Our results generalise those obtained previously in [2].M. Among other uses this reduction allows one to study the eleven-dimensional geometries corresponding to the lifting of the four-dimensional BPS AdS black hole and domain-wall solutions [13] of gauged supergravity. Unfortunately the equations seem not to allow an explicit solution in terms of elementary functions.3. Cvetiˇc et al. three dilatons and three axions. and the other a pseudoscalar. This truncation retains the fields necessary for describing a . The first example is the S 7 reduction of eleven-dimensional supergravity.) Substituting into (73). namely  (83) Y(2) = eiφ3 c dξ ∧ σ3 + 12 sc2 σ1 ∧ σ2 − isc2 σ3 ∧ dφ3 . / Nuclear Physics B 584 (2000) 149–170 167 harmonics has a structure that is particularly naturally adapted to our parameterisation of the sphere. one can solve the equations of motion in this two-scalar sector to obtain a supersymmetric domain wall solution. This corresponds to the stationary point of (62) with [4] X = 2−1/3 . Our second example is a truncation of five-dimensional maximal gauged supergravity. One simple oxidation that we can perform is to take the D = 5 solution corresponding to the second (non-trivial) supersymmetric stationary point of the potential. we find that the internal 5-sphere metric d sˆ52 at this stationary point is given by   3 c2 2c2 2 2 2 σ + σ σ2 + d sˆ5 = 7/3 dξ 2 + 2 2 2(1 + s 2 ) 1 3 + 5s 2 3  2  s 2 (3 + 5s 2 ) c2 + − σ dφ . we can oxidise the metric in any solution of the two-scalar truncation of five-dimensional maximal gauged supergravity back to a solution of type IIB supergravity in D = 10. to a subsector in which two spin-0 fields are retained. (85) 3 3 3(1 + s 2 )2 3 + 5s 2 4. we have obtained the metric ansatz for two examples of Kaluza–Klein sphere reductions.

this can include the renormalisation-group flow [21] associated with the second supersymmetric extremum. K68 → 12 (K68 + K57 − K13 − K24 ). K34 → 12 (K34 + K12 − K56 − K78 ). K56 → 12 (K56 + K12 − K34 − K78 ). Cvetiˇc et al. The metric reduction ansatz that we obtain here allows one to study the ten-dimensional geometries corresponding to the lifting of solutions of the five-dimensional theory. We can describe S 7 as the unit sphere in R 8 . K58 → 12 (K58 + K14 − K23 − K67 ). K48 → 12 (K48 + K37 − K15 − K26 ). K23 → 12 (K23 + K14 − K58 − K67 ). K27 → 12 (K27 + K18 − K36 − K45 ). (86) Appendix B In this appendix. K57 → 12 (K57 + K13 + K24 + K68 ). K67 → 12 (K67 + K14 − K23 − K58 ). K37 → 12 (K37 + K15 + K26 + K48 ). we present the explicit form of the SO(8) triality rotation that we used in Section 2. with N = 2 supersymmetry and SU(2) × U (1) invariance. In principle. K78 → 12 (K78 + K12 − K34 − K56 ). Tuan Tran and Nick Warner for discussions. K46 → 12 (K46 + K35 − K17 − K28 ). K17 → 12 (K17 − K28 + K35 − K46 ).168 M. K24 → 12 (K24 − K13 + K57 − K68 ). K18 → 12 (K18 + K27 + K36 + K45 ). in addition to the maximally-supersymmetric one with SO(6) invariance [4]. with 8 real coordinates xI : xI xI = 1.1 in order to simplify the Kaluza–Klein metric reduction ansatz: K12 → 12 (K12 + K34 + K56 + K78 ). Krzysztof Pilch. K38 → 12 (K38 + K16 − K25 − K47 ). K28 → 12 (K28 − K17 + K35 − K46 ). K15 → 12 (K15 − K26 + K37 − K48 ). (87) . Acknowledgement We are grateful to Jim Liu. although the explicit form of this five-dimensional solution is not known. K36 → 12 (K36 + K18 − K27 − K45 ). Appendix A In this appendix. K47 → 12 (K47 + K16 − K25 − K38 ). K26 → 12 (K26 − K15 + K37 − K48 ). K16 → 12 (K16 + K25 + K38 + K47 ). / Nuclear Physics B 584 (2000) 149–170 second supersymmetric vacuum in D = 5. K14 → 12 (K14 + K23 + K58 + K67 ). K13 → 12 (K13 − K24 + K57 − K68 ). K45 → 12 (K45 + K18 − K27 − K36 ). K35 → 12 (K35 + K17 + K28 + K46 ). we collect some results on the geometry of the 7-sphere. K25 → 12 (K25 + K16 − K38 − K47 ).

z4 = x7 + ix8 . ∂φ4 (94) These are the four commuting U (1) generators. ∂φ1 K34 = ∂ . B 281 (1987) 211. From the coordinate transformations above. and in the S 7 reduction ansatz obtained in [2]. z3 = x5 + ix6 . K56 and K78 are simply of the form: K12 = ∂ . for example. (91) where (89) implies that 4 X µ2i = 1. φi ) coordinates are precisely the ones used for describing higher-dimensional rotating black holes in [24]. References [1] B. that 2 2 2 2 + K14 + K23 + K24 = (µ1 ∂µ2 − µ2 ∂µ1 )2 + K13 µ22 µ21 µ1 µ22 ∂2 + 2 φ1 ∂φ22 (95) with analogous results for the other five combinations. (90) We can parameterise these complex coordinates as z1 = µ1 eiφ1 . ∂φ3 K78 = ∂ . de Wit. φ3 ) and (µ4 . z2 = µ2 eiφ2 . ∂φ2 K56 = ∂ . it has a manifest SO(8) symmetry. Cvetiˇc et al.M. ∂x2 ∂µ1 µ1 ∂φ1 (93) with analogous expressions involving (µ2 . x6 ) and (x7 . After some algebra we find. ∂xJ ∂xI (88) We can also describe S 7 as the unit sphere in C 4 . z3 = µ3 eiφ3 . with 28 Killing vectors KI J given by KI J = x I ∂ ∂ − xJ . z2 = x3 + ix4 . respectively. . etc. x4 ). (92) i=1 These (µi . φ4 ) for the pairs (x3 . It is easy to see from this that the four Killing vectors K12 . (x5 . (µ3 . Nucl. We also note that the Killing-vector bilinears in the top 3 lines in (14) are also relatively simple. x8 ). with 4 complex coordinates zi : z¯ i zi = 1. (89) We can relate these complex coordinates to the previous real ones as follows: z1 = x1 + ix2 . K34 . Nicolai. φ2 ). / Nuclear Physics B 584 (2000) 149–170 169 As such. Phys. The consistency of the S 7 truncation in D = 11 supergravity. H. when expressed in terms of the µi and φi coordinates. z4 = µ4 eiφ4 . ∂x1 ∂µ1 µ1 ∂φ1 ∂ cos φ1 ∂ ∂ = sin φ1 + . it is straightforward to establish that the real derivatives ∂/∂xI that appear in the Killing vectors (88) are given by sin φ1 ∂ ∂ ∂ = cos φ1 − . It is convenient to write them as ∂φ1 .

Sati. Sadrzadeh. Phys. Cvetiˇc. [24] R. hep-th/9812035. B 208 (1982) 323. J.J. Lü. H. Freedman.P. Ann. C. R. H.J. Phys. K. N. [19] I. Duff.J. hep-th/9912132. Pilch. S. Pope. [21] D. Nucl.P. H. Nucl. Anti-de Sitter black holes in gauged N = 8 supergravity. Phys. Lett. Nastase. D.P.C. N = 8 supergravity. 172 (1986) 304. [14] M. Warner. H.N.N. K.P. [8] M. Romans. Liu. New vacua of gauged N = 8 supergravity in five dimensions. Lü. K. Four-dimensional N = 4. Perry. [15] P. A. Pope. [18] I. States and curves of five-dimensional gauged supergravity. Lett. Nastase. Gauged six-dimensional supergravity from massive type IIA. L. hep-th/9901149. Larsen. C. P. B 154 (1985) 268. H. P. Vaman. Lett. Continuous distributions of D3-branes and gauged supergravity.S. SU(2) × U (1) gauged supergravity from type IIB. Renormalization group flows from holography.N. Pope. B 108 (1982) 285. H. Kraus. Nucl. Cvetiˇc. van Nieuwenhuizen. hep-th/9909121. Martinez-Acosta. B 558 (1999) 96. Phys. Freedman. Nicolai. Pilch. [13] M. C. Nucl. M. hep-th/9909203. supersymmetry and a C-theorem. N. Phys. [20] M. de Wit.J. hep-th/9910252. Tran. Symmetric potentials of gauged supergravities in diverse dimensions and Coulomb branch of gauge theories. B 475 (2000) 261. Pilch. L. Gubser. P. Phys. hep-th/9904017. N.Z. N. Cvetiˇc et al. Khavaev. Phys. . [12] B. de Wit. JHEP 9903 (1999) 003. Sfetsos. M-branes and algebraic curves. Black holes in higher-dimensional spacetimes. hep-th/9905075.P.170 M. in press. K. S.T. Nicolai. Brandhuber. Sfetsos. Bakas. B 554 (1999) 237. [3] M. D. [5] M. T. The Coulomb branch of gauge theory from rotating branes. hep-th/9906221. [17] A. Liu. hep-th/9908116. Phys.N. C. Cvetiˇc. Duff. B. Phys. Gunaydin. P. Warner. Vaman. Pope. Phys. Consistency of the AdS7 × S4 reduction and the origin of self-duality in odd dimensions. Kim. N. [9] H. D. [23] H. Gubser. C. N = 8 supergravity with local SO(8) × SU (8) invariance. J. [22] H. Myers. Mass spectrum of ten-dimensional N = 2 supergravity on S 5 . D 32 (1985) 389. Lu. Phys. Lett. Five-dimensional N = 4. Brandhuber. The embedding of gauged N = 8 supergravity into D = 11 supergravity. H. Rev. van Nieuwenhuizen. D. Cvetiˇc. Trivedi. Lü. S. Embedding AdS black holes in ten and eleven dimensions. hepth/9909041. H. Rev. hep-th/9911238.P.T. in press. Consistency of Kaluza–Klein sphere reductions of symmetric potentials. [10] B. Tran.J.A. [7] H. Gauged N = 8 supergravity in five dimensions. Lett. Warner. Lü. Phys.Z. B 255 (1985) 29. K. Rev. J. Pope. A. Consistent nonlinear KK reduction of 11-D supergravity on AdS7 × S 4 and selfduality in odd dimensions. / Nuclear Physics B 584 (2000) 149–170 [2] M. Pope. van Nieuwenhuizen.J. hep-th/9903214.N. Warner. S. hep-th/9811120. K. Lü. Cvetiˇc. Nonstandard compactification with mass gaps and Newton’s Law. Bakas. Nicolai.A. Lü. hep-th/0002056. Domain walls of gauged supergravity. Warner. 83 (1999) 5226. C. SO(4) gauged supergravity from D = 11. On the nonlinear KK reductions on spheres of supergravity theories. Vaman. Sfetsos. [11] B.S. Gubser. F.X. T. Nucl.N. [16] D. hep-th/0002028. [6] H. [4] A. B 469 (1999) 96. M. de Wit. Phys. hep-th/9906194. Hoxha. Nastase. H. Romans.

the quantum-corrected supersymmetry transformation laws of the supergravitons are directly identified in closed form. which exhibit an intricate interplay between supersymmetry and gauge (BRST) symmetry.c. E-mail: kazama@hep3. As an application. just to mention only the direct comparison with eleven dimensional supergravity. Ward identity 1. Tokyo153-8902.V. PACS: 11.u-tokyo.V. including the normalization. All rights reserved.elsevier. 11. accepted 26 May 2000 Abstract Supersymmetric Ward identity for the low energy effective action in the standard background gauge is derived for arbitrary trajectories of supergravitons in Matrix theory. [1] and later re-interpreted by Susskind [2] in the framework of discrete light-cone quantization. one of the keys should be the understanding of the structure and the role of ∗ Corresponding author. Japan Received 20 March 2000.  2000 Elsevier Science B. In our formalism.4] and that for the two-body potential between arbitrary fermionic as well as bosonic objects at 1-loop [5] can be cited as highly nontrivial and remarkable. Meguro-ku. we explicitly compute the transformation laws for the source-probe configuration at 1-loop and confirm that supersymmetry fixes the form of the action completely. In particular. 11.-w. originally proposed by Banks et al. Muramatsu 1 Institute of Physics.Nuclear Physics B 584 (2000) 171–196 www.c.jp 1 tetsu@hep1.nl/locate/npe On the supersymmetry and gauge structure of Matrix theory Y.Db Keywords: Supersymmetry. Despite such impressive pieces of evidence as well as general supportive arguments [6. Introduction By now a considerable amount of evidence has been accumulated for the Matrix theory for M-theory. All rights reserved. Komaba.9 . Kazama ∗ .30.u-tokyo.ac. the deep reason and the mechanism of the agreement are yet to be fully understood. complete agreement for the multi-graviton scattering (including the recoil effects) at 2-loop [3.jp 0550-3213/00/$ – see front matter  2000 Elsevier Science B.25. T.7].Db. to the lowest order in the derivative expansion. Evidently.25. PII: S 0 5 5 0 .3 2 1 3 ( 0 0 ) 0 0 3 6 7 . Gauge symmetry. Matrix theory.ac. University of Tokyo.

in which the supersymmetry and the BRST symmetry are intertwined. Perhaps less familiar is the generalized conformal invariance [8–10]. one would like to obtain the Ward identity in so-called the “background gauge”. Further. T. this was not to be: one must carefully disentangle the dependence on the background field of the effective action by making use of BRST Ward identities. the study of how the symmetries govern the structure of the effective action becomes quite nontrivial. in which all the calculations have been performed. which must be deeply connected. the agreement of the multi-body scattering amplitudes strongly suggests that the eleven-dimensional Lorentz invariance is present in a highly nontrivial manner in the Matrix theory. as SUSY and gauge symmetry are known to be intimately intertwined offshell. This prompts us to resort to the conventional means. one must allow the background fields to have arbitrary time-dependence. the form of the effective action depends on the choice of the gauge as well as on the definition of the fields. This in turn inevitably leads to the necessity of examining the gauge (BRST) symmetry. since the supergravity interactions between various objects arise only after introducing the corresponding backgrounds and integrating out the quantum fluctuations around them. the realizations of these symmetries are in general modified for the effective action of interest. This is essentially due to the lack of completely off-shell consideration. although not yet clearly identified. The well-known symmetries present both in the Matrix theory and in the supergravity theory are the global Spin(9) invariance. considered to be the most powerful among the ones listed above. Muramatsu / Nuclear Physics B 584 (2000) 171–196 the symmetries. As we shall elaborate in some detail in Section 3. the supergravity possesses the general coordinate invariance while the Matrix theory has the Yang–Mills type gauge symmetry. another important ingredient of Matrix theory. In fact it has been claimed that a large degree of SUSY. the CPT invariance and the invariance under 16 supersymmetries. What makes the off-shell analysis difficult is that we do not as yet have the off-shell unconstrained superfield formulation in the case of N = 16 supersymmetry. However. the generalization of the conformal symmetry that plays the major role in the AdS/CFT correspondence of Maldacena [11]. Besides. there is always a vast degree of ambiguities as one can add total derivatives. Besides these symmetries. In this article. the above-mentioned symmetries of the Matrix theory are easily recognized in the original action. present in the theory imposes strong restrictions on the form of the effective action and leads to a number of “non-renormalization theorems” [12–17].172 Y. The result is a somewhat complicated Ward identity. Kazama. Such an analysis has not been performed in the past. A notable feature of our Ward identity is that one can read off the effective quantum-corrected SUSY transformations in closed form and this should serve as a starting point of various truly off-shell investigations. upon close examinations one finds that such assertions in the existing literature can still be challenged. Although the derivation of such a Ward identity is expected to be a text-book matter. we shall focus on the supersymmetry (SUSY). . In order to be able to check against the available explicit result for the effective action. Finally. to understand precisely to what extent SUSY is responsible in determining the effective action. namely the Ward identity for the symmetries in question. However. N = 16. being an off-shell quantity. Except for the eleven dimensional Lorentz invariance. For these reasons.

we proceed in Section 4 to the derivation of the SUSY Ward identity for the effective action in the background gauge. Θ] + Θ γ [Xm . 2 Fermions are not transformed. T. . Xn ]2 2 4  i T g T m + Θ [Dt . n. Action and its symmetries Let us begin by recalling the action of the Matrix theory and its symmetries. which agrees with the explicit calculation [18] including the normalization.Y. e S0 ≡ −iS0 . By carefully separating the different origins of the dependence on the background field. the gauge field A. to emphasize the necessity of completely off-shell analysis. denote the spatial directions. The basic action of the Matrix theory can be written as  Z 1 g2 S0 = Tr dt [Dt . and the fermionic fields. Θα. Section 6 is devoted to a short summary and discussions of future problems. while the Greek letters such as α. . Then in Section 3 we make some important remarks on the determination of the effective action and its symmetries. running from 1 to 9. 2 action e S0 by τ ≡ it. We find that at this order the effective action is indeed fully determined by the requirement of supersymmetry. . it is convenient to define the theory by going to ˜ and the the “Euclidean formulation”. we explicitly work out in Section 5 the SUSY transformation laws to the lowest order in the derivative expansion at 1-loop and show that the effective action to the corresponding order is completely determined by the Ward identity. Introduce the Euclidean time τ . The rest of the paper is organized as follows: in Section 2. we recall the symmetries of the original action of Matrix theory and the BRST symmetry associated with the background gauge fixing. Xm ]2 + [Xm .3) . . respectively. the gauge.1) 2 2 (2. γ n } = δ mn . we obtain the desired Ward identity together with the closed expressions for the effective SUSY transformation laws. to the corresponding order. The 16 × 16 γ -matrices γ m are real symmetric and satisfy {γ m . Kazama. Aij (t). A˜ ≡ −iA. (2. This constitutes the main result of this work.2) Dt = ∂t − igA. Θ] . As an application. Having clarified the issue. where N × N hermitian matrices Xijm (t). (2. which at the same time serves to set our notations. β = 1 ∼ 16 are used for the SO(9) spinor indices. we compute the transformation laws to the lowest order in the derivative expansion at 1-loop and analyze the restriction imposed by SUSY on the effective action for a background with arbitrary time-dependence. To facilitate the quantum computations. The middle Latin indices m.ij (t) stand for the bosonic. Muramatsu / Nuclear Physics B 584 (2000) 171–196 173 As a simple application. 2.

δ A˜ =  T Θ. δK g = 3τg. δK τ = −τ .5) Besides the obvious Spin(9) symmetry.7) (2.174 Y. this action is invariant under the following transformations: 1. (2. Besides these well-established symmetries. and allow g to depend on τ to a linear order. δΛ Θ = ig[Λ. 3. Supersymmetry transformations with a spinor parameter α : δ Xm = −i T γ m Θ. formulation in terms of unconstrained superfields is not known and hence the algebra closes only up to the equations of motion in general. T m T (2. Muramatsu / Nuclear Physics B 584 (2000) 171–196 Then the action and the covariant derivative become  Z 1 g2 e S0 = Tr dτ [Dτ . T.4) (2. Generalized conformal transformations: if we rescale the fields Xm and A˜ by a factor S0 is invariant of g. Kazama. Xm ]γ m + [Xm . (2.6) 2. (2. γ n ] is real anti-symmetric. . 2 δK Θα = 0.8) where γ mn ≡ 12 [γ m . imposes a useful restriction on the form of the effective action.10].9)  where Λ = −2i  γ λXm −  λA˜ . e under a generalization of the conformal transformations [8. When the system possesses symmetries other than the supersymmetry. In particular. (2.10) and similarly for the other fields. such as the Matrix theory. δK Xm = 2τ Xm . Xm ]. Thus it is expected that the proper understanding of the supersymmetry of the Matrix theory must include the analysis of these nontrivial features. Λ]. Θ] . 2 (2. the invariance under the special conformal transformation defined by ˜ δK A˜ = 2τ A. Θ]. In fact in the present case it is well-known that it involves the gauge symmetry with a field-dependent ˜ gauge function. Xn ]γ mn . For example on A. δΛ Xm = ig[Λ. Gauge transformations with a gauge parameter matrix Λ: δΛ A˜ = [Dτ . δλ ]A˜ = −2i Tλ ∂τ A˜ + δΛ A. such as Xm → Xm /g. ˜ [δ . Xm ]2 − [Xm .11)  = an infinitesimal bosonic parameter. Xn ]2 2 4  1 T 1 T m + Θ [Dτ . the remarkable agreement between the 11dimensional supergravity calculations and the 2-loop Matrix theory calculations for multibody scattering processes [3] strongly suggests that the Matrix theory actually possesses 11-dimensional Lorentz symmetry in a highly nontrivial manner. 2 2 ˜ Dτ = ∂τ − ig A.   g δ Θ = i [Dτ . for a system with N = 16 supersymmetry. Moreover. the supersymmetry algebra may close not only on the space– time translation but also on the generators of such additional symmetries. Θ] − gΘ γ [Xm .

C¯ and b are. δB Ym = −ig[Xm . To quantize the theory. after integrating over the b field. δB b = 0. In the M-theory interpretation of the Matrix theory. Xm . respectively. T. not satisfying any equations of motion. are implemented in the quantum effective action of the supergravitons. Xm = In fact essentially all the existing explicit calculations have been performed in this gauge. To avoid this problem. it is important to take Bm (τ ) and θα (τ ) as arbitrary backgrounds. we need to fix the gauge. C. Xm . respectively. C]. This leads.Y. We shall denote them by Bm and θα . (2. δB C = igC . g As was already emphasized in the introduction and will be further elaborated in the next section. we em in place of Bm and write the gauge-fixing will tentatively use a different function B function as  m  e . Θ}. we shall focus on how the first two of these symmetries. It is specified by the gauge-fixing function of the form   (2. 2 (2. C]. (2. e = −∂τ A˜ + i B (2. Muramatsu / Nuclear Physics B 584 (2000) 171–196 175 In this article.14) G = −∂τ A˜ + i B m . can be readily adapted to any choice of gauge. the anti-ghost and the Nakanishi–Lautrup auxiliary fields. Although our derivation of the Ward identity. and separate them from the quantum parts Ym and Ψα as 1 Bm + Ym . as it will become clear. to the familiar gauge-ghost action . the ghost.15) G em = Bm . we will set B The corresponding ghost action can be readily obtained by the standard BRST method.12) g 1 (2. respectively. The background fields are not transformed. Later at an appropriate stage. δB Ψ = ig{C. to be presented in Section 4.13) Θα = θα + Ψα . the actual computations are extremely cumbersome except in the standard background gauge. the naive use of this gauge leads to a subtle but important complication in deriving the correct Ward identity. The BRST transformations for the quantum part of the fields are given by δB A˜ = [Dτ .17) Sgg = δB Tr dτ C¯ G i 2 We will henceforth set the gauge parameter α to be 1. the coordinates and the spin degrees of freedom of these supergravitons are represented by the diagonal backgrounds for Xm and Θα . Kazama. Then the combined gauge-ghost action e Sgg is generated by    Z 1 e e− αb . Only in this way we can unambiguously determine how much restrictions are imposed by the supersymmetry on the effective action for these background fields.16) δB C¯ = ib. However. which are intimately intertwined.

19) Here. 0). (3. for clarity of discussions we shall consider the so-called “source-probe configuration”. the spin of the source is neglected. can be dangerous and misleading. θα = diag(θα . In the case of U (N + 1) gauge group.18) 3. the eikonal approximation. which point to the necessity of off-shell analysis. Although many of the remarks will apply for general backgrounds. v · b = 0. one must make sure that they are logically consistent for one’s aim. Besides. Muramatsu / Nuclear Physics B 584 (2000) 171–196 Z  m  1 e . As already emphasized in the introduction. As a simple illustration. bm = impact parameter. Of course the on-shell S-matrix elements do not depend on these factors. . 0. it is necessary to perform an off-shell analysis with (A) ∼ (C) properly taken into account. Because of this. . T. 3 Thus. Xm 2 dτ −∂τ A˜ + i B 2 Z  m  ¯ τ [Dτ . − i Tr dτ C∂ e Sgg = Tr (2. consider the 1-loop eikonal phase shift [19] for v  1 given by v3 v5 3 v7 + 0 × 10 − + O(v 9 ). As usual. Kazama. realizations of the symmetries of the effective theory are in general modified nontrivially. C] − g C¯ B e . it is defined as the situation where a probe supergraviton interacts with N supergravitons situated at the origin that act as a heavy source. [Xm . one often needs to make some approximations. . In fact the answer depends on the form of the effective Lagrangian assumed. the primary feature of Matrix theory is that it generates the supergravity interactions among various objects only after (i) introducing the corresponding backgrounds and (ii) integrating over the quantum fluctuations around them. .. . being an off-shell quantity. some of the often used approximations are not appropriate. quantum-corrected) on-shell condition itself requires the knowledge of the off-shell effective action. . Remarks on the determination of effective action and its symmetries Before starting the derivation of the off-shell supersymmetry Ward identity. the determination of the full (i. Γe1e = − 3 This was clearly demonstrated in [18]. The background fields representing this situation are Bm = diag(rm . For the present purpose. . (B) the (re)definition of the fields. 0. and (C) the freedom of adding total derivatives. In doing so.176 Y. we wish to make some important remarks on the determination of the effective action and its symmetries. 6 b b 2 b14 rm = vm τ + bm . rm represents the coordinate of the probe and θα its spin content. where one tries to reconstruct the effective action from the eikonal phase shift. . where the agreement between the supergravity and the Matrix theory calculation was achieved with the recoil corrections. However.e. Now since an exact analysis is practically impossible. C] . For example. 0. in order to understand the symmetry structure of the effective theory fully. 0). 0. the form of the effective action is affected by (A) the gauge choice. .

. one may combine this with the usual loop expansion. δλ ] = 2λT ∂τ . −(15/16) = A + (1/7)B + (1/21)C. is to use off-shell backgrounds with arbitrary τ -dependence 4 and to classify terms by derivative expansion according to the “order” defined by 1 order = # of ∂τ + # of fermions. Although the eikonal phase shift vanishes. They can be roughly classified into two categories. devised by Paban et al. field redefinitions affect the form of the SUSY transformations and hence they must be considered as a pair. The situation at O(v 6 ) is even more striking. 2 (3. For example. then the effective action that reproduces this phase shift is uniquely determined to be   Z 15 v 4 v6 9009 v 8 10 e + 0 × 11 − + O(v ) . the explicit computation reveals that there are 5 non-vanishing independent structures present in the effective action [20]. T. often referred to as SUSY non-renormalization theorems. even restricting to O(v 4 ). Related discussion can also be found in [18]. 4 This was emphasized in the context of generalized conformal symmetry in [21]. namely [δ . r). Then demanding that the closure is canonical. S1 = dτ − 16 r 7 r 4096 r 15 However. Kazama. and hence 5 parameters remain undetermined. Having emphasized the importance of off-shell considerations. we now make some related comments on the general arguments on the restrictions imposed by supersymmetry. it is unclear why the closure should be canonical off-shell and further it is not obvious if the O(v 0 ) transformation laws must be of the standard form in a particular basis adopted. not affected by the total-derivative ambiguities. The first type of argument. r7 r7 r On the other hand. the most general form allowed for the effective action contains 6 independent structures. relies on the closure property of SUSY transformations. δ r m = −iγ m θ. Although the argument is quite simple and plausible.20) If necessary. they show that there cannot be a correction to the transformation laws at O(v2 ) and hence the O(v 2 ) effective action is tree-exact. [12].Y. In general. Muramatsu / Nuclear Physics B 584 (2000) 171–196 177 If one assumes the effective Lagrangian to be of the form L = L(v. It should be clear from these illustrations that the only logically consistent procedure. at O(v∈ ) they first make a choice of the definition R of the fields so that the action takes the form dτf (r)v 2 and take the O(v 0 ) SUSY transformation laws in that basis to be the standard ones without any correction. there is only one condition. δ θα = i(/v)α . after eliminating total-derivative ambiguities:  Z v4 v 2 (v · r)2 (v · r)4 e + C S1 = dτ A 7 + B r r9 r 11  2 2 v (a · r) (a · r) a2 +D +E +F 5 . required for the correct phase shift.

the diagonal and the off-diagonal. θ 8 term in this case. This motivates us to the study of the Ward identity. In view of the importance of precise understanding of the role of supersymmetry and its connection with gauge symmetry.22) . δij + Y g Ψα. 2 2 )) is SUSY-complete without a θ 4 term.14). one may form a super-invariant not containing An example already occurs at O(v 2 ). it is desirable to perform an unambiguous off-shell analysis with arbitrary backgrounds.178 Y. in the manner Ym. in particular the v 4 term. δij + Ψ g and introduce the sources only for the diagonal fields: (4. is not renormalized beyond 1-loop. Muramatsu / Nuclear Physics B 584 (2000) 171–196 The other type of argument is known as the SUSY completion method.ij . This method appears efficient. An attempt at filling this gap was made in [16]. To make use of the well-established method.21) By showing that the top form. Put differently. we shall now derive the SUSY Ward identity for the effective action Γe in the standard background gauge. to be described in the next two sections. at O(v 4 ) one expects relations of the form v4 ← v3 θ 2 ← v2 θ 4 ← v θ 6 ← θ 8 . but some care is needed in drawing firm conclusions. One problem is that sometimes the chain starting from the top form stops at an intermediate stage.i bα.ij = yˆm. (2. which makes use of the chain of relations produced among terms with different number of θ ’s by SUSY transformations. where the tree-level expression Rthe top 2form. T. the bosonic contribution at O(v 6 ) nonetheless exists [20].i bm. Later it was recognized [18]. Kazama. that the higher derivative terms neglected in this analysis can actually be absorbed into the tree-level Lagrangian by a suitable field redefinition. More nontrivial example ˙ dτ ((v /2g ) + (θ θ/2g 6 is seen at O(v ): although θ 12 term was shown to vanish [13] at 1-loop. Unfortunately.ij = ψˆ α. For example. the problem still persists: by such a field redefinition the higher derivatives are simply shifted into the SUSY transformation laws and one must reanalyze the issue with such modifications. however. one wishes to infer the non-renormalization of all the other terms in the chain. Thus one sees that although the existing analyses are highly plausible they are not airtight. In this work all the connections in the chain (3. (3. 4. SUSY Ward identity for the effective action in the background gauge Having argued the importance of off-shell analysis for arbitrary trajectories. Again one must be cautious in accepting this conclusion: in this analysis v and θ were taken to be τ -independent and hence the assumed form of the effective Lagrangian was not the most general one allowed in the proper derivative expansion with arbitrary backgrounds. let us further split the quantum fluctuations Ym and Ψα into two parts. used exclusively in the actual computations.21) were examined and it was concluded that SUSY is indeed powerful enough to fix the effective action at this order up to an overall constant.ij . which appeared to resurrect the validity of the analysis made in [16].

30). δJm.i (τ ) dτ (Jm.i ψα.i (τ )i + hδ ψˆ α. (4.25) in terms of the generating functional Γe.i (τ ) Sgg i from the gauge-ghost part.i ψˆ α. The first step is to note that the supersymmetry transformations (2.i(τ ) ≡ e δW . Define as usual the classical fields and Γe by ym. the contribution to the Ward identity from the variation of the source action can be written as   Z δ Γe δ Γe hδ yˆm. e η] = Dµ exp(−e Z[J.i + ηα.i ). This is undesirable since what we wish to understand is how the supersymmetry acts on the effective action Γe. a direct calculation yields As for the contribution hδ e a rather complicated expression.Y.25) Here and in what follows. Fortunately.26) We now rewrite this identity (4. Muramatsu / Nuclear Physics B 584 (2000) 171–196 e Ss = Z 179 dτ (Jm. 0 = hδ e (4. Kazama. By making the change of integration where W variables corresponding to the supersymmetry transformations. e [J. one obtains the primitive form of the Ward identity Sgg i + hδ e Ss i.27) (4. Sgg i can be written as starting from (2. hδ e .i (τ ) ≡ Z e δW . hOi = R Stot Dµe−e (4. η] − ψα. δψα.i (τ ) (4.23) The Euclidean generating functionals are defined by Z e [J.i yˆm.16) on all the fields.i (τ )i .i ym.28) Then the sources are expressed in terms of Γe as Jm. which constitutes an inhomogeneous term in the Ward identity regarded as a functional integro-differential equation for Γe.29) Therefore. Stot ≡ e ˜ ¯ Dµ ≡ DADY DΨ DCDC. η] is the one for the connected functions. Therefore. which is 1PI (1-particle-irreducible) with respect to the diagonal fields.i (τ ) = − δ Γe . (4.30) Ss i = dτ − hδ e δym.i ). ψ] ≡ W [J.i (τ ) ηα. Stot ) = exp(−W (4. as can be checked straightforwardly.i (τ ) = δ Γe .17).24) e S0 + e Sgg + e Ss . T. (4.i + ηα. δym.i (τ ) δψα. η]). δηα.8) commute with the BRST transformation (2.7) and (2. hOi for an operator O means R e DµOe−Stot .i (τ ) Γe[y. it was noted long ago [22] in the context of four-dimensional super Yang– Mills theory that one can reexpress such a term in a form similar to (4.

36) Putting all together. and consider the generating functional with a source j for the operator O : Z (4. and then expressing the source J in terms of Γe. This expression.34) 0 = D φ (−1)|φ| J δB φ − j δB O e−(Stot +J φ+j O ) .180 Y.i (τ )i + hδB yˆm. in the presence of external sources. e dτ C¯ δ G (4.i (τ ) δψα. .32) is a fermionic composite operator. we get  Z e hδ Sgg i = − dτ  δ Γe δ Γe ˆ hδB yˆm.31) (4.33) Z[J. − δψα. δym. it becomes proportional to the sources. being an expectation value of a BRSTexact form. we arrive at the following SUSY Ward identity expressed solely in terms of the derivatives of Γe:  Z  δ Γe hδ yˆm. and hence to the functional derivatives of Γe. We get Z  (4. setting j = 0. j ] = D φe−(Stot +J φ+j O ) . respectively.i (τ )O i . This procedure is indeed valid for the fermions since ψˆ α and θα always appear in the combination ψˆ α + θα in the original action and it is a simple matter to prove that this gets converted to ψα + θα in Γe.i (τ )i − hδB ψα.i (τ )O i . T.i (τ ) Normally it is now a simple matter to convert this into the desired Ward identity for the effective action as a functional of the backgrounds B m and θα : one would rewrite the derivatives with respect to ym and ψα into those with respect to B m and θα and then set ym = ψα = 0. Kazama. Muramatsu / Nuclear Physics B 584 (2000) 171–196 *   Z 1 ¯ e e C G− hδ Sgg i = δ δB Tr dτ i *   Z 1 ¯ e C G− = δB δ Tr dτ i where 1 O ≡ Tr i Z 1 b 2 1 b 2 + + = hδB O i. one easily obtains the following BRST Ward identity: Z δ Γe hδB φ(τ )O i.i (τ ) (4. where |φ| is 0 (1) if φ is bosonic (fermionic). (4.35) hδB O i = − dτ δ φ(τ ) In this way.i (τ )   δ Γe ˆ ˆ (4. However. Now make a change of variables corresponding to the BRST transformation.i (τ )O i + hδB ψα. vanishes in the ordinary vacuum. By differentiating with respect to j once. Let us collectively denote by φ and J the basic fields and the corresponding sources.37) hδ ψα.i (τ )O i 0 = dτ δym.

i . we must disentangle the two different types of B m dependence buried in the standard background gauge formulation and construct the Ward identity which takes both types of dependence into account.41) em. we get from (2. From the point of view of the Ward identity above. Since B and the variation with respect to it commutes with the BRST transformation.i stands for the diagonal elements B and B ii e as δ W e /δ B em. to start out with a different symbol B Now the problem we face is the following. is   δ Γe δ Γe ∆Γe = + . Although this is a technical rather than a conceptual problem. In this way we get the identity  Z δ Γe δ Γe 0 hδB yˆn.38) where Om.i = X  C¯ j i Ym.37) and in the relation (4. In order to be able to apply the Ward identity em to the case of the standard background gauge. (4.40) above.i (τ ) δ B e Substituting (4.j m.i e δ B (τ ) (4. (4. where Bm and yˆm appear together. (4. While most of the B m dependence comes from bm . T. This is why we chose em for this field.i (τ )i = − dτ e δyn. In the limit B Bm .39) j em . Muramatsu / Nuclear Physics B 584 (2000) 171–196 181 This is not so for the bosonic field.j (τ 0 ) δ Bm.i (τ ) δBm.Y. On the other hand the expectation value of the right-hand side can be treated in exactly the same way as we treated δB O . which has often been neglected.i since it is a parameter field. it is again a manifestation of the gauge theory nature of the Matrix theory. we need to express the dependence on B m again in the form of the functional derivative of Γe with respect to B .40) we then get . Kazama.i (τ ). em appears only in the gauge-ghost action The problem can be solved as follows.i (τ )  δ Γe 0 ˆ hδ (τ )O (τ )i . the total variation of Γe with respect to Bm . In other words.ij − C¯ ij Ym. which is equal can be expressed in terms of the generating functional W e e to δ Γ /δ Bm.j (τ 0 )Om.j i . B m in the the splitting Xm = (1/g)(Bm + yˆm ) + Y gauge-ghost sector (in the standard background gauge) is not accompanied with yˆm .40) ψ + B α. The expectation value of the left-hand side em. which we denote by ∆Γe/∆Bm to avoid confusion.i δψα.i (τ ) B=B ∆Bm.17) δe Sgg = δB Om. it is an independent extra parameter field which should be distinguished from the bona fide background field. m.j (τ 0 ) Now let us replace ym and ψα with Bm and θα respectively and then set ym = ψα = 0 em → in the previous Ward identity (4.

j (τ 0 )Om. T.j (τ 0 ) dτ 0 (4.i and ∆ θα.j (τ 0 ) Z δ Γe hδB ψˆ α. however.mi (τ. respectively.44) (τ.mi (τ. τ 0 ) dτ 00 + dτ 0 Tmi. the BRST . the total substituting this expression into (4. are much more complicated.k (τ 00 )O i. i. the result can be put in the desired form   Z δ Γe δ Γe + ∆ θα.43) By inverting this relation.i (τ ) ∆Bn.182 Y. which denotes the usual functional derivative of the effective action computed in the standard background gauge (i. they are represented by hδ yˆn. In our notation. ∆ Bm. we can express the partial variation δ Γe/δBm in terms of the total variation ∆Γe/∆Bm : Z ∆Γe δ Γe −1 = dτ 0 Tmi.k (τ 00 ) where T −1 is the inverse of T .7) and (2.i (τ ) Z δ Γe Tnj .46) ∆ θα.e. (4.i (τ )O i Z −1 (τ 00 .47) (As said before y = ψ = 0 is understood. Naively. One can see. τ 0 ) δBm. the supersymmetry and the gauge (BRST) symmetry are nontrivially intertwined.e. Kazama.42) with Tnj . (4.j (τ 0 ) Z Z δ Γe −1 hδB ψα. one might expect that the effective transformation laws are obtained as the expectation values of the original transformation laws (2.j (τ 0 )i. τ 0 ) ≡ δmn δij δ(τ − τ 0 ) − hδB yˆn.i (τ ) . that the corrections to the naive laws all involve δB .k (τ 00 ) − dτ 0 dτ 00 Tmk.8).. − dτ 0 δθα. the correct Ward identity in the standard background gauge is obtained by em → Bm is taken.nj (4. (4.j i and hδ ψˆ α. τ ) hδ yˆn.j (τ ) δθα. (4.i (τ )i.46) and (4.j (τ 0 )i + hδB yˆn. • As expected.) With this understood.j (τ 0 )O i . τ 0 ) δBn. (4. Note the following features: • We have succeeded in putting the Ward identity in the form where the supersymmetry transformation laws for the effective action are cleanly identified in closed forms..nj + δB yˆm.i (τ ) δBn.37). The actual transformation laws.i (τ )i − hδB ψˆ α. τ 0 )hδB ψˆ α.j (τ 0 )ihδ yˆm. Once the limit B variation ∆Γe/∆Bm can be identified with δ Γe/δBm .i (τ )On.i (τ )i.i (τ ) = dτ 0 Tmi.i (τ ) = hδ ψˆ α.i (τ ) where the effective SUSY transformation laws are given by Z  −1 (τ 0 .i .nj (τ.47). Finally. with Bm in the gauge-fixing term.j (τ 0 )Om.nj δθα.) This is the main result of this section.k (τ 00 )On. Muramatsu / Nuclear Physics B 584 (2000) 171–196 ∆Γe = ∆Bm.i i in ∆ Bm.45) 0 = dτ ∆ Bm.

0. θα = diag(θα . . . UI∗ ≡ UI 1 . albeit somewhat formal at this stage. We shall denote the time derivative of the coordinate by vm ≡ ∂τ rm . 0. Explicit calculations and implications We now compute the effective SUSY transformation laws explicitly and study the implications of the Ward identity. . Muramatsu / Nuclear Physics B 584 (2000) 171–196 183 transformation. which we display again for convenience: Bm = diag(rm .1. it is convenient to introduce the following notations: for any matrix U we define UI ≡ U1I .   2 g m ˜ mn hA · Ym∗ − A˜ ∗ · Ym i + γαβ hYm · Yn∗ − Ym∗ · Yn i β . the transformation laws derived above are exact. 0). which we abbreviate as U · V .49) hδ yˆm i = −i T γm θ. we shall restrict ourselves to the source-probe situation. 0. are more complete for this configuration. 5. not special to the standard background gauge adopted here. To facilitate the computations. In particular. • As has already been remarked. . T. as it does at the tree level: We have so far not been able to produce a proof.  δB ψˆ α = ig 2 C · Ψα∗ − C ∗ · Ψα . Kazama.) We shall call such index I the off-diagonal matrix vector index.  Z  1 ¯ ˙ C11 θα + ψ˙ˆ α + C¯ · (∂τ + /r)αβ Ψβ∗ O = −iα dτ g (5. we will carefully examine the structure of this Ward identity at the 1-loop level and draw implications. since the existing calculations of the effective action itself.Y. .52) .48) rm represents the coordinate of the probe and θα its spin content. Source-probe situation Although the Ward identity derived in the previous section is valid for any background. 0. .50) δB yˆm = −ig 2 (Ym · C ∗ − Ym∗ · C).19). . (The symbol ∗ here does not stand for complex conjugation. Thus it is far from obvious that the anti-commutator of the effective transformations would close solely on the translation generator. With this convention. Then the “11” component of a product of two matrices becomes (U V )11 = PN+1 ∗ ∗ I =2 UI VI . (5. the basic quantities appearing in the Ward identity take the following forms: (5. 0).51) (5. hδ ψˆ α i = i v/αβ − ig 2 γαβ 2 (5. 5. to be compared later. where I = 2 ∼ N + 1. The background fields representing this situation for the gauge group U (N + 1) were already described in (3. there is no inherent distinction between the tree level contribution and the quantum corrections. this is a universal feature. Since the quantization of the system inevitably requires a gauge fixing. In the next section. .

55) (5. One-loop at order two In this article. this means that we need to compute ∆ rm and ∆ θα to order 0 at 1-loop. the relevant multi-body expectation values become  2 m ˜ ˆ A(τ )Ym∗ (τ ) − A˜ ∗ (τ )Ym (τ ) hδ ψα (τ )i = i(/v)α (τ ) + g N γαβ  i mn ∗ ∗ (5. we always have the structure hUI (τ )VJ∗ (τ 0 )i = δI J hU (τ )V ∗ (τ 0 )i. we shall perform the calculation of the simplest nontrivial contributions to the effective transformation laws. “a correction to O(θ 0 )” for ∆ rm refers to a term of the form θf (r). i.54) 5. With these remarks.56) The expectation values of the composite operators themselves simplify considerably at this order. It is easy to check that at 1-loop only the 2-point functions contribute. i (5. Muramatsu / Nuclear Physics B 584 (2000) 171–196  + C¯ ∗ · (∂τ − /r)αβ Ψβ + C¯ I J Θ˙ J I .2. δI J ’s then contract to produce a factor of N .20). the expressions (4. (5. In what follows. ∆ θα (τ ) = hδ ψˆ α (τ )i − hδB ψˆ α (τ )O i Z 1 − dτ 0 hδB ψˆ α (τ )Om (τ 0 )i (γ m θ )(τ 0 ). it is more convenient to refer only to the number of additional θ ’s relative to the tree level contribution.46) and (4. At 1-loop order. Kazama.58) . since the terms involving the BRST variation δB start only at 1-loop. where U and V ∗ can be regarded as single-component fields. (5. where the transformation parameters α are considered to be of order −1/2. the number of derivatives plus twice the number of fermions. T. Moreover.e. 2 Z ¯ 0 )i(∂ + /r)αβ hΨβ∗ (τ 0 )Ym (τ )i hδB yˆn (τ )O i = α g 2 N dτ 0 hC ∗ (τ )C(τ  − hC(τ )C¯ ∗ (τ 0 )i(∂ − /r)αβ hΨβ (τ 0 )Ym∗ (τ )i . while for ∆ θα we need θ 2 contributions as well. Because rm and θα are of order 0 and 1/2. Such contributions are further classified according to the number of θ ’s involved. respectively. Since the tree action is already of order 2. Om = C¯ ∗ · Ym − C¯ · Ym∗ .47) for ∆ rm and ∆ θ simplify to Z 1 1 m m ∆ r (τ ) = (γ θ )(τ ) + dτ 0 hδB yˆm (τ )On (τ 0 )i (γ n θ )(τ 0 ) i i + hδB yˆm (τ )O i. Here the order is defined as in (3. namely those which govern the order 2 part of the 1-loop effective action.184 Y. For instance. since there are no mixing between the fields with different off-diagonal matrix vector indices in such propagators.53) (5. we need to consider ∆ rm to linear order in θ . which goes with the coupling g 2 ..57) + γαβ Ym (τ )Yn (τ ) − Ym (τ )Yn (τ ) β (τ ).

= −hC ∗ (τ )C(τ (5. + gβ (5.60) where ∂ stands for the derivative with respect to τ .59) (5. 5. We will now evaluate these expressions to the relevant order by perturbation theory.Y. The propagators for the massless fields ψˆ α and C11 . Ψα (τ )Ψβ (τ 0 ) = hτ |Dαβ (5. The part of the Lagrangian quadratic in the fields without θ is of the form  L = Ym · (−∂ 2 + r 2 )Ym∗ + A˜ · (−∂ 2 + r 2 )A˜ ∗ − 2ivm A˜ · Ym∗ − A˜ ∗ · Ym 1 ˆ ψˆ + Ψα · (∂ + /r)αβ Ψβ∗ + 2 ψ∂ 2g + i C¯ · (−∂ 2 + r 2 )C ∗ + i C¯ ∗ · (−∂ 2 + r 2 )C + i C¯ 11 (−∂ 2 )C11 . The simplest is the ghost propagator. C¯ in L as if they were singlevector indices.61) Following the remark already made on the trivial dependence on the off-diagonal matrix ˜ Ψα and C.64) (5. (5. we can expand D ± in powers of the velocity v(τ ) as . C¯ 11 turn out to be needed only at O(θ 2 ).62) where ∆ ≡ (−∂ 2 + r 2 )−1 . A. T. Muramatsu / Nuclear Physics B 584 (2000) 171–196 hδB ψˆ β (τ )O i = −α g 2 N Z − gβ Z Z 185 ¯ 0 )i(∂ + /r)αγ Ψγ∗ (τ 0 )Ψβ (τ ) dτ 0 hC ∗ (τ )C(τ  − hC(τ )C¯ ∗ (τ 0 )i(∂ − /r)αγ Ψγ (τ 0 )Ψβ∗ (τ ) ˙ (τ 0 ) b dτ 0 CΨ ∗ (τ )C¯ 11 ψ β ˙ (τ 0 ) . The propagators for the Ψα system are given by − |τ 0 i. we may treat the fields Ym .66) From the relation −(∂ ± /r)(∂ ∓ /r) = (−∂ 2 + r 2 )(1 ± ∆/v ). Corrections at O(θ 0 ) Let us begin with the corrections at O(θ 0 ). First we need to compute the tree-level propagators. component.65) where D ± ≡ (∂ ± /r)−1 .2.63) (5. b dτ 0 C ∗ Ψ (τ )C¯ 11 ψ β ¯ )i hδB yˆn (τ 0 )Om (τ )i = ig 2 N Ym (τ 0 )Yn∗ (τ ) h(C ∗ (τ 0 )C(τ  ∗ 0 + Ym (τ )Yn (τ ) h(C(τ 0 )C¯ ∗ (τ )i . Ψα (τ )Ψβ∗ (τ 0 ) = hτ |Dαβ ∗ + |τ 0 i. to be discussed in the next subsection. Kazama.1. (5. which can be directly read off as ¯ )C ∗ (τ 0 )i = hC¯ ∗ (τ )C(τ 0 )i hC(τ ¯ 0 )i = −hC(τ )C¯ ∗ (τ 0 )i = ihτ |∆|τ 0 i.

Ym (τ )A˜ ∗ (τ 0 ) = − A(τ (5. Below. dτ 0 hδB yˆm (τ )On (τ 0 )i (γ n θ )(τ 0 ) = −iγm θ i 2 r3 hδB ψˆ β (τ )O i = 0. hA(τ ˜ )Ym∗ (τ 0 ) = hτ |2i∆vm ∆|τ 0 i + O(v 3 ). As an example. T. we will only need the v-independent part. 2 r3 (5. ˜ )A˜ ∗ (τ 0 )i = hτ |∆|τ 0 i + O(v 2 ). 2 r3 (5. (5. the normal-ordering is trivial and we get hδB yˆn (τ )O i = −2i Tg 2 Nγn θ hτ |∆2 |τ i = −i T γn θ 1 g2 N .68) The only other 2-point functions appearing at this order are hΨα (τ )Ym∗ (τ 0 )i and hΨα∗ (τ )Ym (τ 0 )i. ∂] = 2∆(r · v)∆ etc. it becomes hδB yˆ n (τ )O i = −i T g 2 Nhτ |∆[(∂ + /r)D + + (∂ − /r)D − ]γ n θ ∆|τ i = −2i T g 2 Nhτ |∆(−∂ 2 + r 2 )∆γ n θ ∆|τ i.76) (5. Substituting the expressions for the various propagators already computed. 1 1 g2 N . let us consider hδB yˆn (τ )O i given in (5.58). we readily obtain hY m (τ )∗ Y n (τ 0 )i = δmn hτ |∆|τ 0 i + O(v 2 ). Kazama. we have a mixing term with an arbitrary coefficient function vm (τ ) and it can only be resolved in the derivative expansion. (5.77) .70) (5. A useful list of formulas so obtained are collected in the appendix of [23].73) In a similar manner. Expanding in powers of the mixing term. which can be readily computed by inserting the vertices of the form  m Ym · Ψβ∗ − Ym∗ · Ψβ .69) −θα γαβ The results are − (γ m θ )β ∆|τ 0 i. and then use Baker–Campbell–Hausdorff and the Gaussian integration formulas to evaluate it.186 Y. we shall give a sample calculation of this sort and then simply list the results for the needed expectation values.67) To the order of interest..75) (5. Muramatsu / Nuclear Physics B 584 (2000) 171–196 D ± = −(∂ ∓ /r)(1 ∓ ∆/v)−1 ∆ = −(∂ ∓ /r)(∆ ± ∆/v ∆ + · · ·). The essence of this method is to first rearrange the order of a product of various operators and functions into the standard form ∼ f (τ )∂ m ∆n . (5. hδB yˆm (τ )O i = −iγm θ g2 N 1 .71) With this preparation. the calculations of the various expectation values in the Ward identity can be performed efficiently to the desired order with the use of the so-called “normal ordering method” developed in [20]. Ψα (τ )Ym∗ (τ 0 ) = −hτ |Dαβ ∗ + (γ m θ )β ∆|τ 0 i. Ψα (τ )Ym (τ 0 ) = hτ |Dαβ (5. Now as for the Ym –A˜ system. we obtain the following results: Z hδB yˆn (τ 0 )Om (τ )i = 2g 2 Nδnm hτ 0 |∆|τ i2 . using the commutation relations such as [∆.74) (5.72) To the order of interest.

Corrections at O(θ 2 ) Now we move on to the corrections at O(θ 2 ). One can see from the calculations outlined above that exactly half Sgg i and the other half was produced of the quantum correction for ∆ r m came from hδ e from the procedure of taking into account the dependence on the extra parameter field we em .Y. as we have done. namely from −ig[A.84) .83) Although it appears quite complicated. (5. Muramatsu / Nuclear Physics B 584 (2000) 171–196 N vm .80) ∆ r m = −i T γ m θ 1 + 3 . originally called B 5. As shown there.83). 187 (5. 7 types of diagrams contribute. the other hand.3. there are nontrivial corrections to the tree level laws.b).79) Making use of these formulas. we need to compute such contributions only for ∆ θα . we find the O(θ 2 ) corrections to ∆ θα to be 2 ∆θ θα = 3ig 2 N −rl (θ γ mnl θ )(γ nm )α + 2rl (θ γ ml θ )(γ m )α − 4rl (γ m θ )(θ γ ml )α 16r 5  + 4(θ )(/rθ )α − 4(/rθ )θα . On SUSY transformation law.82) hψˆ α (τ )ψˆ β (τ 0 )i = g 2 hτ | |τ 0 i. (5. Applying this to (5. one would not get any corrections to all orders. Xm ]γ m .81) ∆ θ = i/v 1 − 3 . r Already at this order. it can be drastically simplified with the use of the SO(9) Fierz identities [5] described in the Appendix B. What we want is the effective SUSY transformation operating on the effective action Γe and this can only be obtained through the analysis of the Ward identity. Kazama. The procedure is entirely similar to the O(θ 0 ) case but the calculations are more involved and we relegate the details to the Appendix A. Two of them. After some calculations. involve genuine 3 point vertices as well as massless propagators given by 1 1 hC11 (τ )C¯ 11 (τ 0 )i = ihτ | 2 |τ 0 i. diagrams (B2-a. A˜ · Ym∗ − A˜ ∗ · Ym = i r3 Ym · Yn∗ − Ym∗ · Yn = 0. The relevant identity is 0 = −rl (θ γ mnl θ )(γ nm )α + 2rl (θ γ nl θ )(γ n )α + 4rl (γ n θ )(γ nl θ )α + 4rl (θ )(γ l θ )α + 12rl (γ l θ )θα . ∂ ∂ They are singular in the infrared. we get (5. Since θ 2 is of order 1.78) (5. r   g2 N (5. T. we find the effective SUSY transformation laws at O(θ 0 ) to be   g2 N (5. but such singularities cancel in the end result. For ∆ θ the entire correction came from the expectation value of the non-linear part of the original ˜ Ym ]γ m contained in [Dτ . the same procedure is not applicable for ∆ rm : if one naively took the expectation value of the basic transformation law δ Xm = −iγ m Θ.

1. Muramatsu / Nuclear Physics B 584 (2000) 171–196 3ig 2 N (/rθ )θα . (5.90). Kazama. (5.92) The transformation law for this new field is then  ˜ β + O(g 4 ). From (5. (5. Implication of the Ward identity Having found the transformation laws. [∆1 . is . By a simple calculation. ∆ θ˜α = i(/v)α + g 2 i(/v)α F (r) + ∆ F (r)θ˜α + Gαβ (r. the effective supersymmetry transformations laws are found to be   g2 N m T m (5. it is easy to convince oneself that the most general structure at order 2. ∆2 ]rm = 2˙rm (2 1 ) + O(g 4 ). to the order we are considering. (5. one immediately finds [∆1 . this is not a feature guaranteed by the general analysis.188 Y. the desired field redefinition can be read off as  θ˜α ≡ 1 + g 2 F (r) θα . 5.93) part of this expression vanishes exactly. On general grounds. r   g2 N 3ig 2 N (/rθ )θα .85) r5 Let us summarize the results. Closure of the effective algebra and field redefinition It is of interest to examine the closure property of the transformation laws obtained above. One way to appreciate this is to consider a field redefinition which makes the form of ∆ r m to be the same as the one at the tree level. As we remarked at the end of Section 4. ∆2 ]θα = 2θ˙α (2 1 ) + O(g 4 ).89) Thus to the 1-loop order.3. up to total derivatives. the subscript denoting the number of loops.90) (5.86) ∆ r = −i γ θ 1 + 3 .4.91) (5.88) (5. θ )β . the closure turned out to be precisely canonical. (5.87) ∆ θα = i(/v)α 1 − 3 − r r5 2 ∆θ θα = − 5. are  1 ∆ r m = (γ m θ ) 1 + g 2 F (r) . T. Let us write the effective action up to 1-loop as Γe = Γe0 + Γe1 . we are now ready to analyze the consequence of the Ward identity on the structure of Γe. The tree-level action is given by  2  Z v θ θ˙ + . θ) It is quite nontrivial that the O(g 2 ) (5. i ∆ θα = i(/v)α + g 2 Gαβ (r.94) Γe0 = dτ 2g 2 2g 2 As for Γe1 . To order 2 at 1-loop. the most general form of ∆ r m and ∆ θα .

The simplest way to proceed is as follows. Moreover. This then proves A = 0. Now note that while the four-fermion structures in the second line of (5. 6 Thus we find F = 0.96) dτ F 9 r k −2r 2 (γ l θ )(θ γ lm θ ) + 7r l r j (γ j θ )(θ γ lm θ ) (θ γ mk θ ). we have found that the order 2 contribution at 1-loop for the effective action in the background gauge is completely determined by the requirement of supersymmetry and takes the form − 5 The structure of the form ∼ r l r k (θ γ lmn θ )(θ γ mkn θ ) can be expressed in terms of the last term in (5. Kazama.99) r5 which cancels the first term of (5. T. 6 We have also checked this numerically. They read Z  i (5.Y. (5. Since the Fierz identities can only relate structures with the same number of free indices.98). demand that the O(θ ) part of ∆ Γe vanish. the ones in the last line carry three free indices.98) 3iN rm (γ m θ )(θ θ˙ ). First.98) are all proportional to A. that the second line does not vanish by itself. expressions in these two lines cannot cancel each other. 5 Now we demand that ∆ Γe vanish to the order of interest. the 1-loop level transformation applied to Γe0 produces (5. .97) Γe1 = dτ A 3 + (A + N) 3 + 2 r r r5 where A remains undetermined. look at the O(θ 3 ) part of ∆ Γe. The remaining terms in (5. which can only be produced from the last term of (5. look at the O(θ 5 ) terms. The contributions arising from the tree level transformation of Γe1 are 3iN ˙ rm (γ m θ )(θ θ) r5  3iA + 5 2rm (γ m θ )(θ θ˙ ) − vn (γ m θ )(θ γ mn θ ) − rm (γ n θ˙ )(θ γ mn θ ) 2r  15iA + 7 rm rn vl (γ n θ )(θ γ ml θ ) + (γ m θ )(θ γ nl θ ) . 4r On the other hand. Muramatsu / Nuclear Physics B 584 (2000) 171–196 Γe1 = Z 189  2 v (v · r)2 Θ T θ˙ Θ T /rθ˙ dτ A 3 + B +C 3 +D 4 5 r r r r  Θ T γ mn rm vn θ r l r k (θ γ lm θ )(θ γ mk θ ) +E +F . it is easy to check.95) where A ∼ F are numerical constants. Finally. Next.95) via a Fierz identity and hence can be omitted. contracted with an arbitrary vector rm (τ ) or vm (τ ). r It can be checked that the integrand does not vanish by any of the Fierz identities. (5.95) by the tree-level part of the transformation ∆ rm . using the Fierz identities given in the Appendix A. r5 r7 (5. It is straightforward to show that this reduces the allowed form of Γe1 to be  2  Z v θ θ˙ 3A θ γ mn rm vn θ . In summary.98) have only one “free index”.

. Nevertheless. can be eliminated from the effective action by partial integration and the analysis is essentially the same as in the existing literature. Nonetheless. Summary and discussions In this paper. it is gratifying that already at order 2 it has enabled us to see explicitly how the supersymmetry and the gauge symmetry intimately work together to dictate the form of the effective action. There will be a considerable number of independent structures allowed in the most general effective action.e. it appears feasible to determine whether SUSY alone is enough to fix the form of the effective action at order 4 for arbitrary trajectories. in particular the supersymmetry. with the direct calculation performed in [18].100) This indeed agrees. r3 (5. including the overall normalization. determine the lowenergy effective action of Matrix theory?” Our work was motivated by the observation that the existing analyses are incomplete in that off-shell trajectories with arbitrary timedependence have not been fully considered.87). T. i. from order 4.190 Y.. where complete elimination of higher derivatives will no longer be possible. To find out whether the remarkable . which cannot be determined by the closure property alone. They exhibit an intricate interplay with the gauge (BRST) symmetry of the theory. we have derived the exact supersymmetric off-shell Ward identity for Matrix theory as a step toward answering the “old” yet important unsettled problem: “to what extent do the symmetries. such as the acceleration etc. An important aspect of our Ward identity is that it allows the quantum-corrected effective supersymmetry transformation laws to be directly identified in closed form. Another important direction into which to extend our present work is to apply our Ward identity to genuinely multi-body configurations. It can be easily checked that the normalization is directly linked to the magnitude of the O(θ 0 ) quantum corrections in (5. Muramatsu / Nuclear Physics B 584 (2000) 171–196 Γe1 = Z dτ N θ θ˙ . we computed the explicit form of these transformation laws at 1-loop to the lowest order in the derivative expansion. Kazama.86) and (5. with an aid of computerized calculation. As an application. The full significance of our off-shell Ward identity should become apparent starting from the next order. This is as expected since at this order the higher derivatives. and examined if the invariance under them determines the form of the effective action to the corresponding order. 6. As we shall discuss in the concluding section. Even the proper listing of them requires careful analysis due to the total derivative ambiguities and the existence of nontrivial Fierz identities. a preliminary investigation indicates that. a feature not properly appreciated previously. the power of our off-shell Ward identity can only be fully utilized at the next order in the derivative expansion. such as r¨m and θ¨α . confirming the earlier result [12]. We found that the answer is affirmative. where the most general form of the action unavoidably contains many terms with higher derivatives.

(A. Appendix A. India). Kazama.4) Hereafter in this appendix. however.2) (A. except in the final expression. An analysis based on the general closed form expressions for the transformation laws should shed light on this intriguing question. Acknowledgement We would like to express our special gratitude to Y. Calculation of ∆ θαA (τ ) The explicit expression is . from Japan Ministry of Education. Muramatsu / Nuclear Physics B 584 (2000) 171–196 191 agreement with supergravity in such a situation [3] is due to supersymmetry alone would certainly deepen our understanding of the Matrix theory further. At the more formal and structural level. We hope to be able to report on these and other related issues in the near future.K. we shall omit the overall factor of N . At this order. we exhibit some details of the calculations of O(θ 2 ) terms in ∆ θα at 1-loop order.Y.56)) ∆ θα (τ ) = ∆ θαA (τ ) + ∆ θαB (τ ) + ∆ θαC (τ ). acknowledges the warm hospitality extended to him by the organizers at the third international symposium on Frontiers of Fundamental Physics (Hyderabad.1) where ∆ θαA (τ ) ≡ hδ ψˆ α (τ )i. i (A. what we need to evaluate is (see (5. Z 1 C ∆ θα (τ ) ≡ − dτ 0 hδB ψˆ α (τ )Om (τ 0 )i (γ m θ )(τ 0 ). As we have shown. ∆ θα will refer only to the O(θ 2 ) part. we should mention that a further study should be made on the issue of the closure property of the effective SUSY transformations. we have not been able to answer whether this persists at higher orders and loops. Calculations of O(θ 2 ) terms in ∆ θα In this appendix. The relevant Feynman diagrams are shown in Fig. Also. So far. Y. where a preliminary version of this work was presented. 1. ∆ θαB (τ ) ≡ −hδB ψˆ α (τ )O i. T. at the lowest order the closure turned out to be canonical.3) (A. This work is supported in part by Grant-in-Aid for Scientific Research on Priority Area #707 “Supersymmetry and Unified Theory of Elementary Particles”. Okawa for a number of clarifying discussions and his interest in our work. 10044061. 09640337. Science and Culture. and Grant-in-Aid for International Scientific Research (Joint Research) No. Grant-in-Aid for Scientific Research No.

8) Performing the normal-ordering. this can be computed as Z ˜ )A˜ ∗ (τ 0 )iθα (τ 0 )(γ n θ (τ 00 ))β hΨα (τ 0 )Ψβ∗ (τ 00 )i ˜ )Ym∗ (τ ) = i dτ 0 dτ 00 hA(τ A(τ ×hYn (τ 00 )Ym∗ (τ )i = −ihτ |∆θα (∂ + /r)αβ (γ m θ )β ∆2 |τ i. Using the tree-level in (A. ∆θαA = g 2 AY 2 (A. this becomes ˜ )Ym∗ (τ ) = −i(θ /rγ m θ )hτ |∆3 |τ i = − 3i (θ /rγ m θ ) = 3i rl (θ γ ml θ ). A(τ 16r 5 16r 5 (A. neglecting the terms which generate derivatives. Schematic depiction of Feynman diagrams relevant for the calculations in this appendix. respectively. 2 ∗ ˜ m − Ym A˜ ∗ (γ m )α + i g Ym Yn∗ − Yn Ym∗ (γ mn )α . T.5) To compute the expectation values above to O(θ 2 ). This generates the diagrams (A1) and (A2) in Fig. 1. Consider for example the term hA(τ propagators given in (5.68). we need to perform the second order perturbation using the O(θ ) vertices Z  ˜ α∗ − Ψα A˜ ∗ .9) . 1.5).7) SY Ψ = − dτ (γ m θ )α Ym Ψα∗ − Ym∗ Ψα . e = −i dτ θα AΨ (A. for the first and the second term ˜ )Ym∗ (τ )i.6) SAΨ ˜ Z  e (A.64) and (5.192 Y. (A. Kazama. Muramatsu / Nuclear Physics B 584 (2000) 171–196 Fig.

Again neglecting the derivatives produced in the process of normal ordering we obtain  3ig 2 2(θ )(/rθ )α − 2(/rθ ) θα − 2rl (γ m θ )(θ γ ml )α . In this way one finds ∆θαA =  3ig 2 2rl (θ γ ml θ )(γ m )α − rl (θ γ mnl θ )(γ nm )α . Kazama. C11 and C¯ 11 in the intermediate steps. ¯ b b ∆ θα = −gβ dτ CΨ (τ )C11 ψ β (τ ) + gβ dτ 0 C ∗ Ψ (τ )C¯ 11 ψ β ∆θαB1 = (A.1. in distinction to all the other contributions. ∂ (A. This gives the contribution Z 1 ∆ θαB2−a = 2ig 2 dτ 0 dτ 00 hτ |∆|τ 00 ihτ 00 | |τ 0 ir n (τ 00 )hτ 00 |∆(γ n θ )α ∂∆|τ i(θ )(τ 0 ). the one from (B2) involves propagation of massless fields ψˆ α . (B1) and (B2) in Fig.19) .Y. (A. For (B1). inserting V1 .6) and SY Ψ (A. e Insertions of the vertices e SAΨ ˜ (A. T. The original expression is.18) V5 ≡ − 2 dτ ψ g First. Muramatsu / Nuclear Physics B 584 (2000) 171–196 193 Likewise. b (A. to the order of interest. Z ˙˜ − C¯ A˜˙∗ C − C¯ A˜ ∗ C˙ .7) twice generate the diagrams (B1-a) and (B1-b). 5 16r (A. we get the diagrams of the type (B2-a). Z Z B2 0 ∗ 0 ˙ ˙ (τ 0 ) .14) V1 ≡ −ig dτ r m C¯ ∗ C11 Y m + r m CY Z (A. V2 and V5 . we need to use the following five types of vertices: Z  ¯ m∗ C11 . we have Z B1 2 ¯ 0 )i(∂ + /r)βγ Ψγ∗ (τ 0 )Ψα (τ ) dτ 0 hC ∗ (τ )C(τ ∆ θα (τ ) = β g  (A.10) Calculation of ∆ θαB (τ ) This receives contributions from two classes of diagrams.15) V2 ≡ dτ θ γ n (Ψ Y m∗ − Y m Ψ ∗ ). (A. (A.17) V4 ≡ i dτ θ (Ψ A˜ ∗ − AΨ Z 1 ˙ θ.11) − hC(τ )C¯ ∗ (τ 0 )i(∂ − /r)βγ Ψγ (τ 0 )Ψα∗ (τ ) .16) V3 ≡ g dτ C¯ ∗ C˙ 11 A˜ + C¯ ∗ C11 A 11 11 Z ˜ ∗ ). (A. one easily finds that −hA˜ ∗ (τ )Ym (τ )i gives exactly the same contribution.12) 5 16r Now.13) To extract O(θ 2 ) contributions. The evaluation of the diagram (A2) proceeds in an entirely similar manner.

the final result is ∆ θ α = 3ig 2 N 16r 5 −rl (θ γ mnl θ )(γ nm )α + 2rl (θ γ ml θ )(γ m )α − 4rl (γ m θ )(θ γ ml )α  + 4(θ )(/rθ )α − 4(/rθ ) θα . 4! 1 Q = rm  a1 a2 a3 a4 b1 b2 b3 b4 m (θ γ a1 a2 a3 a4 θ )(λγ b1 b2 b3 b4 m ). we record the SO(9) Fierz identities which are crucial in simplifying the O(θ 2 ) part of ∆ θα at 1-loop.4) (B.23) Summary Adding up all the contributions and reinstating the factor of N . Using the vertex (A. It takes the form Z ∆ θαC (τ ) = g 2 dτ 0 hC(τ )C¯ ∗ (τ 0 )i Ym (τ 0 )Ψα∗ (τ )  ¯ 0 )ihYm∗ (τ 0 )Ψα (τ )i  T γ m θ (τ 0 ). let us introduce the following quantities for n = 0. E¯ n = rm (γ a1 ···an θ )(θ γ an ···a1 m λ). Muramatsu / Nuclear Physics B 584 (2000) 171–196 Similarly.24) Appendix B.6) . 1.21) Calculation of ∆ θαC (τ ) Finally. 3 (repeated indices are summed): En = rm (γ a1 ···an m θ )(θ γ an ···a1 λ). use of V3 . T. F¯n = rm (θ γ 1 P = rm  a1 a2 a3 a4 b1 b2 b3 b4 m (γ a1 a2 a3 a4 θ )(θ γ b1 b2 b3 b4 m λ).1) (B. a1 ···an m an ···a1 θ )(λγ ). Adapting the notations of Taylor and Raamsdonk [5]. ∆ θαB2 = ∆ θαB2−a + ∆ θαB2−b = 2ig 2 (θ )hτ |∆3 |τ i(/rθ )α = 3ig 2 2(θ )(/rθ )α . 2. 4! (B. SO(9) Fierz identities In this appendix. Fn = rm (θ γ a ···a a ···a m n n 1 1 θ )(λγ ). Kazama.7) and proceeding similarly to the previous calculations. + hC ∗ (τ )C(τ (A.194 Y. the contribution of which is worked out to be ∆ θαB2−b = 2ig 2 hτ |∆(θ )∆r m (γ m θ )α ∆|τ i − ∆ θαB2−a . V4 and V5 generates the diagrams of the type (B2-b). consider ∆ θαC . we obtain ∆θαC =  3ig 2 −2(/rθ ) θα − 2rl (γ m θ )(θ γ ml )α . (A. 16r 5 (A. (A.5) (B.22) which is represented by the diagram (C).3) (B. 5 16r (A.2) (B.20) Therefore.

namely F0 = F¯0 = F¯1 = F3 = Q = 0. Fischler. hep-th/9709220. one easily finds the identity 0 = F2 − 2F1 − 4E¯ 1 + 4E¯ 0 + 12E0 . 2 From these relations. S. five of them actually vanish.17) which was used in Section 5. Removing λ. hep-th/9610043. M. we get 0 = −rl (θ γ mnl θ )(γ nm )α + 2rl (θ γ nl θ )(γ n )α + 4rl (γ n θ )(γ nl θ )α + 4rl (θ )(γ l θ )α + 12rl (γ l θ ) θα .16) P = (F2 + F¯2 ). References [1] [2] [3] [4] [5] [6] T. L.20) of [5] should be +.13) E¯ 2 = − F1 − (F2 + F¯2 ) − F¯3 . Because θ Γ θ = 0 for any symmetric matrix Γ . 4 4 24 21 3 3 (B. Theor. Phys. Adv. W. F2 . Phys. hep-th/9808188.11) E¯ 1 = − F1 + (F2 − F¯2 ) − F¯3 . Okawa. (B. (B. T. Muramatsu / Nuclear Physics B 584 (2000) 171–196 195 where λ is an arbitrary spinor.14) E3 = − F1 + (F2 − F¯2 ) + F¯3 . Kazama. B 541 (1999) 163. A. Banks. hep-th/9806108.15) E¯ 3 = − F1 − (F2 − F¯2 ) + F¯3 . Phys. 0 . 8 8 48 3 1 1 (B. Yoneya. 4 4 8 15 (B. 7 The sign in front of 48E − in Eq. Shenker.10) E1 = − F1 − (F2 − F¯2 ) − F¯3 . 4 4 24 7 1 1 (B. hep-th/9704080. Susskind. Rev. Y. B 538 (1999) 67.V.12) E2 = F1 − (F2 + F¯2 ) + F¯3 . hep-th/9812239. 2 (1998) 51. Phys. T. Nucl. (B. Y. L. Sen. (B. Raamsdonk.7) Since there are nine independent Fierz identities [5]. JHEP 04 (1999) 013. 4 4 8 21 3 3 (B. Then the remaining quantities can be expressed in terms of them as 7 1 1 1 E0 = F1 − (F2 + F¯2 ) − F¯3 . 8 8 48 7 1 1 (B.8) 16 32 96 1 1 1 (B. T. F¯2 and F¯3 . Yoneya. D 55 (1997) 5112. Taylor. Nucl. 16 32 96 3 1 1 (B. Okawa. W.H. only four structures are independent.Y. Susskind. which we take to be F1 . Math.9) E¯ 0 = − F1 − (F2 + F¯2 ) + F¯3 .

Lowe. B 535 (1998) 335. Yoneya. T.A. Theor. hep-th/9810075. hepth/9608024. Stern. Pouliot.R. Phys. Kazama. JHEP 11 (1998) 009. D. Nucl. Nucl. hep-th/9904042. S. B 558 (1999) 349. Seiberg. J. B 485 (1997) 85. Jevicki. 1999. Muramatsu / Nuclear Physics B 584 (2000) 171–196 N. hep-th/9805069. hep-th/9808039. Adv. D. Maldacena. B 534 (1998) 137. 2 (1998) 231. Rev. Nucl. Paban. Phys. Shin. B. S. Y. Douglas. Rev. S.H. Phys. Phys. Lett. Nucl. hep-th/9903049. H. Freedman. M. Y. H. M. T. Kazama. hep-th/9903025. H. Yoneya. Y. Phys. Rev. D 12 (1975) 2286. Okawa. Phys. hep-th/9901034. Nicolai. Hata. Y. B 452 (1999) 45. A. A. 79 (1997) 3577. Shenker. Kiem. M. Stern. H. D 59 (1999) 066001. Moriyama. Paban. Sethi. Hyun. D 60 (1999) 126006. 81 (1998) 5072. Y. JHEP 06 (1998) 012. Moriyama. Phys. Talk at YITP Workshop. Sethi. Jevicki. hep-th/9903022. S. hep-th/9810146. July. Rev. Kabat. Rev. Phys. Lett. S. T. . Stern. Math. A. hep-th/9806028. Lett. Phys. hep-th/9711200.196 [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] Y. S. de Wit. Hata. D. Okawa. Jevicki. J. Phys. S. hep-th/9710009. Phys. Phys. Nucl. S. B 552 (1999) 447. hep-th/0001106. Plefka. hep-th/9805018. Sethi. Kazama. M. S. Kyoto. JHEP 06 (1999) 004. P.Z. Yoneya. T.

80. accepted 6 June 2000 Abstract We consider chiral condensates in SU(2) gauge theory with broken N = 2 supersymmetry.  2000 Elsevier Science B. All rights reserved. PII: S 0 5 5 0 . University of Minnesota. Gatchina. A. We interpret this phenomenon as a deconfinement of electric and magnetic charges at the Argyres–Douglas point. The solution is described in terms of Riemann surfaces and the Coulomb branch parametrizes the moduli 0550-3213/00/$ – see front matter  2000 Elsevier Science B. PACS: 11. These results are consistent with those obtained by the ‘integrating in’ procedure. MN 55455. Vainshtein b .V. Matter and gaugino condensates are determined by integrating out the adjoint field.elsevier. Russia Received 25 April 2000. St. We then calculate monopole.Nuclear Physics B 584 (2000) 197–215 www.7 . then new wave of development was initiated by Seiberg. Yung c a Institute of Experimental and Theoretical Physics.-q.V. The only nonperturbative input is the Affleck–Dine–Seiberg (ADS) superpotential generated by one instanton plus the Konishi anomaly. Petersburg 188350. monopole 1. 11. The key feature of the N = 2 theory is the existence of the Coulomb branch where the vacuum expectation value of the adjoint scalar serves as a modulus [5]. Gorsky a .30. chiral condensates.nl/locate/npe Deconfinement at the Argyres–Douglas point in SU(2) gauge theory with broken N = 2 supersymmetry A. Introduction The derivation of exact results in N = 1 supersymmetric gauge theories based on low energy effective superpotentials and holomorphy was pioneered in [1–3]. Moscow 117259. and charge condensates using the Seiberg–Witten approach. The matter sector contains an adjoint multiplet and one fundamental flavor. Minneapolis. 14. dyon. USA c Petersburg Nuclear Physics Institute. We show that the monopole and charge condensates vanish at the Argyres–Douglas point where the monopole and charge vacua collide.15. including a reproduction of the Seiberg–Witten curve from the ADS superpotential.Pb.3 2 1 3 ( 0 0 ) 0 0 3 4 9 . A.Hv Keywords: Supersymmetric gauge theories. Russia b Theoretical Physics Institute. see [4] for review. All rights reserved. Additional input was provided by the Seiberg–Witten solution of N = 2 supersymmetric gauge theories with and without matter [5].

This field is a partner to the gauge fields in the N = 2 supermultiplet.198 A. Recently. Our strategy is as follows: First. Namely. while the gluino condensate can be evaluated using the Konishi anomaly which relates the two condensates (see Ref. We then compare the condensate of the adjoint matter with points in the u plane corresponding to the vanishing of the discriminant defined by Seiberg–Witten solution in N = 2 theory and find a complete match. Difference with pure SQCD is due to the tree level nonrenormalizable term generated by the heavy adjoint exchange. such as the gluino e condensate hTr λλi and the condensate of fundamental matter hQQi. we integrate out the adjoint matter to get SQCD-like effective superpotential for the fundamental matter. The simplest way to break N = 2 supersymmetry (SUSY) down to N = 1 amounts to giving a nonvanishing mass µ to the chiral N = 1 superfield in the adjoint representation.9. Generically. the latter can be found in SQCD using the effective superpotential. / Nuclear Physics B 584 (2000) 197–215 space of their complex structures. mentioned above.23] and can be viewed as an independent confirmation of this method. At small values of µ the theory is close to its N = 2 counterpart while at large µ the adjoint matter decouples and the pure N = 1 theory emerges. The only nonperturbative input in this effective superpotential is given by the Affleck–Dine–Seiberg superpotential generated by one instanton [1]. some points concerning formation of the condensate and identification of the relevant field configurations were clarified in [15–18] The brane picture provides another approach to the problem. To obtain the condensate in pure N = 1 Yang–Mills theory one has to start with the massive SQCD and use the holomorphy to decouple massive matter. Gorsky et al. memory of the structure of the Riemann surfaces remains. Similarly to SQCD. Different vacua are distinguished by the values of chiral condensates. The key point concerning the brane configurations is that to break SUSY down to N = 1 one has to rotate the N = 2 picture. the vanishing of the discriminant of the Riemann surface defines the set of vacua in the corresponding N = 1 theory [5–13]. Our results for matter and gaugino condensates are consistent with those obtained by the ‘integrating in’ method [8. The brane configurations for N = 1 theories with different matter content are known [19. A trace of the massive adjoint remains in the effective theory in the form of nonrenormalizable quartic terms [6] in the superpotential which are suppressed by 1/µ.22. [14] for a review). the effective superpotential together with the Konishi relations unambiguously fixes condensates of fundamental and adjoint matter as well as the gaugino condensates in all three vacua of the theory. What is specific to our approach is . The emerging theory at large µ is close to supersymmetric QCD (SQCD) but does not coincide with it. The superpotentials calculated from brane configurations are in correspondence with field theory expectations. However only configurations which correspond to the vanishing of the discriminant can be rotated which means that at any value of the adjoint mass these points remain intact. Although in the N = 1 theory the degeneracy on the Coulomb branch is lifted by the superpotential. In this paper we consider an N = 1 theory with both adjoint and fundamental matter and limit ourselves to the most tractable case of SU(2) gauge group with one fundamental flavor and one multiplet in the adjoint representation.20] and the recipe for calculating minima of the superpotentials has been formulated [21].

that it was before the notion of the AD point was introduced [24].A. This paradox is a part of more general problem: whether there is an uniquely defined theory at the AD point. The same continuity also leads to vanishing tension for domain walls interpolating between colliding vacua when we approach the AD point. while the condensation of charges provides confinement of monopoles in the charge vacuum. 1 We interpret this as deconfinement of both electric and magnetic charges at the AD point. Our result resolves this paradox. Note. in the N = 1 theory it is the point in parameter space where two vacua collide. Argyres and Douglas conjectured this in their consideration of SU(3) theory [24] Let us recall that the condensation of monopoles ensures confinement of quarks in the monopole vacuum [5]. As shown by ’t Hooft [28] it is impossible for these two phenomena to coexist. The paper is organized as follows. Our key result is that both monopole and charge condensates vanish at the AD point. In Section 2 we dwell on the calculation of matter and gaugino condensates. Indeed. and charge condensates following the Seiberg–Witten approach. The continuity of chiral condensates in the AD point we find shows that these condensates are not playing this role. charge and dyon condensates are considered in Section 3. We discuss if these domain walls could serve as a signal of two vacua in the AD point. while monopole.. 1 Vanishing of condensates for coalescing vacua was mentioned by Douglas and Shenker [27] in the context of SU(N ) theories without fundamental matter for N > 3. It is believed that the theory in the AD point flows in the infrared to a nontrivial superconformal theory. This extension includes not only the mass term of adjoint but also breaking of N = 2 in Yukawa couplings. considering effective superpotentials near singularities on the Coulomb branch of the N = 2 theory.e. i. The notion of the AD point continues to make sense even when the N = 2 theory is broken to N = 1. . holomorphy allows us to extend our results to the domain of the “hard” N = 2 breaking. Gorsky et al. These points were originally introduced in the moduli/parameter space of N = 2 theories as points where two singularities on the Coulomb branch coalesce [24–26]. In particular. The question is whether there is any physical quantity which could serve as an order parameter differentiating these two vacua. / Nuclear Physics B 584 (2000) 197–215 199 that we start from the weak coupling regime where the notion of an effective Lagrangian is well defined. when two vacua collide the Witten index of the emerging effective theory at the AD point is fixed. We subsequently determine monopole. we consider the AD point where the monopole and charge vacua collide at a particular value of the mass of the fundamental flavor. In Section 4 we briefly discuss a definition of the theory at the AD point and the related problem of domain walls. namely there are two bosonic vacuum states. Our next step is the study chiral condensates in the Argyres–Douglas (AD) points. Our results are discussed in Section 5. and then use holomorphy to extend results for chiral condensates into strong coupling. dyon. It provides evidence that the theory at the AD point remains superconformal even after strong breaking of N = 2 to N = 1. This apparently leads to a paradoxical situation in the AD point where the monopole and charge vacua collide. Again.

In the classical approximation this integration reduces to the substitution   1 1 α γ α fg α Qβf Qg − δβ Qγf Qg . 2. To obtain an effective theory similar to SQCD we integrate out the adjoint field Φ implying that mΦ  mQ . (5) Λ31 = 4MPV g0 V= The coefficient µ2 Λ31 /4 in the ADS superpotential is equivalent to Λ5SQCD in SQCD. The scale parameter Λ1 is given in terms of the mass of Pauli–Villars regulator MPV and the bare coupling g0 (plus the vacuum angle θ0 ) as   8π 2 3 exp − 2 + iθ0 . 2) describing one flavor. Having in mind normalization appropriate for the N = 2 case we choose for bare 0 0 = 1/g02 . The effective superpotential then is Weff = mV − (− det h) 2 µ2 Λ31 V + 4µ 4V (3) where the gauge and subflavor invariant chiral field V is defined as 1 α f Q Q . (2) Φβ = − √ h 2 2 2µ which follows from ∂W/∂Φ = 0. 2 is explicit.. β = 1. mΦ = µ/ZΦ . β = 1. mQ = m/ZQ . The general renormalizable superpotential for this theory has the form. Effective superpotential and condensates We consider a N = 1 theory with SU(2) gauge group where the matter sector consists of the adjoint field Φβα = Φ a (τ a /2)αβ ( α. and two fundamental fields Qαf (f = 1. Ref. 2 f α 2 (1) Here the parameters µ and m are related to the masses of the adjoint and fundamental fields. a = 1. Unbroken N = 2 SUSY appears when µ = 0 and det h = −1. e. and parameters ZΦ summation over color indices α. Gorsky et al. 2. The matrix of Yukawa couplings hfg is symmetric. . by the corresponding Z factors in the kinetic terms. 3). / Nuclear Physics B 584 (2000) 197–215 2. The factor µ2 in the coefficient reflects four fermionic zero modes of the adjoint field. (4) 2 f α The first two terms in Eq. (3) appear on the tree level after substitution (2) into Eq. Matter and gaugino condensates 2. W = µ Tr Φ 2 + m α f 1 Q Q + √ hfg Qαf Φβα Qβg . see. ZQ = 1.g. (1) while the third nonperturbative one is the ADS superpotential. What is the effect of quantum corrections on the effective superpotential? It is well known from the study of SQCD that perturbative loops do not contribute and nonperturbative effects are exhausted by the Affleck–Dine–Seiberg (ADS) superpotential generated by one instanton [1].29] for details.1.200 A. [17.

The charges of the fields and parameters of the theory under these two U(1) symmetries are shown in Table 1.   (− det h) Λ1 3/2 . (8) v=µ 4m Then Eq. When det h is nonvanishing we have three vacua. m− 2 µ 4 v (6) (7) This equation shows. (7) it is also clear that the dependence on µ is given by the scaling v ∝ µ. adjoint and fundamental fermions). Thus. At h = 0 it vanishes and we are back to the known SQCD case with two vacua and a Higgs phase for small m. To see the dependence on the other parameters let us substitute v by the dimensionless variable κ defined by the relation s Λ31 κ. it is given by a power of µ equal to half the UJ (1) charge of the condensate. / Nuclear Physics B 584 (2000) 197–215 201 The only term in the superpotential (3) which differentiates it from the SQCD case is the second term which is due to tree level exchange by the adjoint field. m and det h as background fields and identify nonanomalous R symmetries which prove the dependence discussed above. m and det h follows from the R symmetries of the theory. v = hV i. we can use holomorphy to extend results to arbitrary values of µ. The dependence of v on µ is linear as we discussed above. there are three U(1) symmetries in the theory associated with the three fermion fields (gaugino. . marked by the vevs of the lowest component of V . when rewritten in terms of κ. In the quantum theory one can organize two nonanomalous combinations (a symmetry is nonanomalous if it does not transform the scale Λ1 . This explains the zero charge of the coupling h with respect to this symmetry. 1 =0 κ2 is governed by the dimensionless parameter σ . the field V has UJ (1) charge equal to 2 which ensures that v ∝ µ. The particular dependence of condensate v on the parameters µ. that although the second term in the superpotential (3) seems to be suppressed at large µ it turns out to be of the same order as the ADS term. Classically. From Eq. Namely.A. associated with regulators). in particular. Gorsky et al. These vevs are roots of the algebraic equation dWeff /dv = 0 which has the form   Λ31 µ 2 (− det h) v − = 0. In particular. σ= 4 m 1 − σκ − (9) (10) We see that the two parameters m and det h enter only as m(− det h)−2/3 . The first of these symmetries UJ (1) is a subgroup of the global SUR (2) group related to the N = 2 superalgebra [5]. Following Seiberg [30] one can consider µ. The symmetry UJ (1) fixes the µ dependence of condensates. (7).

The left hand side is a total derivative in superspace so its average over any supersymmetric vacuum vanishes. As a consequence. Second. (7). (11) where QJ . Eq. Here is an example. / Nuclear Physics B 584 (2000) 197–215 Table 1 Nonanomalous U(1) symmetries Fields/parameters Φ Q W θ m µ h UJ (1) charges UR (1) charges 0 1 1 −1 1 1 1 1 0 4 2 0 0 3 The second nonanomalous symmetry UR (1) is similar to the R symmetry of Ref. 2 f α 2 8π 2 2   1 Tr W 2 = 0. Firstly. we find an expression for gluino condensate [31] . 4 ∂Q 8π 2 (13) where T (R) is the Casimir in the matter representation. (10). [1] extended to include the adjoint field. and fX is an arbitrary function of the dimensionless parameter σ defined by Eq. To verify this interesting mapping we need to determine the vev (12) u = hU i = Tr Φ 2 . one can relate them to two vacua of SQCD in the Higgs phase. This parameter is neutral under both U(1)’s. after the substitution in (2) and comparing with Eq. which can be accomplished using the set of Konishi anomalies. The important benefit of the consideration above is that in a theory with N = 2 SUSY strongly broken by large µ and det h 6= −1 we can still relate chiral condensates with those in softly broken N = 2 where det h = −1 and µ is small. In our case this results in two relations for the condensates. For two finite roots one can suggest dual interpretations.2 = ±1 and the third one goes to infinity as κ3 = 1/σ . But this limit should bring us to the monopole and dyon vacua of softly broken N = 2 SYM. (8) is an example of the general relation (11) with QJ = 2.   1 fg 1 Tr W 2 m α f α β Q Q + √ h Qαf Φβ Qg + = 0. (14) 2µ Tr Φ 2 + √ hfg Qαf Φβα Qβg + 2 8π 2 2 From the first relation.202 A. for det h = −1 (which is its N = 2 value) one can make σ small by taking the limit of large m. The naming of vacua refers to the particle whose mass vanishes in the corresponding vacuum. Generic equation for an arbitrary matter field Q looks as follows (we are using the notation of the review [14]): Tr W 2 ∂W 1 2 + T (R) D JQ = Q . UR (1) charges of the field X. (9) are κ1. Gorsky et al. taking h = 0. QR are the UJ (1). When σ → 0 two roots of Eq. QR = −2 and fV = κ(σ )/2. dX is its dimension. for a given chiral field X d −(QJ /2)−(QR /4) hXi = µQJ /2 mQR /4 Λ1X fX (σ ).

y 2 = 0 and dy 2/dx = 0. (21) This is in agreement with the master parameter σ which contains the product m−3/2 det h and the nonanomalous U(1) symmetries we discussed above. (16) u= 2µ 2µ 4 v 4 κ Now we see that in the limit of large m two vacua κ = ±1 are in perfect correspondence with u = ± Λ20 for the monopole and dyon vacua of N = 2 SYM. (17) at det h = −1. Gorsky et al. / Nuclear Physics B 584 (2000) 197–215 s= hTr λ2 i hTr W 2 i µ2 Λ31 . This means . i. In other words.e. q    3 2 mΛ31  1 3 µ Λ1 3 1 (mv + 3s) = mv + = κ+ . Combining the two relations in (14) we can express the condensate u in terms of v. Moreover. the correspondence with N = 2 results can be demonstrated for the three vacua at any value of m. where some fundamental fields become massless.A. 8 2 We see that this is a particular case of the N = 1 Eq. 8 2 (17) The three roots of this equation are the vevs of Tr Φ 2 in the corresponding vacua. Λ40 = mΛ31 is the correct relation between the scale parameters of the theories. (19) x 3 − ux 2 + Λ31 mx − Λ61 = 0. (20) after the rescaling u = (− det h)1/3 u0 . 1 1 1 3x 2 − 2ux + Λ31 m = 0. v = (− det h)−1/3 v 0 . breaking of N = 2 by Yukawa couplings does not influence consideration of the chiral condensates modulus the rescaling (21). the points in the u-plane where the discriminant of the curve vanishes are defined by two equations.. The consideration above shows that the only nonperturbative input needed to determine the chiral condensates is provided by the one-instanton ADS superpotential. Moreover. m = (− det h)2/3 m0 . (9) to derive the following equation for u. when det h is not equal to its N = 2 value (−1) Eq. To this end we use the relation (16) and Eq. Indeed. 4 64 Singularities of the metric. How does this look from N = 2 side? The Riemann surface governing the Seiberg– Witten solution is given by the curve [5] 1 1 (18) y 2 = x 3 − ux 2 + Λ31 mx − Λ61 . see [32]. 9 27 (− det h)u3 − m2 u2 − (− det h)mΛ31 u + m3 Λ31 + 8 (− det h)2 Λ61 = 0. 4 64 4 which lead to 9 27 (20) u3 − m2 u2 − mΛ31 u + m3 Λ31 + 8 Λ61 = 0. For the third vacuum at large m the value u = m2 /(− det h) corresponds on the Coulomb branch to the so called charge vacuum. (17) coincides with Eq. = − = 16π 2 16π 2 4v 203 (15) P This is consistent with the general expression [TG − T (R)]hTr λ2 i/16π 2 for the nonperturbative ADS piece of the superpotential (3).

Moreover. when the mass mV of the field V . (− det h)  m2 1 + O σ2 . it contains no light monopole degrees of freedom near the monopole vacuum point at small µ. the superpotential (3) gives a complete description of the low energy physics. What we see as an advantage of our approach it is that. Indeed. Gorsky et al.D = ± Λ31 m 1 + O(σ ) . u = hTr Φ 2 i. mV = 2m(2 − 3σ κ). in regard to these condensates the exact Seiberg–Witten solution of N = 2 is equivalent to the ADS superpotential. In the charge vacuum:  2µm 1 + O σ2 . Holomorphy then allows for continuation of these results for the condensates to strong coupling. Matter and gaugino condensates in the limit of large mass f Here we summarize the results for matter. such as mΦ = g 2 µ and mW = |g 2 v|1/2 . and gaugino. The Konishi relations help to determine the condensates of heavy fields in this phase..D = ±µ 4m q  uM.9] by the ‘integrating in’ procedure introduced in [23]. v = hQαf Qα /2i. sC = 8m vC = In the monopole and dyon vacua: s  Λ31 1 + O(σ ) . At strong coupling the superpotential (3).34]. within a certain range of parameters.204 A.2. can be viewed as an independent proof of the ‘integrating in’ procedure.e. Our approach which is based on ‘integrating out’. uC = (− det h)  µΛ31 (− det h) 1 + O σ2 . we are in the weakly coupled Higgs phase and enjoy full theoretical control. The relations for the condensates we have derived are not new. does not describe the low energy physics. (10) is small. / Nuclear Physics B 584 (2000) 197–215 that any reference to the N = 2 limit is not crucial at all.9] of the ‘integrating in’ procedure and can be viewed as a shorthand equation that gives the values of the condensates. This can be achieved in the limit of large m if the Yukawa coupling is fixed. vM. plus the Konishi relations. At strong coupling the superpotential (3) is equivalent to the effective superpotential [8. 2. (22) is much less than the other masses. For example. there is no single local superpotential which could describe mutually nonlocal degrees of freedom which become light in different regions of the moduli space of the theory. they were obtained in [8. condensates in the limit where the parameter σ defined by Eq. i. (23) . or by taking det h to be small otherwise. s = hTr λ2 /16π 2 i. like other versions of the Veneziano– Yankielowicz Lagrangians [33.

. Let us consider this transition when. On the Coulomb branch of the N = 2 theory these points correspond to a non-trivial conformal field theory [25]. Small mass limit The limit of massless fundamentals m → 0 corresponds to σ → ∞. (23). Here we study the N = 1 SUSY theory. Argyres–Douglas points When the mass m changes from large to small values we interpolate between the two quite different structures of vacua shown above. from Eqs. Gorsky et al.  µΛ1 e2πik/3 1 + O σ −2/3 (k = 0. Collisions of two vacua still occur in this theory. 4 (25) 2. but much larger. (24) we see that the dyon vacuum is at negative u. values of u. det h = −1 and m is real and positive and changes from large to small values. become massless at these points.4.D = ± µ Λ31 m 1 + O(σ ) . s = µΛ21 (2 det h)1/3 e−2πik/3 1 + O σ −2/3 4 v= (26) Note that the massless limit exists due to the nonvanishing Yukawa coupling. for definiteness. 2 205 (24) The upper sign refers to the monopole vacuum. When h → 0 we are back to the runaway vacua of massless SQCD. / Nuclear Physics B 584 (2000) 197–215 q  1 sM. ±1).3. The point in the parameter manifold where the two vacua coincide is the AD point [24]. ±1). the monopole vacuum is at positive u. 2. 1/3 (2 det h)  3 (k = 0. ±1). while the lower one is for the dyon vacuum. In this limit the three vacua are related by a Z3 symmetry [5]. where N = 2 is broken by the mass term for the adjoint matter as well as by the difference of the Yukawa coupling from its N = 2 value. u = Λ21 (2 det h)1/3 e−2πik/3 1 + O σ −2/3 8  1 (k = 0. In the next section we study what happens to the confinement of charges in the monopole point at non-zero µ once we approach the AD point. At large positive m all the vacua are situated at real values of u. As discussed above we can interpret these vacua also as the two vacua of the Higgs phase in SQCD. To this end we need to consider the limit of small det h and m  ΛSQCD with the identification 1 Λ5SQCD = µ2 Λ31 . say charges and monopoles. When m diminishes then at some point the monopole and charge vacua collide on the real axis of u and subsequently go more off to complex values producing the Z3 picture at small m. In the SU(2) theory these points were studied in [25]. and the charge vacuum is also at positive.A. In this subsection we find the values of m at which AD points appear and calculate the values of the condensates at this point. Mutually non-local states.

let us work out the AD values of m. 3 mAD = ωΛ1 (− det h)2/3 . ±1). The second term breaks N = 2 supersymmetry down to N = 1.    3 Λ3 µ 2 µ (− det h) v − 1 = 0. Dyon condensates In this section we calculate various dyon condensates at the three vacua of the theory. where AD is a neutral chiral field (it is a part of the N = 2 dual photon multiplet in the N = 2 theory) and U = Tr Φ 2 considered as a function of AD . Following Seiberg and Witten [5] we write down the effective superpotential in the following form.206 A. (28) related by Z3 symmetry. we calculate the monopole condensate near the monopole point. / Nuclear Physics B 584 (2000) 197–215 First. holomorphy allows us to find these condensates starting from a consideration on the Coulomb branch in N = 2 near the singularities associated with a given massless dyon. generalizing the consideration in [25]. the charge condensate near the charge point and the dyon (nm . (− det h)1/3 3 uAD = ω−1 Λ21 (− det h)1/3 . . Near this point the effective low energy description of our theory can be given in terms of N = 2 dual QED [5]. √ e (30) W = 2 MMA D + µU. (7) the derivative of its lefthand-side should also vanish. ne ) = (1. sAD = ω 4 vAD = ω (29) 3. 3. 4 −1 1 µΛ21 (− det h)1/3 . 1) condensate near the point where this dyon is light. Although we start with small values of the adjoint mass parameter µ. our results for condensates are exact for any µ as well as for any value of det h. It includes a light monopole hypermultiplet interacting with a vector (dual) photon multiplet in the same way as electric charges interact with ordinary photons. Coalescence of two roots for v means that together with Eq. (27) m− 2 µ 4 v v This system is consistent only at three values of m = mAD . 4 ω = e2πin/3 (n = 0. Gorsky et al. −(− det h) + Λ31 = 0. The condensates at the AD vacuum are µΛ1 . Namely.1. As discussed above. Monopole condensate Let us start with calculation of the monopole condensate near the monopole point.

the Varying this superpotential with respect to AD .e. i. / Nuclear Physics B 584 (2000) 197–215 207 e we find that AD = 0. and µ du .A. Gorsky et al.. M and M monopole mass vanishes.

.

(31) MM . e = −√ .

2 daD aD =0 The condition AD = 0 means that the Coulomb branch near the monopole point. Let us work out the r. The non-zero value of the monopole condensate ensures U(1) confinement for charges via the formation of Abrikosov–Nielsen–Olesen vortices.1 this limit can be viewed as a RG flow to pure Yang–Mills theory with the identification Λ40 = mΛ31 . At the monopole vacuum. when u = uM . r 3 (35) e − e0 = u2M − mΛ31 . where the monopole mass vanishes. (36) MM 1 4 To test the result let us consider first the limit of a large masses m for the fundamental matter. u = uM . (18) we use the form y 2 = (x − e0 )(x − e− )(x − e+ ). (34) du 4 e − e0 The value of e − e0 (equal at u = uM to (1/2) d2(y 2 )/dx 2 ) is fixed by the equation d(y 2 )/dx = 0. we have √ I 2 dx daD = . two branch points e+ and e− coincide. (36) gives √ e = 2 iµΛ0 . Then Eq. In this limit we have uM = Λ20 . (31) determines the value of monopole condensate. . (31) to determine the µ and m dependence of the monopole condensate. From the exact Seiberg–Witten solution [5]. of Eq. √ i 2 daD (uM ) = √ . e+ = e− = e and the integral (32) is given by the residue at x = e. Below we consider aD as a function of u. The value of u. (38) MM which coincides with the Seiberg–Witten result [5]. As in Section 2. (33) The integration contour γ in the x plane circles around two branch points e+ and e− of y(x). shrinks to a single vacuum state at the singularity while Eq. This ensures monopole condensation and charge confinement in the monopole point at large m. 4 Substituting this into the expression for the monopole condensate (31) we get finally 1/4  e = 2iµ u2M − 3 mΛ3 . (37) where Λ0 is the scale of the N = 2 Yang–Mills theory. at the monopole vacuum was determined in the previous section.h.s. (32) du 8π y(x) γ Here for y(x) given by Eq.

At the AD point all three roots of y 2 become degenerate. √ e+ = e− = e0 . In particular. so the monopole condensate which is proportional to e − e0 naturally vanishes. The UR (1) charge of MM √ e Indeed. (7) for v.. (28) and (29). (36). can be easily generalized to arbitrary det h by e is equal to one. Thus we interpret the AD point as a deconfinement point for both monopoles and charges. Gorsky et al. . the coefficient of the 2 MMAD term in the superpotential (30) which is equal to one in the N = 2 limit remains the same when N = 2 is broken down to N = 1. however. When m and u are substituted by mAD and uAD from Eqs. We first consider m above its AD value (28) and then continue our results to values of m below mAD . / Nuclear Physics B 584 (2000) 197–215 Notice. we get e = 0. in the limit m = 0 we recover Z3 symmetry.2. we assume det h = −1. It is instructive to rewrite the result (36) for the monopole condensate in terms of v.    1/2 µ3 Λ31 1/2 2 e = i (− det h)v 2 − 4µs . (39) MM = i (− det h)v − v where we also show a form which uses the gluino condensate s. i. The question is: what happens when monopole and charge points collide in the u plane? The monopole condensate at the AD point is given by Eq. As a result we see that the Eq. when the value of v is large and the nonperturbative term in (39) can be neglected.e. and the square of the monopole condensate reduces to the gluino condensate. (40) MM AD We see that the monopole condensate goes to zero at the AD point. means of U(1) symmetries considered above. Our derivation makes it clear why it happens. The result. Now let us address the question: what happens with the monopole condensate when we reduce m and approach the AD point? The AD point corresponds to a particular value of m which ensures coalescence of the monopole and charge singularities in the u plane. 3. In the next subsection we calculate the charge condensate in the charge point and show that it also goes to zero as m approaches its AD value (28). that in the derivation above N = 2 was not broken by the Yukawa coupling. the monopole √ e condensate reduces to that of the quark. Another interesting limit is the SQCD one when h → 0. (39) dominates. It is interesting to observe that at m  Λ1 . It follows from U(1) symmetries together with decoupling of fundamental matter at large m. Near the monopole point we have condensation of monopoles and confinement of charges while near the charge point we have condensation of charges and confinement of monopoles. As shown by ’t Hooft these two phenomena cannot happen simultaneously [28]. hMMi → iv − det h. Charge and dyon condensates In this subsection we use the same method to calculate values for the charge and dyon condensates near the charge and dyon points respectively. It follows from the expression (16) for u and Eq. In this limit the nonperturbative term in Eq. (36) for the monopole condensate remains valid for arbitrary det h.208 A.

/ Nuclear Physics B 584 (2000) 197–215 209 Let us start with the charge condensate. (42) W=√ Q + Q+ A + mQ 2 Minimizing this superpotential we get condition (41) as well as . the partner of photon in the N = 2 supermultiplet. At µ = 0. det h = −1 and large m the effective theory near the charge point √ (41) a = − 2m on the Coulomb branch is N = 2 QED. Once we hypermultiplet Q add the mass term for the adjoint matter the effective superpotential near the charge point becomes 1 e e+ Q+ + µU. Q+ of charged particles in the effective electrodynamics.A. Here a is the neutral scalar. The massless fields form one e+ . Gorsky et al. Half of the degrees of freedom in color doublets become massless whereas the other half acquire a large mass 2m.

√ e+ Q+ = − 2 µ du .

.

As with the monopole condensate. (44) to find the charge condensate at the AD point. (46) Q+ Q+ = v − v(− det h) (− det h) This expression differs from the one for the monopole condensate only by a phase factor. is natural. similar to the hMMi condensate × hQ considered above. As it was mentioned we interpret this as deconfinement for both charges and monopoles. At large m uC = m2 /(− det h). (23). i.e. and e+ Q+ = Q  2µm 1 + O σ2 . The coincidence of the charge condensate with the quark one at large v. The difference is due to nonperturbative effects and is similar to the difference between a 2 /2 and u on the Coulomb branch of the N = 2 theory. We choose to consider − det h e e+ Q+ i because it has the UR (1) charge equal to one.. . da a=− 2 m Now. 1 4 (43) (44) √ where we include a generalization to arbitrary det h. By uC we denote the position of the charge vacuum in the u plane. Q √ . Using Eqs. we get  1/4 √ e+ Q+ = 2µ u2C − 3 mΛ3 − det h Q . So we can use Eq. at weak coupling. In particular.  1/2  1/2 µ3 Λ31 4µs 2 2 e = v − . (28) and (29) we see that the charge condensate vanishes at the AD point in the same manner the monopole condensate does. (− det h) (45) Holomorphy allows us to extend the result (44) to arbitrary m and det h. following the same steps which led us from (31) to (36). we can also relate the charge condensate with the quark vev v. At strong coupling the difference is not small. the charge condensate vanishes at the AD point while the quark condensate remains finite. see Eq.

(48) it is clear that the absolute values of all three condensates are equal because the values of the three roots ui are on the circle in the u plane. 1 (53) Tr λ2 = λ+ λ− + λ3 λ3 . (48). Photino and gaugino condensates The gaugino condensate hTr λ2 i we found in the previous section can be viewed as a sum of the condensates for charged gauginos and the photino. Gorsky et al. (44) we have ζC = −i. ζM = −i. 3. see (48). see (26). condensates are still given by Eq. while the dyon condensate remains non-zero.210 A. which stands for charge. / Nuclear Physics B 584 (2000) 197–215 Now let us work out the dyon condensate. 2. More generally let us introduce the dyon field Di . M. (50) In fact one can fix the charge phase factor by imposing the condition that the charge condensate should approach the value 2mµ in the large m limit. For the monopole condensate at real values of m larger than the mAD Eq. but the charge and monopole phase factors can change. D . From Eq. In the limit m = 0 we should recover the Z3 -symmetry for the values of condensates. (52) 3.  (47) Di = (− det h)1/4 Q+ . i = 1. 1) dyon. Imposing the requirement of Z3 symmetry at m = 0 we can fix the unknown phase factors ζC and ζM below the AD point using the value (51) for dyon. monopole and (1. The arguments of the previous subsection which led us to the result (36) for monopole ei Di i condensate give for hD  1/4 ei Di = 2iζi µ u2i − 3 mΛ3 . This gives ζC = i. (36) gives ζM = 1. (48) D 1 4 where ui is the position of the ith point in the u plane and the ζi are phase factors. (49) while for the charge condensate from Eq. 2 In gauge invariant form the photino condensate can be associated with (54) (Tr W Φ)2 . 2 The dyon phase factor (51) does not change when we move through the AD point because the dyon condensate does not vanish at this point. Below the AD point. (51) At the particular AD point we have chosen the monopole and charge condensates vanish. For the dyon the phase factor is ζD = i. 2 Note that the quantum numbers of the “charge” and “monopole” are also transformed. see [35] .3.

In the case of BPS domain walls their tension is given by the central charge [32]. Indeed. A natural possibility to consider is a domain wall interpolating between colliding vacua. the monopole vacuum. it is very similar to the Witten index. Let us start with the Coulomb branch in the N = 2 theory. at the AD point we encounter the problem of not having a uniquely defined vacuum state. This does not prove the absence of a relevant order parameter so the quest can be continued. It is unlikely that the number of states with zero energy will change when we reach the AD point. . However. 4. Gorsky et al. It was shown in [36] that the perturbation linear in AD does not break N = 2 in the effective QED. the continuity of the chiral condensates we obtained above shows that they are no longer parameters which differentiate the two states once we reach the AD point. An immediate consequence of this observation is that the photino condensate continues to vanish. (30). where in the expansion of U as function of AD it is sufficient to retain only linear term. All gaugino condensates vanish in N = 2 for a simple reason: λ2 is not the lowest component in the corresponding N = 2 supermultiplet. When the perturbation µ U which breaks N = 2 is added to the superpotential the gaugino condensate is proportional to µ. The Argyres–Douglas point: how well is the theory defined As discussed in the Introduction. the term µ U in the superpotential does not break N = 2 SUSY in the effective QED. Consider.A. The corresponding effective superpotential is given by Eq. for example. However. / Nuclear Physics B 584 (2000) 197–215 211 We argue here that the photino condensate vanishes so that the gaugino condensate is solely due to the charged gluino. when the mass parameter m approaches its AD value mAD we deal with two vacuum states which can be distinguished by values of the chiral condensates.

.

(55) Tab = 2.

Weff (va ) − Weff (vb ).

The central charge here is expressed via values of exact superpotential (3) in corresponding vacua. In the conclusion of this section let us review briefly the brane construction of N = 1 vacua. b label the colliding vacua. Brane configurations responsible for N = 1 theories were suggested in [19. therefore its thickness is infinite at the AD point.. The brane . where a. T ∝ (m − mAD )3/2 when m → mAD .20] and a derivation of domain wall tensions from analysis of Riemann surfaces (which is similar to the calculation of the masses of BPS particles in N = 2 theories) can be found in [21]. If such a domain wall were observable at the AD point it could serve as a signal of two vacua. The continuity of the condensate v shows that the domain wall becomes tensionless at the AD point. Gauge theories are realized on brane worldvolumes. that this domain wall is not observable in continuum limit. The crucial point is that the wall is built out of massless fields. This makes it impossible to observe this tensionless wall in any physical experiment of a limited spatial scale. We argue. however.

212 A. The adjoint field shows up only as an extra (as compared with SQCD) nonrenormalizable term quartic in the fundamental fields. / Nuclear Physics B 584 (2000) 197–215 configuration for the N = 1 theory with one flavor and SU(2) gauge group is described by Riemann surface embedded into three dimensional complex space C 3 parametrized by three variables t. v and w. to an SQCD-like effective superpotential. Since the tension is defined by integration of the holomorphic 3-form around the resolved singularity the tensionless wall has a geometrical interpretation as the M5 brane wrapping this vanishing cycle. The nonperturbative part is given by the ADS superpotential generated at the one instanton level. Actually. We interpret this as a deconfinement of electric and magnetic charges . charge and dyon condensates departing from the Coulomb branch of the N = 2 theory. The most interesting phenomenon occurs at the AD point: when the monopole and charge vacua collide both the monopole and charge condensates vanish. The tension of the walls. Conclusions The approach of this work is similar to that used in SQCD. (56) w+ + w− + mµ = 0. Λ. The Argyres–Douglas points introduced in N = 2 theories are shown to exist in the N = 1 theory as well. (57) Let us consider the geometry of the brane configuration near the AD point. Results for chiral condensates of matter and gaugino fields are continued into the range of a small adjoint mass where we find a complete matching with the N = 2 Seiberg– Witten solution.38] that the AD point corresponds to a singular Calabi–Yau 3-manifold which is resolved if one adds particular perturbation. µw t = µ−2 w(w − w+ )(w − w− ). Although the bulk of our results for matter and gaugino condensates overlaps with what is known in the literature we think that our approach clarifies some aspects of duality in N = 1 theories. 5. m. The embedding is given by the following equations v+m= (w − w+ )(w − w− ) . 1 2 w− w+ = −(µΛ1 )3 . It was shown recently [37. Namely. This resulted in explicit relations between these condensates and those of the fundamental matter. in some range of parameters. the curves can be considered as fibered over the complex m plane and the AD singularities correspond to the appearance of vanishing cycles in the fiber in a manner quite similar to the Seiberg–Witten solution of N = 2 theories where vanishing cycles correspond to massless BPS particles. Gorsky et al. we integrate out the adjoint field which leads. 2 with free parameters µ. This superpotential describes the low energy theory at weak coupling where we have full theoretical control. which have the interpretation of the M5 branes wrapping three-cycle with the boundaries on the Riemann surface above can be calculated by integrating the holomorphic three-form over this cycle Z T = dv ∧ dw ∧ d(log t). We then analyze monopole.

Vanishing of condensates signals that the theory at this point becomes superconformal. The integrable system responsible for N = 2 SQCD was identified with the nonhomogenious XXX spin chain [39–41]. It would be very interesting to find a similar connection between spin chain Hamiltonians and superpotentials in the N = 1 SQCD. Hanany. and A. A. Shifman for helpful discussions. A. This has been confirmed by direct calculation in the pure N = 2 gauge theory [42] as well in the theory with a massive adjoint multiplet [43]. The most interesting problem in the generic situation involves Seiberg IR duality of the electric SU(Nc ) theory with Nf flavors and the magnetic SU(Nf − Nc ) theory. After perturbation to the N = 1 theory the Hamiltonian of the integrable system is expected to coincide with the superpotential of corresponding N = 1 theory. However. It was recognized that N = 2 theories are governed by finite-dimensional integrable systems. A. Argyres. thanks J. are thankful to ITP for hospitality and support from NSF by the grant PHY 94-07194. The work of A. Acknowledgements Authors are grateful to P. Ritz. is supported in part by DOE under the grant DE-FG02-94ER40823.G. A. the XXX spin chain would have additional symmetries similar to superconformal ones. and A. Nf case for a separate publication. is supported in part by the grant INTAS-97-0103. K. is supported in part by Russian Foundation for Basic Research under the grant 99-02-16576. Rudaz.V.A. and A.Y. In this paper we considered only the SU(2) theory with one flavor postponing the generic Nc .Y. and M. A. We hope to discuss these points in more details elsewhere. participated in the SUSY99 program organized by the Institute for Theoretical Physics at Santa Barbara. The approach of the present paper can be applied to this case as well. One more point to be clarified is the meaning of the AD point within approach based on integrability. A. Konishi. . thank the Theoretical Physics Institute at the University of Minnesota where this work was initiated for support. Part of this work was done when two of the authors.G. S. Ambjorn for hospitality at the Niels Bohr Institute where part of this work has been done.Y. / Nuclear Physics B 584 (2000) 197–215 213 at the AD point. A. and A. A. In our approach we see straightforwardly that the one-instanton generated ADS superpotential is the only nonperturbative input needed to fix all chiral condensates. Let us mention a relation to finite-dimensional integrable systems. In generic case of nondegenerate fundamental masses we expect deconfinement at the AD points.V. Marshakov.G. Since the quark mass is identified as a value of spin [39–41] one might expect that at particular spin values corresponding to the AD mass.V. since Higgs branches do not disappear at the AD points [25] we do not expect deconfinement to occur in this case [44]. Gorsky et al.Y. A degeneracy in fundamental masses leads to the appearance of Higgs branches. The general nature of this statement is seen from our derivation which relates polynomial coefficients in the Seiberg–Witten curve to the ADS superpotential..

B 391 (1997) 107. A.V. Phys. Konishi. Phys. B 396 (1997) 64. hep-th/9603051. Nucl. Khoze. ’t Hooft. B 516 (1998) 175. Phys. Nucl. Argyres. hep-th/9611097. Amsterdam–Singapore. Strassler. Novikov. [12] T. Nucl. Shenker. Rev. B 218 (1983) 493. H. hep-th/9707005. B 138 (1978) 1. [11] T. [31] V. [40] A. B 416 (1998) 319. M.R. A. JHEP 0001 (2000) 015. hep-th/9707120. Phys. Rabinovici. [34] G. Kitao. 1999. Taylor. hep-th/9402044. hep-th/9701009. Rev. V. Phys. hepth/9905015. Nucl. World Scientific. North Holland/World Scientific. V. [36] A. Nucl. hep-th/9902018. B 426 (1994) 19. Seiberg. Giveon. [39] A. Ritz. hep-th/0001036. Schwimmer. hep-th/9509130.A Shifman. Gorsky et al. B 379 (1996) 121. Veneziano. hep-th/9509066. Seiberg.). [30] N. [24] P. Phys. Nucl. hep-th/9707247. D 49 (1994) 6857. F. [9] S. Kutasov. B 471 (1996) 430. [2] V. p. [20] E. Phys. Phys. B 507 (1997) 658. hep-th/9706082. hep-th/9909073. Yang.A. Ferrara (Ed. hep-th/9503163. A. H. Supersymmetry. Phys. Nucl. . [5] N. Phys. Ito. Intriligator. B 461 (1996) 71. Phys. Nucl. K. Vol. Affleck. hep-th/9706145. Eguchi. S. Intriligator. Oh. Terashima. Morozov. Intriligator. Yung. Gukov. A. Henningson. Phys. [10] S. Seiberg. B 399 (1997) 75. 1 (1998) 1. M. N. hep-th/9612128. [28] G. E. Shifman (Ed. [35] A. Forge. Novikov. K. Gorsky. Phys. Nucl. Yankielowicz. Lett. B 430 (1994) 485. E. Nucl. Phys. Yang. B 511 (1998) 264. M. Seiberg.I. A.). Lett. N. Hori. hep-th/9510222. hep-th/9911112. Vainshtein. [21] S. B 459 (1996) 160. 485. Shifman.R. Nucl. 606. hep-th/9407106. S. E. 2. M. Zakharov. E. JHEP 0002 (2000) 004. Phys. S. Seiberg. [26] T. Lett. Davies. [13] K. [29] A. Dvali. Nucl. Theor. [32] G. Math. Giveon.A.A. Shapere. Bilal. M. Phys. B 336 (1994) 409. A. N. B 459 (1996) 455. hepth/9407087. B 137 (1984) 187. Y. Vainshtein. Witten. N. Intriligator. Phys.P. S. Phys.A. Nucl. Lett. [33] G. Nucl. B 447 (1995) 271. Phys. B 380 (1996) 75. Vafa. Nucl. Nucl. Phys. Phys.M. [18] G. [4] K. Yankielowicz. H. R. Rabinovici. A. Phys. Witten. Witten. Vainshtein. Nucl. B 513 (1998) 87. hep-th/9403198. hep-th/9603002. hepth/9603140.R. B 559 (1999) 123. Terashima.I. Phys. [25] P. Leigh. B 229 (1983) 407. BC 45 (1996) 1. Oz. [7] K. Phys. Adv. S. Lett. Hanany. Phys. Lett. [22] K. Nucl. Phys. Nucl. B 113 (1982) 231. M. Hori. hep-th/9408155. [3] M. Phys. Ooguri. hep-th/9505062. [15] N.I. hep-th/9906053. Carlino. Zaffaroni. Singapore. Proc. Sin.C. M. BPS structure of Argyres–Douglas superconformal theories. in: M. Konishi. Hollowood. Veneviano. Nucl. Seiberg. hep-th/9706109. Nucl. hep-th/9607151. Phys. Phys. Nucl. V. T. [27] M. Phys. Lett. B 296 (1988) 445. Gorsky.H. Ferrari. Mattis. K. S. [8] S. C. Seiberg. Shifman. Khoze. Phys. B 517 (1998) 409. B 402 (1997) 290. S. T.M. Phys. Forge. Phys. A. Phys. Lett. Nucl. Argyres. B 353 (1995) 79. Shifman. Douglas. M. hep-th/9408099. Mironov. Mironov. B 431 (1994) 484. hepth/9910182. Plesser. Vainshtein. [38] A. A. Phys. A. Zakharov.C. 1. K. B 431 (1994) 551. [16] N. [19] K. Nucl. Nam. Lett. p.I. [37] A. A. N. 1987. A. M. [23] K. Elitzur. K. Dine. B 448 (1995) 93. S. A. V. [17] A. Gustavsson. Davies.214 A. Terao. Lett. Suppl. Elitzur. Nucl. A. [6] D. Phys.V. Kitao. Phys. / Nuclear Physics B 584 (2000) 197–215 References [1] I. Murayama. B 463 (1999) 551. N. Lett. Seiberg. B 260 (1985) 157. Phys. A. D 50 (1994) 1092. Intriligator. Marshakov. hep-th/9511154. [14] M. hep-th/9504080. B 566 (2000) 311. Vol. Phys. Yang. Lett. Phys.G. B 153 (1979) 141. hep-th/9707244. Vainshtein. hep-th/9604096. ITEP Lectures on Particle Physics and Field Theory. B 485 (1997) 38. S. Reprinted in: S. Shifman.J. Douglas.

Katz. Gorsky et al. Nucl. N. A. Dorey. B 497 (1997) 196. C. B 550 (1999) 513. Phys. Phys. JHEP 9907 (1999) 021. B 562 (1999) 191. Nucl. hep-th/9902030. Gorsky. A.A. / Nuclear Physics B 584 (2000) 197–215 [41] [42] [43] [44] A. Mironov. 215 . S. hep-th/9611090. Yung. Vafa. hep-th/9906011. Nucl. Phys. hep-th/9906243.

00173 Roma..Nuclear Physics B 584 (2000) 216–232 www. Stefano Kovacs b .elsevier.15.Gh.Bt.Jv.60. Renormalization 1. On the one hand. 1 On leave of absence from Institute for Nuclear Research and Nuclear Energy. 11.T.38.N.10. 11.nl/locate/npe Anomalous dimensions in N = 4 SYM theory at order g 4 Massimo Bianchi a .B.  2000 Elsevier Science B. accepted 19 May 2000 Abstract We compute four-point correlation functions of scalar composite operators in the N = 4 supercurrent multiplet at order g 4 using the N = 1 superfield formalism. 11. I.A.Hi.Hf. Wilberforce Road.10.15. Università di Roma “Tor Vergata”. 12. in particular N = 4 supersymmetric Yang–Mills theory (SYM) in D = 4. insights gained in such finite theories may prove to be useful in theories with (partial) supersymmetry breaking. Introduction and summary of the results The conjectured AdS/CFT correspondence [1–4] has renewed the interest in conformal field theories (CFTs).25. one may ask whether it is possible to recognize some systematic pattern in the perturbative N = 4 field-theoretic results. and has raised the issue of finding to what extent high order perturbative computations are feasible in the weak coupling regime. Superdiagrams. We confirm the interpretation of short-distance logarithmic behaviours in terms of anomalous dimensions of unprotected operators exchanged in the intermediate channels and we determine the two-loop contribution to the anomalous dimension of the N = 4 Konishi supermultiplet. The interest is twofold. given the explicit analytic expressions of certain four-point amplitudes in the AdS context. Stanev a.1 a Dipartimento di Fisica. All rights reserved. Supersymmetry. Italy b D. 11. Sofia.V. 1. On the other hand. CB3 0WA.M. UK Received 30 March 2000. Cambridge. – Sezione di Roma “Tor Vergata”.V. 12. Bulgarian Academy of Sciences.6 .F. Via della Ricerca Scientifica. Superconformal invariance. PACS: 11. BG-1784. Yassen S. Giancarlo Rossi a .N. AdS/CFT correspondence.P.Bx Keywords: Anomalous dimension. All rights reserved. 0550-3213/00/$ – see front matter  2000 Elsevier Science B.3 2 1 3 ( 0 0 ) 0 0 3 1 2 . PII: S 0 5 5 0 . University of Cambridge. Bulgaria.

we will write down the action. such as the one presented here. no such non-renormalisation theorems are believed to hold for higher-point functions. In order to illustrate our point we will concentrate in most of our presentation on the simplest of the six independent four-point functions of chiral primary operators [12–14] (see below). A detailed comparison of the two classes of results is problematic because of the different regimes in the (g 2 . We confirm the interpretation of the short-distance logarithmic behaviour in terms of anomalous dimensions of unprotected operators exchanged in intermediate channels and we extract the two-loop contribution to the anomalous dimension of the N = 4 Konishi supermultiplet. Explicit computations have confirmed the validity of these theorems in the context of the AdS/CFT correspondence [5–7]. we find the same result for the four-point function on which we focus our attention. although we adopt a different approach. Their approach is based on the less familiar N = 2 harmonic superspace formalism. It is remarkable that.13. Identities that have been proved to hold both in perturbation theory in [12. For protected operators non-renormalisation theorems hold that prevent quantum corrections to the lowest order contribution. A recent interesting paper by Eden. The plan of the paper is as follows. contrary to expectations. Nevertheless. N ) parameter space that are explored by the two approaches. Schubert and Sokatchev has reported on similar computations at order g 4 from a different vantage point and confirmed the validity of the above identities [29]. Similar analyses have been performed in the AdS context in the supergravity limit [25]. 20–24]. up to an overall constant that was not fixed in [29]. have been recently completed in [26] in the supergravity approximation with the help of the results of [27]. All such computations show logarithmic behaviours at short-distances that allow an interpretation in terms of anomalous dimensions of unprotected operators in long supermultiplets exchanged in intermediate channels [14]. We would like to stress that. explicit AdS computations in the scalar sector that are amenable to a perturbative analysis in the field-theoretic weak coupling regime.M. In this paper. dimension-two gauge-invariant scalar composites in the N = 4 supercurrent multiplet. we compute at order g 4 four-point correlation functions of the lowest dimension scalar composite operators in the N = 4 supercurrent multiplet using the N = 1 superfield formalism. Except for extremal [8– 10] and sub-extremal correlators [11]. for the sake of completeness. / Nuclear Physics B 584 (2000) 216–232 217 In CFT the dependence of two.28] and at the one-instanton level in [14] should be enough to determine all four-point functions from the one we consider. Bianchi et al. i. the N = 1 superfield approach we have pursued is not prohibitively complicated and it is of wider applicability. propagators and vertices in the N = 1 superfield formalism and . In particular we have in mind applications to some interesting finite N = 1 superconformal theories that are promising candidates for realistic theories after soft supersymmetry breaking. Motivated by the presence of string corrections to the AdS effective action [15–19]. A particularly interesting class of higher-point functions are the four-point functions of the lowest chiral primary operators.and three-point functions of scalar primary operators on the external insertion points is completely fixed by conformal symmetry.e. instanton computations have been carried out in [14. Perturbative computations involving such operators have been performed at order g 2 in [12–14]. In Section 2.

4). Φ I and ΦI† transform in the representations 3 and 3 of the global SU(3) “flavour” Q-symmetry. ϕ i .13] and [29] dub diagrams of this order as (` + 1)-loop calculations. .218 M.13. the spinors of the chiral multiplets ψ I have charge −1/2 and the three complex scalars φ I have charge +1. namely   1 1 φI† = √ ϕ I − iϕ I +3 . all in the adjoint representation of the gauge group G. 3. 2. [12. . form the vector multiplet. V . 2. .29] or to the more familiar N = 1 formalism pursued in [14]. In this formulation only an SU(3) × U (1) subgroup of the original SU(4) R-symmetry group is manifest. in the N = 4 Konishi supermultiplet. In Section 3 we draw the relevant Feynman superdiagrams and sketch the computation for the simplest possible four-point function. Aµ . Since no off-shell formulation is available that manifestly preserves N = 4 supersymmetry. 2. λ = ψ 4 . . are the spinors of the chiral multiplets. in order to compute quantum corrections one has to resort either to the N = 2 harmonic superspace approach pursued in [12. are combined into three complex fields. (1) φ I = √ ϕ I + iϕ I +3 . Three of the Weyl fermions. 4g 2 Notice that we call “`-loops” calculations that are of order g 2` . The authors of Refs. Aµ . N = 4 SYM in N = 1 superspace The field content of N = 4 SYM [30. In the N = 1 formalism the fundamental fields can be arranged into a vector superfield. Φ I (I = 1. (2) − √ 3! 2 where Wα is the chiral superfield-strength of V  1 (3) Wα = − D 2 e−2gV Dα e2gV . Under the axial U (1) R-symmetry the vector Aµ is neutral. Bianchi et al. while V is a singlet. / Nuclear Physics B 584 (2000) 216–232 identify the six independent four-point functions of chiral primary operators. 6). respectively. Φ K − d2 θ¯ εI J K ΦI† ΦJ† . and three chiral superfields. ψ A (A = 1. K1 . We confirm the one-loop results of [14] and extract the two-loop contribution 2 to the anomalous dimension of K1 . together with the vector. The fourth spinor. the gaugino λ has charge +3/2. four Weyl spinors. In Section 4 we discuss the final result and interpret the dominant logarithmic behaviours at short distance in terms of the anomalous dimension of the lowest dimensional operator. ψ I . 2 2 that are the lowest components of the superfields Φ I and ΦI† . 3). The six real scalars. The (Euclidean) action in the N = 1 superfield formulation reads Z  Z  Z 1 d2 θ W α Wα + h. + d2 θ d2 θ¯ e−gV ΦI† egV Φ I S = −2 tr d4 x 16 Z  Z    g † d2 θ εI J K Φ I Φ J .31] comprises a vector. and six real scalars. ΦK . 2. ϕ i (i = 1. .c. We finally comment on possible extensions of our results to finite N = 1 supersymmetric theories that arise after soft breaking of N = 4 supersymmetry.

C I J = tr(Φ I Φ J ). x4 ) = C 11 (x1 )C11 where the gauge-invariant composite operators  CI†J = tr φI† φJ† C I J = tr(φ I φ J ). The superfield propagators in the conventions of Eq. In the following we will carry on the calculations using N = 1 formalism in coordinate superspace. θ¯i )Vb (xj . as it makes all corrections to the propagators of the fundamental superfields vanishing at order g 2 [32. Bianchi et al. θ¯j ) = 4π 2 xij (7) δab δ(θij )δ(θ¯ij ) . As usual we decide to take Z Z   1 (5) tr d4 x d2 θ d2 θ¯ (D 2 V )( D 2 V ) . θi . ΦI†a (xi . θj . (10) G(H ) (x1 . x2 . Sgf . θj . Va (xi .35]. Sgf = 16α where α is a gauge parameter. The choice α = 1 greatly simplifies all the computations. . Notice that a gauge fixing term. 2 a.M. Expanding the exponentials e±gV in (2) gives  Z g2 a a 4 2 2¯ S = d x d θ d θ V a Va − ΦIa† ΦaI − igfabc Φ † I V b Φ I c + fabe fecd Φ † I V b V c Φ I d 2    ig † † abc I J K ¯ IJK † − √ f ΦaI ΦbJ ΦcK δ(θ ) + · · · . dim(G). θ¯j ) = − 2 8π xij2 (8) where xij = xi − xj . (6) are δI J δab (ξ +ξ −2ξ )·∂ 1 e ii jj ji j 2 . (4) where T a are the generators in the fundamental representation of the gauge group. (11) are the lowest components of the (anti-)chiral superfields  CI†J = tr ΦI† ΦJ† . . / Nuclear Physics B 584 (2000) 216–232 219 The trace over the colour indices is defined by 1 tr(T a T b ) = δ ab . b = 1. θij = θi − θj . has to be added to the classical action (2). x3 . θ¯i )ΦbJ (xj . We will not display here ghost terms since they do not contribute to the Green functions that we will consider at the order we work. µ µ (9) In the next section we will describe the calculation of the order g 4 perturbative correction to the four-point correlation function † † (x2 )C 22(x3 )C22 (x4 ) . (12) . ξij = θiα σα α˙ θ¯jα˙ . θi . that is more suitable for the study of CFTs. (6) εI J K Φa Φb Φc δ(θ ) − ε 3! 2 where fabc are the structure constants of the gauge group and we have displayed only the terms that are relevant for our subsequent computations. . .

x2 . The Q20 belong to the real representation 20 of SU(4) and are defined as   δ ij ij ϕk ϕ k . (16) where s is one of the two independent conformally invariant cross-ratios r= 2 x2 x12 34 2 x2 x13 24 . the lowest scalar components of ij the N = 4 current supermultiplet. x4 ) = C 11 (x1 )C 11(x2 )C11 We stress that the above five linearly independent four-point functions have been shown to be functionally related to (10) both in perturbation theory [12. x2 . .13. G(2)(x1 . x2 . Bianchi et al. under SU(4) → SU(3) × U (1). x3 . x3 . z3 . Superdiagrams and calculations The four-point function G(H ) (x1 . x4 ) = C 11 (x1 )C22 † † (x3 )C22 (x4 ) . x4 ) = C 11 (x1 )C 22(x2 )C11 † † (x2 )C11 (x3 )C 11 (x4 ) . x3 . (13) Q20 = tr ϕ i ϕ j − 6 ij The most general four-point function of the Q20 G(Q) (x1 . z2 . x4 ) = C 11 (x1 )C11 † † (x3 )C11 (x4 ) . x4 ) = Qi1 j1 (x1 )Qi2 j2 (x2 )Qi3 j3 (x3 )Qi4 j4 (x4 ) (14) can be expressed as a linear combination of G(H ) . defined in Eq. (15) G(4)(x1 . s= 2 x2 x14 23 2 x2 x13 24 . x2 . x3 . x4 ). and the following five independent four-point functions † † (x2 )C 11 (x3 )C11 (x4 ) . x2 . For instance. x2 . G(1)(x1 . 3.220 M. G(V ) (x1 . x3 .29] and nonperturbatively at the one-instanton level [14]. x2 . z4 ) = C 11 (z1 )C11 viz. G(3)(x1 . x2 . x3 . x4 ) = sG(V ) (x1 . x4 ) = C 11 (x1 )C11 † † (x2 )C 22 (x3 )C11 (x4 ) . one finds G(H ) (x1 . (10). / Nuclear Physics B 584 (2000) 216–232 ij In turn C I J and CI†J appear in the decomposition of Q20 . x3 . x2 . x3 . (17) that can be constructed out of four points. x3 . x4 ) defined in (10) is the lowest component of the supercorrelation function † † (z2 )C 22 (z3 )C22 (z4 ) (18) Γ (H ) (z1 .28.

z3 . x2 . z2 . z4 ). G(H ) (x1 . x4 ) = Γ (H ) (z1 . x3 .

The potential corrections to the propagators at order g 4 would µ . θiα . i i (19) where zi = (xi . θ¯i α˙ ) as usual.θ =θ¯ =0 . The N = 1 superfield approach greatly simplifies the calculations. The choice of external flavours guarantees that there are no connected supergraphs at tree level.

21 superdiagrams with one internal vector line and 10 superdiagrams with two internal vector lines. This explains why we have not explicitly displayed the corresponding couplings in (6). . 2. Bianchi et al. only contribute to disconnected supergraphs 3 since the choice α = 1 in the gauge-fixing term makes all propagator corrections vanish at order g 2 [32–36].M. due to combinatorial factors and colour and flavour contractions. 1. We have only displayed superdiagrams that do not vanish by colour contractions. The numbering of the chiral lines refers to the SU(3) flavours. Diagrams with two vector lines. There are 4 superdiagrams with only chiral lines. 2 and 3. is greatly simplified if fake different coupling constants are introduced 3 Order g 4 corrections to two-point functions of operators belonging to ultra-short supermultiplets vanish [37]. / Nuclear Physics B 584 (2000) 216–232 221 Fig. Notice that none of them involves cubic or quartic pure vector vertices. 1. Fig. The computation of the overall weights. Diagrams with only chiral lines. The connected superdiagrams contributing to (18) are reported in Figs.

. 3. / Nuclear Physics B 584 (2000) 216–232 Fig.222 M. Diagrams with one vector line. Bianchi et al.

where ∆ is the conformal dimension of the operator inserted at x1 . Exploiting the conformal invariance of the full correlation function we send one of the external points (say x1 ) to infinity after multiplying the correlator by (x12 )∆ . x 2 and x 2 .e. This increases the power of x1 in the denominator making it larger than 4 with the consequence that terms generated in this way vanish in the limit. (22) Note that C10 is zero due to Grassmann integration over the θ ’s of the interaction vertices. This amounts to put to zero the θ variables of the external insertions. Only at the very end one imposes the relations among the couplings that guarantee N = 4 supersymmetry. The final result is a complicated combination of double. In general. i. The ∞. On the other hand. The generalization to an arbitrary gauge group amounts to substituting g 2 N with g 2 C2 (A) and N 2 − 1 with dim(G). . As a result one finds 4 for the superdiagrams with only chiral lines g4 2 2 N (N − 1)(2A1 + A2 + A3 + A4 ). The general strategy for computing the resulting integrals is the following. (21) and for the superdiagrams with two vector lines Ctot = g4 2 2 N (N − 1) 2 × (2C1 + 2C2 − 2C3 − 2C4 + C5 − C6 + 2C7 + 2C8 − C9 ). substitutions x23 23 14 24 24 13 34 34 12 Taking the limit x1 → ∞ considerably simplifies the computation for two reasons. triple and quadruple integrals of convolutions of massless scalar propagators and derivatives thereof. one has to carefully subtract divergent terms that behave like powers of log(x12 ) as x1 → 2 . 8 for those with one vector line Atot = (20) g4 2 2 N (N − 1)(2B1 + 2B2 + B3 + B4 − B5 − 2B6 + 2B7 + 2B8 + B9 4 + B10 − B11 − 2B12 + B13 + B14 + B15 + B16 Btot = − + B17 − B18 − B19 − B20 − B21 ). This trick also simplifies the θ integrals. x 2 → x 2 x 2 and x 2 → x 2 x 2 . Note that this simple prescription only works for operators (like C) with protected conformal dimension (∆ = 2 in the case at hand). before taking the limit. / Nuclear Physics B 584 (2000) 216–232 223 for each interaction term in the action. For operators that acquire anomalous dimensions. C2 (A) being the quadratic Casimir of the adjoint representation. In this limit the final answer will depend on the three variables x23 24 34 complete x1 dependence is dictated by conformal invariance and can be recovered after the 2 → x 2 x 2 . On the one hand it lowers the number of propagators to be integrated.. independent of the coupling constant g 2 . Bianchi et al. one can integrate by parts some of the derivatives in the chiral propagators and make them act on propagators involving the point x1 . As a result one is left with only double integrals except for the diagrams B21 and C6 where the remaining integral 4 In what follows we will only display group theory factors relevant for the case G = SU(N ). the integration over the θ variables of the interaction vertices is greatly simplified by our choice of computing the lowest component of the supercorrelation function (18).M.

g(2. An important simplification arises by observing that in the limit x1 → ∞ certain linear combinations of diagrams contributing to (21) and (22) actually vanish. For the purpose of the present investigation. s) = √ ln(r) ln(s) − ln p 2     2 2 − 2Li2 . B9 = 0. k) ≡ f (i. (28) − 2Li2 √ √ 1+r −s+ p 1−r +s + p with Li2 (z) = ∞ n X z n=1 n2 and p = 1 + r 2 + s 2 − 2r − 2s − 2rs.13] 1 (31) j f (i. B3 − 2B6 + B13 + B17 − B19 = 0. The explicit expression of the function f (i. 2B2 − B5 − B20 = 0. .224 M. f (i. A useful expression for the function g is [14]  2 2  x x24 π2 . Precisely one finds B1 = 0. Bianchi et al. thus diagrams that were related by rotations (or reflections) to one another must be separately computed. 2 2 x2 x2 xi5 xj25 xi6 k6 56 (24) where in the notation used in [12] f (i. k) has been obtained in [38]  2 2  π 4 (2) xij xik . j . k). however. we only need the following identities [12. s) involves polylogarithms of up to fourth order. / Nuclear Physics B 584 (2000) 216–232 is (naively) a triple one. k) = 2 2 xi5 xj25 xk5 and Z f (i. j. C1 − C3 = 0. j. k) = dx5 dx6 . B15 − B18 = 0. The whole result can be expressed in terms of only two functions Z dx5 (23) g(i. 2C4 − C5 + C9 = 0. xij . j. 3. j. k) = −4π 2 2 g(i. k) = 2 Φ xj k xj2k xj2k (29) (30) where Φ (2) (r. i. . j. (27) β1 β2 + rβ0 β1 + sβ0 β2 The result of the integration in (27) can be expressed as a combination of logarithms and dilogarithms as follows:    √  r +s−1− p 2 1 B(r. A drawback of our procedure is that the limit x1 → ∞ breaks many of the permutation symmetries of the integrand. As already observed C10 = 0 due to the θ integration even before taking the limit. j. 4) = 2 B 34 2 x2 x23 x23 23 (25) (26) where B(r. j. C2 + C8 = 0. s) = dβ0 dβ1 dβ2 δ(1 − β0 − β1 − β2 ) 1 . k). s) is a box-type integral given by Z B(r.

2. Bianchi et al. 3) + x24 + x34 + x23 [g(2. x 2 → x 2 x 2 and x 2 → x 2 x 2 . k). 2 8 2 2(4π ) 4x34 (36) (37) Summing the above three contributions yields g 4 N 2 (N 2 − 1) 2 x1 →∞ 4(4π 2 )8 x34    1 2 2 2 × f (2. 3. at short distances (i. 4) − 4f (4. 4)]2 . the actual x1 dependence is recovered by performing in the right 2 → x 2 x 2 . 4) − 2f (3. 3. 1. j. j. 5) g(2. Logarithms and anomalous dimensions As in previous computations at order g 2 [12–14] and at the one-instanton level [14]. (33) can in turn be proved using the identity  2   2 2 2 3 g(2. 4) + f (4. g(2. 2. 3. Btot = − Ctot = 2 g 4 N 2 (N 2 − 1) 1  2  x24 g(2. 4) . 2. 3. 4) . 3 2 2 x25 x34 Eq. (35) Atot = 2 8(4π 2 )8 x34 g 4 N 2 (N 2 − 1) 1 2 4(4π 2)8 4x34  2 2  2 2 × g(2. 3. B8 and B11 .. B7 . (26). 4) + f (2.M. 3. in the limit in which pairs of points are taken to coincide) one finds linear and quadratic logarithms that are not incompatible with the finiteness of N = 4 SYM in the superconformal phase. x2 . 4) . k) = −4π 2 xj2k 2 xij2 xik g(i. 2.e. 4) x24 + x34 − x23 − 4f (3. Indeed. The total contributions of each of the three classes of diagrams of Figs. 2. 4) = x34 2 x24 (34) which can be verified by resorting to the explicit expression of g(2. x4 ) ≡ lim x14 G(H ) (x1 . 3) . . x3 . the function (38) shows the expected short-distance logarithmic behaviour. 3. 4) . 4) + g(2. 2 and 3 is     2 g 4 N 2 (N 2 − 1) 2  f (3. (38) 4 G(x2 . 3. 4. / Nuclear Physics B 584 (2000) 216–232 i f (i. 225 (32) The latter identity has to be used in both directions to simplify the integrals appearing in the diagrams A2 . 4) given by Eq. The hand side of (38) the substitutions x23 23 14 24 24 13 34 34 12 final result is in perfect agreement with the expression of the corresponding four-point function of hypermultiplet bilinears computed by means of N = 2 harmonic superspace techniques in [29]. 2. x3 . 4) + f (3. 3. The triple integrals that appear in the computation of the diagrams B21 and C6 can be simplified with the help of the relation Z 2  2 2 dx5  2 x24 = −4π (33) g(3. x4 ) = As already explained. 3. 4.

e. ∂w ) (42) QA (x)QB (y)OK (w) . +g 2 1 (41) where after renormalisation we have set A∆ = a∆ µ−2γ . From these formulae one can extract the corrections to both the OPE coefficients and the anomalous dimensions of the operators OK . In (0) general the operators OK may have anomalous dimensions. although the exact expression (39) is conformally invariant. / Nuclear Physics B 584 (2000) 216–232 To illustrate this point in a simple fashion we will follow closely [14] and consider as an example the two-point function of a primary operator of scale dimension ∆ † A∆ . QC . (0) (0) where ∆K is the tree-level scale dimension. Descendants are implicitly taken into account by the presence of derivatives in the Wilson coefficients. (42) is determined by conformal invariance to be of the form QA (x)QB (y)OK (z) = (x − y)∆A +∆B −∆K (x CABK (g 2 ) . Bianchi et al. the operator under consideration has an anomalous dimension. C’s. QB . at each order in g 2 Eq. The three-point function in the right hand side of Eq.226 M. QD are protected operators. (43). − z)∆A −∆B +∆K (y − z)∆B −∆A +∆K (43) Eq. i. (40) g2 yields The perturbative expansion of (39) in powers of    † a∆ 1 − g 2 γ1 log µ2 (x − y)2 O∆ (x)O∆ (y) = (0) (x − y)2∆     2  2   4 1 2 2 2 2 γ log µ (x − y) − γ2 log µ (x − y) +··· . The same procedure . Similar considerations apply to generic n-point Green functions as well. like the two-point function (41). = ∆ +∆ −∆ C D K (z − w) K where K runs over a complete set of primary operators. Now suppose that ∆ = ∆ + γ ..e. To simplify formulae we assume that QA .. Assuming the validity of the OPE. i. (39) O∆ (x)O∆ (y) = (x − y)2∆ where A∆ is an overall normalisation constant possibly depending on the subtraction (0) scale µ. γK . they have no anomalous dimensions. with ηI J K the perturbative correction to the OPE coefficients. An expansion like (42) is valid in the other two channels as well. (41) contains powers of logarithms that seem to even violate scale invariance. so that ∆K = ∆K + γK . can be expanded in power series in g 2 giving rise to logarithmic terms. Similarly CI J K = CI J K + ηI J K . In perturbation theory γ = γ (g 2 ) is expected to be small and to admit an expansion in the coupling constant g 2 γ (g) = γ1 g 2 + γ2 g 4 + · · · . a four-point function can be expanded in the schannel in the form QA (x)QB (y)QC (z)QD (w) X CCD K (z − w. Thus.

k). j. x4 ) −→ 4 x2 16(4π 2 )6 x23 x23 x4 →x3 34 (46) We now compare Eq. 2 6 4 2 x23 x4 →x3 (4π ) x23 x34 (47) where the coefficients a0 and a1 are given by 1 a1 = −Nπ 2 . 4) 2 G(x2 . k) makes it difficult to directly analyse the short distance behaviour of the function (38). (31). The latter yields  2    x34 g 4 (N 2 − 1) 2 2 2 G(x2 . x3 . x3 . 3 They represent the tree-level contribution of the Konishi scalar K1 to the s-channel expansion of the function G(H ) and its finite one-loop correction. that is clearly not renormalized. j. one obtains  4 g 4 N 2 (N 2 − 1) X ∂j g(2. 4) · ∂j g(2. Thus we find it more convenient to compute 2 G(x2 . which can be expressed in terms of only the much simpler function g(i. Using Eqs. x3 . x4 ). The coefficients a0 and a1 are determined by the one-loop analysis performed in [14] that gives !   2 x34 g 2 (N 2 − 1) 1 γ1 log + a a G(x2 . 3. 2 G(x2 . Only operators in the singlet of the manifest SU(3) × U (1) may be exchanged [14]. Let us analyze the function G(H ) in the limit x4 → x3 that exposes the s-channel. x4 ) = 8(4π 2 )8 j =2    1 1 1 (45) − 3(4π 2 ) 2 2 + 2 2 + 2 2 g(2. 4) . x4 ) −→ a0 γ1 log 2 + a0 (γ1 + 2γ2) + 2a1 γ1 . the leading contribution comes from the operator K1 = 1 : tr(ϕ i ϕ i ) : 3 (44) (0) that has naive dimension ∆ = 2 and is the lowest component of the long Konishi supermultiplet. Barring the identity operator.and four-point functions as well as to their limits for x1 → ∞. x3 .M. x4 ). (48) a0 = (4π 2 )2 . the relative complexity of the explicit expression for f (i. x4 )|1-loop −→ 0 1 2 x2 2 2 (4π 2 )6 x23 x23 x4 →x3 34  2    x34 1 g 2 (N 2 − 1) 1 2 log π N − 1 . 3. / Nuclear Physics B 584 (2000) 216–232 227 applies to derivatives of the three. (49) = 2 x2 2 2 (4π 2 )6 x23 x23 34 . x3 . x23x34 x24 x34 x23x24 The leading behaviour of (45) in the limit x4 → x3 is given by   2   x34 g 4 N 2 (N 2 − 1) 1 3 log 2 − 5 . 3. (46) with the result of the OPE analysis of the order g 4 contribution to the function 2 G(x2 . In the case under consideration. x3 . Bianchi et al. (32) and (34).

Bianchi et al. one has to resort either to the N = 1 superspace approach pursued here or to the less familiar N = 2 harmonic superspace approach pursued in [29]. At the quantum level. We find it convenient to analyse the four-point correlator G(V ) . The constant term corresponds to a combination of the one. A careful OPE analysis. First of all the very fact that it was possible to compute in closed form a four-point function of protected composite operators at order g 4 shows that N = 4 SYM is both non-trivial and calculable.228 M. No off-shell approach is known that preserves N = 4 supersymmetry. based on the bonus symmetry proposed in [55]. theories that can be handled with N = 1 superfield techniques. The two possible operators of naive dimension four in the 84 mix at one-loop [14]. that led to drastic simplifications in the computations reported in [29]. (47) shows that the coefficient of the logarithmic term is related to the square of the one-loop anomalous dimension of K1 . confirms the non-renormalisation of the 105 operators. For the g 4 contribution to the anomalous dimension of the Konishi supermultiplet one obtains 3g 4 N 2 . The lowest dimensional operators that can be exchanged in both channels belong to the 105 and to the 84 representations of the SU(4) R-symmetry.13]. We thus conclude that the operator D 4 identified in [14] as the combination orthogonal to K84 .43. Comments Let us briefly comment on the bearing of the results of this paper. Comparison of Eq.44]. The introduction of supersymmetric mass-terms gives rise to interesting. implies that this operator b84 . One-loop [12.and double-trace operators of dimension four in the 105 are expected to be protected [7. as it is obtained from the above functional relation. saturates the logarithmic behaviour in this channel. in some cases confining. (46) with Eq. K84 . defined in (15). (38) under the exchange of x2 and x4 . Single.20] and two-loop [29] computations give sG(V ) = G(H ) . (50) 16(4π 2)2 The analysis of the other two non-singlet channels (x24 → 0 and x23 → 0) is more subtle. This is in agreement with the fact that it belongs to a supermultiplet satisfying a generalized shortening condition [45–53].and two-loop anomalous dimensions and the known one-loop correction to the OPE coefficient. / Nuclear Physics B 584 (2000) 216–232 We stress that the quadratic logarithmic term in G(H ) contributes a linear logarithmic term to 2 G(H ) in Eq. γ2-loop = γ2 g 4 = − 5. (47). is a superconformal descendant of K1 and as such has the same anomalous dimension. combined with the symmetry of the factor in brackets in the right hand side of Eq. One of them. is protected also at order g . Alternatively one might consider other finite N = 1 gauge theories some of which are conjectured to be . one-instanton [14. and agrees with 3N the known value γ1 = 16π 2 [39–42]. The OPE analysis of the g 4 contribution to G(V ) . In order to disentangle the two SU(4) representations it is necessary to consider another four-point function. The coincidence of all known results in the two approaches gives independent support to some N = 2 harmonic superspace identities [54].

In fact protected operators belonging to supermultiplets that satisfy generalized shortening conditions [45–53] are typically mixtures of single. Phys. but at the same time yields a negative two-loop contribution to the anomalous dimension.e. / Nuclear Physics B 584 (2000) 216–232 229 dual to type IIB superstring on AdS5 × S 5 /Γ . hep-th/9807098. We have shown and confirmed in the present paper that their mixing with single trace operators is not suppressed in the large N limit at finite ’t Hooft coupling. Gunaydin and J. Adv. Math.67]. One would like to clarify their rôle in the operator algebra in relation to the “string exclusion principle" that is expected to drastically reduce the spectrum of AdS excitations at finite “radius". Minwalla. Gubser. Acknowledgements We would like to thank E. This result certainly requires further investigation and some cross-checks [65]. D. and S.N. Maldacena.S. W.Z.M. Witten. These gauge theories govern the dynamics of the light degrees of freedom of stacks of coincident D3-branes at special orbifold singularities [57– 61]. By OPE one would then extract anomalous dimensions and couplings of unprotected operators such as those belonging to Konishi-like long supermultiplets. hep-th/9711200. The OPE analysis of the two-loop computations confirm the oneloop result.S.F.M. Phys. Skiba. Another related issue is the role of multi-trace operators that are expected to be dual to multiparticle states in the AdS description. Adv. that is dual to the large N and strong ’t Hooft coupling limit. Three-point functions of chiral operators in D = 4 N = 4 SYM at large N . [5] S. B 428 (1998) 105. 323 (2000) 183. References [1] J. Gauge theory correlators from non-critical string theory. M. Large N field theories. The fact that the one-loop anomalous dimension of the Konishi multiplet has been known for some time [39–42] to be positive was somewhat reassuring in this respect. 2 (1998) 253. where Γ is a discrete subgroup of an SU(3) subgroup of the SU(4) R-symmetry. Field theory tests for correlators in the AdS/CFT correspondence. N. Theor. The large N limit of superconformal field theories and supergravity. In all such cases the N = 1 formalism pursued in this paper might allow one to compute correlation functions of protected composite operators. Ferrara. [2] S. M. Phys. Math.. Phys.N. [6] E. Section at Università di Roma “Tor Vergata” for hospitality and financial support.S. . Aharony. S. Morales for useful discussions. Lett.and multi-trace operators. S. Phys. D’Hoker. Adv.S. J. would like to thank the Physics Department and I.64]. 2 (1998) 697. I. Theor. This is the case of D3-branes at generalized conifold singularities [62]. Lee. 2 (1998) 231. Ooguri. whose supergravity dual replaces S 5 with less trivial Einstein spaces [63. hep-th/9905111. Gubser. D 59 (1999) 045008. Other N = 1 finite theories can be obtained by perturbing N = 2 theories by massterms. A. Bianchi et al.R. [3] E. Rept. [4] O.K. Freedman. Oz. i.F. Rev. Anti De Sitter space and holography. S. Theor. Phys. Y. hep-th/9806074. Ya. H. D’Hoker. at finite N [66. Klebanov. Polyakov. Rangamani. Seiberg. Maldacena. Math. These are expected to be dual to genuine string excitations and as such should decouple from the operator algebra in the supergravity limit. string theory and gravity.

type IIB D-instantons. Erdmenger. Sokatchev. Non-renormalisation of next-to-extremal correlators in N = 4 SYM and the AdS/CFT correspondence. Eden. [27] G.C. V. [24] N. D-particle bound states and the D-instanton measure. [13] B. S. M. Lett.C. Bianchi et al. hep-th/9804170. Dorey. hepth/9811172. Phys. Yang–Mills instantons in the large-N limit and the AdS/CFT correspondence. Supersymmetric Yang–Mills theories. Green. Phys. M.P. West. hep-th/0002170. Howe. hep-th/0001138. C. hepth/0003096. Scalar quartic couplings in type IIB supergravity on AdS5 × S 5 . P. D’Hoker. [22] N. hep-th/9910016. Hollowood.C. P. B 498 (1997) 195. B 466 (1999) 20. hep-th/9908160. C. hep-th/9911222.C. hep-th/9901128. H. [25] E. B 557 (1999) 355. Four-point functions in N = 2 superconformal theories. Three-loop four-point correlator in N = 4 SYM. Arutyunov. High Energy Phys. [18] M. Green. Lett. Phys. L. Vanhove. J. L.V. Sokatchev.C.S. Bianchi. Mattis. Matusis. Gutperle. Four-point functions of lowest weight CPOs in N = 4 SYM4 in supergravity approximation. Frolov. Sokatchev.V. Phys. hepth/9701093. Mathur. Vandoren. [21] M. J.J. hep-th/9710151. [14] M. Extremal correlators in fourdimensional SCFT. J.P.V. Bianchi. strings and M-theory. Frolov. Phys. [20] M. D 60 (1999) 105038. Nucl.D. Nucl. Rossi. S. G. [10] B. Green. A. Perez-Victoria. Kwon. [30] L. Nucl. [23] N. Howe. Simplifications of four-point functions in N = 4 supersymmetric Yang–Mills theory at two loops. S. hep-th/9910150. Bianchi. hep-th/9810243.B. S. V. Arutyunov. Rastelli. Correlators of short multi-trace operators in N = 4 supersymmetric Yang–Mills theory.J. D’Hoker. Green. Instantons in supersymmetric Yang–Mills and D-instantons in IIB superstring theory. D. M. M. J. B 472 (2000) 323. Skiba. Mathur. High Energy Phys. Hollowood. Schubert. M. J. Eden. S. T. Kovacs. P. S. Eden. Multi-instanton calculus and the AdS/CFT correspondence in N = 4 superconformal field theory. S. hep-th/9811060. J. 9801 (1998) 005. Phys. Four-point functions in N = 4 supersymmetric Yang–Mills theory at two loops. S. Lett.B. Phys. P. Non-renormalisation of extremal correlators in N = 4 SYM theory. hep-th/9808157. Green. M.S. B 468 (1999) 102. Banks. [9] M. High Energy Phys. Phys. hep-th/9807033. Schubert. B 552 (1999) 88.B.S. Stanev. P. P.D. [16] M. hep-th/9906188. Gutperle. Nucl. E. Operator product expansion of N = 4 SYM and the 4-point functions of supergravity. D-instantons. Kovacs. Multi-instantons and Maldacena’s conjecture. G. West. Mattis. Phys. hepth/9912210. [8] E. J.B. . hep-th/9907088. E. West. M. C. P. [12] B. B 121 (1977) 77. On the logarithmic behaviour in N = 4 SYM theory. [29] B. A. Yang–Mills instantons vs. [15] M. 05 (1998) 002.230 M. P. 06 (1999) 023. [19] T. Schubert. High Energy Phys. Brink. High Energy Phys. Phys. Effects of D-instantons. Extremal correlators in the AdS/CFT correspondence. [17] M. Phys. S.B. Sokatchev. A. Lett. Lett. Khoze. S. Vandoren. E. Scherk. Kovacs. D. Freedman. Eden. Schwarz. Sokatchev. Vandoren. E. Bianchi. [26] G. Kovacs. V. [11] J. Khoze. B 408 (1997) 122. Khoze.H. B 421 (1998) 149. 08 (1998) 013. S. Matusis. 08 (1999) 020. M. Howe. hep-th/9906051. Dorey. Sixteen-fermion and related terms in M-theory on T 2 . Schubert. P. M.Z. Rossi.B. E. hep-th/9711107. Howe. Eden. Rastelli. Non-perturbative effects in AdS5 × S 5 string theory and d = 4 SUSY Yang–Mills. / Nuclear Physics B 584 (2000) 216–232 [7] W. Ya.S. C. Freedman. Pickering.C. [28] B. Green. B 442 (1998) 145. hep-th/9912250.S. Lett. Mattis. Dorey. hep-th/9704145. Gutperle. T. Rev. West.P.Z.

B 544 (1999) 737. Scherk. Nucl. multiplet shortening and the AdS5 × S 5 correspondence. W. M. Nonperturbative formulas for central functions of supersymmetric gauge theories. M. Johansen. [37] S. 2|4) multiplets in the AdS/CFT correspondence. Explicit construction of nilpotent covariants in N = 4 SYM. hep-th/9905112. Sokatchev. KK excitations on AdS5 × S 5 as N = 4 “primary” superfields. Nucl. Zero three-loop beta function in the N = 4 supersymmetric Yang– Mills theory. A. Phys. [52] M. Nucl. [44] L. V. Phys. Gliozzi. West. [48] L. [50] S. Minic. B 534 (1998) 96. S. |N ) superalgebras and multiplet shortening. Rahmfeld. [32] S. Ferrara. B 122 (1977) 253. [40] D. Grisaru. D.T. E. Nucl. Ferrara. B 541 (1999) 369. Ussyukina. Nucl. Gunaydin. M. A perturbative re-analysis of N = 4 supersymmetric Yang–Mills theory. Kallosh. Grisaru. hep-th/9711035. A.M. D. hep-th/9905020. Anselmi. Grisaru. B. B 298 (1993) 363. 2|N ) and harmonic superspace analyticity.Z. A. Rev. Lett. B 79 (1974) 413. [56] K. Lett. Two-point functions of chiral operators in N = 4 SYM.E. Novel supermultiplets of SU(2. Claus. Short representations of SU(2. [36] S. 12 (1999) 006. Freedman. [34] W. Phys. Freedman. Ferrara. [53] P. Zanon. [51] M. Phys. Sokatchev. R. J. Nucl. Nucl. J. P. Howe. hep-th/0002141. 2.T. hep-th/9708042. hep-th/9810226. C. The N = 4 quantum conformal algebra. Caswell. Intriligator. Erlich. Santambrogio. hep-th/9912007. D 57 (1998) 7570. hep-th/9812067. Y. Anselmi. Phys.A. hep-th/9806042. 4d doubleton conformal theories.I. [43] L. S. Lett. Phys. Davydychev. Intriligator. [47] S. Johansen. Petkova. D. Johansen. Exact results for three and four point ladder diagrams with an arbitrary number of rings. D. hep-th/9910197. A. M. Lett. Nucl. Ferrara. E. Shortening of primary operators in N extended SCFT4 and harmonic-superspace analyticity. hepth/9809192.B. Phys. [33] M. Phys. . 45 (1980) 1063. B 551 (1999) 575. Math. Phys. Andrianopoli. Bonus symmetry of N = 4 super Yang–Mills correlation functions via AdS duality. Zanon. PhD thesis. CPT and IIB string on AdS5 × S 5 . Gunaydin. Minic. 05 (1999) 019. Ferrara. Phys. J. [49] S. M. Universality of the operator product expansion of SCFT4 .I. Olive. Lett. Kovacs. D. Kovacs. Nucl. Supergauge invariant Yang–Mills theories. Phys. Ferrara. Anselmi. A. [42] D. [38] N. B 430 (1998) 248. Zumino.K. Phys. Nucl. Siegel. hep-th/9811047. hep-th/9803171. Zagerman. M. hep-th/9608125. supergravity and the dual spinor model. B 182 (1981) 125. [45] V. hep-th/9912168. Rev. Phys. 48 (1999) 145. Dobrev. Zero value of the three-loop β function in N = 4 supersymmetric Yang–Mills theory. [46] S. Superconformal field theories. B 526 (1998) 543. Supersymmetry. B. W. Ferrara. High Energy Phys. Phys. B 162 (1985) 127. E. B 559 (1999) 165. hep-th/9910011. Roˇcek. Bonus symmetry and the operator product expansion of N = 4 super Yang–Mills. Zupnik. High Energy Phys. Ferrara. Sokatchev. J. 2|4) and the AdS5 /CFT 4 duality. Phys. [35] S. hep-th/9908163. [39] D. [41] D.I. / Nuclear Physics B 584 (2000) 216–232 231 [31] F. Penati. [55] K. Anselmi.Z. Zaffaroni. Phys. Positivity constraints on anomalies in supersymmetric gauge theories. hepth/9902047. talk given at “Strings ’99”.C. Freedman. Zunger. B 394 (1997) 329. Skiba. hep-th/9908171. Andrianopoli. All positive energy unitary irreducible representations of extended conformal supersymmetry. N = 4 supersymmetric Yang–Mills theory and the AdS/SCFT correspondence. B 538 (1999) 531 (Erratum). Andrianopoli. Zagerman. 2|4). J.Z. Superspace representations of SU(2. On short and long SU(2. Lett. S. [54] P. Gunaydin. A.S. D. Bianchi et al. D. Supertwistors as quarks of SU(2.A. D. Schubert. Phys.

Gauge theories from orientifolds and large N limit. R. Witten. [67] A. Theor. Silverstein. C. in preparation. A. D’auria. Bershadsky. S. hep-th/9905226. B 523 (1998) 59. Phys. [64] A.S. Ferrara. Phys. Kachru. D 61 (2000) 066001. 17 (2000) 1017. hep-th/9804085. Lett. Class. Phys. Lawrence. Nucl. Bershadsky. Phys. Dall’Agata. E. J. D. / Nuclear Physics B 584 (2000) 216–232 [57] S. C. Kovacs. Kakushadze. Maldacena. [66] J. hep-th/9802183. 80 (1998) 4855. Nucl. N. hep-th/9807080. B 536 (1998) 199. hep-th/9806194. Rev. 2 (1998) 733. Adv. Comments on string theory on AdS3 . N. A. hep-th/9803214. Phys. hep-th/9910066. Math. G. G. Johansen. Bianchi. D’auria. Giveon. [60] A. Kutasov. B 536 (1998) 141. Stanev. Superconformal field theories from IIB spectroscopy on AdS5 × T 11 . Quant. Nucl. Phys. Nucl. Large N limit of orbifold field theories. hep-th/9803249. [59] Z. AdS3 black holes and a stringy exclusion principle. Rossi and Ya. E. Ceresole. B 529 (1998) 157. String expansion as large N expansion of gauge theories. Grav. Z. Phys. G. Vafa. [58] M.C. [61] M. Ceresole. Rev.232 M. Seiberg. 4-D conformal theories and strings on orbifolds. Vafa. Klebanov. R. S. hep-th/9803015. Superconformal field theory on three-branes at a Calabi–Yau singularity. [65] M. hep-th/9803076. Dall’Agata. . Strominger. 12 (1998) 005. High Energy Phys. Phys. [62] I. Ferrara. Nekrasov. S. Bianchi et al. Kakushadze. B 533 (1998) 199. [63] A. Nucl. Spectrum of type IIB supergravity on AdS5 × T 11 : predictions on N = 1 SCFTs. On conformal field theories in four dimensions.

1. 0) conformal theory describing N coincident M5 branes.elsevier. All rights reserved. for large N . Comparison of the 2-point and 3-point correlators of the stress tensor of (2. 0550-3213/00/$ – see front matter  2000 Elsevier Science B. 0) supersymmetric conformal field theories in 6 dimensions are the free tensor multiplet theory describing low energy dynamics of a single M5 brane. Columbus. The known results for R-anomaly suggest that the total scale anomaly extrapolated to N = 1 should be the same as the anomaly of a single free (2. 0) tensor multiplet.3 2 1 3 ( 0 0 ) 0 0 3 8 0 . A way to study the latter theory is provided by its conjectured duality [1] to M-theory (or.A. 0) theory and in the free tensor multiplet theory) is shifted to 4N 3 − 3N.V. and still largely mysterious interacting (2.1 . Introduction Two known maximally (2. 0) theory as predicted by the AdS7 × S 4 supergravity [2.nl/locate/npe R 4 terms in 11 dimensions and conformal anomaly of (2.V. This R 4 term is argued to be part of the same superinvariant as the P-odd C3 R 4 term known to produce O(N) contribution to the R-symmetry anomaly of (2. Imperial College.ohio-state. All rights reserved. accepted 6 June 2000 Abstract Using AdS7 /CFT 6 correspondence we compute a subleading O(N) term in the scale anomaly of (2.2 Department of Physics. Moscow. PII: S 0 5 5 0 .edu 2 Also at Blackett Laboratory.3] to those in the free tensor multiplet theory shows [4–6] that they differ only by the overall coefficient 4N 3 . 3 The same is true also for the correlators of R-symmetry currents [7].Nuclear Physics B 584 (2000) 233–250 www. Tseytlin 1. OH 43210-1106. 11-d supergravity corrected by higher derivative terms) on AdS7 × S 4 background. 0) theory describing N coincident M5 branes. 3 The remarkable 1 tseytlin@mps. USA Received 16 May 2000. 0) theory A. 0) theory. A proposed explanation of this agreement is that the coefficient 4N 3 in the anomaly (which was found previously to be also the ratio of the 2-point and 3-point graviton correlators in the (2. The Ohio State University. While the leading O(N 3 ) contribution to the anomaly is determined by the value of the supergravity action. the O(N) contribution comes from a particular R 4 term (8-d Euler density invariant) in the 11-dimensional effective action.  2000 Elsevier Science B. London and Lebedev Physics Institute.

0) theory may have special “protected” form. This was.14]. which. allowing one to interpolate between N  1 and N = 1 cases. These R 3 corrections produce extra O(N) terms in the conformal anomaly of the boundary (2. in fact. Since this conclusion is sensitive to numerical values of coefficients in the 11-d low energy effective action we shall start with a critical review of what is known about the structure of R 4 terms in type IIA string theory in 10-d and their counterparts in M-theory. one may expect that anomalies and 2. We shall argue that the 11-d action contains a particular R 4 term. of course. The analogy between the d = 4 and d = 6 cases should not. and for N = 1 is equal to the R-symmetry anomaly corresponding to the single tensor multiplet [13. one should then expect to find a similar O(N) correction to the O(N 3 ) supergravity contribution [8] to the (2. observed for the R-symmetry anomaly of the (2. 0) theory obtained from the 11-d action containing the standard supergravity term plus a higher-derivative C3 R 4 term [12] is given by the sum of the leading supergravity O(N 3 ) and subleading O(N) terms. this would be a resolution of the “4N 3 ” puzzle. 0) theory [11]: the anomaly of the (2. the one-loop term in type IIA theory is a combination of two different R 4 structures. Still. given that the (2.and 3-point correlators of currents of the (2. the coefficient 4N 3 in the ratio of the (2. which is equal to 1 for N = 1. 0) theory [8] and free theory of a single tensor multiplet [9]. 0) theory should belong to the same d = 6 supermultiplet [11. Our aim below is to discuss a mechanism of how this may happen. Since the conformal and R-symmetry anomalies of the (2.15]. 0) theory and tensor multiplet scale anomalies may be shifted to 4N 3 − 3N . upon compactification on S 4 .10] led to the ratio N 2 . leads to a special combination of R 3 terms in the effective 7-d action.A. 0) conformal theory. .234 A. Tseytlin / Nuclear Physics B 584 (2000) 233–250 coefficient 4N 3 was originally found in [5] in the comparison of the M5 brane world volume theory and the D = 11 supergravity expressions for the absorption cross-sections of longitudinally polarized gravitons by N coincident M5 branes. We shall argue that they should be organized into two different N = 2A superinvariants — J0 and I2 (containing P-odd B2 tr R 4 term) in a way different than it was previously suggested (Section 2). Since the latter is equal to 1 for N = 1. 0) conformal anomaly. While the type IIB theory effective action contains the same J0 ∼ R 4 invariant at the tree and one-loop levels. being an interacting conformal fixed point without a free coupling parameter. The reason why the coefficient 4N 3 was puzzling in [5] was analogy with the d = 4 case: a similar comparison of the gravitational and world-volume absorption cross-sections in the case of D3-branes [5. As a result. 0) theory should have a different structure than SYM theory. The corresponding tow D = 10 superinvariants “lifted” to D = 11 represent the leading R 4 corrections to the 11-d supergravity action (Section 3). This agreement in the d = 4 case was later understood [6] as being a consequence of nonrenormalization of the conformal anomaly and thus of the 2-point stress tensor correlator in N = 4 SYM theory. with simple dependence on N . The same coefficient 4N 3 appears also as the ratio of the scale anomalies (or Weyl-invariant parts of conformal anomalies) of the interacting (2. be taken too seriously. This O(N) correction should originate from a higher-derivative R 4 term in the 11-d action which should be a part of the same superinvariant as C3 R 4 term (just like the secondderivative supergravity terms R and C3 F4 F4 are).

2 · 3! L1 = b1 J0 + b2 K.  1 H32 + · · · + b0 α 03 J0 . The relevant terms in the tree + one-loop type IIA string theory effective action can be written in the form S = S0 + S1 . and upon reduction to 8 spatial 10 10 dimensions mn.5) In the notation we are using the numerical coefficients are 2 = (2π)7 gs2 α 04 .6) b1 = (2π)4 1 . · 32 · 213 (2. For other notation see also [16]. 3 · 211 b2 = −12b1 = − (2. paying special attention to explicit values of numerical coefficients.7) 4 We use Minkowski notation for the metric and  tensor.A. S0 = − 1 Z d 10 x 2 2κ10 1 S1 = − 2πα 0  √ −G e−2φ R − Z d 10 x √ −G L1 . so that   = −10!.. 0) theory using the method of [8]. R 4 terms in 10 dimensions Let us start with a review of the structure of the R 4 terms in the effective actions of type IIA superstring in 10 dimensions and the corresponding terms in M-theory effective action in 11 dimensions. 4 · 2! (2.. (2π)4 · 3 · 211 1 .1) (2. → −28 8 . 2κ10 b0 = 1 ζ (3).3) J0 = J1 − J2 ≡ t8 · t8 RRRR − 1 10 · 10 RRRR. Tseytlin / Nuclear Physics B 584 (2000) 233–250 235 These terms should be supplemented by proper F4 = dC3 dependent terms as required by supersymmetry and chosen in a specific “on-shell” scheme not to modify the AdS7 × S 4 solution of the D = 11 supergravity.A. (2. 4 (2.4)   2 1 K = 10 B2 tr R 4 − tr R 2 . 4 · 2! (2. in Section 4 we discuss higher derivative corrections to the 7-d action of S 4 compactified theory which follow from the presence of the R 4 terms in D = 11 action. In Section 5 we compute the corresponding O(N) contributions to the scale anomaly of the (2. Assuming that. . and draw analogy between the total O(N 3 ) + O(N) result and the expression for the Rsymmetry anomaly found in [11]. mn.2) where Hmnk = 3∂[m Bnk] and 4 J0 = J1 + J2 ≡ t8 · t8 RRRR + 1 10 · 10 RRRR.. 2..

the coefficients in (2. 4 1 n n D D R 4 = ±8!δ[m1 1 . 6 1 The invariant J2 = 4·2! 10 · 10 RRRR which will play important role in what follows is the D = 10 extension of the integrand of the Euler invariant in 8 dimensions 1 J2 = E8 . but our overall normalization of this term is different (by factor 25 compared to [20]). First.. i. Since the D = 10 supersymmetry algebra does not close off shell. e. E8 = (D − 8)! (2.e. e.g.29]. D = 10 supersymmetry only. and this assumption passes some compactification tests [28. Φ = φ + · · · + θ 4 R + · · · written in terms of N = 1 or N = 2B [36.38–40] can be chosen as J0 . 7 The expansion of E8 near flat space (gmn = ηmn + hmn ) starts with h5 terms (see.8) where ± correspond to the case of Euclidean or Minkowski signature. . [21]). Tseytlin / Nuclear Physics B 584 (2000) 233–250 The tree and one-loop coefficients of the well-known J1 = t8 · t8 RRRR term 5 can be determined from the 4-graviton amplitude [17–19].35] (i.e. [33]). X1 . Its coefficient b2 can be fixed by considering compactification to 2 dimensions [30]. so the leading-order corrections to effective action evaluated on the supergravity equations of motion should be invariant under the standard supersymmetry transformations.37] on-shell superspace superfield Φ). 1 6 Note that the total coefficient of the t · t RRRR term in S is thus − (ζ (3)/gs2 + π 2 /3). The sigma-model approach implies [22–26] that E8 does appear in S0 ..A. δm88 ] R m1 m2n1 n2 · · · R m7 m8n7 n8 . 7 The Euler number in 8 dimensions is χ = R 8 √ 1 d x g E8 .. so that its coefficient cannot be directly determined from the on-shell 4-graviton amplitude.5) can be established [30] following similar calculations of anomaly-related terms in the heterotic string [31. .32]. and its value is in agreement with the coefficient required by 5-brane anomaly cancellation [12] (see also below). The 8 8 (2π )7 ·3·211 α 0 2 2 2 2 0 −4 relative combination ζ (3)/gs + π /3 is the same as in [19] (where g = (2κ10 ) (2α ) = 16π 7 gs2 ) and in [20]. The presence of the P-odd one-loop term K (2.g. X2 5 The more explicit form of this term is J = 24t [tr R 4 − 1 (tr R 2 )2 ]. . The low-energy effective string action should be supersymmetric. A basis of the three independent N = 1 invariants [16. If one first restricts consideration to N = 1. then one can use the classification of possible bosonic R 4 parts of on-shell non-linear N = 1 superinvariants given in [38–40]. that (up to usual field redefinition ambiguities) the tree-level type II string R 4 term is indeed proportional to J0 (2.4) of the J1 and J2 terms. the full off-shell effective action should be invariant under “deformed” supersymmetry transformations (see. where R = (R ab ) and t tr R 4 ≡ mn 1 8 8 4 tr(16Rmn Rrn Rml Rrl + 8Rmn Rrn Rrl Rml − 4Rmn Rmn Rrl Rrl − 2Rmn Rrl Rmn Rrl ). 8 Remarkably. d 16 θ Φ 4 .3).7) are indeed consistent with what is known about the structure of possible R 4 superinvariants. (4π )4 ·3·27 8 The string S-matrix is invariant under on-shell supersymmetry. The structure of the kinematic factor (t8 + 12 8 )(t8 + 12 8 ) in the one-loop type IIA 4-point amplitude with transverse polarisations and momenta suggests [27–29] that the one-loop R 4 terms in D = 10 type IIA theory should be proportional to the oppositesign combination J0 (2. the h4 term in t8 t8 R 4 is the bosonic part of the on-shell linearized R superspace invariant [34.236 A..

10 Once the dilaton dependence of J terms is taken into account. 10 We would like to argue that it is I2 and not I1 that is the true N = 2A superinvariant.9) 1 1 X2 = t8 tr R 2 tr R 2 − 10 B2 tr R 2 tr R 2 . Then the tree + one-loop J0 terms in the type IIA theory will be exactly the same as in the type IIB theory. as L1 = b1 J0 + b2 K = b1 (J1 − J2 − 12K) = b1 (−J0 + 2I1 ).e. Tseytlin / Nuclear Physics B 584 (2000) 233–250 J0 = t8 · t8 RRRR + 1 10 · 10 RRRR.4) with ±6K (2. 4 4 One may try to combine these N = 1 invariants to form potential N = 2A superinvariants.A.11) = t8 · t8 RRRR − 610 B2 tr R 4 − tr R 2 .5). (2.13) The 1-loop term L1 (2. As a result. + − 3 (2π)4 · 3 · 213 α 0 gs2 9 In [38–40] where non-linear extensions of N = 1 on-shell R 4 superinvariants were constructed the transformation of the dilaton prefactor was ignored. Since t8 t8 R 4 = 24t8 [tr R 4 − 14 tr R 2 tr R 2 ]. For a discussion of supersymmetry of e−2φ R + f (φ)J0 action in type IIB supergravity theory see [41]. (2.28]. i. It is natural to conjecture that f (φ)J0 terms should combine into an N = 2A superinvariant (invariant under deformed supersymmetry). 9 A non-trivial question is which of I1 and I2 can be actually extended to an invariant of N = 2A supersymmetry. it is the Euler term J2 = 14 E8 and not J1 = t8 t8 RRRR that is the “superpartner” of the B2 -dependent term K (2. .e. The form of the 1-loop correction L1 that admits a superextension is then (2. one may consider two candidate invariants which contain combinations of J1 (2.   ζ (3) π 2 1 J0 . (2. (2. 4 or   1 I2 = J0 − 24 X1 − X2 = J2 + 6K 4   2 1 1 10 · 10 RRRR + 610 B2 tr R 4 − tr R 2 .A. one was not able to make a distinction between J0 terms appearing at the tree and 1-loop levels. i.2) with b2 = −12b1 can thus be represented as a combination of two different R 4 superinvariants [27.15) or as The J0 -term should represent a separate N = 2 invariant.13). (2. Namely.15) and not (2.14) L1 = b1 (J0 − 2I2 ). 4 · 2! 237 (2.14).5)..   1 I1 = 24 X1 − X2 = J1 − 6K 4   2 1 (2..10) X1 = t8 tr R 4 − 10 B2 tr R 4 .12) = 4 · 2! 4 I1 + I2 = J0 .3) or J2 (2. one will not be able to freely switch between 0 I1 and I2 using (2.

At the same time. and the superpartner of the 11 C3 R 4 term should be the D = 11 analog of J2 = 14 E8 (see Section 3). The t8 t8 RRRR term appears [20. one concludes that J1 and K cannot be parts of the same superinvariant. [20. for example. i. the Chern–Simons type terms like 10 B2 R 4 cannot appear in the UV divergent part of one-loop effective action.4). cf. and the corresponding 5-point amplitudes should be related by global supersymmetry. and thus can be rewritten in a manifestly gauge invariant but ∂ nonlocal form). one may use R 4 mn = 0 to simplify the structure of R invariants as the graviton legs in the string amplitudes they correspond to are on mass shell.1). (2.27.46]) these terms cannot be related by supersymmetry. ignoring other fields. (2. e. let us add few comments about the structure of the D = 10 effective action (2.. but 11 C3 R 4 term [12] can have only a finite coefficient (with a non-perturbative dependence on κ11 on dimensional grounds). Similar argument can be given in the context of D = 11 theory. Indeed. the tree-level effective action (2.28. the weak-field expansions of both E8 and K start with 5-order terms. 11 Known examples of induced CS terms have finite coefficients and originate from IR effects (they appear from 1-loop contributions containing 12 massless poles.15) proportional to the superinvariant I2 . In addition to the R 4 terms given explicitly in (2. . it is hard to imagine how the linearized on-shell “W 4 ” N = 2 superspace invariant corresponding to h4 term in t8 t8 RRRR may have a non-linear extension containing P-odd term K. Tseytlin / Nuclear Physics B 584 (2000) 233–250 with the type IIA theory action containing in addition one extra one-loop contribution (2. The well-known field redefinition ambiguity [18. Λ = fixed. cf. (2. 11 This can be proved directly by using the background field method: all one-loop UV divergent terms must be manifestly invariant under 2-form gauge transformations and as well as diffeomorphisms.16) J0 = 3 · 2 R 2 The field redefinition ambiguity allows one to choose the action in a specific “scheme” where only the Weyl tensor part of the curvature appears in J0 . Since. in the one-loop 4-graviton amplitude. At the same time.43–45] as a cubic UV divergence (with a particular value of the UV cutoff being fixed by duality considerations [45]).238 A. terms involving two and more powers of H3 = dB2 (which were not included in the discussion of superinvariants in [38–40]). 12 In particular. Thus (contrary to some previous suggestions in the literature..38–40.3) and (2. it may contain also other Ricci tensor dependent terms as well as terms depending on other fields (cf.g. a proper time cutoff is expected to preserve supersymmetry at the level of one-loop UV divergences.48] allows one to change the coefficients of “on-shell” terms.e. α 0 → 0.49])   1 hkmn 8 hmnk rsp q rsp q Rpmnq Rh R rsk + R Rpqmn Rh R rsk + O(Rmn ). 12 For example.2)).1) may contain other Rmn dependent terms in addition to the full curvature contractions present in J0 (see [24–26.A. A more serious argument against t8 t8 RRRR being a “superpartner” of 10 B2 [tr R 4 − 1 2 2 4 (tr R ) ] is the following.27.2). Before turning to a detailed discussion of the D = 11 terms. [47]). The D = 10 type II supergravity is known to contain a oneloop quadratic Λ2 UV divergence proportional to t8 t8 RRRR (this can be seen [42] by taking the field theory limit.

(2. Tseytlin / Nuclear Physics B 584 (2000) 233–250 239   1 J0 → Jˆ0 = 3 · 28 C hmnk Cpmnq Ch rsp C q rsk + C hkmn Cpqmn Ch rsp C q rsk . we conclude that the effective action of the D = 11 theory should contain two distinct R 4 superinvariants: (i) J0 with t8 t8 R 4 as its part.11)) 1 J0 = t8 · t8 RRRR + E8 .17). its coefficient cannot develop dilaton dependence without breaking B2 gauge invariance.A.27–29] to D = 11 (with its tree-level part giving vanishing contribution).3) does not receive higher than one-loop perturbative string corrections.e. (2. Taking the limit gs → ∞ this term can then be lifted to a corresponding superinvariant in D = 11 theory. S0 = − 1 2 2κ11 d 11 x  √ −g R − d 11 x √ −g (J0 − 2I2 )..1) (3.45].2) Here Fmnkl = 4∂[m Cnkl] and the two R 4 superinvariants are (see (2. i. (2. and the coefficient of the I2 term (its C3 R 4 part) is precisely the one implied by the M5 brane anomaly cancellation condition [12]. its value cannot be renormalized from its coupling-independent one-loop value [16]. 3. it can be also lifted [20. For example.8). in particular. Assuming that the coefficient of the J0 invariant (2. R 4 terms in 11 dimensions Since the invariant I2 in (2. With this separation.9).3) (3.2)) S = S0 + S1 .15) contains the P-odd CS type part K. both in 10 and in 11 dimensions (see Section 3).A. the coefficient in front of the J0 term is then in agreement with the 4-graviton amplitude (with the Mtheory cutoff [45]). the D = 11 action is then (cf. Z Z S1 = −b1 T2  1 1 F42 −  C F F 11 3 4 4 . In view of the above discussion. I2 = E8 + 211 C3 tr R − tr R 4 4 (3. and (ii) I2 which is a sum of the E8 and 11 C3 R 4 structures. The resulting presence of the t8 t8 R 4 term in the M-theory effective action is indeed in agreement with what follows directly from the low-energy expansion of the 4-graviton amplitude in D = 11 supergravity [27. (2. (2.4) . 3!    1 1 4 2 2 .1). Explicitly. 2 · 4! 6 · 3! · (4!)2 (3. 4 E8 = 1 11 · 11 RRRR. in order to avoid corrections to certain highly symmetric leading-order solutions. in type IIB theory the (scale of) AdS5 × S 5 solution is not modified by the R 4 terms [50] only in the scheme [51] where they have the form (2.17) 2 That freedom of choice of a special scheme is crucial.

T2 = 1 3 2πl11 2 = (2π)2/3 2κ11 −1/3 . one may derive the corresponding d = 7 supergravity action. Tseytlin / Nuclear Physics B 584 (2000) 233–250 The constant b1 = 1 (2π)4 ·32 ·213 is the same as in (2.7) 4.2) may influence the existence of the AdS7 × S 4 solution and expansion near it. (2π )4 ·3·26 14 In addition to F dependent terms (which may contain up to 8 powers of F ) there are also ∂F dependent 4 4 4 terms which accompany t8 t8 R 4 part of J0 in the 4-point S-matrix [56] (as suggested by the analysis of tree-level 4-point scattering amplitudes in 11-d supergravity). 2πα 0 (3. 4 (3.2) in the effective action of 11-d theory may contain also other O(Rmn ) and O(F4 ) terms. These derivative terms vanish on AdS7 × S 4 background. AdS7 × S 4 solution and compactification on S 4 The D = 11 supergravity admits the well known AdS7 × S 4 solution with F4 flux N through S 4 [54]. C3 = 3! mnk dx ∧ R 1 n k ab ab m n 4 4 dx ∧ dx . 0) theory. (3. so that the 10-d invariant T 4 2 1 B2 ∧ tr(∧R) in (2. which gives the O(N 3 ) contribution [8] to the conformal anomaly in the corresponding boundary conformal (2. R = 2 Rmn dx ∧ dx . Let us consider how the presence of the R 4 terms in the 11-d effective action S1 (3.5) T1 = 1 = 2πR11 T2 . 2πR11 √ √ R11 = gs α 0 . where in the form notation B2 = 2 Bmn dx ∧ dx . it was argued in [55] that these corrections cannot modify the maximally supersymmetric AdS7 × S 4 solution.6) The subleading O(T2 ) term (3. .2) (supplemented with F4 dependent terms as required by supersymmetry 14 and chosen in a special “on-shell” scheme analogous but R 13 Note that B and C are canonically normalized. Thus S1 (3. Using the on-shell superspace description of 11-d supergravity and assuming that all local higher-order corrections to the equations of motion can be written again in terms of the basic on-shell supergravity superfield. Compactifying on S 4 .A.2) contains T2 C3 ∧ tr(∧R) with the coefficient 4 · 3! · 2 b1 = 1 which is the same as in [12]. Note that in the exterior form notation I2 may be written as I2 e0 ∧ e1 ∧ · · · ∧ e10 2 = 11 e ∧ e ∧ e ∧ R ∧ R ∧ R ∧ R 3  + 25 · 3!C3 ∧ tr(R ∧ R ∧ R ∧ R) −  1 tr(R ∧ R) ∧ tr(R ∧ R) . l11 = (2πgs )1/3 α 0 . The invariant J0 (supplemented with appropriate F4 dependent terms) may be considered as a non-linear extension of the linearized “R 4 ” superinvariant in on-shell D = 11 superspace [52].240 A. κ10 = .7) and the 10-d and 11-d parameters are related as follows (T1 and T2 are the string and the membrane tensions) 13 2 κ11 2 9 2 2κ11 = (2π)5 l11 . The P-even part of the second superinvariant starting with I2 (3. It should be possible to see explicitly that adding the J0 term in (3.4) may also include extra O(F4 ) terms.2) goes R 3 1 1C 4 m n m into the 11-d one T2 C3 ∧ tr(∧R) .

while not changing the vacuum solution.. Since the R-symmetry and conformal anomalies should belong to the same 6-d supermultiplet. however. essentially as in the case of AdS5 × S 5 solution of type IIB theory corrected by J0 term [50] (taken in the form (2. modify the expression for the conformal anomaly of the boundary conformal theory. where Wabcd (x. 19 It is known that the C3 R 4 part of I2 (3. it is natural to expect that the “superpartner” of the C3 R 4 term. E8 contains “reducible” curvature contractions like ((Rmnkl )2 )2 + R(Rmnkl )3 + · · · and thus may. ∂F4 = 0) will vanish. 18 It is important to stress for what follows that in the above discussion we treated J (3. i.4) gives a subleading O(N) correction to the R-symmetry anomaly of the (2. The argument of [55] should certainly apply to the first subleading correction to the 11-d supergravity equations of motion coming from R 4 terms in the action. we shall assume that. θ ) is the on-shell supergravity superfield [52]. the E8 term in I2 . and thus should lead only to C 4 terms upon compactification to d = 7. As for the E8 part of I2 . now defined in 7 dimensions. 16 The vanishing of the first variation is equivalent to the vanishing of the first correction to the 11-d supergravity equations of motion γ abc DWabcd = 0 due to the supercovariant constancy of W [55].3) as a whole. the radii of its factors) is not modified by the J0 correction can be also represented as a consequence of the fact that upon compactification of the 11-d theory on S 4 with F4 flux the J0 term (taken in the special “on-shell” scheme) reduces to the Weyl tensor dependent C 4 term (2. 16 The fact that the AdS7 × S 4 solution (and. without 0 splitting it into t8 t8 R 4 and E8 parts. as in the case of J0 .2). in addition to the one coming from the J0 term (2. One may view J0 as originating from a restricted superspace integral of W 4 . will give subleading O(N) corrections to the conformal anomaly of the boundary CFT.A. 19 These terms will give also another O(N ) correction to the entropy of (2. in particular. It is only that particular combination of R 4 terms that takes the “irreducible” form (2. [49]). . second and third variation over the metric evaluated on AdS7 × S 4 + F4 -flux background (Rmn ∼ (F42 )mn . Then J0 ∼ (R + F4 F4 )4 and its first. in principle.17) found in [51]. 17 The tree + one-loop J term in type IIB theory leads to the same C 4 term (2. n < 4. 0) theory. supplemented with proper F4 -dependent terms.16) (cf. should lead to an O(N) correction 15 Note that in contrast to AdS × S 5 space with equal radii the 11-d space AdS × S 4 space with radii 1 and 1 7 5 2 is not conformally flat. 0) theory describing multiple M5 branes.17) in 10-d theory) does not change the leading-order AdS7 × S 4 solution.17). 0) theory [11. At the same time.17) in the 5-d effective action 0 obtained by compactifying the type IIB theory on S 5 with F5 flux.e. Upon reduction on S 4 it leads to O(N) CS terms in d = 7 action [11]. Tseytlin / Nuclear Physics B 584 (2000) 233–250 241 not equivalent 15 to (2. 17 This term produces an O(N) correction [51] to the leading N 3 term [57] in the entropy of (2. lead to O(R n ). The main point is that upon compactification on S 4 the E8 term in (3.A. also does not modify the leading-order AdS7 × S 4 solution. terms upon compactification to d = 7. As in the AdS5 × S 5 case in type IIB theory. this C 4 term does not.17)).12].4) in (3.4) should produce additional R 3 higher-derivative terms in the 7-d effective action which. there exists an “on-shell” scheme in which this term. 18 Let us now discuss the second invariant I2 (3. which has the structure W = F4 + · · · + θ θ (γ · · · γ R + γ · · · γ F4 F4 + γ · · · γ DF4 ) + · · · (γ · · · γ stand for products of gamma matrices). It is easy to see that its P-odd part 11 C3 [tr R 4 − 14 (tr R 2 )2 ] does not modify the AdS7 × S 4 solution.

3) coming from E8 .8). M 11 = M 7 × M 4 .e. to determine the terms in the 7-d action that originate from the E8 part of the invariant I2 in (3..6) and [51.5) .3) = 2 7 7 RRR + 4 7 7 RR. We shall start with E8 and compute it in the case when the 11-d space is a direct product.A.2) and E2n (M d ) = 0 for d < 2n. but it is easy to see that a potential C 2 term in the 7-d action (coming from E4 in (4. Choosing the normalization in which the radii of AdS7 and S 4 are 1 and L = 1/2 so that 2 2 Vol(S 4 ) = 8π3 L4 = π6 . T2 = 2 8 5 9 5 2 2 3 π π 3π 2κ11 2 π L 2κ7 2κ11 R The relevant [N 3 (R − 2λ) + NC 3 ] terms in the 7-d action are then Z Z √ √ N3 γN S (7) = − 3 d 7 x −g (R + 30) + 2 11 3 d 7 x −g Eˆ 6 + · · · . L L A remarkable property of the E8 invariant is that it does not produce a correction to the cosmological or Einstein term in the 7-d action. 3π 3 ·2 ·π where the explicit form of the Eˆ 6 ∼ C 3 correction term is (cf. as in (2. (2. (d − 2n)! d > 2n.4) (4. = = .3) but with the curvature tensor R of M 7 replaced by its Weyl tensor C part. for E8 supplemented by F4 and Rmn dependent terms so that it does not produce a modification of the leading-order AdS7 × S 4 solution) then the resulting terms in the 7-d action will be the same as in (4. = = .  E2n M d ≡ 1 d · d R n . Since we do not know the F4 (and Rmn ) dependent terms which supplement E8 to a superinvariant. (4.242 A. It is easy to see that      (4. This is what we are going to suggest below.1) E8 M 7 × M 4 = 4E2 M 4 E6 M 7 + 12E4 M 4 E4 M 7 . In what follows we shall consider only on the E6 (M 7 ) ∼ C 3 + · · · term in (4. Tseytlin / Nuclear Physics B 584 (2000) 233–250 to the conformal anomaly of the boundary 6-d theory.8)) (4.3)) can lead only to terms in the conformal anomaly which vanish when the 6-d boundary space is Ricci flat.57]) N3 2N 3 1 Vol(S 4 ) N3 2N 1 . Next. (3.2) we shall use the following heuristic strategy. In the case when M 4 is a 4-sphere of radius L (RS 4 = 12/L2 ) and M 7 has curvature R we get  3 · 25  32 · 27  E4 M 7 E8 M 7 × S 4 = 2 E6 M 7 + L L4 3 · 26 3 · 25 (4. where.5). and assuming that the value of the quantized F4 flux is N . we shall assume that when the same reduction is repeated for the analog of E8 term in a special “on-shell” scheme (i. we get (see (3. The reason is that we shall compute the corresponding contribution to the scale anomaly of the boundary theory only modulo Rmn -dependent terms.

assuming topology of 6-space is trivial) and ignore terms which depend on Rmn . In the present case of (4. (4.6) and RAdS7 = −42)  3 Z  Z Z N γN dρ p 1 dρ √ (7) 6 6 −g(x.3)  20 Similar computation of subleading corrections to conformal anomaly of 4-d boundary conformal field theories (with N < 4 supersymmetry) coming from R 2 curvature terms in 5-d effective action were discussed in [46. = −6!δ[m 1 (4. ρ) − 2 6 3 −g (2I1 + I2 ) + · · · .3) the numerical coefficient γ is γ = 1.2) Since [8] Z   √ dρ p −g(x. Tseytlin / Nuclear Physics B 584 (2000) 233–250 243 Eˆ 6 = (E6 )Rmn =0 = 7 7 CCC n n1 . (4. Conformal anomaly of (2.5). . I2 = Cab mn Cmn pq Cpq ab .8) but we shall keep it arbitrary.e. 6 ρ4 (5. 5. We shall follow the same method as used in [8] in computing the leading N 3 term in the anomaly. δm66 ] C m1 m2n1 n2 C m3 m4n3 n4 C m5 m6n5 n6 = −32(2I1 + I2 ).A. 20 We shall compute only the O(N) contribution to the scale anomaly (which is the same as integrated conformal anomaly..6) where I1 and I2 are defined as I1 = Camnb C mpqn Cp ab q . 0) theory Let us now determine the contribution of the C 3 correction term in the 7-d action (4. . 4 evaluate the action on the solution g = g0 (x) + ρg2 (x) + · · · . S = d x 3π 3 ρ4 3 ·2 ·π 2 ρ   (5.5) which originated from the E8 part of the I2 superinvariant in the 11-d action (3.4) we set the radius of AdS7 to be equal to 1) 1 (5. and compute its variation under the Weyl rescaling of the 6-d boundary metric. .5) (for example. concentrate only on the Weyl-invariant non-total derivative C 3 terms (“type B” part) in the 6-d conformal anomaly. given the uncertainties in the above derivation of the correction term in (4. i. The anomaly is essentially determined by the coefficient of the logarithmic divergence produced by the integral over ρ [8].2) to the conformal anomaly of the d = 6 boundary conformal theory.5) we find (using (4.7) As follows from (4.1) ds 2 = ρ −2 dρ 2 + ρ −1 gij (x. ρ) dx i dx j .58– 61].A. (4. To obtain the conformal anomaly one is to solve the 7-d equations for the metric (as in (4. the presence of (F4 )2 (Rmnkl )3 terms in I2 would shift the value of γ ). ρ) = −g0 a0 (x) −3 + · · · − a6 (x) ln  + · · · .

the separation of the leading N 3 Weyl anomaly of the (2. 3π 3 3 ·2 ·π (5.0) + A(2.62] as follows: E(6) . in terms of the invariants E6 .6) was given in [64–66] and is equivalent to the Weyl invariant I3 used in [9] and here. (5.5) (5.0) = − N3 γN · 2a6 + 2 6 3 (2I1 + I2 + · · ·). contrary to what was assumed in [8] (this can be easily checked by computing it for the metric of a sphere S 6 : one finds that while I1 (S 6 ) = I2 (S 6 ) = 0. one has the following relations (cf. 23 Note that modulo terms that vanish for Rmn = 0 and total derivative terms.7) and I3  I3 = Cmnbc ∇ 2 C mnbc + O(Rmn ) + O ∇i J i . I3 = I3 − 3 (2I1 + I2 ) − 12 E6 + O(∇i J ).0) = −   4N 3 E6 + 8(12I1 + 3I2 − I3 ) + O ∇i J i .0) vanishes for Rmn = 0.A. 22 We use this opportunity to point out that the curvature invariant I3 = −5I30 as defined in [8.7) E6 = −32(2I1 + I2 ) + O(Rmn ). I1 . as it should [8]. Tseytlin / Nuclear Physics B 584 (2000) 233–250 the anomaly is given by the sum of the O(N 3 ) and O(N) terms 21 3 N A(2.0) = AN (2. I3 with E6 . I30 (S 6 ) 6= 0). in fact. and omit other parts that depend on the Ricci tensor of g0 . I2 . I1 . 0) tensor multiplet found in [9]: 21 To obtain the O(N ) contribution we evaluate the C 3 term in the 7-d action on the leading-order solution for 2 action in d = 5). In fact. (4. The correct separation was given in [9] and is used here. I2 are evaluated for the boundary metric g0 . . 3 2 5 (4π) · 3 · 2 where E6 = 6 6 RRR.6) which form the basis of 3 Weyl invariants are the same as used in [9]. They are related to the 1 invariants used in [8.244 A. (5. I2 and I3 in [8] are equal to 33 ·2 11 E6 .63] is 3 AN (2. Eq.4) Here a6 and I1 . up to a covariant total derivative term. 0) theory [8] into type A and type B parts was not presented correctly in [8]. a 22 Our curvature tensor R a bmn = ∂m Γbn − · · · has the opposite sign to that of [8]. (5. 3 The result of [8] for the leading-order contribution AN (2. type B (Weyl invariant) and scheme-dependent (covariant total derivative) terms [62. I2 and I3 used in [9] and here.5) is to be compared with the expression for the conformal anomaly for the free (2. and dots stand for O(N) Rmn -dependent and total derivative terms we are ignoring. Since I3 of [8] or I30 is a mixture of the true Weyl invariants I1 . The proper third Weyl invariant of type C∇ 2 C (5. The invariants I1 . Note also that [9] was assuming Euclidean signature where E6 is defined as −6 6 RRR.0) written as a sum of the type A (Euler). 3 so that AN (2. separate the C 3 part the metric (5.1) (see [58] for a similar computation in the case of the Rmnkl depending on the 6-d metric g0 . covariant under Weyl transformations. −I1 .62] is not. 23 Note that when R 0 mn = 0 the two invariants — I3 and I3 — coincide. a separation of the conformal anomaly into type A and type B parts becomes ambiguous on a Ricci flat background.6))  I3 = 4I1 − I2 + O(Rmn ) + O ∇i J i . I2 (4. −I2 8 1 0 0 i and −5I3 .

so that we can ignore the integral of the Euler term E6 ) Z Z √ √ Atens. . A(2.0) 3 2 (4π) · 3 (5. 24 . since the N 3 contribution (5. = − (4π)3 · 32 d 6x √ −g0 (2I1 + I2 ). 25 Note that if we were comparing the full local conformal anomalies evaluated for R mn = 0 then.A.0) = − 4(N 3 − N) (4π)3 · 33 Z d 6x √ −g0 (2I1 + I2 ) + (4 − γ )NAtens.0) = − d 6 x −g0 (2I1 + I2 ).8) which follows directly from reduction of E8 (4.0) = γN (4π)3 · 32 Z 1 Atens. (5.5). Since we have found the O(N) correction to the anomaly of the (2.13) Thus if the true value of γ is 3 instead of the naive value 1 (4.A.7) to express I3 in terms of I1 and I2 . (5.0) = AN (2.4) only modulo Rmn -dependent and total derivative terms.8) have exactly the same form. A(2. ignoring possible F4 -dependent (F42 R 3 ) terms in the 11-d superinvariant I2 . Using (5. we would need γ = 3/4 in order to reproduce the non-zero Rmn = 0 value of the tensor multiplet anomaly (5. 0) theory in (5.0) (2. concluding that γ = 1 gives already the desired coefficient 4N 3 − 3N in (5.8) As was concluded in [9]. . 6 1 2 3 i (4π)3 · 32 · 25 4 245 (5. or scale anomalies (assuming that the d = 6 space has trivial topology. This N = 1 relation should be expected.0) reproduces the scale anomaly (5.11) of a single (2.8) by the N = 1 limit of the O(N ) term in (5. Z d 6x √ −g0 (2I1 + I2 ).0). (5.0) = d 6 x −g0 A(2. the Weyl-invariant (type B) parts of the leading (2.5) and the tensor multiplet anomaly (5. (4π)3 · 32 AN (2. we find from (5.10) (5.9) (5. 25 24 In the original version of the present paper we mistakenly used the basis of type B invariants including I of 3 [8] instead of the correct invariant of [9] and as a result got the O(N ) term with extra coefficient 3. Though we are unable to show that γ = 3 does follow from the d = 7 reduction of the 11-d superinvariant I2 containing P-odd C3 R 4 term. = −    7 1 i E + 8(12I + 3I − I ) + O ∇ J . The total scale anomaly of the (2.12).5). A(2.12) Equivalently.4).5) vanishes in this case. we are able to compare only type B anomalies.4) is then Z √ 4N 3 − γ N 3 N + A = − d 6 x −g0 (2I1 + I2 ). we find it remarkable that the required value of γ differs from the naive value 1 simply by factor of 3. then A(2. up to the overall factor 4N 3 in (5. 0) theory anomaly (5. 0) tensor multiplet. = d 6 x −g0 Atens. given that a similar correspondence is true for the R-symmetry anomalies [11] (see below).5).4) and (5. 0) theory following from (4.8) Z √ 4N 3 N3 A(2.3).11) (5. . Tseytlin / Nuclear Physics B 584 (2000) 233–250 Atens.

They satisfy the descent relations d(δΓ ) = δI7 .4) (cf.14) also as A(2.0) = −  N3 − N  E6 + 8(12I1 + 3I2 − I3 ) + O ∇i J i + NAtens. equivalently.A. 2 p2 (F ) = −   1 1 tr F¯ 4 − tr F¯ 2 ∧ tr F¯ 2 .246 A. 0) theory and free tensor multiplet theory.8)) 26 1 · 32 · 25      9 3 3 i × 4N − N E6 + 4N − 3N · 8(12I1 + 3I2 − I3 ) + O ∇i J ..14]   2 1 1 tens. R) = 4 3 · 24 with (here F 2 ≡ F ∧ F .17) I8 (F. that the value of the 7-d action (4. 4 2 i F. (5. Tseytlin / Nuclear Physics B 584 (2000) 233–250 Making a natural conjecture that the same relation Atens. and the claim of [59] that the coefficient of the type A (Euler) term in the anomaly of a generic effective theory is determined only by the value of the action on the AdS solution.43.2) is [11] 26 The shift of the coefficient of the E term in the conformal anomaly seems to imply a contradiction between 6 our assumption that the R 4 terms in the 11-d action (3.0) (5.0) = − (4π)3 or. one can make a prediction about the complete structure of the O(N) term in the (2. Atens.0) )N=1 should be true between the full expressions for the conformal anomalies of the (2. (5.0) = −   N E6 + O(Rmn ) + O ∇i J i . (5. p2 (F ) − p2 (R) + p1 (F ) − p1 (R) . The 1-loop effective action Γ for a free 6-d tensor multiplet in a background of 6-d Lorentz curvature R and SO(5) R-symmetry gauge field F has local SO(6) and SO(5) anomalies. = − 3 7 (4π) · 3 · 2 It is useful to compare the above expressions (5.) p1 (F ) = 1 ¯2 tr F .5) evaluated on the AdS7 solution is not changed). (5. (5.5). 0) theory and tensor multiplet. . I8 = dI7 . (5. A(2. (4π)3 · 3 · 27 (5.1) with the R 4 correction term (3.2) do not change the scale of AdS7 × S 4 solution (i. 0) theory describing multiple M5 branes derived (by assuming that the total M5-brane anomaly + inflow anomaly should cancel) from the 11-d supergravity action (3.18) F¯ = 2π The corresponding anomalies of the interacting (2.12).16) in agreement with the fact that for Rmn = 0 the conformal anomaly of the tensor multiplet becomes [9.44]   1 E6 + O ∇i J i .e.7). 0) theory anomaly A(2. etc.15) 3 2 3 (4π) · 3 · 2 Using (5. .14) with the previously known results for the R-symmetry anomalies of the interacting (2. with the 8-form anomaly polynomial I8 being [13. (5.14) 4 A(2. = (A(2. we can rewrite (5.

Klebanov. Maldacena. hep-th/9711200. 0) theory is the same as the anomaly of a single tensor multiplet.R. B 413 (1997) 41.A. R) = 1 3 · 24  247   2 1  2N 3 − N p2 (F ) − Np2 (R) + N p1 (F ) − p1 (R) . Frolov. Howe. Yu. This work was supported in part by the DOE grant DOE/ER/01545. A. Osborn.19) 4 Here the O(N 3 ) term comes [67] from the CS term in supergravity action (3. (5. K. JHEP 07 (1998) 023. Gubser. Manvelian.20) N 3 − N p2 (F ) + NI8tens. Intriligator. A. hep-th/0001041. 2 (1998) 231.2). [2] H.CLG 974965 and PPARC SPG grant PPA/G/S/1998/00613. hep-th/9804083. Tseytlin. Shifman for helpful discussions and comments. [6] S.R. R). Klebanov. EC TMR grant ERBFMRX-CT96-0045. Frolov for a collaboration at an initial stage and many useful discussions.4). Obukhov. Phys. Bastianelli. Skenderis. Tseytlin. Bastianelli. Tseytlin. Theor. Klebanov. Frolov. This is obviously consistent with the fact that Rsymmetry and conformal anomalies should be parts of the same 6-d supermultiplet. . K.A. (5. World-volume approach to absorption by non-dilatonic branes. The large N limit of superconformal field theories and supergravity. R. This is the same type of a relation we have established above (cf. (2. Nucl. [3] G. (3. We would like also to acknowledge J.A. with the crucial O(N) contribution coming from the P-even E8 part of the superinvariant I2 (3. A. String theory and classical absorption by threebranes. hep-th/0003017.A. Phys. S. Nucl. Nucl. Conformal anomaly of (2. Lett. H. Phys. [8] M. [10] I. Liu. hepth/9806087. Rev. NATO grant PST. Nucl.C. S. [9] F.0) I8 (F. I. Petkou and M. (F. Phys. 0) tensor multiplet in d = 6 AdS/CFT correspondence. hep-th/9901121. 3 3·2 Thus for N = 1 the anomaly of the (2. B 508 (1997) 245. 0) tensor multiplet in six dimensions and AdS/CFT correspondence. Harvey. D = 4 super Yang–Mills. [5] S. hep-th/9708005. B 496 (1997) 231.  1 (5. [4] F. Absorption by branes and Schwinger terms in the world volume theory. Petkou. B 499 (1997) 217. Three-point Green function of the stress-energy tensor in the AdS/CFT correspondence. and D = 4 conformal supergravity. 99-0590.S. Metsaev. Three-point correlators of stress tensors in maximallysupersymmetric conformal theories in d = 3 and d = 6. S. R) = Acknowledgements We are grateful to S. D 60 (1999) 026004.1) and the O(N) term [12. A. Tseytlin. INTAS grant No. Phys. hep-th/9703040.R.S. JHEP 02 (2000) 013. D = 5 gauged supergravity. Arutyunov.14] — from the P-odd C3 R 4 part of the superinvariant I2 (3. [7] R. Adv. I. Henningson.A. hepth/9911135.A. B 533 (1998) 88. Math. Phys. P. References [1] J. Phys.4). Frolov. The holographic Weyl anomaly. A note on R-currents and trace anomalies in the (2. T. hep-th/9702076. Tseytlin / Nuclear Physics B 584 (2000) 233–250 (2.13)) for the scale anomalies. Equivalently.0) I8 (F. A. Gubser.

Anomaly cancelling terms from the elliptic genus. Howe. Green. Lerche. R. strings and M-theory. M. Zumino. Nucl. [17] M. Phys. Four-loop divergences for the N = 1 supersymmetric nonlinear sigma model in two dimensions. Sigma model superstring corrections to the Einstein–Hilbert action. A. [18] D.E. M. More on sigma model beta functions and low-energy effective actions. Townsend. B 328 (1989) 439. B 238 (1984) 181. P. Phys. G. D. B 507 (1997) 571. [37] M. B 578 (2000) 139. B. S. [30] C. Narain. Park. Green.E. B. Nucl. [35] B. Interconnections between type II superstrings.K. Witten. G. P.A. B 408 (1997) 122. Antoniadis. Phys. Phys. Five-brane effective action in M-theory. Ferrara. Lett. van de Ven. Nucl. Higher order sigma model counterterms and the effective action for superstrings. Zanon. Pioline. Witten. [15] P. B 181 (1986) 63.J. Zanon. 22 (1997) 103. A.P. Alvarez-Gaume. q) string instantons. [19] N. C. [29] I.T. [28] E. Non-abelian tensor-multiplet anomalies. J. Rept. Phys. Vafa. Eleven-dimensional origin of string/string duality: A one-loop test. Nucl. Minasian. 2. B 198 (1982) 252. [34] R. B 177 (1986) 347. hep-th/9707013. A. JHEP 09 (1998) 004. Phys. B.B. Zanon.C. Grisaru. B 178 (1986) 199. Lett. West. Gross. A. Lett. B 467 (1996) 383. B 277 (1986) 1. D-instantons.S. Phys. Pope. Tseytlin / Nuclear Physics B 584 (2000) 233–250 [11] J.E. Supersymmetry in six dimensions. Gravitational anomalies. Nucl. Supersymmetrical dual string theory. Liu. [27] J.F.J. Zanon. Phys. Phys.S. Freeman. Y. Superstring modifications of Einstein’s equations. A. Warner. hep-th/9704145.T. Tseytlin. Phys.B. Sohnius. [13] L. Vanhove. Phys. [22] M. Phys. E. Heterotic–type I superstring duality and low-energy effective actions. [12] M. Nucl. D. B 452 (1995) 261. Rev. Witten. de Roo. On R 4 threshold corrections in type IIB string theory and (p.N. Phys.D. B 299 (1988) 91. [20] M. Phys. Stelle.248 A. K. Phys. Nucl. Nucl.N. Russo. B 173 (1986) 423. Lett. van de Ven. The quartic effective action of the heterotic string and supersymmetry. Phys. Phys. K.E. Nucl.T.T. Schwarz. J. Phys. Nucl. Tollsten. D. One loop amplitudes and effective action in superstring theories. .H.E. R 4 couplings in M. Sierra. hep-th/9512081.B. hepth/9610234. hep-th/9707018. Nucl. Lett. Nilsson. Phys. [36] P. Vertices and trees. D. Howe.A. Grisaru. Nucl. Harvey. B 234 (1984) 269. B 287 (1987) 457. P. Phys. M theory and N = 4 Yang–Mills.S. hepth/9505053. Phys. Nilsson. [23] M. Strings and superspace. Geom. Duff. Tseytlin.A. J. Green. Kallosh. [14] E. Supersymmetrization of ζ (3)R 4 in superstring theories. B 508 (1997) 509. Minasian. 137 (1986) 109. N.and type II theories on Calabi–Yau spaces. Bergshoeff. Phys. Kiritsis. hep-th/9903124. A. Nilsson.S. [24] M. R. One-loop four graviton amplitude in eleven-dimensional supergravity. D 35 (1987) 4038. hep-th/9707134. E. Nucl. [32] W. D = 10 supergravity.N. A one-loop test of string duality.A. Nucl. B 221 (1983) 331. [21] B. hep-th/9506126. R. Nucl. Grisaru. hep-th/9808060. Phys. Witten. B 447 (1995) 261. [33] E. [25] M.A. Schellekens.K. B 277 (1986) 409. B 289 (1987) 609.G. Minasian. Phys. Lerche. Sakai. Phys. [31] W. Schellekens. E. Four loop beta function for the N = 1 and N = 2 supersymmetric nonlinear sigma model in two dimensions. The complete N = 2. Scripta T 15 (1987) 118. Gravity theories in more than four dimensions. Heterotic string loop calculation of the anomaly cancelling term. [26] Q. Tanii. Moore. Nucl. [16] A.

H. P. [54] K. Tseytlin. B 409 (1997) 177. Green. B 405 (1993) 326. Green. Partouche. hep-th/9808061. A. 17 (2000) 1129. JHEP 9805 (1998) 002. Odintsov. Supersymmetry constraints on type IIB supergravity. Phys.A. [48] A. . too). Lett. Brink. One-loop corrections to string theory effective actions. Supersymmetry and string effective actions. Kehagias. Non-perturbative effects in AdS5 × S 5 string theory and d = 4 SUSY Yang–Mills. Phys.D. PhD Thesis. Tree amplitudes and two-loop counterterms in D = 11 supergravity. On the conformal anomaly from higher derivative gravity in AdS/CFT correspondence. B 422 (1998) 109. [59] C. Grav. J. The supersymmetric effective action of the heterotic string in ten dimensions. S. Myers. B 227 (1983) 252. Odintsov. Formulation of eleven-dimensional supergravity in superspace. Phys. hep-th/9706175. [41] M. A. Blau. Nucl. H. World Scientific. [49] R. hep-th/9804170. B 280 (1992) 39. Pilch. A. hep-th/9903033. hep-th/9805041. Diff. [53] L. D 59 (1999) 046006. Tseytlin. Conformal invariants. van Nieuwenhuizen. A. Rev. Pasti.A. Tseytlin. Nucl. Lett.S. Subleading corrections and central charges in the AdS/CFT correspondence. A. hep-th/0002241. Phys. Phys. Diffeomorphisms and holographic anomalies. Gubser. Nucl. hep-th/9604089. Tseytlin. Present state of quantum supergravity.R. [52] E. B 309 (1993) 279. [39] M. 1994. Parker. de Roo. Markov et al. Phys. Yankelowicz. hep-th/9210099. hep-th/9904179. Anselmi. Phys. Geom. Phys. S. Klebanov. Metsaev. Entropy of near-extremal black p-branes. Lett. Schwimmer. 1984. October 23–25. C. M. Vanhove.A. Groningen. [58] M. Moscow. [47] A. Suelmann. Ferrara. Cremmer. Graham. B 475 (1996) 164. [46] D.D. [57] I. P. [40] J.R. S. Nojiri.R. [62] L. hep-th/9910267. [65] T. A. I. hep-th/9710023. [63] S. Kehagias. D. H. The exact quartic effective action for the type IIB superstring.A. Ambiguity in the effective action in string theories.A. E. hep-th/9812092. Lett. B 298 (1988) 109. P. Compactification of D = 11 supergravity on S 4 (or 11 = 7 + 4. Phys. Weyl cocycles. p. Grav. Schwimmer. Bonora. Fefferman. Tseytlin. M. Phys. [43] E.S. [56] S. hep-th/9910113. Fradkin. hors serie 95 (1985). Narain.B. S. Rajaraman. [45] M.K.B. Wiedemann. Eleven-dimensional supergravity on the mass-shell in superspace. Invariants of conformal Laplacians. A. Phys. Phys. Tseytlin. D 58 (1998) 125003. Quant. [51] S. K. Gava.S. B 289 (1987) 701. Coupling constant dependence in the thermodynamics of N = 4 supersymmetric Yang–Mills theory. Quantum properties of higher-dimensional and dimensionally reduced supersymmetric theories. of Third Seminar on Quantum Gravity. A. 303. Phys. [60] S.B. hepth/9805156. Deser. Proc. M. Theisen.A. S. Phys. Lett. [61] S. Imbimbo. Rev. Nojiri. S. A. [64] C. Wiedemann. Phys. On subleading contributions to the AdS/CFT trace anomaly. Sethi. One-loop in eleven dimensions. [42] R. B 176 (1986) 92. Class. Nucl. Geometric classification of conformal anomalies in arbitrary dimensions. JHEP 9909 (1999) 018. Phys. Howe. Deser.A. B 534 (1998) 202. 3 (1986) 635. Quant. B 91 (1980) 384. [44] E. Superstring gravity and black holes. Lett. [55] R. B 91 (1980) 61. Gutperle.A. hep-th/9302047.S. S. H. de Roo. Asterisque. [50] T. P. Townsend. Kallosh. (Eds. Tseytlin / Nuclear Physics B 584 (2000) 233–250 249 [38] M. B 242 (1984) 377. A. in: M. A. Banks. Seminara. Supersymmetric R 4 actions in ten dimensions. Weyl anomaly from Weyl gravity. B 455 (1999) 155. Vacua of M-theory and string theory. Suelmann. Nucl. P.). Nucl. Class. Suelmann. 25 (1987) 199. Bregola. Nucl. Lett. 1985. Fradkin. Green.A. Rosenberg. Lett. Klebanov. Phys.

Adv. Freedman.T. Grav. Minasian. Fradkin. hep-th/9803205.A. Johansen. Phys. hep-th/9704108. B 134 (1984) 187. Conformal anomaly in weyl theory and anomaly free superconformal theories. Moore. hep-th/9708042.250 A. 2 (1998) 601. Class. [68] E. Erdmenger.A. Nonperturbative formulas for central functions of supersymmetric gauge theories. Conformally covariant differential operators: properties and applications. Tseytlin / Nuclear Physics B 584 (2000) 233–250 [66] J. R.Z. Lett. J. G. 14 (1997) 2061. Grisaru.S. B 526 (1998) 543. [67] D.A. Theor. Tseytlin. A. [69] D. Harvey. Gravitational anomaly cancellation for M-theory fivebranes. Phys. Phys. Anselmi. Nucl. M. D. Math. A. Freed. . Quant.A.

PII: S 0 5 5 0 . Italy c Mehta Research Institute of Mathematics and Mathematical Physics Chhatnag Road. PACS: 11.25.uk 1 panda@mri. accepted 13 June 2000 Abstract We investigate the classical geometry generated by a stable non-BPS D-particle. University of Cambridge. UK b The Abdus Salam International Centre for Theoretical Physics Strada Costiera 11.cam. Jhoosi Allahabad 211019. CB3 0WA Cambridge.in 2 See [4–6] and references therein. All rights reserved.nl/locate/npe The spacetime life of a non-BPS D-particle Eduardo Eyras a. 2 For a single D-brane the regimes of validity of these two descriptions are complementary. we recover the noforce condition at the critical radius and the v 2 behaviour of the probe. Sudhakar Panda b.Sq Keywords: Branes.elsevier. By using the non-BPS D-particle as a probe of the background geometry of another non-BPS D-particle. and at the critical radius we find a BPS-like behaviour up to v 4 corrections for long and short distances. assuming that the no-force persists for the complete geometry we derive part of the classical solution for the non-BPS D-particle. Moreover. Wilberforce Road.1 a Department of Applied Mathematics and Theoretical Physics. All rights reserved. On the one hand as a conformal field theory described by open strings with Dirichlet boundary conditions [1–3]. Introduction BPS D-branes enjoy a double life.3 2 1 3 ( 0 0 ) 0 0 3 7 6 .∗ . Non-BPS 1. We consider the boundary state of a stable non-BPS D-particle in the covariant formalism in the type IIB theory orbifolded by (−1)FL I4 . Trieste.c. At short distances we find no force at order v 2 for any radius.25.Nuclear Physics B 584 (2000) 251–283 www.eyras@damtp. India Received 30 March 2000. and on the other hand as classical solitons of supergravity. I-34014.V. 11.V. We calculate the scattering amplitude between two D-particles in the noncompact and compact orbifold and analyse the short and long distance behaviour. Projecting the boundary state on the massless states of the orbifold closed string spectrum we obtain the large distance behaviour of the classical solution describing this non-BPS D-particle in the non-compact and compact cases.  2000 Elsevier Science B. 0550-3213/00/$ – see front matter  2000 Elsevier Science B.ac.ernet.X . The conformal field theory description is valid at ∗ Corresponding author. E-mail: e.-w.

These branes have an exact conformal field theory description. these solutions are unstable and probably do not remain valid classically. In order to find the appropriate conditions for constructing a classical solution for a nonBPS D-brane we must find out which properties of the D-branes are such that (1) they assure the validity of a classical supergravity description and (2) they are not necessarily related to fractional supersymmetry. However. There it was shown that at a particular critical radius of the compact orbifold the non-BPS branes develop a Bose–Fermi degeneracy at one loop. Eyras.24–31]. . On the other hand. D-branes can be studied in more general situations. for a large number of branes N the system can be well described by a classical solution at weak coupling [7]. which is the same as having a no-force condition at all distances [6. 3 See [21] and references therein. Stable non-BPS D-branes have been found in different theories [11–16. Another key property is the fact that we can superpose an arbitrary number of parallel D-branes. On the other hand. it seems natural that in order to find a classical solution for a non-BPS D-brane. These properties certainly comprise stability. The critical radius is in fact the value beyond which the non-BPS D-brane becomes unstable and can decay into a pair of BPS D-branes. which ensures that the state survives at strong coupling. whose consistency conditions do not rely on spacetime supersymmetry [11–14. for instance. for which spacetime supersymmetry is not preserved. whereas the supergravity solution corresponds to a strongly interacting gravitational system. S. it is natural to wonder whether the non-BPS D-branes also enjoy a double life and have a description in terms of a classical solution of a certain effective (super)gravity theory. On the other hand. the entropy calculation of black-holes using D-branes [8. this brane should be stable and enjoy a no-force property. which can exist independently of supersymmetry. It has been recently suggested in [20] that the unstable non-BPS branes of type II theories might have a gravitational counterpart described by a gravitational sphaleron. This is a consequence of the BPS property. Moreover. Having such a precise description for the non-BPS D-branes in the conformal field theory side.11]. only non-BPS D-branes in a certain orbifold of type II theories are known to enjoy both stability and the no-force property [33].23]. Therefore. The two main properties of the BPS D-branes that assure this consistent dual behaviour are the fact that BPS-branes preserve a fraction of supersymmetry.22.9]. hence they are stable under variations of the string coupling. However. and the fact that one can consider a superposition of a large number of D-branes in such a way that they still preserve a fraction of supersymmetry.252 E. 3 These are unstable solutions with finite energy which interpolate between two (possibly distinguishable) vacuum configurations. being the lightest object carrying a certain quantum number [10. for instance. which corresponds to strong coupling.16]. Panda / Nuclear Physics B 584 (2000) 251–283 weak coupling. we expect that consistent classical solutions will be related to non-BPS branes which have some stability properties similar to the BPS D-branes. This is the case of the non-BPS D-branes [10–19]. stability can be based on grounds different from supersymmetry. These properties make possible.

We analyse the interaction potential between two non-BPS D-particles in relative motion. Remarkably. our assumptions will be only acceptable for distances much larger than the string scale. In fact. the static force is moreover vanishing. 6 The same results were obtained earlier in [43] using different techniques. we find no matching between the v4 terms in the open and closed string description. it has been recently proposed in [46] that even at 1-loop in the open string theory. charged under the U (1) field of the D5–O5 system [11]. 5 For a recent review see [37]. On the other hand. is charged electrically under a twisted R–R 1-form field and it is the strong-weak coupling dual of a non-BPS state in the orientifold Ω · I4 of type IIB. In this paper we implement the same technique to obtain the long distance behaviour of the non-BPS D-particle geometry. at a critical radius of the compact orbifold this D-particle meets all the requirements suggesting the existence of a classical solution. S. Moreover. there might be many different possible geometries with the same asymptotic structure. the no-force is removed in favour of another vacuum configuration in which the branes attract each other. That is. Eyras. open strings loops might spoil this property. Unlike for BPS branes. A BPS Dp-brane is described by a boundary state with two parts. and it was shown that the asymptotic behaviour of the corresponding classical solution is precisely recovered. although the Bose–Fermi degeneracy occurs at any distance between the non-BPS D-branes [33]. the NS–NS part and the R–R part. the long distance behaviour of the classical massless fields generated by the D-brane was computed in 6 [44. For these reasons.45]. as it happens for BPS branes [23]. and of the R–R massless fields (R–R (p + 1)-form potential). . up to v 4 corrections. respectively. They generate the asymptotic behaviour of NS–NS massless fields (metric and dilaton). in principle. In order to restrict these geometries we will assume that the no force condition also takes place when one consider the complete geometry at the critical radius. we take into account the extra pieces that one would need to add to the asymptotic behaviour in order to recover the full form of metric.E. at short distances. The coupling of the non-BPS D-particle to a twisted R–R vector field is the origin of the stability of this non-BPS D-particle [12]. under the 4 The gravity duals of BPS branes in orbifolds have been considered before in [34–36]. hence the non-BPS D-particle presents a BPS-like behaviour. Moreover. which is also the range of validity of the classical solution for a BPS D-brane [47]. we find no force at order v 2 as for BPS D-branes. One difficulty about recovering the full solution for the non-BPS D-particle from its asymptotic form is that. This is given by the asymptotic form of a metric and a dilaton propagating in the bulk. and a twisted R–R 1-form potential propagating in the orbifold fixed plane. We use the technique of the boundary state in the covariant formalism 5 [38–42] to describe the non-BPS D-particle. Using the non-BPS D-particle as a probe in the background of another non-BPS D-particle we recover the no-force behaviour at the critical radius. This D-particle is a truncated D-brane [30]. Using the boundary state for BPS D-branes. Panda / Nuclear Physics B 584 (2000) 251–283 253 In this article we study the particular case of a stable non-BPS D-particle in type IIB string theory 4 orbifolded by (−1)FL I4 . At the critical radius.

Extending this property for the complete geometry of the non-BPS D-particle source. ds10. for the non-compact and compact cases. n < −3 denoted by · · · which remain to be determined. since it is only there where one can consider a superposition of a large number of D-particles N .E where we do the split µ = (0. 2π α gmn (y) = 1 + 2aΩ4 |y|3 The expressions are given in terms of a parameter a. Moreover. 2π α g00 (y) = − 1 + 2aΩ4 |y|3  1a 6  κ6 T0 2 0 −1 1 + · · · δmn . and the explicit form of the complete covariant boundary state for the D-particle in Appendix B. In Section 4 we recover the no-force property at the critical radius and derive part of the complete classical solution. S. C0 = 1 + 3 4aΩ4 |y| Here T0 is the tension of the D-particle and is related to its charge Q0 by T0 = Q0 π 2 α 0 . m. In Section 3 we evaluate the asymptotic behaviour of the massless fields excited by the non-BPS D-particle. by using the assumptions mentioned above. For completeness.254 E. By using the assumptions mentioned above we are able to find the form of two of the components of the metric at the critical radius:  − 7 a 6  κ6 T0 2 0 −1 1 + · · · . . we also include the derivation of the zero-mode of the twisted R–R boundary state in Appendix A. 2π 2 α 0 eφ = 1 + 2aΩ4 |y|3  − 4 a 3 κ6 Q0 1 (1) + · · · − 1. We expect that the solution is consistent classically at the critical radius only. Panda / Nuclear Physics B 584 (2000) 251–283 assumptions presented above we are able to obtain part of the complete metric. hence at the critical radius the fields above are given in terms of a single function. dilaton and twisted R–R 1-form generated by the D-particle. Eyras. This article is organised as follows. agreeing with the vanishing of the v2 in the interaction potential. We find a diagonal metric with SO(5) × SO(4) symmetry (in Einstein frame): 2 = g00 (y) dt 2 + gmn (y) dy m dy n + gij (y) dx i dx j . and I4 acts on the i directions. We also evaluate the amplitude for D-particles in relative motion and analyse the long and short distance behaviour of the interaction. the D-particle cannot probe the precise geometry in these directions. we are able to find more properties of the classical geometry. for the non-compact and compact cases. In Section 2 we carry out the construction of the covariant boundary state for the non-BPS D-particle in the orbifold of type IIB generated by (−1)FL I4 . The velocity dependent part of the brane action multiplies a flat metric. We find no expression for gij since. and some possible extra dependence in |y|n . i) according to the orbifold symmetry. the form of the dilaton and twisted R–R potential at the critical radius are found to be:  a −1 1 κ6 T0 + · · · . as will be explained in Section 4.

r ∈ Z + 1/2. n ∈ Z. On the other hand. 9. is longitudinal to the fixed plane.1) twisted R: αri . α = 0. . The operator FL is the spacetime fermion number of the left-sector. 7. This has been carried out in the light-cone gauge in [12]. 8. In the non-compact case.e. the twisted groundstate will be given by these zero-modes. Panda / Nuclear Physics B 584 (2000) 251–283 255 2. and Xi . The closed string spectrum of the orbifold theory consists of an untwisted sector. i = 6. Eyras. given by type IIB states which are invariant under (−1)FL I4 .E. or equivalently 4 scalars under SO(1. . 1) supersymmetry in 6 dimensions. 9. . αn . transverse. . ψni . which is stable but non-BPS [10]. It corresponds to the S-dual of a massive particle-like state in the system of a D5-brane on top of an orientifold 5-plane. (2. n ∈ Z. hence (−1)FL changes the sign of the of the R–R groundstate. . If we split the coordinates as Xµ = (Xα . Let us consider first the non-compact theory. 9) symmetry down to SO(1. the orbifold contains a fixed plane at x 6 = x 7 = x 8 = x 9 = 0. . In fact. . 5). . the supersymmetries are generated by two Weyl supercharges of different chirality. i = 6. the oscillators of the twisted sector have the following modding:  α  α ψr . αn . i. the massless twisted sector can be identified with the worldvolume degrees of freedom of a NS-5 brane of type IIB [11. .48. the twisted R–R sector gives rise to a vector under SO(1. without having any action on the oscillators. The operator I4 corresponds to a reflection in the directions x i . . hence the orbifold breaks the SO(1. type IIB in Minkowski space orbifolded by (−1)FL I4 . The D-particle boundary state is made up of an untwisted NS–NS part and a twisted R–R part 7 [12]: 7 We use the subindex NS and R in the boundary states for short. The D-particle boundary state In this section we construct the boundary state for the stable non-BPS D-particle in type IIB orbifolded by (−1)FL I4 . where Xα . r ∈ Z + 1/2. 2. 7. 9. which can be identified with the U (1) gauge field in the worldvolume of a NS-5 brane on top of the orbifold fixed plane. and a twisted sector localised on the orbifold fixed plane. This corresponds to (1. Stable non-BPS D-particle in IIB/(−1)FL I4 The stable non-BPS D-particle of type IIB string theory orbifolded by (−1)FL I4 was described in [10–12]. Similarly. twisted NS: αri .49].. 5. the stability of the D-particle is due to its charge under a twisted R–R 1-form. 8. n ∈ Z. S. Here we use the covariant formalism for the boundary state. Xi ). . 1. ψri . r ∈ Z + 1/2.1. From the point of view of the fixed plane. r ∈ Z + 1/2. 5). Its stability is due to the fact that this particle is the lightest state charged under the U (1) gauge field of the D5–O5 worldvolume theory. The twisted NS–NS sector gives a vector of SO(4). 5. which has not been implemented before for non-BPS D-branes in orbifolds. . . and 6 fermionic zero-modes in the R-sector in the directions α = 0. There are 4 fermionic zero-modes in the NS sector in the directions i = 6.  α  α ψn . Since the intercept vanish for both sectors. n ∈ Z. 5) × SO(4).

9. G=− m=1/2 ∞ X (γ−m βm + β−m γm ). .U . ηiNS. 2 where the GSO projector in the NS–NS sector is given by |D0iNS.U is very similar to the NS–NS boundary state for type II branes. Eyras. For the bosonic untwisted boundary state the most general conditions invariant under the orbifold symmetry are |D0.U takes the form [50]: T0 |D0X i |D0gh i |D0ψ . . For the right movers these operators are analogously defined. (2. There is a bosonic and a fermionic part. These are explicitly given in Appendix B. i p = 1. ηiR. ηiNS ).U − |D0.7) ∞ X 1 n=1 n # α−n · S · α˜ −n |k = 0i. .T is defined as the GSO invariant combination of boundary states |D0.256 E.T . ηiNS .5) m=1/2 and β. .8) Note that the D-particle position in the directions i = 6. from which we can deduce that |D0X i = δ (5) (qˆ − y ) δ p p (4) " (qˆ ) exp − i (2. . ηiNS. the only piece modified by the orbifold with respect to the type II case is the zero-mode part of |D0X i. The other pieces of the untwisted boundary state are not modified by the orbifold and are the same as for the type II case. since there are nine Dirichlet directions for the case of the D-particle we have 8 Sµν = −δµν . γ are the superghosts. is to be determined later. . Xp |τ =0 |D0X i = y p .6) 2 where T0 . −iNS. X |τ =0 |D0X i = 0.4) 4 with F and G the (worldsheet) fermion and superghost number operators. 5. The twisted R–R boundary state that we denote by |D0iR. ηiNS ).U = ∞ X F= ψ−m · ψm .U + |D0iR. Panda / Nuclear Physics B 584 (2000) 251–283 |D0i = |D0iNS. +iNS. ηiNS |D0sgh. i = 6. .3)  1 ˜ ˜ 1 − (−1)F +G 1 − (−1)F +G . +iNS. 9 is restricted by the orbifold symmetry to be on the fixed plane. (2. . .U = PGSO.U . (2.U |D0. The boundary state |D0.2) The NS–NS boundary state is defined as the GSO-invariant combination of boundary states |D0.T and is given by [12]: 8 We take the Minkowski metric to be mostly plus. ηiNS. (2. . The state |D0. and also a ghost and a superghost part (|D0gh i and |D0sgh .U = (2. . . Moreover. (2. respectively: PGSO. ηiNS. a constant related to the tension of the D-particle. S. .U = ∂τ X0 |τ =0 |D0X i = 0. (|D0X i and |D0ψ . with η = ±1. In fact. which turns out to be [12]:  1 |D0.

12) have zero-modes along the fixed plane only. hence the corresponding pieces have the same form as for a type II R–R boundary state.T |D0sgh. ψm 1 j  (2. Q0 6= T0 . The fermionic overlap conditions now read: β  α − iη S α β ψ˜ −m |D0ψ .12) 2 where Q0 is a normalisation factor related to the charge density of the brane and to be determined below. respectively: PGSO. ηiR .T + |D0. As a consequence (2. The (super)ghosts are not affected by the orbifold [51].T = 0. the twisted bosonic part takes the form: " ∞ # ∞ X1 X 1 i β j (5) p p α α δαβ α˜ −n + α δij α˜ −n |k = 0i. (2. (2.T is given by: Q0 |D0X iT |D0gh i|D0ψ . which are explicitly given in Appendix A. the state (2. +iR. m ∈ Z. ηi(0) |D0ψ . ηiR. ηiR.T . The twisted R–R boundary state |D0.14) ψri − iη S i j ψ˜ −r |D0ψ .12) have fermionic zero-modes on the fixed plane only.T = n=1 n=1/2 The fermions in the twisted R–R sector have integer modding along the fixed plane and half-integer modding along the orbifolded directions. Eyras. Here Ψ (and Ψ˜ ) represent the zero-mode parts of the GSO-projectors.T . G = −γ0 β0 − F j i δ ψ ψ−r ij r .T |D0. ηiR. ηiR.T = PGSO. unlike the BPS D-branes.T = 0. ηi(0) R.T = − exp −iη R.13) |D0X iT = δ (qˆ − y ) exp n −n n −n |D0. Panda / Nuclear Physics B 584 (2000) 251–283  1 |D0. ηiR. Accordingly.10) 4 with F and G the (worldsheet) fermion and superghost number operators in the twisted R-sector. +iR. (2.T = 257 (2.14) and is constructed upon the twisted R–R groundstate: . r ∈ Z + . 2 In this sector the GSO-operator is given by: |D0iR. Since in the twisted R–R sector the bosons have integer modding along the orbifold fixed plane and half-integer modding along the orbifolded directions.E.T is determined from the zero mode overlap conditions in The zero-mode part (2. −iR. 2 From them we can deduce: " # ∞ ∞ X X β j α i ψ−m δαβ ψ˜ −m − iη ψ−m δij ψ˜ −m |D0. m=1 m=1/2 (2.15) |D0. ηiR. Notice that since this D-particle is non-BPS.T = ∞ P β α η ψ ψ−m αβ m ∞ P (−1)r=1/2 (−1) = Ψ (−1)m=1 ∞ X (γ−m βm + β−m γm ). S. (2.9)  1 ˜ ˜ 1 + (−1)F +G 1 − (−1)F +G .11) m=1 and similarly for the right movers.

258 E. η2 iR.T hD0. η2 iR. to express the above closed string channel result in the open string channel. implicitly given in (2. with α0 Da = 4π Z |z|61 (2.11). Panda / Nuclear Physics B 584 (2000) 251–283 f |D0. Having constructed the covariant boundary state for the non-BPS D-particle. (2. ηi(0) R.T hD0. We use the same regularisation as (0) (0) in [50]. ηiR we give below the regularised result: (0) (0) R.T = lim ρ→1 (0) (0) 2F0 +2G0 |D0. the superghost zero modes produce infinite number of terms with any superghost number contribution. Making a change of variables according to |z| = e−π` and d2 z = −πe−2π` d` dφ we find: −9/2 V1 α 0 π 2π 2 α 0 hD0|D|D0i = 16 Z∞ y2 d` `−9/2 e− 2π α0 ` 0    4 −π` )f34 (e−π` ) f 8 (e−π` ) − f48 (e−π` ) 2 2 0 2 f2 (e .1 .T |D0sgh .17) d2 z L0 −a L˜ 0 −a z z¯ . we determine next its tension and charge using the open–closed string consistency condition for boundary states.16) where a and b are spinor indices of SO(1. Defining |D0. The interaction between two D-particles separated by a distance y in the fixed plane is given by the amplitude between two boundary states located at a relative distance y with respect to each other.18) where a = 1/2 in the untwisted NS–NS sector and a = 0 in the twisted R–R sector. Note that we have considered both closed string and open string to have same periodicity in the spatial direction of the worldsheet.T R. Open–closed string consistency then requires that the result must be equal to the 1-loop amplitude of the open strings stretched between the two D-particles. Also care must be taken in computing the matrix elements involving the zero modes of the superghosts and fermions in the twisted R–R sector.19) where F0 and G0 are the zero-mode parts of the operators F and G. η1 |D0. η1 |ρ = −4δη1 η2 . The explicit form of Mab and its derivation are given in Appendix A. We can make a worldsheet transformation. with the insertion of a closed string propagator hD0|D|D0i. (2. hence a regularisation is needed.T = |D0ψ .20) − (Q × (T0 )2 3 ) α ` 2π 0 f14 (e−π` )f44 (e−π` ) f18 (e−π` ) where V1 is the (infinite) length of the D-particle worldline and fi are functions of e−π` defined in the usual manner. However. Eyras. Moreover. The open string states on a non-BPS D-particle . 5). ηi(0) R. in general. |z|2 (2. This is because. (2. ` = 1/τ .T = Mab |aiT |biT . one needs to take the appropriate moddings in the expression for L0 and L˜ 0 in the twisted sector. ηiR. S. care must be taken to allow only states invariant under the orbifold projection to propagate in the loop [11].

23) T0 = 8(α 0 )3/2 π 7/2 . (2. Moreover. τ 3/2 f14 (e−πτ )f24 (e−πτ ) f18 (e−πτ ) (2. In fact. such that the 2 factor is compensated. √ 2 times bigger than the type IIA BPS D-particle. define an electric charge with respect to the (twisted) R–R field as √ we can √ µ0 = 2 Q0 = 8π 2πα 0 . open–closed string consistency allows us to fix the normalisation of the boundary state: √ Q0 = 8π πα 0 . the open string 1-loop amplitude is given by: Z∞ A = 2V1   dτ 1 TrNS−R 1 − (−1)F +G (1 + I4 )e−2πτ L0 2τ 4 0  1 1 + (−1)F +G (1 − I4 )e−2πτ L0 4   Z∞  −2πτ L dτ 1 F +G 0 TrNS−R 1 − (−1) I4 e . On the other hand. = 2V1 2τ 2  + (2. S.22) 0 As mentioned above. 10 The unstable non-BPS D-particle of type IIB has a tension . Accordingly.21) 0 where F and G are the (worldsheet) fermion and superghost number operators for the open string. charge. 9 This can be derived from a D0 anti-D0 system in type IIA following the prescription given in [16. their charges are different.E. We obtain the following 1-loop amplitude: −1/2 1 A = V1 8π 2 α 0 2   Z∞ f44 (e−πτ )f34 (e−πτ ) dτ − y 2 τ0 f38 (e−πτ ) − f28 (e−πτ ) 2π α − 4 × e . there is an extra factor 1/2 in the open string amplitude coming from the projection operator. In analogy with the BPS D-branes. 10 However.17]. The tension of the non-BPS D-particle is given by T0 and Q0 is related to its electric √ Using that in ten dimensions κ10 = 8π 7/2gs (α 0 )2 . we find the mass per unit volume of the D-particle: M0 = T0 1 = √ . The tachyon is projected out in the trace and this renders the non-BPS D-particle stable [12]. states with Chan–Paton factor σ1 have opposite GSO projection and are odd under I4 . the symmetry (−1)FL does not have any action on the open string oscillators.24) which agrees with the mass of the D-particle found in [11]. However. Eyras. and that for the orbifold κorb = 2 κ10 . in the orbifold √case. µ0 is exactly twice the charge of the BPS D-particle of type IIA in six dimensions. States with Chan–Paton factor 1 have usual GSO projection and are even under I4 . It is interesting to observe that the tension T0 of this non-BPS D-particle is the same as that of a BPS D-particle of type IIA theory in ten dimensions. Panda / Nuclear Physics B 584 (2000) 251–283 259 are labelled by Chan–Paton factors 9 1 and σ1 . κorb g 2α 0 (2.

i = 6. i. . The only piece modified in the boundary state with respect to the non-compact case is the bosonic part and the overall normalisation. The amplitude between two non-BPS D-particles in the closed string channel is given by: |D0. Using the worldsheet duality ` = 1/τ and the Poisson resummation formula r X − 2π (mR)2 X − π α0 ` n 2 2 e 2 R = R e α0 ` . 9: " ∞ # 9 X Y X1 i ni  i q ˆ α−n · S · α˜ −n |k = 0.25) e n Ri exp − |D0X i = δ (5) qˆ p − y p n i=6 ni ∈Z The states n=1 |ni i = exp(iqni Rnii )|n = 0i are normalised as hni |n0i i = Φi δni n0i . that we choose x i = 0. (2. The coordinates x i . The D-particle in the compact orbifold Let us consider now the boundary state of a non-BPS D-particle in the case of the compact orbifold. ηiNS |D0sgh . 7.27) We consider the new overall normalisation factor as a constant times the normalisation for the uncompactified case: T0 N |D0X i |D0gh i |D0ψ .26) where Φi is the self-dual volume [52] which satisfies lim Φi = 2πRi . Eyras. 8. ηiNS . . The bosonic part is modified to include Kaluza–Klein modes in the directions i = 6. Ri (2. for simplicity.260 E. The boundary state is now constructed upon a groundstate with zero winding and zero momentum. 0 `α n∈Z one obtains m∈Z (2. we recover the non-compact boundary state [53]. (2.. S. located at x i = 0. i = 6. type IIB on T 4 /(−1)FL I4 . 9. We consider a non-BPS D-particle located on one of the fixed planes. . 8. (2. These two modifications will only affect the NS–NS part of the boundary state.2. 9. Panda / Nuclear Physics B 584 (2000) 251–283 2. are periodic and there are 16 fixed planes instead of one. n = 0i.e. 7.28) 2 The factor N can be obtained by open–closed string consistency and is such that in the decompactification limit Ri → ∞. .30) .U = −5/2 V1 α 0 π 2π 2 α 0 hD0|D|D0i = 16 Z∞ 0 ( × (T0 ) N 2 2 9 Y i=6 − (Q0 )2 Φi y2 d` `−5/2 e− 2π α0 ` ! 9 X Y i=6 ni ∈Z f24 (e−π` )f34 (e−π` ) f14 (e−π` )f44 (e−π` ) 0 e − π α2 ` ) ni Ri 2 f38 (e−π` ) − f48 (e−π` ) f18 (e−π` ) . ηiNS. lim Φi = Ri →∞ Ri →0 2πα 0 .29) (2. πRi .

49].31) − (Q0 )2 44 f1 (e−πτ )f24 (e−πτ )  Open–closed string consistency imposes that this amplitude must be equal to the open string 1-loop amplitude for the compactified case Z∞ A = 2V1    1 dτ TrNS−R 1 + (−1)F I4 e−2πτ L0 2τ 2 0 −1/2 1 = V1 8π 2 α 0 2 × ( 9 Y X i=6 mi ∈Z Z∞ dτ − y 2 τ0 e 2π α τ 3/2 0 e τ − 2π (mi Ri )2 α0 f38 (e−πτ ) − f28 (e−πτ ) f18 (e−πτ ) −4 f44 (e−πτ )f34 (e−πτ ) f14 (e−πτ )f24 (e−πτ ) ) . i = 6. which is expected since the compactification is done in the transverse directions of the D-particle. 9. (2.24–29. (2. This also fixes the normalisation factor N to be: !−1/2 9 Y 2πRi Φi . we can recover the untwisted boundary state for the non-compact case. (2.E.35) Ri = 2 there are four states. Eyras.32) This gives the same value for Q0 as in the uncompactified case. 7. S. (2.33) N= i=6 It is easy to check that with this normalisation and using Φi = 2πRi . in the decompactified limit. Panda / Nuclear Physics B 584 (2000) 251–283 −5/2 V1 α 0 π hD0|D|D0i = 2π 2 α 0 16 ( Z∞ 261 dτ − y 2 τ0 e 2π α τ 3/2 0 ! 9 2 Y 9 Y X − 2π τ (m R )2 f 8 (e−πτ ) − f 8 (e−πτ ) 2 2 2 3 2 × (T0 ) N Ri Φi e α0 i i 0 8 (e−πτ ) α f 1 i=6 i=6 mi ∈Z ) f 4 (e−πτ )f34 (e−πτ ) . the tension depends on the compactification radii. The open strings on the D-particle have now winding modes:    9  X 1 wi Ri 2 1 2 + 0 N− . These are the modes of the tachyon field that at the critical radius correspond to the . which become massless [16. At this point we can recover the vanishing of the amplitude at the critical radius [33]. for which only one wi 6= 0. At a particular critical value of the radii r α0 .34) M = α0 α 2 i=6 The groundstate with zero winding is the tachyon and is projected out by the orbifold symmetry. 8. (2. Thus the net effect of the compactification is a renormalisation of the tension of the D-particle and as expected.

interacting with another non-BPS D-particle at rest. In the NS–NS untwisted sector. vi = 1 − v 2 " × exp − i=2 ∞ X n=1 # 1 α−n · S(v) · α˜ −n |k = 0i. We consider first the non-compact orbifold. (2. n where the matrix S(v) is given by (2. Eyras. 2. For definiteness we consider a D-particle moving in the direction X1 with a velocity v.38) which vanishes. by the abstruse identity and using the expressions for T0 and Q0 . This amplitude can be obtained by using a boosted boundary state [54] for the non-BPS D-particle. at any distance y. In the closed string channel this can be seen by using the relations: X 2 e−π`n = f1 (e−π` )f32 (e−π` ).37) i=6 Hence we can write the amplitude as −5/2 V1 α 0 π 2π 2 α 0 hD0|D|D0i = 16 Z∞ y2 d` `−5/2 e− 2π α0 ` 0 4 −π` ) f3 (e × 4 4 −π` −π` f1 (e )f2 (e )f44 (e−π` )  2   T0 8 −π` 8 −π` 2 8 −π` ) − f4 (e ) − (Q0 ) f2 (e ) .39) . f3 (e × π 2 α0 (2. Long and short distance interactions In this section we consider the interaction amplitude for non-BPS D-particles in relative motion.3. S. In the open string channel it is simpler to demonstrate this by using the above identities. Panda / Nuclear Physics B 584 (2000) 251–283 marginal deformation which takes the D-particle to a bound state of a D-string and an antiD-string [11]. N= 2π 2 α 0 (2. The boosted boundary state include some v-dependent modifications that we write explicitly below. This will give a velocity dependent potential which we will compare with the usual BPS case.262 E. the bosonic part of the boundary state becomes ! 9 p Y  p δ(qˆ ) δ qˆ 1 + v qˆ 0 |D0X .36) Moreover. At this critical radius a Bose–Fermi degeneracy takes place. n∈Z f4 (e−π` )f2 (e−π` )f3 (e−π` ) = √ 2. which translates to the fact that the 1-loop amplitude vanishes [33]. the normalisation factor N at the critical radius becomes !−1/2 9 Y 1 Φi .

(2.44) −∞ We can extract the long range interaction potential taking the limit ` → ∞ in the integrand of (2. Finally.43). (2.m. For the other pieces we do the same substitution S → S(v). in the following way: Z∞ hD0. n = 0i.42) r2 dt e− 2π α0 ` −∞ 0 6 −π` 6 −π` Θ3 (u. n  # α−n · S(v) · α˜ −n |k = 0. Similarly in the R–R twisted sector. and then performing the integral in `.40) 1 − v2 1 − v2 and S(v) = S for the other components. for the R–R zero-mode part there is as well a v-dependent modification: f |D0. The other pieces in the NS–NS untwisted boundary state have the usual form except for a substitution of the matrix S by S(v). i`)f12 (e−π` )f44 (e−π` ) (2. the boosted bosonic boundary state takes the form: S00 (v) = S11 (v) = − |D0X . the cylinder amplitude takes the following form: u=  2 0 −1/2 hD0.’ we indicate that one takes the corresponding twisted moddings of the twisted sector. Θ1 (u. i`)f3 (e ) − Θ4 (u. Eyras. vi(0) R. v|D|D0i = 8π α  × Z∞ sin πu d` ` − 29 Z∞ (2. For convenience we also define the following variable 1−v 1 ln . Panda / Nuclear Physics B 584 (2000) 251–283 263 1 + v2 2v . 2πi 1 + v Finally.41) and the matrix Mab (v) is given in Appendix A. i`)f4 (e ) − 4`2 6 −π` Θ1 (u. r(t)).E. i`)f1 (e ) Θ2 (u. S. We define the distance between the particles as r 2 = b2 + t 2 v 2 (1 − v 2 )−1 . i`)f22 (e−π` )f34 (e−π` )  . η. i`) are the Jacobi Θ-functions. v|D|D0i = dt U(v. viT = p 1 − v2 " 5 Y δ(qˆ p ) δ qˆ 1 + v qˆ 0 p=2 × exp − X1 t. For slow velocities.20). In order to study the interaction we define the interacting potential of the scattering U . (2. where with ‘t. S10 (v) = S01 (v) = − . b is the impact parameter: b2 = y22 + · · · + y52 .T = Mab (v)|aiT |biT . and we have used the value of the tension and charge of the D-particle.43) where Θs (u. where t is the proper time of the moving particle along which we also integrate.m. For v = 0 we recover the amplitude for the static interaction given in (2. we find the following expansion in powers of v: .

Note that the pieces originally from the NS–NS and R–R sectors multiply now in (2. iτ )f26 (e−πτ ) Θ1 (−iuτ. For BPS branes the v 4 term in the open string channel coincides with the v4 term from the closed string description [47]. as expected since the D-particle is non-BPS.23]. The first term is a linear atractive force coming from the stretched strings.47) which for slow velocities and after integrating in τ becomes (to order Uopen = 4 (2πα 0 )3 Γ (7/2) v 4 r + . As we will see below for the compact case.48) Surprisingly. Using the modular properties of the Θ-functions we can write (2. the velocity dependent corrections start at order v 2 .43) in the open string channel:  2 0 −1/2 hD0.46) We can derive the interaction potential at short distances by taking the limit τ → ∞ in the integrand and then performing the integral in τ . even though at the critical radius the static potential and v 2 terms will be suppressed.46) the same powers of τ . the v4 correction has precisely the same form as the v 4 term for a scattering of two BPS D-particles. Moreover. iτ )f36 (e−πτ ) − Θ2 (−iuτ. In this region the open string description dominates.264 E. iτ )f16 (e−πτ ) −4 Θ4 (−iuτ.45) to order v 4 . We can analyse this amplitude for very short distances. S. Hence the BPS-like behaviour will not extend beyond order v 2 . as for BPS D-branes. This time we obtain:  2 0 −1/2 Uopen = −i 8π α Z∞ sin πu r2 dτ τ −1/2 e− 2π α0 τ 2 + 2 cos(−2πiuτ ) − 8 cos(−iπuτ ) . Panda / Nuclear Physics B 584 (2000) 251–283 (2πα 0 )3 Uclosed ' (4π)1/2     1 1 1 Γ (3/2) Γ (7/2) 1 4 Γ (7/2) v × 8 1 + v2 + v4 − + . This is something we could expect from the fact that the particle breaks all supersymmetries. hence we can expect that the interaction potential will be very different from the closed string case. Moreover. . This does not occur in the non-compact orbifold. 2 24 r7 (2πα 0 )2 r 3 8 r7 (2. as for the potential between two BPS D-branes of different dimensionality [22. iτ )f42 (e−πτ )f34 (e−πτ ) Θ1 (−iuτ. iτ )f12 (e−πτ )f24 (e−πτ )  . there is no v 2 correction to the potential. 2πα 0 r7 (4π)1/2 v 4 ): (2. v|D|D0i = −i 8π α  × Z∞ sin πu dτ τ 0 − 12 Z∞ r2 dt e− 2π α0 τ −∞ Θ3 (−iuτ. Eyras. the matching of the v 4 terms will not occur either. There is a static force. sin(−iπuτ ) 0 (2. and as for the non-BPS D-particle of type I [55]. (2.

Panda / Nuclear Physics B 584 (2000) 251–283 265 We extend next the analysis of the cylinder amplitude for the case of relative motion of the D-particles in the compact orbifold.E.23].52) 1+ v + v Uclosed ' 3 r 2 6 4 8 i=6 We see that for generic radii the potential has v 2 corrections.49) |D0X . as for BPS D-branes [22. and the other pieces have the usual form except for a substitution of the matrix S by S(v). and the D-particle would not be stable. i`)f12 (e−π` )f44 (e−π` )  (2. S. those radii do not make the amplitude (2. as before. Notice that the static and v2 terms of the potential also vanishes for other values of the radii. i`)f16 (e−π` ) ) Θ2 (u. Eyras. However.29). the boosted bosonic boundary state and fermionic zero-mode have the same form as for the noncompact case.50) where the matrix S(v) is as given before in (2. On the other hand.51) For v = 0 we recover the amplitude for the static interaction given in (2.40). − Θ1 (u. which is typical for potentials between two BPS D-branes of different dimensionality [22. and occurs also for the non-BPS D-particle of type I [55]. interacting with another non-BPS D-particle at rest. i`)f22 (e−π` )f34 (e−π` ) . the cylinder amplitude takes the following form: r hD0. In the NS–NS untwisted sector. i`)f46 (e−π` ) Θ1 (u.29) vanish. n = 0i. i`)f36 (e−π` ) − Θ4 (u. vi = 1 − v 2 p=2 × 9 Y X e i iqˆ ni Rn i ! " exp − i=6 ni ∈Z ∞ X 1 n=1 n # α−n · S(v) · α˜ −n |k = 0. this would require some of the radii to be below the critical radius. In the R–R twisted sector. For slow velocities (u ' (iπ)−1 v) and after integration in `. the long range interaction potential to order v 4 is given by ! ) (  02 Y 9 4πα 0 v4 1 2 1 4 α −1 Ri − 1 − . (2. v|D|D0i = 2 sin πu π 2 α0 ( × × Z∞ d` ` 0 9 0 2 Y α 4  i=6 Ri−1 − 25 Z∞ r2 dt e− 2π α0 ` −∞ 9 Y X 0 e − π α2 ` ni Ri 2 ! i=6 ni ∈Z Θ3 (u. Using the same notation as before. at the critical radius they start at v 4 . the bosonic part of the boundary state becomes ! 5 p Y  δ(qˆ p ) δ qˆ 1 + v qˆ 0 (2. We consider again one non-BPS D-particle moving in the direction X1 with a velocity v.56]. The boosted boundary state include the v-dependent modifications on top of the modifications due to compactification. . Moreover. (2.

iτ )f12 (e−πτ )f24 (e−πτ ) )  Θ1 (−iuτ. the v 4 terms does not match with the closed string result. At order v 4 the open strings begin to describe the geometry of non-BPS D-particle very differently from the closed strings. Notice also that the r-dependence of the v 4 term is as for the uncompactified case. at the critical radius Uopen takes a very BPS-like form. as we announced before. v|D|D0i = −i 2 0 π α ( × −∞ 0 1 4 − 9 Y X e− 2π τ α0 (mi Ri )2 ! Θ3 (−iuτ. At the critical radius. we expect an equivalence between the (super)gravity and worldvolume description of the non-BPS D-particle. Carrying out the integral in τ we find: ! 9 q  4 (2πα 0 )3 Γ (7/2) v 4 1X 2 2 R 2 − α 0 /2 r + 4π + .266 E. (2.53) To extract the leading contribution at short distances and low velocities we take the limit τ → ∞ in the integrand. From this analysis we conclude that the long and short range interactions of the nonBPS D-particle are quite different for generic radii of the orbifold. Panda / Nuclear Physics B 584 (2000) 251–283 In order to study the short-distance behaviour we write the scattering amplitude in the open channel: r Z∞ Z∞ r2 2 − 12 sin πu dτ τ dt e− 2π α0 τ hD0.55) r − Uopen = i 2πα 0 4 r7 (4π)1/2 i=6 As for the non-compact case we find no v 2 corrections. to order v 2 . 3. the boundary state of a D-brane encodes the long distance behaviour of the corresponding classical solution of supergravity. However. Spacetime description As shown in [44]. In the particular case of the critical radius the static and v 2 terms of the interaction potential are absent in the open and closed string description. iτ )f42 (e−πτ )f34 (e−πτ ) Θ1 (−iuτ. this classical .54) i=6 At low velocities the expression simplify. The static part of the potential vanishes when all the radii are equal to the critical radius. and the v 4 corrections are again the same as for the BPS D-particle in ten dimensions. At large distances. S. iτ )f16 (e−πτ ) . iτ )f26 (e−πτ ) i=6 mi ∈Z Θ4 (−iuτ. (2. which are relevant very close to the critical radius Ri > Rc : r Z∞ r2 2 sin πu dτ τ −1/2 e− 2π α0 τ Uopen = 2 0 π α 2i sin(−iπuτ ) 0 1 X − 2π R0 i e α × cos(2πiuτ ) − 4 cos(iπuτ ) + 1 + 2 9 2 ! τ +πτ . This time one must also include contributions with one unit of winding number mi = 1. (2. Eyras. Remarkably. iτ )f36 (e−πτ ) − Θ2 (−iuτ.

−1) picture. This implies that string states which depend on the momentum in the orbifolded directions are only invariant if they are symmetric with respect to I4 : k i → −k i . only those which are invariant under the orbifold symmetry will appear in the theory. In the untwisted sector. (3. The non-compact case Given a certain massless closed string state |ϕi. the D-particle will excite gravitons gµν and dilatons φ in the untwisted sector and a R–R 1-form field C (1) in the twisted sector. we use the following notation for the NS–NS groundstate in the (−1. we also find a twisted R–R 1-form which is restricted to the fixed plane and whose dependence on the spatial coordinates on the fixed plane is the correct one for a particle in 6 dimensions.45]: hP(ϕ) |D|Di. The projection operators in the untwisted NS–NS sector are given in the (−1. −1) picture: ] −1 . these will include gravitons that propagate off the orbifold fixed plane which are symmetric in k i . C (1) ' κ6 2 δC (1) . In this section we implement this technique on the stable non-BPS D-particle to obtain the asymptotic form of the solution. The fluctuations δhµν . At large distances. generated by a D-brane is determined by computing the overlap of the boundary state with the state |P(ϕ) i with the insertion of a closed string propagator [44. (3. normalised as hϕ|ϕi = ϕ 2 . The fluctuations around this background is exactly reproduced by the boundary state. |ki ≡ |k/2i−1 |k/2i The projectors for the dilaton and the metric are given by: 11 We use the same normalisation as in [44]. In particular. From the NS–NS part one obtains the asymptotic behaviour of the dilaton and the metric.1) using the state |D0i constructed previously. 8. and from the R–R part one obtains the asymptotic behaviour of the R–R form potential under which the brane is charged. and whose dependence is the expected one for a particle in 10 dimensions. (3. We consider first the case of the uncompactified orbifold. The asymptotic behaviour of the classical field ϕ. are given by (3. i = 6. 9. For simplicity. We obtain a metric and a dilaton in the bulk.3) . S.2) gµν ' ηµν + 2κ10 δhµν .1. one can define a projection operator hP(ϕ) | associated to this state. which depend on all the coordinates transverse to the D-particle.E. such that hP(ϕ) |ϕi = ϕ [45]. 3. Eyras.1) Since the D-particle has an untwisted NS–NS part and a twisted R–R part. 7. We describe first the non-compact case. On the other hand. δφ and δC (1) . these fields will be of the form of a fluctuation around a flat background 11 : √ √ φ ' κ10 2 δφ. Panda / Nuclear Physics B 584 (2000) 251–283 267 solution tends to a flat background configuration.

k i | + hk⊥ . 13 For simplicity we omit the subindex T of the twisted groundstate.6) (3. diag . 4 2  ν  1 µν µ µ ν ψ˜ 1/2 + ψ1/2 ψ˜ 1/2 hP(h) | = hk⊥ . −3/2) picture associated to the twisted R–R 1-form field 13  1 ^ k/2i−3/2 .5) Π± = (18 ± γ ). Eyras. In the R–R sector. −3/2). with k⊥ the spatial components of the momentum along the fixed plane directions.4) where k = (k⊥ . 5). We find the following results for the asymptotic fields in momentum space: 3 V1 δφ(k) = hP(φ) |Da=1/2|D0iNS. where the matrix γ is defined in Appendix A. where the R–R vertex operator couples to the potential instead of the field strength [50. . 2  × (Cγ α Π− )ab cos β0 γ˜0 + (Cγ α Π+ )ab sin β0 γ˜0 such that hP(C)α |C (1) i = Cα(1) . . Moreover. . 8 k   1 T0 V 1 7 1 . S.U = 2 k2 4 4 4 12 See [44] for an explanation about this. The form of this projector is similar to the case of the type IIA BPS D-particle. 2  × (Cγ α Π− )ab cos β0 γ˜0 + (Cγ α Π+ )ab sin β0 γ˜0 . Panda / Nuclear Physics B 584 (2000) 251–283  ν 1 µ hP(φ) | = √ hk⊥ . with 1 (3. with the difference that the twisted R–R groundstate carries spinor indices of SO(1. 5) spinor indices. k i ). −k i | ψ1/2 4 1 − hP(φ) | √ (ηµν − k µ `ν − k ν `µ ). . b as the SO(1. 2 2 (3. The NS–NS (R–R) fields have non-zero overlap with the NS–NS (R–R) boundary state only. k⊥ /2| hP(C) α | = − √ −1/2 hb. The corresponding projector is given by 1 ^ k⊥ /2|−3/2 ha. δhµν (k) = hP(h) µν |Da=1/2|D0iNS.U = T0 √ 2 . 57]. |C (1) i = √ Cα(1) (Cγ α Π+ )ab cos γ0 β˜0 + (Cγ α Π− )ab sin γ0 β˜0 |a.7) . we write the following quantum state in the (−1/2.268 E. k/2| hC (1) | = −Cα(1) √ −1/2 hb. . 2 the chirality projector in 6 dimensions. ` is such that `2 = 0 and k · ` = 1. k/2i−1/2 |b. −k i | ψ1/2 ψ˜ 1/2 (ηµν − kµ `ν − kν `µ ). 2 whose conjugate state is given by: 1 ^ k/2|−3/2 ha. This (1) state is normalised as hC (1) |C (1) i = Cα C (1) α . (3. k i | + hk⊥ . the most natural picture is the asymmetric 12 picture (−1/2. In analogy to the type II case described in [50] and using a.

however. . although the D-particle is stuck on the fixed plane. since the momentum dependence is restricted to the fixed plane spatial directions. √ 7 |x| 14 2 Ω8   1 T0 7 1 1 diag . δhµν (x) = 14 Ω8 |x|7 4 4 4 δφ(x) = (3. and Ω4 is the volume of a unit 4-sphere surrounding the particle inside the fixed plane. with d − 1 non-zero components. 2 k⊥ (3. Thus. Panda / Nuclear Physics B 584 (2000) 251–283 269 For the twisted R–R field we only have a contribution along the worldline direction of the D-particle. .. namely: Q0 V1 δC0(1) (k⊥ ) = hP(C) 0 |Da=0 |D0iR.2. we have d = 6. i. longitudinal to the fixed plane.10) In the R–R sector. For the NS–NS sector we have d = 9. Consider one of the compact orbifolded directions: x9 ∼ x9 + 2πR9 . For a generic momentum K. δC (1) (y) = − √ 3 |y| 3 2 Ω4 (3. one should make sense of a superposition of them. The compact case In order to consider a spacetime description for a stable non-BPS D-brane. . which turns out to be possible only in the compact orbifold at the critical radius. . type IIB on T 4 /(−1)FL I4 . We derive first the asymptotic form of the metric and dilaton in the compactified case using an approximation in the background fields. . This is a consequence of the fact that the D-particle can emit massless untwisted fields off the fixed plane which are symmetrised in the momenta along the orbifolded directions. that we will compare with the boundary state calculation.T = − √ 2 . that the untwisted fields do see the entire spacetime.e. hence we find the following asymptotic behaviour for the metric and dilaton: 1 3T0 . . (3. In this section we extend the analysis of the metric and dilaton of the D-particle to the case in which the space transverse to the orbifold fixed plane is a 4-torus.8) Note that δC0(1) depends only on the momenta transverse to the D-particle.E. the associated metric and dilaton background solution in the uncompactified case will extend also in the orbifolded directions. . 3. Eyras. S. in a regime where it remains valid classically. We see.9) dt d(d−1)x (d − 3) Ωd−2 |x|(d−3) K 2 where x are d − 1 spatial coordinates transverse to the D-particle and Ωd−2 is the area of a unit sphere surrounding the D-particle. we have: Z eiK·x V1 = . and the asymptotic form of the R–R 1-form is found to be: Q0 1 .11) where now y indicates the spatial directions in the fixed plane only. We now make use of the following Fourier transformation in order to translate the results into position space. Note that the power of |y| is exactly the power expected for a particle in 6 dimensions.

. √ √ ¯ φ ' κ10 2 δφ ≡ κ6 2 δ φ.18) 1/2 .12) 7 2 |x| (r + (x9 − 2πnR9 )2 )7/2 n∈Z with |x|2 = r 2 + x92 . Eyras. 6 Ω4 4 4 4 |y| (3. these integrals satisfy the properties nIn = (n − 1)In−2 . Within this approximation we obtain the following asymptotic behaviour for the dilaton and metric for the compactified case: 9 Y 1 T0 (2πRi )−1 3 .17) i=6 We rewrite the fluctuations using the 6-dimensional gravitational constant gµν ' ηµν + 2κ10 δhµν ≡ ηµν + 2κ6 δ h¯ µν .e. . δφ(y) ' √ |y| 2 2 Ω4 i=6 δhµν (y) '   9 1 T0 Y 7 1 1 . . Thus we can write: X 1 1 ' . (3.16) i=6 with y now indicating the spatial directions in the fixed plane. S. we obtain 14 We use that for any dimension D. (3. Changing variables we can write X n∈Z 1 1 ' 2 2 7/2 (r + (x9 − 2πnR9 ) ) 2πR9 r 6 Z+∞ du −∞ 1 I5 1 = . Using the relation 14 between κ10 and κ6 : κ10 = κ6 Q9 i=6 (2πRi ) (3. (1) (1) δC ≡ δ C¯ .13) (3. We can approximate the sum by an integral assuming that the distance in the non-compact directions is much larger that the size of the compact one. . G D D−d = GD /Vd . (3.15) If we do this approximation for each of the compactified directions of the orbifold we obtain 9 Y 1 1 ' I I I I (2πRi )−1 3 . 2 3 4 5 |x|7 |y| (3. (2πRi )−1 3 diag . r  R9 . . .14) 0 Moreover.270 E. D with Vd the volume of the compact space. for n > 2.. and for any lower dimension D − d. and Ωd = 2Id−1 Id−2 · · · I1 I0 . 2 7/2 (1 + u ) 2πR9 r 6 where we use the notation: Zπ In = dθ sinn θ. and r 2 = x12 + · · · + x82 . Panda / Nuclear Physics B 584 (2000) 251–283 A D-particle sitting at the origin of this S 1 can be seen from the point of view of the covering space as an infinite array of equally spaced D-particles. 2κ 2 = 16π G . i.

. with the difference that there appear extra normalisation coefficients. Φi hk⊥ . .11). Eyras.9) in the NS–NS sector to derive the spacetime behaviour: Y 1 T0 ¯ (2πRi )−1/2 3 .U = √ k⊥ 2 2 i δ h¯ µν (k) = hP¯(h) µν |Da=1/2|D0iNS. Instead of splitting the projectors in terms of 6-dimensional fields.26) in the amplitude. (2πRi )−1/2 3 diag . . and the momentum dependence is only in the Q −1/2 in the normalisation spatial directions of the fixed plane. The type of calculation we perform in this case is similar to the non-compact case.19) i=6 The asymptotic behaviour of the twisted R–R 1-form stays the same as for the non-compact case (3. ni = 0| ψ1/2 2 9 i=6  1 − hP¯(φ) | √ ηµν − (δ µ α δ ν β + δ µ β δ ν α )k α `β . We compare now this approximation with the asymptotic behaviour derived from the boundary state. . |y| 2 2 Ω4 i=6 δ h¯ µν (y) '   9 1 T0 Y 7 1 1 .U   1 T0 Y 7 1 −1/2 V1 = (2πRi ) diag . we can write the projection operators for the massless states as follows:  1 Y −1/2 ν ˜µ ψ1/2 ηµν − (δµ α δν β + δµ β δν α )kα `β . 2 2 4 4 4 k⊥ (3. . 6 Ω4 4 4 4 |y| (3.20) where now k α ηαβ `β = 1 and `α `β ηαβ = 0. .E. The massless closed string states excited by the D-particle correspond to the massless Kaluza–Klein modes. Note that the factors of i Φi Q of the boundary state and the massless states cancel with a factor i Φi coming from the normalisation (2. δ φ(y) = √ |y| 2 2 Ω4 i .21) i We can make use of the Fourier transformation (3. As before. we keep the ten-dimensional notation in order to compare it with the non-compact results. . . and with the approximation made previously. since the momentum is quantised in the orbifolded directions. We find the following results: 3T0 Y V1 ¯ (2πRi )−1/2 2 . ni = 0| ψ1/2 hP¯(φ) | = √ 2 2 i=6 9 hP¯(h) | = µν  1 Y −1/2 µ ν ˜µ ν ψ1/2 + ψ1/2 ψ˜ 1/2 Φi hk⊥ . . Using the same notation as before and using bars for the objects in the compact case. 2 2 (3. . S. δ φ(k) = hP¯(φ) |Da=1/2 |D0iNS. the NS–NS fields will have non-zero overlap with the NS–NS boundary state. The closed string fields have an infinite number of Kaluza–Klein modes. . Panda / Nuclear Physics B 584 (2000) 251–283 271 9 Y T0 1 ¯ δ φ(y) ' √ (2πRi )−1/2 3 .

which as we will see in the next section. 1. If we label the coordinates as before Xµ = (Xα . diag . 9. In this section we recover the no-force condition using the non-BPS D-particle as a probe of the geometry of another non-BPS D-particle. . and set to zero all fields which are odd under (−1)FL I4 . Xi ). On the other hand. . δC (1) (y) = − √ 3 2 Ω4 |y|3 (3. the Bose–Fermi degeneracy for non-BPS D-branes at the critical radius described in [33] represents a remarkable example in which the no-force property takes place without being BPS. . 5 labels the orbifold fixed plane directions. . S. The no-force property is a consequence of the vanishing of the one-loop amplitude between the D-branes. Note that the asymptotic behaviour derived from the compactified boundary state (3. The µ worldvolume theory of this D-particle is given by scalars b−1/2|0iNS ⊗ 1.11).272 E. the tachyon and the fermions with Chan–Paton factor σ1 [16]. a tachyon |0iNS ⊗ σ1 . µ = 1. This was verified at the level of the effective action and background geometries for BPS branes in [23]. Panda / Nuclear Physics B 584 (2000) 251–283 δ h¯ µν (y) =   7 1 1 1 T0 Y (2πRi )−1/2 3 diag . 6 Ω4 |y| 4 4 4 (3. at the critical radius there is an accidental Bose–Fermi degeneracy. 9. . . . and 16 fermions in the Ramond groundstate of each of the Chan– Paton sectors 1 and σ1 .1. One starts with the action of a non-BPS D-particle in type IIB. . implies a no-force in a brane probe. No-force condition at the critical radius A composite of branes preserving the BPS property satisfies a no-force condition between the constituents. At the critical radius we can then consider a large number of branes N . where α = 0. . . 4. . we consider the effective action for the non-BPS D-particle.22) i The twisted R–R sector remains the same as for the uncompactified case (3.o. Eyras. Accordingly. . describing the motion of the D-particle on the orbifold fixed plane. plus 16 fermionic d. . 7. we can write the following action for the D-particle: . .f. The action of the D-particle Following the prescription given in [58] one can construct the action for the D-particle in the orbifold of type IIB we are considering. 8. . .16). . the non-BPS D-particle is stable and contains 5 physical bosonic degrees of freedom. .22) exactly reproduces the behaviour of the 6-dimensional fluctuations obtained before (3. In particular. The orbifold (−1)FL I4 eliminates the scalars in the directions i = 6.23) 4. Firstly. hence the fluctuations take the following form: 1 T0 N ¯ . δ φ(y) = √ 2 0 2 2 Ω4 (2π α ) |y|3   1 1 T0 N 7 1 . δ h¯ µν (y) = 6Ω4 (2π 2 α 0 ) |y|3 4 4 4 Q0 N 1 . The asymptotics of the dilaton and the graviton are now certainly comparable with that of the twisted R–R 1-form.19) using the approximation given in (3. .

(4.7) Skin = − 2 0 2π α κ6 However. Moreover. χ i . 2 2 (4. In the compact case. (4.U = T0 V1 h00 . since the D-particle carries charge under the twisted R–R 1-form C (1) . Eyras. (4.1) in Einstein frame (gµν = e−φ/2 Gµν ) after rescaling the √ dilaton as φ = κ10 2 ϕ. and at the critical radius.1).2) can be also derived by projecting the boundary state of the D-particle onto the projectors defined previously. these extra modes will not play any role in our discussion below. after the rescaling C (1) = κ6 2 C (1) .5) √ which reproduces (4.3) whereas the coupling to the metric is given by µν J(h) = ηµν hP(h) |D0iNS. the winding modes of the tachyon become massless. we can include a WZ coupling 15 . 273 (4. which amounts to the relation of the 6-dimensional gravitational constant with the 10-dimensional one. in the effective action [49. it does feel the transverse geometry through the renormalised tension. These extra modes appear as new degrees of freedom. The boundary state reproduces the couplings of the following action: Skin = − Z 9 q T0 Y (2πRi )−1/2 dτ e−φ |∂τ Xα ∂τ Xβ Gαβ |.E. J(C) = Cα(1) hP(C) 2 (4. Panda / Nuclear Physics B 584 (2000) 251–283 Skin = − T0 κ10 Z q dτ e−φ |∂τ Xα ∂τ Xβ Gαβ |. the coupling of the D-particle to the massless closed string fields can be obtained as above. which involves the 1-form potential on the fixed plane: Z Q0 (4. this time using the compactified boundary state constructed in Section 2. These two results reproduce the couplings of (4. The coupling to the dilaton is given by 3T0 J(ϕ) = hP(φ) |D0iNS.2). analogously to the BPS case considered in [44]. S. the coupling to the twisted R–R 1-form is given by: Q0 (1) α |D0iR.4) where hµν is the symmetric and traceless helicity tensor of the graviton.1) and (4.6) i=6 Thus although the D-particle does not couple to the components Gij (4. Finally. the open strings on the D-particle carry winding modes. Furthermore.T = √ V1 C0 .58]: Z q  T0 dτ e−φ |∂τ Xα ∂τ Xβ Gαβ | 1 − Gτ τ ∂τ χ i ∂τ χ i .2) ∂τ Xα Cα(1).2. SWZ = 2κ6 The couplings of the stable non-BPS D-particle to the background fields given by Eqs. .U = √ V1 . The net result is a change of the factor in front of the kinetic term.1) where Gµν is the background string metric. κ6 (4. 15 Wess–Zumino couplings for stable non-BPS D-branes have been recently considered in [59].

one can check whether the Bose–Fermi degeneracy takes place. . The potential then takes the form: ! 9 Y Q20 2 2 −1 (2πRi ) − T V(y) = 4 3Ω4 |y|3 0 i=6 !  0 2 Y 9 α 4πα 0 Ri−1 − 1 . 2 2 2 i=6 We can insert now the asymptotic form of the solution for the compactified case given by (3. and choose the static gauge: X0 = τ.11) i=6 where we have defined the effective static potential V: 3 p Q0 (1) T0 Y (2πRi )−1/2 e− 4 φ |g00 | + C .8) We identify the embedding scalars with the transverse coordinates of the background geometry and expand the action in derivatives 16 :   q p 1 m n α β |∂τ y ∂τ y gαβ | = |g00 | 1 − ∂0 y ∂0 y gmn + · · · . . 17 We neglect the constant parts of the potential. since these would not generate any force. C0(1)(y). φ(y). Plugging into this potential the background geometry generated by another D-particle. (4. Panda / Nuclear Physics B 584 (2000) 251–283 4. Since from the boundary state we obtain the asymptotic form of the D-particle solution.13) V(y) ' T0 (2πRi )−1/2 √ δ φ¯ + δ h¯ 00 + √ δC0(1) . there will be a force term in the field equations for y m .22) and (3.10) The action can then be approximated by 3 Z Z 9 e− 4 φ T0 Y (2πRi )−1/2 dτ √ ∂0 y m ∂0 y n gmn .274 E.2.11). (4. . only the zeroth component of the twisted R–R 1-form is turned on: ∂τ Xα Cα(1) = C0(1) . D-particle at the critical radius We consider a non-BPS D-particle probe moving in the background of another nonBPS D-particle in the compactified orbifold. S. 5. S ' dτ V(y) + κ6 2 |g00 | (4. (4.18) 17 :   9 Y 3 Q0 (4. . Eyras. We split the embedding coordinates as Xα = (X0 . (4.14) = |y|3 2 i=6 16 We use the Einstein frame since we want to use in the action the results given by the boundary state computation. for a D-particle background. Xm ). we rewrite the potential V in terms of the fluctuations (3. m = 1.12) i=6 If this potential is not constant or zero for a given background g00 (y). .9) 2g00 Moreover. κ6 2κ6 0 9 V(y) = − (4.

the velocity corrections start at order v 4 and there is no v 2 corrections to the potential in for any radius. We can then deduce part of the form for g00 . Panda / Nuclear Physics B 584 (2000) 251–283 275 which coincides with the static potential p derived from the cylinder amplitude in (2. 4. Thus we recover the behaviour given by the open strings in (2. in our case.11).52) and vanishes for the critical radius (Ri = α 0 /2). However. is flat [23]. C0 (y) = 1 + 4aΩ4 |y|3 (4. In the case of the compact orbifold.38). as seen in (2. No-force can occur for V either constant or zero [23]. this translates into the fact that the metric on the moduli space. The dots indicate other possible contributions with dependence |y|n .48) and (2. 2π α g00 (y) = − 1 + 2aΩ4 |y|3  a κ6 T0 1 (2πα 0 )−1 3 + · · · .22).3. for the non-compact orbifold this metric takes the form:  3  3 e− 4 φ gmn ' δmn 1 − √ κ10 δφ(x) + κ10 δh00 (x) + 2κ10 δhmn (x).10).55). (4. it is easy to see that the metric is flat (= δmn ).17) holds. respectively. S.19) and (3.3. Eyras.12). as happens for BPS branes. and C (1) :  − 7 a 6  κ6 T 0 2 0 −1 1 + · · · . We can then restrict the possible classical geometries by imposing the no-force at the level of equation (4.11). we obtain the same flat metric for any value of the radii. eφ (y) = 1 + 2aΩ4 |y|  − 4 a 3 κ6 Q0 1 (1) + ··· − 1.16) |g00 | 2 2 Substituting the asymptotic fluctuations (3. (4. .15) √ |g00 | 2 2 Using the asymptotic fluctuations given in (3. According to the interaction potential Uopen derived in Section 2.E. These are in fact the factors of the NS–NS and R–R contributions of the potential (4. From the probe point of view. In the asymptotic limit. and at the critical radius equation (4.12).18) These functions are given in terms of a parameter a yet to be determined. multiplying the velocity dependent piece of the action (4. Hence we conclude that the classical solution must be such that at the critical radius the following equality holds 3 p (1) (4. and are such that asymptotically they become the fluctuations in (3. hence we recover the no-force at the critical radius using the background geometry generated by the non-BPS D-particles. the no-force condition at the critical radius must enforce the relation T0 = (π 2 α 0 )Q0 . The classical geometry of the non-BPS D-particle In this section we assume that the no-force condition will persist at the full non-linear level of the field equations. φ. this metric factor takes the form:   3 e− 4 φ 3 ¯ gmn ' δmn 1 − √ κ6 δ φ(y) √ + κ6 δ h¯ 00 (y) + 2κ6 δ h¯ mn (y).17) e− 4 φ |g00 | = C0 .

7). Therefore. this is certainly broken to SO(5) × SO(4) by the orbifold when we get closer to the position of the brane. the form of gij (y) is expected to be different from gmn (y). Accordingly. |g00 | (4. We find a metric and a dilaton propagating in the bulk.20) 2π α gmn (y) = 1 + 3 2aΩ4 |y| where the dots represent the same contributions as in (4. although the asymptotic behaviour of the metric in the non-compact case (3. We have used the technique of the boundary state to obtain the asymptotic form of the classical solution for the non-compact and compact versions of the orbifold.276 E. it seems that gij is out of the reach of the present analysis. We can make one further assumption if we consider that the metric factor in the velocity dependent piece of the action (4. we make use of the expressions for g00 (y) and φ(y) given above to obtain:  1a 6  κ6 T0 2 0 −1 1 + ··· δmn .19) Note that the functions describing the dilaton and the metric must be the same at the critical radius. which are subleading in the asymptotic limit (|y| → ∞). On the other hand. From it we have extracted the long and short distance interactions.10) presents an SO(9) symmetry. 5. Comments In this paper we have investigated the description of a stable non-BPS D-particle in terms of a classical solution. The twisted R–R 1-form has in both cases the same asymptotic form with the usual dependence for a particle in six-dimensions. In the non-compact case the bulk fields have the dependence expected for a particle in ten dimensions. A possible way to include it in the analysis is by a coupling to the extra massless states χ i in (4. Moreover. On the other hand.18) at the critical radius. but which are relevant when we come closer to the brane. whereas in the compact case they have a dependence typical of a particle in six dimensions. Eyras.11) remains flat for the complete geometry. we have calculated the cylinder amplitude for nonBPS D-particles in relative motion. the v 4 corrections do not match. S. We expect these extra terms to appear since harmonicity. (4. which is a direct consequence of supersymmetry. Panda / Nuclear Physics B 584 (2000) 251–283 n < −3.19). This is in fact already suggested by the asymptotics in the compact case (3. Finally. Moreover. at . We impose then the following relation: 3 e− 4 φ √ gmn = δmn . and a twisted R–R 1-form propagating in the fixed plane. we have recovered the no-force property at the critical radius using the asymptotic behaviour. but the open string description presents no v 2 terms for all radii. For generic radii these contain v 2 corrections in the closed string description. Moreover. may not be present in the complete solution. Using the non-BPS D-particle as a probe in the background of another non-BPS D-particle. like for BPS D-branes. it is only at this radius where we actually expect to find a consistent classical geometry. unlike for BPS branes.

On the other hand. up to the v4 corrections.1) The bulk action involves the metric and the dilaton. hence in Einstein frame we have:   Z p 1 1 10 2 (5. 1. x i = 0). . x i = 0).2) d x |det(gµν )| R − (∂φ) . This assumption allows us to derive part of the classical solution.10) and . d6 y |det(g˜αβ )| (5. The form of the asymptotic behaviour we have found suggests that the classical solution for the non-BPS D-particle will be a solution to the field equations derived from an action involving 10-dimensional bulk fields and 6-dimensional matter fields constrained on an orbifold fixed plane: Stotal = Sbulk + Splane . although the asymptotic behaviour of the metric (3. Here m is a factor related to the tension of the NS-5 brane. ˜ α ) = φ(y α .E. the fact that there is a coordinate system in which the metric can be written in terms of harmonic functions is very much related to residual supersymmetry. we can derive the fixed plane action from the effective action of the type IIB NS-5 brane in Einstein frame: 3 ˜ Z q e− 2 φ Fαβ F αβ . to which the non-BPS D-particle does not couple. S. Moreover. β = 0. we expect that extra terms may appear in the solution at the classical level. Moreover. The twisted R–R 1-form has been identified with the U (1) gauge field on the NS-5 brane worldvolume F . (5. It is straightforward to see that the asymptotic fields (3. which reproduces the asymptotic behaviour and the no-force property. Moreover. since they correspond to the twisted NS–NS sector. which do not match in the open and closed descriptions. this is expected to be broken to SO(5) × SO(4) by the orbifold when we get closer to the position of the brane. This particular coupling of the twisted fields with the bulk metric is obtained by expanding the NS-5 brane kinetic term in powers of the worldvolume fields. We have assumed that the no-force property of a brane probe holds for the full background geometry. the embedding scalars of the NS-5 brane has not been included. Eyras. 5 are the indices along the fixed plane. only the metric and dilaton are relevant. hence subleading terms which become relevant at short distances escape from this analysis. . Sbulk = 2 2 2κ10 Since the twisted sector is provided by a type IIB NS-5 brane hidden in the orbifold fixed plane.3) Splane = m 4 x i =0 where the tilde denotes the restriction of the bulk fields to the position of the fixed plane: g˜αβ (y α ) = gαβ (y α . .10) presents an SO(9) symmetry. which does not occur in our case. . This fact does not permit to find information about the geometry near the core of the non-BPS D-particle.4) and α. For the case of the D-particle. This is acceptable for distances much larger than the string scale. φ(y (5. and other fields of the massless untwisted sector. Panda / Nuclear Physics B 584 (2000) 251–283 277 the critical radius they present a BPS-like behaviour. This is due to the fact that the boundary state only gives information about the next-to-leading term in the asymptotic limit.

Bergshoeff and M.¯z→0 (A. Dasgupta. therefore we do not expect it would appear as a solution of this action reduced to six dimensions. Perry. (A. P. In order to describe the twisted R–R groundstate we make use of the 8 × 8 gamma matrices of SO(1. 5).1) γ . (A. Panda / Nuclear Physics B 584 (2000) 251–283 (3. is greatful to M.3) such that {γ . P. Stefanski. Liccardo.E. A study of the classical geometry of other stable non-BPS D-branes will be presented in a future publication [60]. Uranga for useful discussions. γ } = 0 and (γ )2 = 18 . Townsend and A. The work of E. Wyllard. Acknowledgements We would like to thank E. M.2) such that {γ α . 5):  α β (A. (A. N. without being supersymmetric. the behaviour of the stable non-BPS D-particle at the critical radius suggests that it probably saturates a BPS type of bound in the effective theory. We define γ = −γ 0 γ 1 γ 2 γ 3 γ 4 γ 5 .E. E.11) are a solution to the weak field limit of the equations of motion of the action Stotal .J. C −1 = C. Note that the non-BPS D-particle is not stable below the critical radius. Furthermore. Eyras. has also enjoyed discussions with C. γ = 2 18 ηαβ .5) The zero-mode part of the twisted R–R boundary state satisfies the following overlap equations in the twisted sector: .E. Appendix A In this appendix we present some details about the zero-mode part of the twisted R–R boundary state and its GSO projection. Gaberdiel. B. This work is also partially supported by the PPARC grant PPA/G/S/1998/00613. de Roo for useful discussions. (γ α )T = −Cγ α C −1 . Lambert. The analysis carried out here can be extended to other stable non-BPS branes. E. Vanhove and N. and can be constructed from the NS-vacuum by means of spin and twist fields as follows: ˜ z)|0i = |aiT |bi fT.4) The twisted R–R groundstate is characterised by left and right spinor indices of SO(1. A. This latter property comes from the fact that the particle couples to a 1-form in the wrong supersymmetric multiplet. C} = 0. S. is supported by the European Community program “Human Potential” under the contract HPMF-CT-199900018. there is a conjugation matrix C = γ 3γ 5γ 0. van der Bruck.278 E. C T = C. T. lim S a (z)Σ(z)S˜ b (¯z)Σ(¯ z. Finally.

1 + iη (A. vi(0) R.T = Mab (v)|aiT |biT . . −1.12):  1 + iηγ 1 . (A. −1). −1. 279 (A. N = Cγ 0 1 − iη f ψα T ha|T hb| 0 (A. Eyras.E. = − √ T hc|T hd| 2 f ψ˜ β = √1 T hc|T hd| f (γ )a c (γ β )b d T ha|T hb| 0 2 we can write the matrix N as follows: 1 + iηγ . −1. η| = T ha|T hb|Nab (A.7) If we define the action of the fermionic zero-modes in the twisted Ramond sector as f T = √1 (γ α )a c (18 )b d |ciT |di f T.9) (γ α )T M − iη S α β γ Mγ β = 0. .8) ψ˜ 0 |aiT |bi 2 the matrix M must satisfy:  (A. η. . .14) (A.T = 0. The solution to this equation is given by: M = Cγ 0 1 + iηγ .T = Mab |aiT |bi (0) (A.15) The boosted version can be written similarly to (A. ηi(0) R. 5 and S α β is the restriction of the matrix S µ ν of the D-particle to the 6-dimensional orbifold fixed plane: S α β = diag(+1. We define f T. S.12) (A. (A. 1. ηiR.16) C γ 0 + vγ m N(v) = √ 2 1 − iη 1−v In order to implement the GSO projection on the zero-mode of the boundary state. C γ 0 + vγ m M(v) = √ 2 1 + iη 1−v On the other hand. we define the zero-mode part of the GSO-projector on the twisted R–R sector as: .6) where α = 0. (A. defining (0) f R.T hD0ψ .11) where the matrix Mab (v) is given by  1 + iηγ 1 . Panda / Nuclear Physics B 584 (2000) 251–283 β ψ0α − iη S α β ψ˜ 0 |D0ψ . ψ0α |aiT |bi 2 β f T = √1 (γ )a c (γ β )b d |ciT |di fT. |D0ψ .13) and using 1 f (γ α )a c (18 )b d .10) The zero-mode R–R boundary state for a non-BPS D-particle moving in a direction m is given by f |D0ψ . −1.

10) and (A.19) Appendix B For the sake of completeness we include in this appendix the explicit definitions of the different parts of the boundary state for the non-BPS D-particle. ! αn · S · α˜ n .1) ! α−n · S · α˜ −n |k = 0i.2) !   c0 + c˜0 f ˜ c−n b−n − b−n c˜−n |1i|1i.T = |D0ψ . (A. Before GSO-projection the NS–NS boundary state is given by: |D0.8).4) r=1/2 The superghost part: ∞ X |D0sgh . (B.3) ! ∞ X ψ−r · S · ψ˜ −r |0i. We include as well the conjugate states. Ψ |aiT |bi f T = (γ )b d |aiT |di fT. ηiNS = exp iη γ−r β˜−r − β−r γ˜−r !  g |−1i|−1i.5) r=1/2 The conjugate states are: hD0X | = hk = 0|δ (5) (qˆ − y )δ p p (4) (qˆ ) exp − i ∞ X 1 n=1  f b0 − b˜0 exp hD0gh | = h2|h2| ∞ X n=1 NS hD0ψ . combining (A. ηiNS = −i exp iη (B. !  n .18) Finally. −ηiR. S.18) we find: (0) (0) (0) Ψ˜ |D0ψ .17) Using this definition and (A. (0) Ψ |D0ψ . ηiNS .T . (B. −ηiR. ηiNS. the action of Ψ and Ψ˜ on the twisted R–R groundstate is found to be: f T = (γ )a c |ciT |bi fT.T = |D0ψ .U = T0 |D0X i |D0gh i |D0ψ .T . ηiR. (A. Ψ˜ |aiT |bi (A. Ψ = −8ψ05 · · · ψ00 . ηiNS |D0sgh. η| = i h0| exp iη ∞ X r=1/2 b˜n cn − c˜n bn ! ψr · S · ψ˜ r . 2 The bosonic part is: |D0X i = δ (5) (qˆ − y )δ p p (4) (qˆ ) exp − i ∞ X 1 n=1 The ghost part: |D0gh i = exp ∞ X n=1 n (B. Panda / Nuclear Physics B 584 (2000) 251–283 Ψ˜ = −8ψ˜ 05 · · · ψ˜ 00 . . 2 The fermionic part: |D0ψ . Eyras.280 E. ηiR. (B.

(B. ηiR = exp iη R .12) n=1 (0) where R hD0ψ . (0) |D0ψ . (B. ηiR.T (0) NS hD0sgh . The ghost part is the same as for the NS–NS boundary state.E. η| = − hD0 . Leigh. the conjugate states are:  X  1 (5) p p α hD0 | = hk = 0|δ ( q ˆ − y ) exp − · S · α ˜ T X n n . New connections between string theories. ηi(0) |D0sgh .8) where ‘t. (0) (B. R. Phys.m. η| is given in Appendix A and (0) ^ −iηβ0 γ˜0 . Panda / Nuclear Physics B 584 (2000) 251–283 g exp −iη NS hD0sgh .m.6) r=1/2 For the twisted R–R sector we have (before GSO-projection): Q0 |D0X iT |D0gh i |D0ψ . Eyras. 281 (B. Finally. Polchinski. η| = h−1|h−1| ∞ X βr γ˜r − γr β˜r !  . ηiR. Dai. η| exp −iη t.  X  (0) ˜ hD0 . ηiR. ηiR. Lett.’ indicates that the sum is performed according to the twisted moddings of the (twisted) R–R sector given in (2. where (B.m. ηiR = eiηγ0 β0 |−1/2i|−3/2i.11) Finally.T |D0sgh . η| = h−3/2|h−1/2| e (B.T and is given in Appendix A. S.13) References [1] J. the superghost part: ! ∞ X  γ−n β˜−n − β−n γ˜−n |D0sgh .1). .9) |D0ψ . M.7) (B. The fermionic part reads:  X  (0) ψ−n · S · ψ˜ −n |D0ψ . ∞ X βn γ˜n − γn β˜n !  . 2 The bosonic part is given by  X  1 (5) p p α−n · S · α˜ −n |k = 0i |D0X iT = δ (qˆ − y ) exp − n t. ηiR.T = − exp iη t.10) n=1 with the zero-mode given by ˜ ^ |D0sgh .T = (B.T . 4 (1989) 2073. |D0. n t. J. ηiR . R hD0sgh . η| exp iη ψ · S · ψ R. η| = R hD0sgh .T ψ ψ n n .m.m.

B 564 (2000) 60. Frau. Eyras.R. Stable non-BPS states in string theory: a pedagogical review. Peet. hep-th/9810214. Sen. Phys. B 499 (1997) 183. Tseytlin. Horowitz. Sen. [16] A. Callan. C. Lett. Bergman. Khuri. [12] O. Sen. Lu. [26] M. Grav. [24] O. Michishita. Dirichlet branes on orbifolds. Lett. Lerda. Stable non-BPS bound states of BPS D-branes. Green. Sen. hep-th/9803194. Witten. JHEP 07 (1999) 022. R. Rev. Nucl. hep-th/9906055. hep-th/0002006. Kraus. Bergman. Phys. Gaberdiel. Maldacena. D-branes and K-theory. Gaberdiel. Comparing D-branes to black-branes. Sen.W. Sen. hep-th/9910109. hep-th/9701137. D 55 (1997) 861. Lett. hepth/9806155. [3] J. Quant. hep-th/9809111. Dirichlet branes and Ramond–Ramond charges. [32] C. Gaberdiel. . D-sphalerons and the topology of string configuration space. JHEP 09 (1998) 023. Microscopic origin of the Bekenstein–Hawking entropy.A. Nucl. hepth/9905006. JHEP 03 (1999) 013. Black hole greybody factors and D-brane spectroscopy. [20] J. Eyras. B 487 (1997) 141. [21] M. hep-th/9510017. Lectures on non-BPS Dirichlet branes presented at the school of the TMRNetwork “Quantum Aspects of Gauge Theories. hep-th/9904207. A non-supersymmetric open string theory and S-duality. Angelantonj. M.282 E. B 472 (1996) 591. SO(32) spinors of type I and other solitons on brane–antibrane pair. Harvey. B 444 (1998) 309. R. Some properties of open-string theories.R. P.B. Dualities of type 0 strings. B 329 (1994) 435. hep-th/9903123. [9] C. Stefanski Jr. Phys. B 360 (1991) 197. [23] A.G. Non-BPS states and branes in string theory. Phys. Rev. hep-th/9810188. Type I D-particle and its interactions. Phys. Class. hep-th/9901014. Stable D8-branes and tachyon condensation in type 0 open string theory. hep-th/9908121. Lett. Sagnotti. [7] J. A. P. Strominger. D-brane approach to black hole quantum mechanics. Non-BPS states in heterotic — type IIA duality. hep-th/9601029. A. Maldacena.. Lifschytz. Phys. Lerda.T. [11] A. Hoˇrava. Gaberdiel. Bergman. Stable non-BPS states in string theory. Strominger. Volkov. Vafa. Rept. Gaberdiel. Lett. Duff. [30] M.M. Russo. Panda / Nuclear Physics B 584 (2000) 251–283 [2] M. S. Stable non-BPS D-particles.J. Nucl. M. JHEP 10 (1999) 005. 75 (1995) 4724. [18] A. No-force condition and BPS combinations of p-branes in 11 and 10 dimensions. hep-th/9812031. Phys.S. 15 (1998) 3291. B 379 (1996) 99. Bergman. P. Phys. BPS D-branes on non-supersymmetric cycles. M. J. B. Black strings and p-branes. [22] G. hep-th/9805019. hep-th/9808141. hep-th/9602043. JHEP 08 (1998) 010. Gaberdiel. A. JHEP 9812 (1998) 019. [31] A. M. Polchinski. 2000. Phys. JHEP 12 (1998) 021. hep-th/0001143. [8] A. Phys. [5] A. hep-th/9509080. [25] O. Lett. Nucl. hep-th/9907094. [4] G. Supersymmetry and Gravity”.R. hepth/9403040.R. J. Stable non-BPS D-branes in type I string theory. Strigazzi. hep-th/9609026.A. [10] A. [27] O. [19] M. [29] E. Torino. B 441 (1998) 133. [28] Y. [15] E. Pointlike states for type IIB superstrings. Phys. [13] A. B 466 (1999) 161. hep-th/9609212. [17] A. The Bekenstein formula and string theory. Phys. [14] A. 259 (1995) 213. JHEP 10 (1998) 021. Phys. Lett. Non-tachyonic open descendants of the 0B string theory. Nucl. String solitons. JHEP 06 (1998) 007.R. D0-branes in SO(32) × SO(32) open type 0 string theory.X. Strominger. Gallot.R. hep-th/9412184. hepth/9712253. hepth/9604156. L. Phys. B 388 (1996) 720. The tale of gravitational sphaleron. [6] M.

Rabinovici. D. Phys. Douglas. A. A. Cangemi. Gauge and gravitational interactions of non-BPS D-particles. Frau. Phys. [44] P. [41] J. Panda. Iengo. Liccardo. Lerda. E. hep-th/9512145. [49] J.L.E. Loop corrections to superstring equations of motion. Phys. [60] E. Lerda. hep-th/9908060.B. Green. Klebanov. Nappi. Fre‘. A. Nucl. Eyras. hep-th/9603194. Callan. S. Di Vecchia. Sciuto. Gravity duals of fractional branes and logarithmic RG flow. Wai. hep-th/0001049. I. The conformal field theory of orbifolds. C. Nucl. Pesando. Gaberdiel. Yost. Non-supersymmetric D-brane configurations with Bose–Fermi degenerate open string spectrum. B 376 (1992) 365. Lett. Boundary states for moving D-branes.R.C. hep-th/9702037. P. Classical p-branes from the boundary state. Di Vecchia. hep-th/980309. S. Duality in string background space. hep-th/9912161. Phys. Lett. Garousi. Gallot.H. hep-th/9906214. Eyras. P. B 485 (1997) 85. R. Phys. in preparation. [51] L. B 296 (1988) 91. [40] C. Bianchi. Consistency of open superstring theories. B. Black holes as D3-branes on Calabi–Yau threefolds. [46] N. Tseytlin. [38] C. Shenker. [55] L. Nucl. “Blowing up” D-branes on non-supersymmetric cycles. Cai. Phys. M. B 507 (1997) 259. [35] I. M. Dp) bound states from the boundary state.A. C. On the effective action of stable non-BPS branes. Strigazzi. Phys. hep-th/9912275. D-branes in string theory. C. Polchinski. D-brane dynamics. Panda / Nuclear Physics B 584 (2000) 251–283 283 [33] M. Lett. hep-th/9909062. M. Scrucca. Phys. [48] D. String loop corrections to beta functions. B 293 (1987) 83. [43] M. Lerda. Nucl.R. Y. Dixon. P. Myers. Callan. Nekrasov. Di Vecchia. Lovelace. hep-th/9511043. Nucl. Sagnotti. Pouliot. hep-th/9906109. Klebanov. Kabat. [56] C. hep-th/9912090. Nucl. Billó. Nucl. Frau. The insertion of boundaries in worldsheets. Russo. Lerda. R. A. Lovelace. Di Vecchia. Phys. A. Pradisi. B 282 (1987) 13. [39] C. D. Bertolini. Sen. B 526 (1998) 199. Microscopic string analysis of the D0–D8 brane system and dual R–R states. JHEP 09 (1999) 004. hep-th/0002159. Sachs. [37] P. B 431 (1998) 22. [34] M. A. Phys. [54] M. Yost. B 322 (1989) 167. Russo. Nucl. Kutasov. Friedan. S.A. E. Billó. [42] M. C. Scrucca. [58] A.R. A. Phys. Martinec. Lambert. A. Supersymmetric world-volume action for non-BPS D-branes. Phys. Russo. hep-th/9911096. P. hep-th/9802088. S. B 400 (1997) 52. A. Yost. hep-th/9608024. [45] P. Nappi. B 475 (1996) 193. Lerda. Nucl. S.R. [53] M. R. B 374 (1996) 37. Scattering of closed strings from many D-branes. Lett. Orbifolds and solitons. JHEP 10 (1999) 008. Phys. S. B 431 (1994) 131. Di Vecchia. Gravity duals of supersymmetric SU(N) × SU(N + M) gauge theories.A. Phys. S. N. Lovelace.A. Toroidal compactification and symmetry breaking in open string theories. Nucl. S. R.A. B 308 (1988) 221. [36] I. Nucl.A.R. A. Pesando. [50] M. JHEP 11 (1999) 008.R. Phys. Giveon. Phys. B 400 (1997) 63. D-branes and short distances in string theory. Nucl. Stefanski. [57] M. Frau. . G. C. Liccardo. Majumder. S. Lett.A.R. Shenker. Non-abelian field theory of stable non-BPS branes. P. Callan.L. Phys. Sen. Frau. A. Sen. P.R. Sciuto. Pesando. C. Nappi. [52] A. I. Russo. B 288 (1987) 525. (F. Veneziano. hep-th/0002061. I. C. hep-th/9707068. [47] M. Superstring scattering from D-branes. S. Adding holes and crosscaps to the superstring. Bachas. Sciuto. G. B 383 (1996) 48. C. I. [59] R. R.L. hep-th/9701190. D.

D-branes are non-perturbative states on which open string can live. and to which various closed strings including Ramond– Ramond states can couple. Type II string theories have two kind of D-branes.3 2 1 3 ( 0 0 ) 0 0 3 6 1 . we then extend the abelian action to nonabelian cases. dilaton or Kalb–Ramond antisymmetric tensor on the worldvolume of a single non-BPS Dp-brane. Introduction Recent years have seem dramatic progress in the understanding of non-perturbative aspects of string theory [1–5]. Birjand.Nuclear Physics B 584 (2000) 284–299 www. Tehran.V.  2000 Elsevier Science B. revised 29 May 2000.V.O. University of Birjand.elsevier. Iran Received 29 March 2000. we find coupling of two open string tachyons to gauge field. PII: S 0 5 5 0 . BPS Dp-branes for p even(odd) [6–8] and non-BPS Dp-branes for p odd(even) [9–13] in IIA(IIB) theory.25.nl/locate/npe Tachyon couplings on non-BPS D-branes and Dirac–Born–Infeld action Mohammad R. accepted 30 May 2000 Abstract By explicit evaluation of certain disk S-matrix elements in the presence of background B-flux. With these studies has come the realization that solitonic extended objects. Box 19395-5746. This action includes non-linearly the dynamics of the tachyon field much like the other bosonic modes of the non-BPS Dp-brane.com 0550-3213/00/$ – see front matter  2000 Elsevier Science B. P. Iran Institute for Studies in Theoretical Physics and Mathematics IPM. All rights reserved. An important object in these investigations has been Dirichlet branes [6–8]. play an essential role. graviton.8 . We then propose an extension of the abelian Dirac–Born–Infeld action which naturally reproduces these couplings in field theory. The non-BPS Dp-branes on the other hand suffer ∗ biruni@iran. other than just strings.-w 1. On the general grounds of gauge and T-duality transformations and the symmetrized trace prescription. Garousi ∗ Department of Physics. All rights reserved. The BPS branes are stable solitons which break half of the space–time supersymmetries and their dynamics are properly described in field theory by Dirac–Born–Infeld(DBI) action [14–16] (see also [17]) and Chern–Simons action [18–20]. PACS: 11.

However. on the other hand.1 we transform the above couplings between commutative fields to their noncommutative counterparts.R. [35]). the negative minimum energy of the tachyon potential plus the positive energy of the brane tension is exactly zero [21–25]. Then we expand this action around a background B-flux to produce various couplings involving two tachyons and one gauge field. graviton. However. The RR fields of the type II theory have no coupling to the non-BPS D-branes through the usual Chern–Simons action. we evaluate the S-matrix elements and check their consistency with the proposed field theory couplings. Hence. Abelian action The worldvolume theory of a single non-BPS D-brane in type II theory includes a massless U(1) vector Aa . In Section 4 we extend our proposed action for describing dynamics of a single non-BPS Dp-brane to the case of the non-abelian theory of N coincident branes using the general grounds of the symmetrized trace. This action has been generalized to the supersymmetric form to include the dynamics of massless fermionic modes of the branes as well [34]. there is a non-vanishing coupling between the RR field and tachyon on the worldvolume of the branes [35]. From these matrix elements we conjecture an extension for the DBI action which includes the tachyon field as well. dilaton or Kalb–Ramond antisymmetric tensor. Appendix contains our conventions and some useful comments on conformal field theory propagators and vertex operators used in our calculations. The Chern– Simons action hence was modified in [36] to incorporate this coupling. This conjecture was studied in [26–33] by explicit calculation of the tachyon potential using the string field theory framework. In Section 2. Consequently. The leading .g.M. The dynamics of massless bosonic excitations of non-BPS Dp-branes are suitablely described by the DBI action in field theory. the non-BPS branes decay to minimum of the tachyon potential. we are interested in generalizing the DBI action to take into account the dynamics of the tachyon field. a tachyonic state T and their fermionic partners (see. 2. and non-abelian gauge and invariance under T-duality transformations. We study this by explicit evaluation of some nontrivial disk S-matrix elements in the first quantized string theory. In Section 3. In the present paper. describing the transverse oscillations of the brane. We do this because the disk S-matrix elements in the presence of the background B-flux with which we are going to compare our conjectured action in the subsequent section are corresponding to non-commutative open string fields [37]. We begin in Section 2 by expressing our conjecture for extension of the DBI action which includes dynamics of the tachyon field. e. Garousi / Nuclear Physics B 584 (2000) 284–299 285 from open string tachyonic mode whose mass causes the brane in a flat background to be unstable. It has been conjectured that at the stationary point of the tachyon potential. there are other terms in the tachyon potential which makes it bounded from below. An outline of the paper is as follows. a set of massless scalars Xi .. the unstable non-BPS branes in flat space–time vacuum should decay to the true vacuum of the theory in which there is no branes.

.. Garousi / Nuclear Physics B 584 (2000) 284–299 order low-energy action for the massless fields corresponds to a dimensional reduction of a ten-dimensional U(1) Yang–Mills theory. Our results indicate that the tachyon should appear in the following extension of DBI action: Z S = −Tp d p+1 σ e−Φ V (T ) p (2) × −det(P [Gab + Bab ] + 2πα 0 Fab + 2πα 0 ∂a T ∂b T ).e.. The conjecture in [21–25] is that the tachyon potential is zero at the minimum of the potential. P [Gab ] = Gµν (1) where in the second line above we have used that fact that we are employing static gauge throughout the paper. Here.g. e. then the action takes a Dirac–Born–Infeld form [14–16]. Note that appearance of the tachyon kinetic term and potential in (2) is similar in form to the kinetic term and potential of the transverse scalar fields in the non-abelian DBI action of N coincident BPS D-branes [38]. To take into account the couplings of the massless open string states with closed strings.e. As usual in string theory. ∂Xµ ∂Xν ∂σ a ∂σ b = Gab + 2Gi(a ∂b) Xi + Gij ∂a Xi ∂b Xj . while the metric and antisymmetric tensors are the pull-backs of the bulk tensors to the D-brane worldvolume. the DBI action may be extended naturally to include massless Neveu–Schwarz closed string fields. there are higher order α 0 = `2s corrections. As long as derivatives of the field strengths (and second derivatives of the scalars) are small compared to `s .R. (22) for T = 0). i. σ a = Xa for worldvolume and Xi (σ a ) for transverse coordinates. we shall evaluate some disk S-matrix elements in string theory and read from them various couplings involving the tachyon field. the metric. we shall evaluate disk amplitudes describing decay of two tachyons to dilaton. Hence. graviton or Kalb–Ramond antisymmetric tensor on the worldvolume of a single non-BPS Dp-brane with background B-flux. Fab is the abelian field strength of the worldvolume ordinary gauge field. i. In this case though the kinetic term appears in the pull-back of the metric under the square root and the potential for scalar fields multiplies the square root in the DBI action(see Eq. In order to extend this action to incorporate dynamics of the tachyonic mode as well. In our conventions the tachyon field is dimensionless. where the tachyon potential is V (T ) = 1 + 2πα 0 m2 T 2 /2 + O(T 4 ) and the tachyon mass is m2 = −1/2α 0 . The amplitude describing the worldvolume coupling of two tachyons to gauge field will . upon tachyon condensation at this point the abelian action (2) of the non-BPS brane becomes zero. assuming the above action (2) and check its consistency with some S-matrix elements. where `s is the string length scale. We now continue backward. In this case one arrives at the following worldvolume action: Z p S = −Tp d p+1 σ e−Φ −det(P [Gab + Bab ] + 2πα 0 Fab . dilaton and Kalb–Ramond filed.. V (T0 ) = 0.286 M.e. To have nontrivial check. i.

the pull back of the Einstein frame metric becomes:     1 κ κ 1 + √ φ ∂a λi ∂b λi + · · · . we begin by expanding (2) for fluctuations around the background Gµν = ηµν . Garousi / Nuclear Physics B 584 (2000) 284–299 287 be evaluated as well. Φ = 0 and T = 0 to extract interactions expected from the proposed action (2). P [Gab ] = ηab 1 + √ φ + 2κP [hab ] + Tp 2 2 where the dots represents terms with two and more closed string fields. Tp With these normalizations.M. As a first step. (2) using  p p  1 det(M0 + M) = det(M0 ) 1 + Tr M0−1 M 2  1 1 −1 −1  −1 2 + ··· − Tr M0 MM0 M + Tr M0 M 4 8 to produce a vast array of interactions. T=p 2πα 0 Tp Aa = 1 p aa . Now with our conventions for string vertex operators (see appendix). . Therefore. Now it is straightforward. one should transform the Einstein frame metric Gµν to the string frame metric gµν via Gµν = eΦ/2 gµν . and linear in massless open and closed string fluctuations. √ Φ = 2 κφ. 2πα 0 Tp 1 Xi = p λi . Bµν = F ab ηaµ ηbν . The appropriate Lagrangian are:   1 2 2 1 ab L2. to expand Eq. the string mode fluctuations take the form gµν = ηµν + 2κhµν . The fluctuations should be normalized as the conventional field theory modes which appear in the string vertex operators.R. We are interested in the interactions quadratic in tachyon.0 = − p 2 2 Tp 2  − V ab fbc V cd ∂d τ ∂a τ = 0. 2 2  c 1 1 2 m (VA )ab fba τ 2 + V ab fba V cd ∂c τ ∂d τ L3.0 = −c m τ + (VS ) ∂a τ ∂b τ . we recall that the graviton vertex operator corresponds to string frame metric. Bµν = F ab ηaµ ηbν − 2κbµν . Hence. 1 τ.

In that way.e. In the above Lagrangians. It is important to note that the antisymmetric matrix VA appears in total derivative terms in the Lagrangian L3.1. 2.0 . (4) c ≡ −det(ηab + Fab ). i. Here we have dropped some total derivative terms in L3. 2 2 (3) where the first and the second subscript in above Lagrangians refers to the number of open and closed string fields. So we continue our discussion for non-vanishing background B-flux. .288 M. Hence.1 vanishes. the coupling of Kalb–Ramond field to tachyon in the second line of L2. Whereas. in order to compare the couplings in (3) with corresponding S-matrix elements.0 which involves only open string fields. open string vertex operators in string theory with background B-flux correspond to non-commutative fields [37].R. there is no nontrivial coupling that confirms the conjectured action (2) is valid or not. The open string fields appearing in the DBI action (2) or (3) are ordinary commutative fields. Garousi / Nuclear Physics B 584 (2000) 284–299 L2. Change of variables In [37] differential equation for transforming commutative gauge field to its noncommutative counterpart was found to be  1 δ Aˆ a (θ ) = δθ cd Aˆ c ∗ Fˆad + Fˆad ∗ Aˆ c − Aˆ c ∗ ∂d Aˆ a − ∂d Aˆ a ∗ Aˆ c . and the graviton and dilaton couplings reduce to the natural coupling of these fields to the kinetic term of the tachyon field. e−Φ Gab ∂a T ∂b T . respectively.1 = −κc    1 V ab (hba − bba ) + √ Tr(V ) − 4 φ 2 2   1 1 (VS )ab ∂a τ ∂b τ + m2 τ 2 × 2 2    1 − V ab hbc − bbc + √ φηbc V cd ∂d τ ∂a τ . In the case that the background B-flux is zero. and VS (VA ) is symmetric(antisymmetric) part of the V matrix above. Aˆ b ]M . one has to transform the commutative fields in (3) to their non-commutative variables. 4  1 δ Fˆab (θ ) = δθ cd 2Fˆac ∗ Fˆbd + 2Fˆbd ∗ Fˆac − Aˆ c ∗ Dˆ d Fˆab + ∂d Fˆab 4   − Dˆ d Fˆab + ∂d Fˆab ∗ Aˆ c .. fab = ∂a ab − ∂b aa and p ab V ab ≡ (η + F )−1 . where the gauge field strength and ∗ product were defined to be Fˆab = ∂a Aˆ b − ∂b Aˆ a − i Aˆ a ∗ Aˆ b + i Aˆ b ∗ Aˆ a = ∂a Aˆ b − ∂b Aˆ a − i[Aˆ a .

i a b ˆ .

fˆ(x) ∗ g(x) ˆ = e 2 θab ∂y ∂z fˆ(y)g(z) (5) .y=z=x .

2  Fab = Fˆab − θ cd Fˆac ∗0 Fˆbd − Aˆ c ∗0 ∂d Fˆab + O(Aˆ 3 ).R. Garousi / Nuclear Physics B 584 (2000) 284–299 289 These differential equations can be integrated perturbatively to find a relation between ordinary fields appearing in (3) and non-commutative fields corresponding to open string vertex operators.M. The result for abelian case that we are interested in is [39]:  1 Aa = Aˆ a − θ cd Aˆ c ∗0 Fˆad − Aˆ c ∗0 ∂d Aˆ a + O(Aˆ 3 ). (6) where the commutative multiplication ∗0 operates as .

.

sin( 12 θab ∂ya ∂zb ) 0 ˆ ˆ f (y)g(z) ˆ .

.

e. Upon integration (6). Then. i. ˆ = 1 . Appropriate transformations for scalar fields such as the tachyon can be read from (6).e. that would produced the desired transformation. Tˆ M + θ cd Fˆca ∗0 ∂b Tˆ + Aˆ c ∗0 ∂a ∂d Tˆ + · · · . This transformation rule was confirmed in [39] by explicit evaluation of S-matrix elements of one massless closed and two open string states.g. one should first using (8) replace ordinary multiplication of two tachyons by the ∗0 multiplication. However. (7) and (8). i. (6) to be fg|θ=0 −→ f ∗0 g|θ6=0 (8) for f and g being any arbitrary open string fields. using (7) the ordinary .e. might be found from a differential equation alike (5) that expresses infinitesimal variation of the quadratic fields in terms of infinitesimal variation of non-commutative parameter. (7) where dots represent terms which involve more than two open string fields. we simply note that the transformation for multiplication rule between two open string fields can be conjectured from the right hand side of Eq.. T = Tˆ + θ cd Aˆ c ∗0 ∂d Tˆ + · · · . They produce couplings between more than three fields upon replacing them into (3) in which we are not interested. ˆ δ Aa . The differential equation (5) expresses infinitesimal variation of linear field. this transforms the linear commutative gauge field in terms of nonlinear combination of non-commutative fields. Instead.. δθ cd . We are not going to find such a differential equations here..    ∂a T = ∂a Tˆ − i Aˆ a . one can transform the commutative Lagrangian (3) to non-commutative counterparts. f (x) ∗ g(x) a b y=z=x 2 θab ∂y ∂z In [39] we verified by explicit calculation of S-matrix elements of one massless closed and two open string states that the transformation (6) is exactly reproduced by perturbative string theory. Now with the help of Eqs. in terms of infinitesimal variation of the non-commutative parameter. In that way. we are interested in transforming quadratic combinations of commutative fields appearing in (3) in terms of non-commutative fields. one would find that not only the fields transform as in (7) but the multiplication rule between fields also undergo appropriate transformation. Such a transformation. Upon integration. in principle. In doing so.

3.1. VS .e. and the antisymmetric part VA appears as a result of transforming commutative fields to non-commutative variables. In our discussion. This is consistent with the conclusion reached in [37]. (10) on the other hand. (9) Lˆ 2. (9) only in the ∗ product terms.0 = −c m2 τˆ τˆ + (VS )ab ∂a τˆ ∂b τˆ + 2 2 4π Tp    1 ab ˆ L2. that is 1 Note that our conventions set θ ab = 4π(V )ab . appears as the metric in (9) and the antisymmetric part appears in the definition of ∗ product in the Moyal bracket. the symmetric part VS appears naturally as a result of expanding the ordinary DBI action around the background B-flux. on the other hand. the antisymmetric matrix (VA )ab appears in Eq. −V 2 2 where ellipsis represent terms which have more than three fields. Scattering calculations The couplings in (9) and (10) should be reproduced by disk S-matrix elements of string theory if the proposed action (2) is going to be valid.. however. The results are   1 1 ic p (VS )ab ∂a τˆ [aˆ b .1 = −κc V (hba − bba ) + √ Tr(V ) − 4 φ 2 2   1 1 2 0 ab 0 (VS ) ∂a τˆ ∗ ∂b τˆ + m τˆ ∗ τˆ × 2 2    φη bc ab cd 0 (10) hbc − bbc + √ V ∂d τˆ ∗ ∂a τˆ + · · · . both symmetric and antisymmetric matrices appear in its different coupling terms.290 M. A . Garousi / Nuclear Physics B 584 (2000) 284–299 tachyon fields should be transformed to their non-commutative counterparts. In the Eq. In the Lagrangian (10). which involves open and closed string fields. i. the difference between ∗0 and ordinary multiplication rules is not just a total derivative terms. In this section.R. 1 It is interesting to note that the symmetric part of the (η + F )−1 matrix. Here we have dropped some total derivative terms which appeared in Lˆ2. 3. Open–open–open couplings In the worldsheet conformal field theory framework. the coupling of two tachyons to a gauge or scalar field is described properly by the correlation of their corresponding vertex operators inserted at the boundary of the disk worldsheet. We begin with the evaluation of the coupling of two tachyons to gauge or scalar fields. we work at the string theory side and evaluate these couplings using the conformal field theory technique. Note that upon inserting the transformation (7) into (3). τˆ ]M + · · · .0 and also replaced ∗0 between two noncommutative tachyons in (9) with ordinary multiplication rule because the deference is some total derivative terms.

The sin(πl) factor above arises basically from two different phase factors corresponding to two distinct cyclic orderings of the vertex operators. We have also normalized the p amplitude by the appropriate coupling factor −c/2π Tp . Using the worldsheet conformal field theory technique. R) symmetry by fixing z3 = i and x1 = ∞. it is not difficult to verify that the S-matrix element (11) is exactly reproduced by the second term in (9). one can evaluate the correlations above and show that the integrand is SL(2. Closed–open–open amplitudes The amplitudes describing decay of two open string tachyons to one massless closed string NSNS mode is given by the following correlation: Z   A ∼ (ε3 ·D)µν dx1 dx2 d2 z :V0 2k1 ·V T . This integral is doable and the result is . [40–42] for the details of the calculations. x2 : µ ν (p3 ·D. x2 :  µ ×:V0 2k3 ·V T .R. π Tp (11) where we have defined l ≡ −2k1 ·V T ·F ·V ·k2 = 2k1 ·VA ·k2 . Gauging this symmetry by fixing the positions of the vertices at arbitrary points. Gauging the SL(2. Garousi / Nuclear Physics B 584 (2000) 284–299 Z A ∼ (ζ3 ·G)µ 291   dx1 dx2 dx3 : V−1 2k1 ·V T . x1 : :V−1 2k2 ·V T . ×:V−1 (p3 . where the vertex operators are given in the appendix. At the same time. λ3 ) = 0 and A(τ1 .M. x3 : . Explicit form of the vertex operators in terms of worldsheet fields are given in the Appendix. τ2 . Using the fact that our conventions set θ ab = 4πVAab . where the integral is taken from −∞ to +∞. vanishing of A(τ1 . We refer the reader to Refs. and s = −p3 ·VS ·p3 = −1/2 − 2k1 ·VS ·k2 . a3 ) = c sin(πl) p (k1 ·VS ·ζ3 ). z¯3 ): . Here again using appropriate worldsheet propagators. R). λ3 ) is consistent with (9).2. τ2 . 3. Each phase factor stems from the second term of the worldsheet propagator (29). z3 ): :V−1 where the closed string vertex operator inserted at the middle and open string vertex operators at the boundary of the disk worldsheet. one arrives at  Z 8ik2 ·V T ·ε3 ·D·V ·k1 A ∼ 2−2s−2 dx (2s + 1) Tr(ε3 ·D) − x −i  T 8ik1 ·V ·ε3 ·D·V ·k2 + x+i × (x − i)s−l (x + i)s+l . τ2 . R) invariant. x1 : :V0 2k2 ·V T . one finds A(τ1 . it is not difficult to perform the correlators above and show that the integrand is invariant under SL(2.

As a check of our calculations. They can be evaluated in field theory as   b eφ3 a a G ea e . Hence. The amplitudes (14) should be reproduced in s-channel of field theory. h3 ) = 4πs 2 Here h3 stands for both graviton and Kalb–Ramond antisymmetric tensors. τ2 .292 M. In fact due to the worldsheet fermions in the tachyon vertex operator. (14) As (τ1 . we have inserted the dilaton polarization (31) into (12) and found that the result is independent of the auxiliary vector `µ . Otherwise our conventions set α 0 = 2. τ2 . τ2 . coupling of any odd number of tachyons is zero. The amplitude (12) has the pole structure at m2open = n/α 0 . (12). 3. They are 2 We explicitly restore α 0 here. the first term of the expansion representing the exchange of massless string states dominate. We will add these terms which have no contribution to the massless poles of field theory to the contact terms in (16). (13) 4πs where dots represent contact terms and the infinite series of massive poles. h3 ) = V V ab aτ1 τ2 A= where the propagator and the vertices can be read from expansion of (2) in terms of noncommutative fields. φ3 ) = √ 2 8π 2s   iκc sin(πl) l Tr(ε3 ·D) − 4k1 ·V ·ε3T ·V ·k2 + 1 ↔ 2. Garousi / Nuclear Physics B 584 (2000) 284–299  iκc Γ (−2s) a1 (s + l) − a2 (s − l) . Massless poles Given the general form of the string amplitude in Eq. (12) 2 Γ (1 − s − l)Γ (1 − s + l) where a1 and a2 are two kinematic factors depending only on the space–time momenta and the closed string polarization tensor A=− a1 = −4k2 ·V T ·ε3 ·D·V ·k1 . V A0s (τ1 .1. one can expand this amplitude as an infinite sum of terms reflecting the infinite tower of open string states that propagate on the worldvolume of D-brane. . the worldvolume of the non-BPS Dp-branes has a Z2 symmetry under which the tachyon changes sign. Making the appropriate explicit choices of polarizations. a2 = (s + l) Tr(ε3 ·D) + 4k1 ·V T ·ε3 ·D·V ·k2 .R. As (τ1 . In the domain where s → 0. In this case the scattering amplitude (12) reduces to iκc sin(πl) (a1 + a2 ) + · · · .2. φ3 ) = V ab aτ1 τ2    b eh3 a a G ea e . 2 This does not have tachyon pole which is consistent with the fact that coupling of three tachyons is zero. one finds some terms which are proportional to s as well. In writing explicitly the above massless poles. We have also normalized the amplitude (12) at this point by the coupling factor −iκc/2π. we find    iκc sin(πl) l Tr(D) + 2 − 4k1 ·V ·V ·k2 + 1 ↔ 2. (15) A0s (τ1 . τ2 .

one finds    iκc sin(πl) s √ − Tr(D) + 2 − 4k1 ·V ·V ·k2 + 1 ↔ 2. the second terms in (17) can be reproduced only if the tachyon kinetic term appears nonlinearly in the determinant under the square root in the DBI action. we have used the covariant gauge VSab ∂a Aˆ b = 0. that is. These contact terms are reproduced exactly by the Lagrangian in (10).2. Replacing above propagator and vertices into (15). 3. (2). Garousi / Nuclear Physics B 584 (2000) 284–299 293 i (VS−1 )ab . l) . h3 ) = 4πl 2 Ac ≡ where again h3 stands for both graviton and Kalb–Ramond antisymmetric tensors. Vφ3 a = √ 2 2 2   p  eh3 a a = Tp κc 1 Tr(ε3 ·D)p3 ·VAa − p3 ·V · ε3T ·V a + V a ·ε3T · V ·p3 . τ2 . .2. V 2π Tp ea G ab = In writing the above propagator. (17) Ac (τ1 . i. + (a1 + a2 ) πl n=1 The factor sin(πl)/(πl) appears for all the contact terms which is consistent with the transformation of multiplication rule in (8). c s p    a Tp κc 1 a a a e Tr(D) + 2 p3 ·VA − p3 ·V ·V + V ·V ·p3 . φ3 ) = 2 8π 2l   iκc sin(πl) s − Tr(ε3 ·D) − 4k1 ·V ·ε3T ·V ·k2 + 1 ↔ 2.e. we now extract the low energy contact terms of the string amplitude (12). iκc sin(πl) (a1 − a2 ). The first terms in (17) by the terms in the first line of (10) and the second terms in (17) by the terms in the second line of (10).M. Note that. This ends our illustration of consistency between disk S-matrix elements and the proposed action (2). τ2 . while the first terms in (17) can be reproduced in field theory by an action in which the tachyon kinetic term appears linearly like the one proposed in [34]. one will find     a1 + a2 sin(πl) a1 − a2 sin(πl) iκc + A= 2 2π s 2 πl ! ∞ sin(πl) X ζ (2n + 1)l (2n+1) + k 2 O(s. Eq. The second term of the first line above is the contact term with minimum number of momentum in which we are interested.R. Expanding the gamma function appearing in this amplitude for s → 0. (16) 4πl Inserting appropriate polarization (see appendix) and adding the residue contact terms of the massless poles (13). Contact terms Having examined in detail the massless poles of string amplitude. V 2  eaτ1 τ2 a = c sin(πl) p k1 ·VSa + 1 ↔ 2. Ac (τ1 .. one finds exactly the string massless poles (14).

that is. respectively. one should start with non-abelian action for D9-branes and then use some sort of T-duality transformations to convert the D9-brane action to non-abelian action for Dp-branes. D ∂σ i 2πα 0 Using the fact that we are always working in the static gauge. Bµν and Fµν are the metric. Non-abelian action In this section we extend the proposed action (2) for a non-BPS Dp-brane to the case of N coincident non-BPS Dp-brane where the worldvolume theory involves a U(N ) gauge theory. T-duality transformations in i = p + 1. one has complete list of T-duality transformations for the fields appearing in (18). Z S = −T9 d 10 σ q   (18) × Tr e−Φ V (T ) −det(Gµν + Bµν + 2πα 0 Fµν + 2πα 0 Dµ T Dν T ) . That is. i. T ]. the gauge fields in those direction to e = T . the gauge trace should be completely symmetric between all non-abelian expression of the form Fµν . the non-abelian action may be constructed from the corresponding abelian case by simply extending the derivative of open string fields to its non-abelian covariant derivative [43. the first term on the right hand side becomes zero because of the assumption implicit in the T-duality transformations that all fields must be independent of the coordinates σ i .44]. Our guiding principle in constructing such a non-abelian action is that the action should be consistent with the familiar rules of T-duality. . Dµ T and individual T appearing in the tachyon potential V (T ).g.. 9 directions of the D9-brane worldvolume converts the D9-brane to Dp-brane.. Therefore. antisymmetric tensor and the non-abelian gauge field strength. T i X are now transverse coordinates of the new Dp-brane. [38]). the covariant derivative of tachyon becomes ei Te = ∂T − i [Xi . . Under this transformation. This guideline has been recently employed by Myers [38] to construct non-abelian DBI action. In this case there is no scalar field corresponding to the transverse direction of D9-branes. ∂σ µ This action is still incomplete without a precise prescription for how the gauge trace should be implemented. . Now adding this transformation to the T-duality transformation of massless fields (see. We expect that a prescription similar to that given for Born–Infeld action [45] should also be given here. To this end. In this way. Now we generalize (18) to the action appropriate for non-BPS Dp-branes for any p. .e. . The new scalar fields A˜ i = Xi /2πα 0 and leaves unchanged the tachyon. Hence. and a trace over the U(N ) representations [45] (see also [17]). e. we apply familiar T-duality transformations rules to the non-abelian D9brane action (18). T ]. Garousi / Nuclear Physics B 584 (2000) 284–299 4. we begin by extending the abelian action (2) to non-abelian action for non-BPS D9-branes.R. and Dµ T = ∂T − i[Aµ .294 M. where Gµν .

T ]Db T E eij − i 0 [Xi . The latter non-abelian field stems from non-abelian Taylor expansion of the closed string fields that appear in the DBI action [46]. Xj ] − 1 0 [Xi . (20) D where now the definition of the pull-back above is the extension of (1) in which ordinary derivative is replaced by its non-abelian covariant derivative. − iEai (Q j i j In Eqs. Xi . T ]Db T Qi j = δ i j − −1 i − iDa X i −1 (21) ) j [X . 2πα 0 eai = 1 Da Xi . T ][Xk . T ](Q−1 )ij Db Xj . E Eia = −E ij Ej a . F F 2πα 0 i eia = − 1 Da Xi . indices are raised and lowered by E ij and Eij . Here E ij denotes the inverse of Eij . If the tachyon field is set to zero. D e2φ = e2φ det(E ij ). Xj ] and individual Xi . Garousi / Nuclear Physics B 584 (2000) 284–299 eab = Eab − Eai E ij Ej b . E eij = E ij . i[Xi . one finds the final T-dual action Z  e S = −Tp d p+1 σ Tr e−Φ V (T )V 0 (T . T ]Db T − iDa T [Xi . i. D eib − Db Xi − i[Xi . This potential term is one for abelian case. eij = − [Xi . this action would get to the result of non-abelian action for N coincident BPS Dp-branes [38].. T ]. Da Xi . eab = Fab . respectively. Xi ) q  (22) × −det(P [Eab + Eai (Q−1 − δ)ij Ej b ] + 2πα 0 Fab + Tab ) . Da T . p where we have defined V 0 (T . Natural extension of this prescription for the trace in the action (22) is that the trace should be completely symmetric between all non-abelian expressions of the form Fab . E 295 eai = Eaj E j i . T ] E 2π α 2π α Manipulating the matrix inside the determinant. The trace is completely symmetric between Fab . Xj ]. i[Xi . 2πα 0 2πα 0 Tab = 2πα 0 Da T Db T − Da T [Xi . Here the matrices Qi j and Tab are defined to be i 1 [Xi .R. Xk ]Ekj − [Xi . . T ]Db T − iDa T [X . T ] and individual T of the tachyon potential. T ](Q−1 )ij [Xj .M. T ][Xj . i[Xi . ˜ ei Te = − i [Xi . Xj ]. Xi ) = det(Qi j ). T ](Q )i Ej b −1 (Q )ij [Xj . the prescription for the gauge trace is studied in [38]. F F 0 2 (2πα ) 2πα 0 (19) where we have defined Eµν ≡ Gµν + Bµν . E ik Ekj = δ i j . (20) and (21).e. Under above T-duality transformations the determinant in (18) becomes ! eaj + Da Xj − iDa T [Xj . Da Xi . one finds     e = det P Eab + Eai (Q−1 − δ)ij Ej b + 2πα 0 Fab + Tab det Qi j det(E ij ). T ] eab + 2π α 0 Fab + 2π α 0 Da T Db T E E e = det . T ]Ekj . In this case. Now replacing (20) into T-dual of (18) and using the transformation for dilaton field (19).

1. . 9. and for the worldvolume directions. . X = 0. Appendix A. Garousi / Nuclear Physics B 584 (2000) 284–299 Acknowledgements I would like to acknowledge useful conversation with R. . one may use a doubling trick to convert the propagators to standard form and give the modification to the vertex operators [40].C.R. . Myers. b = 0. . e.g. 8. . Now Eq. . e. In nontrivial D-brane background the vertex operator remain unchanged while the standard propagators need some modification. c = 0. one usually evaluate correlation function of their corresponding vertex operators with use of some standard conformal field theory propagators [47. early latin indices take values in the worldvolume. To find this matrix. 1. . So we are working on the upper-half plane with boundary at y = 0 which means ∂y is normal derivative and ∂x is tangent derivative. z¯ ) Xν (w. To this end consider the following general expression for propagator of Xµ (z. (25) can be solved for D ab . j = p + 1.. And our index conventions are that lowercase greek indices take values in the entire ten-dimensional space–time.48]. 9. and these equations are imposed at y = 0. e. which yields ηab − D ba − F ab − F a c D bc = 0 (25) = for the orthogonal directions. . for i = p + 1. Perturbative string theory with background field In perturbative superstring theories. Now we have to understand the modification of the conformal field theory propagators arising from these mixed boundary conditions. Alternatively. This work was supported by University of Birjand and IPM.296 M. In trivial flat background one uses an appropriate linear σ -model to derive the propagators and define the vertex operators. . . . z¯ ) fields: µ ¯ = −ηµν log(z − w) − ηµν log(¯z − w) ¯ X (z. . to study scattering amplitude of some external string states in conformal field theory frame. The modifications arising from the appropriate linear σ -model appear in the following boundary conditions [49]: 3 ∂y Xa − iF a b ∂x Xb = 0. Finally. In this appendix we would like to consider a D-brane with constant gauge field strength/or antisymmetric Kalb–Ramond field in all directions of the D-brane. (23) where Fab are the constant background fields.. . p. respectively. and middle latin indices take values in the transverse space. .g. otherwise.. . − D µν log(z − w) (24) where D µν is a constant matrix. i for a. a. 9. p. . The worldvolume (orthogonal subspace) indices are raised and lowered by ηab (N ij ) and ηab (Nij ). 1. i. ν = 0.g. we impose the boundary condition (23) on the propagator (24). w) ¯ − D νµ (¯z − w). . b. that is D ij Dab = 2(η − F )(−1) ab − ηab = 2Vba − ηab . µ. . . . our conventions set `2s = α 0 = 2. −N ij D ia =0 (26) (27) 3 Our notation and conventions follow those established in [40].

x) = ∂Xµ (x) + ik ·ψ(x) ψ µ (x) eik·X(x). (30) where G ab = (ηab + D ab )/2 = V ba for gauge field. Under these replacement. The physical conditions for the massless open string and tachyon are . k : V NS = :VnNS X(x) + X(x). D·ψ(¯z). k : . x) = ik ·ψ(x)eik·X(x). V τ =: Vn (2k ·V T . and p and k are closed and open string momentum. one can write the worldsheet fields in terms of right.and left-moving components.e. ψ(¯ ˜ z). V NSNS = :VnNS X(z). The vertex operator for massless NSNS and NS states and open string tachyon are V NSNS = (ε·D)µν :Vnµ (p. ψ(x) + D·ψ(x). we use the following doubling trick: eµ (¯z) → D µ ν Xν (¯z). ˜ z) → φ(¯z). φ(¯z). Note that the orthogonal property of the D matrix is an important ingredient for writing the propagators in the standard form.R. p : . x) = e−φ(x)ψ µ (x)eik·X(x). z¯ ): V NS = (ζ ·G)µ :Vnµ (2k ·V T . respectively. z): . φ(¯ (28) These replacements in effect extend the right-moving fields to the entire complex plane and shift modification arising from mixed boundary condition from propagators to vertex operators. worldsheet propagator between all right-moving fields take the standard form [50] except the following boundary propagator: iπ Xµ (x1 ) Xν (x2 ) = −ηµν log(x1 − x2 ) + F µν Θ(x1 − x2 ) 2 (29) where Θ(x1 − x2 ) = 1 (−1) if x1 > x2 (x1 < x2 ). Garousi / Nuclear Physics B 584 (2000) 284–299 297 where matrix V is the dual metric that appears in the expansion of DBI action (4). V−1 (k. µ V0 (k.  V NS = : VnNS X(x) + D·X(x). X ψ˜ µ (¯z) → D µ ν ψ ν (¯z). :Vmν (p·D. x): . 2φ(x).. The indices n. where ψ µ is super partner of worldsheet field Xµ and φ is worldsheet superghost field. In terms of these chiral fields. x) = e−φ(x)eik·X(x). In order to work with only right-moving fields. V−1 (k. ψ(z). i. p : . x): .M. φ(¯ ˜ z). m refer to the superghost charge of vertex operators. D µ α D να = ηµν . Using two-dimensional equation of motion. G ij = (ηij − D ij )/2 = N ij for scalar field and G ai = 0 otherwise. The open string vertex operators in (0) and (−1) pictures are  µ V0 (k. ψ(z). closed NSNS and open NS vertex operators are   e z). p : :VmNS D·X(¯z). p : :VmNS X(¯  e ˜ ˜ ψ(x) + ψ(x). The vertex operators under transformation (28) becomes   V NSNS = :VnNS X(z). φ(z). Note that the D µν is orthogonal matrix. φ(x) + φ(x). φ(z).

Suppl. Lovelace. hep-th/9812031. JHEP 9808 (1998) 010. An introduction to nonperturbative string theory. M. JHEP 9810 (1998) 021. hep-th/9611050. hep-th/9702201. B 441 (1998) 133.G. Vafa. B 460 (1996) 351. Mod. hep-th/9808141. B 545 (1999) 233. Phys. B 207 (1988) 169. hep-th/9512077. J. Phys. Thorlacius. Green. A.R. W. hep-th/9510161. hep-th/9607050. Callan. Quant.R. Mod. Maldacena. 68 (1996) 1245. A. Nucl. J. Int. A. C. hep-th/9902105. JHEP 9809 (1998) 023.A. J. Nucl. C. Sen. where the G S is symmetric part of the G matrix. hep-th/9605033. A. Lectures on strings and dualities. .R. Phys. M. Klebanov.H. Theor. Nucl. Proc. Int. S. V. A 11 (1996) 5623. Schomerus. Phys. J. C. hep-th/9802051. Johnson. Sen. (D·G T )ij = −N ij . Callen.G. C. Sen. Duff. Polchinski.A. O. J. D 50 (1994) 622. Mod. Sen. P. hep-th/9607201. M. Phys. A. JHEP 9912 (1999) 027. Phys. D-brane effective field theory from string field theory. Phys. Bergman. Nucl. J. Nappi. Notes on D-branes. hep-th/0001201. Branes within branes . Phys. A. and for massless closed string are p2 = 0 and pµ εµν = 0 where ε is the closed string polarization which is traceless and symmetric(antisymmetric) for graviton (Kalb–Ramond) and 1 (31) ε µν = √ (ηµν − `µ pν − `ν pµ ) . Born–Infeld action. Phys. W. Lett. Polchinski. Nappi. I.J. hep-th/9402113. Leigh. Phys. Taylor. G. hep-th/9602052. hep-th/9801182.A. Gaberdiel. Phys.298 M. Rev. 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] A. hep-th/9608117. Yost. Polchinski.A. Chaudhuri. Polchinski.G. Sen. hep-th/9911116. 14 (1997) 47. `·p = 1 8 for the dilaton.V. hep-th/9812135. A 14 (1999) 4061. (D·G T )ab = G ab . Sen. 2 (1999) 1373. Grav. Horava. Math. Taylor. Abouelsaood. Mod. Lett. Using the fact that D µν is orthogonal matrix. S.M. J. Nucl. Recknagel. Rev. TASI lectures on D-branes. hep-th/9806155. V. JHEP 9812 (1998) 021. hep-th/9908105. Schwarz. hep-th/9811237. Tseytlin. A. hep-th/9912249.W. A 4 (1989) 2767. Moore. C. A. B 422 (1994) 417. Zwiebach. Lett. Callan. one finds the following identities: G ·G T = G S . L. B 308 (1988) 221. gauge theory and M(atrices). B 55 (1997) 1. Tachyon condensation in string field theory. B 280 (1987) 599.R. Samuel. C.R. Phys. Kostelecky. 1 k ·VS ·k = 4 k ·VS ·ζ = 0. Douglas. hep-th/9404008. M. C. A. Class. J. Adv. B. supersymmetry and string theory. hep-th/9805019. Lectures on D-branes. S. A.G. A. Ludwig. Yost. C. Garousi / Nuclear Physics B 584 (2000) 284–299 massless: tachyon: k ·VS ·k = 0 . R. Nucl. M.A. J. Phys. Li. Sen. hep-th/9809111. Harvey. Phys. Sen.R.

Garousi. Garousi. [50] M. JHEP 9909 (1999) 032. Nucl. hep-th/9910053.R. Nucl. 1987. [44] C. [47] M. Billo. Lett. Garousi. Nucl. Klebanov. Kraus. I. S. [34] A. hep-th/9809100. [45] A. S. The tachyon potential in open Neveu–Schwarz string field theory.). B 288 (1987) 173. [31] J. JHEP 9906 (1999) 033. [46] M.C. B 494 (1997) 105. hep-th/9908142.M. B 542 (1999) 73.R. Hashimoto. [32] N. Level truncation and the tachyon in open bosonic string theory. Seiberg. Phys. Proc. Non-BPS States and Branes in String Theory. Sen. Introduction to string and superstring theory II. Sen. J. Craps. [42] A. R. hepth/0001084. W.M. . Tseytlin. From the planck scale to the weak scale: toward a theory of the universe. [36] M.G.C. hep-th/9612120. [37] N. of TASI’s86. Tachyon condensation in superstring field theory. [38] R.A. [48] V. W. JHEP 9912 (1999) 022. [41] A.E. Phys. hepth/0002117. [40] M. Lerche. B 501 (1997) 41. Taylor. Lett. Peskin. Leigh.R. hep-th/9904207. B.A. hep-th/9603194.R. [33] N. B 381 (1996) 437. World Scientific. Lechtenfeld.R. B 475 (1996) 193. Haber et al. Dai. JHEP 9810 (1998) 011. Sen. hep-th/9611214. Klebanov. E. (Eds. Garousi / Nuclear Physics B 584 (2000) 284–299 299 [30] N. O. Witten. Phys. Dorn. Kostelecky. Harvey. Proc. Myers. hep-th/9604065. hep-th/9805078. Polchinski.R. hep-th/9701125. Nucl. [35] A. Phys. Phys. Roose. A 4 (1989) 2073.A. Phys. [43] H. Phys. B 55 (1997) 118. Garousi. Suppl. Nucl. Non-commutative world-volume interactions on D-branes and Dirac–Born– Infeld action.C. A. hep-th/9905157.R. R. hep-th/9909062. Myers. Hashimoto. Moeller. hepth/0002237. JHEP 9812 (1998) 008. Zwiebach. in: H. hepth/0002211. [39] M. hep-th/9711179. Myers. JHEP 9910 (1999) 008. Berkovits. I. R. Mod. D-branes as unstable lumps in bosonic open string field theory. B. Samuel. Berkovits. P. Watamura. Nucl. [49] J. Hull. hep-th/9909214. Phys. F.

E-mail addresses: ghoshal@mri. PACS: 11. since the exact effective action for the tachyon field is not known.Nuclear Physics B 584 (2000) 300–312 www.sen@cern. and that this configuration represents the closed string vacuum without any D-brane [1]. India Received 11 May 2000.in (D. accepted 6 June 2000 Abstract It has been conjectured that an extremum of the tachyon potential of a bosonic D-brane represents the vacuum without any D-brane.1 .ch.in (A.8. Solitons 1.25.ernet. Jhusi.  2000 Elsevier Science B. It has been conjectured that the tachyon potential has a non-trivial extremum where the potential energy of the tachyon exactly cancels the tension of the D-brane. In this paper we point out that in the p-adic string theory introduced in [24–33] (see [34] for a review) we can explicitly check these conjectures. However. the spacetime it describes is the familiar ∗ Corresponding author.nl/locate/npe Tachyon condensation and brane descent relations in p-adic string theory Debashis Ghoshal ∗ .Sq. 11. sen@mri. These conjectures and their generalisations to superstring theories [5–9] have been tested by various methods [2–4. the expression for which is known exactly. Chhatnag Road. there is no direct proof of these conjectures. Introduction and summary The world-volume theory on the D-brane of a bosonic string theory contains a tachyonic mode.elsevier. We show that the tree level effective action of p-adic string theory. All rights reserved. provides an explicit realisation of these conjectures.10–23]. All rights reserved. Ashoke Sen Mehta Research Institute of Mathematics and Mathematical Physics.10.27. and that various tachyonic lump solutions represent D-branes of lower dimension. ashoke. PII: S 0 5 5 0 .V.Lm. It has been further conjectured that various tachyonic lump solutions on the D-brane world-volume represent D-branes of lower dimensions [1–4].V. Sen). D-branes. Ghoshal). It should be emphasised that although the p-adic ‘string’ is an exotic object.ernet. 11.+d Keywords: p-adic string theory. 0550-3213/00/$ – see front matter  2000 Elsevier Science B. Allahabad 211019.3 2 1 3 ( 0 0 ) 0 0 3 7 7 .

it is possible to compute N tachyon amplitudes at tree-level for all N > 3. 1 In the p-adic open string theory. g is the open string coupling constant.34] relating the product of four tachyon amplitudes in p-adic strings for all primes p to that in the bosonic string suggests that they all have the same critical dimension d = 26. which approach the vacuum solution far away from the core of the soliton. This provides strong evidence that these lump solutions can be identified as lower-dimensional D-branes. Section 5 contains some comments on further extension of this work. to the extent that we can compare them with the present knowledge. 1)dimensional space–time.36]. but the so called adelic formula [25. Using p-adic analysis. [24] p-adic string theory was defined as follows. 2.D. Thus this configuration has no physical open string excitations. Consider the expressions for various amplitudes in ordinary bosonic open string theory. Sen / Nuclear Physics B 584 (2000) 300–312 301 one. φ is the tachyon field (after a rescaling and a shift). and ends with speculation on its possible application to the study of tachyon condensation in ordinary bosonic string theory.38]. the exact classical action of the tachyon field and various solutions of the equations of motion are known [26]. This action is given in Ref. In Section 2 we summarise the exact effective action of the tachyon in the p-adic string theory. We are 1 A different type of p-adic string was considered in [35. and that the propagator of the tachyon field describing fluctuations around this background has no physical pole. The paper is organised as follows. This leads to an exact action for the open string tachyon in d-dimensional p-adic string theory. and p is an arbitrary prime number. and the norms of the functions in the integrand by the p-adic norms. We show that the world-volume theory on the solitonic q-brane agrees with the expected world-volume theory on a Dirichlet q-brane in the p-adic string theory. Now replace the integrals over R by integrals over the p-adic field Qp with appropriate measure. (2.1) S = d xL= 2 2 p+1 g p−1 where 2 denotes the d-dimensional Laplacian. . Ghoshal.37. If the original open string theory is formulated in (d − 1. [26]   Z Z 1 p2 1 −12 1 d d p+1 φ d x − φp 2 φ + . Among the known non-trivial solutions is a translationally invariant solution with the property that it is a local minimum of the potential. In Section 3 we analyse the world-volume theory of the solitonic q-brane. written as integrals over the boundary of the world-sheet which is the real line R. A. which in modern language can be regarded as the world-volume theory of a space-filling D-brane. and is naturally identified with the vacuum without a D-brane. and in Section 4 we compare this with the world-volume theory of a Dirichlet q-brane. These rules were subsequently derived from a local action defined on the “world-sheet” of the p-adic string [34. The exact tachyon equation of motion of the p-adic string theory also has classical lump solutions for all codimension > 1. 2 then such a lump solution of codimension (d − q − 1) describes a solitonic q-brane. 2 There is as yet no compelling reason for a critical dimension in p-adic string theory. the known solutions of the equation of motion derived from the action and their properties. Solitonic q-branes of p-adic string theory In Ref.

1 shows the qualitative features of the tachyon potential for different values of p. then our choice of units correspond to [37.6) describes a soliton solution with energy density localised around the hyperplane x q+1 = · · · = x d−1 = 0. . . +). 1.38]. x d−1 . Since the action is symmetric under φ → −φ. (2. Sen / Nuclear Physics B 584 (2000) 300–312 using metric with signature (−. Fig.4) 2g 2 p + 1 • The configuration φ = 0 denotes a configuration around which there is no perturbative physical excitation. . Fig. 1 p− 2 2 φ = φ p . (2. Ghoshal.302 D. +.7) p− 2 ∂η f (η) = f (η) . 3 For p 6= 2. .5) φ(x) = f x q+1 f x q+2 · · · f x d−1 ≡ F (d−q−1) x q+1 . we shall restrict our analysis to solutions with positive φ. • The configuration:     (2. . exp − 2 p ln p (2. which can be identified as the tension Td−1 of the D-(d − 1)-brane configuration.3) Different known solutions of this equation are as follows [26]: • The configuration φ = 1 is the original vacuum around which we quantised the string. Td−1 = −L(φ = 1) = with f (η) ≡ p 1 2(p−1)   1p−1 2 η . is given by 1 p2 . there is also an equivalent solution corresponding to φ = −1. [37. The effective tachyon potential for the p-adic string. . . The equation of motion derived from this action is. The energy density associated with this configuration. This follows from the identity: p 1 2 (2. A. +. .38] ln p . If we denote by (2παp0 )−1 the ‘tension of the p-adic string’ as defined in Refs. By definition we have taken the energy density of this vacuum to be zero. (2.2) αp0 = 2π We have added a constant term to the Lagrangian density L so that it vanishes at φ = 0. . We shall identify this with the vacuum configuration. 3 We shall call this the D-(d − 1)-brane solution.

10) is independent of q.2) . this approaches the formula for the self-dual radius in ordinary bosonic string theory. is given by Z 1 p2 . 1 4 In the linearised approximation this tachyonic mode was discussed in Ref. (2.10) p Tq−1 p2 − 1 p p−1 2π α 0 p In the above equation we have used dimensional analysis and (2. . We shall call this the tachyon field on the solitonic q-brane world-volume. Let us denote by x⊥ = (x q+1 . World-volume theory on the solitonic q-branes Let us now consider a configuration of the type φ(x) = F (d−q−1) (x⊥ )ψ(xk ).8) Tq = − d d−q−1 x⊥ L(φ = F (d−q−1) (x⊥ )) = 2 2gq p + 1 where. This also suggests that the self-dual radius Rsd of the p-adic string theory. 3.1) into (2. [39].7) we get p − 2 2k ψ = ψ p . . (3. with f as in (2. This is a feature of the D-branes in ordinary bosonic string theory. .9) we see that the ratio of the tension of a q-brane to a (q − 1)-brane is 2p   1 p 2 Tq 2π p p−1 ln p − 2 p −1 1 q .8) and (2. x q ) those tangential to it. Ghoshal.4). For ψ = 1 this describes the solitonic q-brane.9) From Eqs. the solitonic q-brane solution. and suggests that the solitonic q-branes of p-adic string theory should have interpretation as D-branes. (2.  gq = g p2 − 1 (d−q−1)/4 2π p2p/(2p−1) ln p . Note that the ratio (2. = = (2. .5).11) Note that as p → ∞. The energy density per unit q-volume of this brane.3) and using (2. thus ψ(xk ) can be regarded as one of the fields on its world-volume. . (3.1) with F (d−q−1) (x⊥ ) as defined in (2. (2.6).5). which can be identified as its tension Tq . . Rsd = p p2 − 1 (2. x d−1 ) the coordinates transverse to the brane and by xk = (x 0 . Fluctuations of ψ around 1 denote fluctuations of φ localised on the soliton.2) to restore factors of αp0 .D. Sen / Nuclear Physics B 584 (2000) 300–312 303 We shall call (2.6). . A. where the tension 2πRsd Tq of a wrapped q-brane is equal to the tension Tq−1 of a (q − 1)-brane. (2. (2. 4 Substituting (3. is given by pp/(p−1) q 0 αp . .

Ghoshal. A.2) gives an exact solution of the full equation of motion (2.  1 (3. we get the following equations of ⊥ motion:  1 p − 2 2k ψ = ψ p + O ξ 2 . For computing amplitudes involving the world-volume fields on the q-brane. can be calculated exactly from the action (3. The above equations can be derived from the effective action:  Z 1 1 p2 1 1 i q+1 ψ p+1 xk − ψp− 2 2k ψ + d Sq (ψ.2) describes the dynamics of the mode ψ on the q-brane world-volume exactly. ⊥ (3.3). 2 2 (3.3). Note that a solution of (3. as it can be changed by rescaling ξ i . Sen / Nuclear Physics B 584 (2000) 300–312 where 2k denotes the (q + 1)-dimensional Laplacian involving the world-volume coordinates xk of the q-brane. d q+1 xk − ψp− 2 2k ψ + = 2 (3.1):  Sq (ψ) = S φ = F (d−q−1)(x⊥ )ψ(xk )   Z 1 1 1 p2 1 ψ p+1 . ξ ) = 2 2 p+1 gq p − 1     1 i − 1 2k i 1 p−1 i i 3 2 −C ξ p ξ − ψ ξ ξ +O ξ . we define the shifted field σ and rescaled fields χ i through the relation gq gq σ . involving only external tachyon states. ψ =1+ p pC and expand the action (3.1) into (2. This gives . The action involving ψ can be obtained by substituting (3.3). ξi = √ (3. Inclusion of these modes correspond to deformation of φ of the form φ(x) = F (d−q−1) (x⊥ ) ψ(xk ) + ∂x i F (d−q−1) (x⊥ ) ξ i (xk ) + · · · .5) p− 2 2k ξ i = ψ p−1 ξ i + O ξ 2 . In terms of scattering amplitudes this means that the tree level S-matrix on the q-brane world-volume.9). (2. (3. and comparing the coefficients of the independent functions (F (x⊥ ))p and (F (x⊥ ))p−1 ∂x i F (x⊥ ) on both sides. Thus Eq.4) Substituting this in Eq.304 D. Of the various other (infinite number of ) modes living on the q-brane world-volume are the (d − q − 1) massless modes ξ i associated with translations of the brane in the (d − q − 1) directions x⊥ transverse to the brane. ψ = 1 corresponds to the solitonic q-brane solution. This does not mean that there are no other modes on the q-brane world-volume.7) χi.3) gq p − 1 2 p+1 where gq has been defined in Eq. rather what this implies is that it is possible to obtain a consistent truncation of the world-volume theory of the q-brane by setting all the modes except ψ to zero.6) in powers of σ and χ i .6) C is a normalisation constant whose value is not important to this order. (2.

According to these authors. p p 12 k 2 − 1 (3. A. we shall content ourselves with the continuation of the relevant formulæ from ordinary bosonic string theory.8) reflecting the fact that σ = 0 is a solution of the equation of motion. The momentum space σ and the χ i propagators computed from this action are given by: 1σ σ (k) = − i 1χ i χ j (k) = − i 1 p−1 . and then replacing the integrals over real line by integrals over the p-adic field. [26].9) The residues at the poles in the σ and the χ i propagators (at k 2 = 2 and k 2 = 0.38]. Sen / Nuclear Physics B 584 (2000) 300–312 p Sq = p−1 305   gq σ p+1 σ 1 p p 1 − 1 2k −1 2 1+ xk − σp σ+ 2 − − 2 2 gq p + 1 p gq 2gq     gq σ p−1 i i 1 i − 1 2k i 1 3 2 1+ − χ p χ + χ χ + O χ . Thus. While this may be the proper way to define D-branes in p-adic string theory.38]. This will help us compare the amplitudes involving σ ’s and χ i ’s in the external legs. with all the norms appearing in the integrand replaced by p-adic norms as in Ref. Now one can either choose Neumann or Dirichlet boundary conditions as in the case of ordinary strings. [37. the computation of the amplitude is identical to the one described in Ref. The generalisation of the Polyakov action is the lattice discretisation of the action for free scalar fields corresponding to the target space coordinates. a ‘world-sheet’ approach to p-adic string has been developed in Refs. for our purposes the world-volume theory of a p-adic D-q-brane is defined by taking the expressions for various amplitudes for an ordinary D-q-brane. (3. Fortunately.38] that Neumann boundary conditions leads to the tachyon amplitudes postulated in [24. It was shown in [37. Comparison with the world-volume theory on a Dirichlet q-brane We shall now compare the world-volume action on the solitonic q-brane with that on the D-q-brane. Ghoshal.8) 2 2 p  Z d q+1 There is no linear term in σ in action (3. 1 2 p p 2 k −1 − 1 1 p−1 δij . described by the vertex operators of the type eik·Xk with momentum k restricted to lie along the world-volume of the D-q-brane. these S-matrix elements can be obtained from an effective action of the form: . We are immediately faced with the question whether it is possible to have Dirichlet branes in p-adic string theory.26]. respectively) have the same values.D. written as integrals over world-sheet coordinates of the appropriate vertex operators. For amplitudes involving the external tachyons. [26]. Thus. [37. 4. this ‘world-sheet’ is a so-called Bruhat–Tits tree — a Bethe lattice with p + 1 nearest neighbours — the ‘boundary’ of which is the field of p-adic numbers Qp . following the analysis there. In principle one should be able to derive these rules from the world-sheet description in Refs.

In deriving (4. These can be easily evaluated. First let us compute this on the solitonic q-brane using the action (3. Sen / Nuclear Physics B 584 (2000) 300–312 1 p2 b Sq (ψ) = 2 gˆq p − 1 Z   1 1 1 ψ p+1 . k3 . A.3) we see that the world-volume actions for the tachyon fields on the solitonic q-brane and the Dirichlet q-brane agree exactly if we choose: gˆq = gq . The Feynmann diagrams contributing to the amplitude hσ σ χχi on the solitonic q-brane. (4.4) Fig.3) + k ·k +1 p 1 4 −1 with the contribution to the four terms in the right hand side of (4. and the answer is:   p−1 1 1 2 p−2 + + k ·k +1 Aij (k1 . . 2. (4. Ghoshal.3) coming from the Feynman diagrams (a). 5 This is related to the problem of computing the tension of the D-q-brane independently. (b). k22 = k32 = 0 . Comparing this with (3.306 D.2) to be true. k2 .8). respectively.3) we have used the mass-shell conditions k12 = k42 = 2. A dashed line denotes the χ propagator and a solid line denotes the σ propagator. (4. 2. (c) and (d). 2. we have a complete agreement between the world-volume theories involving the tachyon fields on the D-qbrane and the solitonic q-brane. d q+1 xk − ψp− 2 2k ψ + 2 p+1 (4.2) independently. 5 But assuming (4. k4 ) = δij gq k ·k +1 1 2 1 3 p p p −1 p −1  1 . (4.1) where the tachyon field ψ is shifted so that ψ = 1 describes the D-q-brane. in Fig. The four Feynmann diagrams contributing to it have been shown in Fig. and hence we cannot verify Eq. and gˆq denotes the coupling constant which characterises the strength of the interaction in the worldvolume theory of the D-q-brane.2) At present there is no independent derivation of gˆ q in terms of g. Next we shall compare the amplitude hσ σ χ i χ j i on the solitonic q-brane and the D-qbrane.

and consider an expansion of the field φ of the form: s d−q−1 √ √  C X i e e(z) + ξ (z) 2 cos( 2 ui ) + · · · . The dots stand for higher momentum modes which will not be required for our analysis. ¯ and compactify 7 (d − q − 1) directions on circles of radii 1/ 2. k2 . This is precisely the integral evaluated in [26].1) (with g replaced by g) ¯ we get the action:    Z 1 e − 1 2z e 2π d−q−1 1 ep+1 1 p2 q+1 2 √ z − ψp ψ ψ+ d 2 g¯ p − 1 2 p+1 2    1 i − 1 2z i 1 p−1 i i e e ξ p 2 e ξ − ψ ξe ξ + O(e −C e ξ 3) + · · · .5) Aˆ ij (k1 .7) This agrees precisely with Eq.9) 6 The notation ∂X is schematic. respectively. but just examine those modes of φ which carry √ either 0 or ± 2 units of momentum in (d − q − 1) of the directions. Using the identity k1 · k2 + k1 · k3 + 2 = −k1 · k4 . Sen / Nuclear Physics B 584 (2000) 300–312 307 Let us now evaluate the same amplitude on the D-q-brane. as care is needed to define the correct vertex operator that corresponds to the analogous one for ordinary string. Substituting this into (2.3) for gˆq = gq .8). (4. To see this. Let ui denote the compact coordinates and zµ the non-compact ones. Inserting the χ i and χ j with the mode χ i on the D-q-brane is given by 6 ∂X⊥ vertex operators carrying momenta k2 and k3 at 0 and 1. agrees with the corresponding amplitude on a Dirichlet q-brane.6) we can express this amplitude as   1 1 p−2 p−1 + + Aˆ ij (k1 . + k ·k +1 p 1 4 −1 (4. k3 . k2 . computed from the action (3. this gives a consistent truncation of the theory at the tree level. 7 Alternatively.1) with g replaced by √ another coupling constant g. A. Ghoshal. 2 2 (4. it is possible to give a general argument showing that an amplitude with two external χ fields and N external σ fields for arbitrary N .D. k4 ) = δij gˆq2 Qp Here | | denotes the p-adic norm and integral over x is over the p-adic field. we can express the amplitude as: Z dx |x|k1 ·k2 |1 − x|k1 ·k3 . (4. . (4. (4. k4 ) = δij (gˆq )2 p p pk1 ·k2 +1 − 1 pk1 ·k3 +1 − 1  1 . The vertex operator associated i ik·Xk e . let us consider the situation where we start with the action (2.8) φ(x) = ψ p i=1 We have restricted φ to be even under ui → −ui for each i. respectively. we can work with the uncompactified theory. and the two σ vertex operators carrying momenta k1 and k4 at x and ∞. k3 . In fact.

9) in a and compute the S-matrix elements around the vacuum ψ i power series in e σ and χ e . It is not easy to establish this in all generality.9) is identical to element involving a χ ei .8) are identical to the corresponding amplitudes in the compactified string theory. 8 On the other hand. looks different from the χ ei vertex operator even after we identify Xk given by ∂X⊥ with Z. This establishes the desired result. Ghoshal. However if√we note that half plane the two point √ on the √ boundary of the upper √ j i (x1 )∂X⊥ (x2 )i are identical. e ξ i and this action looks identical to the one in (3. To summarise.8) around the ψ ej . we can only make this claim for two or less external χ (e χ ) particles. however we can consider a subset of the massive modes on the q-brane and show the agreement between the S-matrix elements on the solitonic q-brane and D-q-brane with at most two of these states on the external leg. In ej e σ N i is given in terms of correlation functions of χ ei . This provides strong evidence that the solitonic q-branes of the p-adic string theory should be identified with Dirichlet q-branes.9) and (3. which in turn are identical to the corresponding ones on the D-q-brane. pC (4. ξ i replaced by ψ the identification xk ∼ z. In presenting this argument we have not been careful about the overall normalisation factors.9) around the ψ = 1 background are identical e = 1 background. operator for e σ is proportional √ The√vertex ik·Z i i ik·Z . It will be interesting to systematically extend this comparison to S-matrix elements involving more than two external χ i states. χ ej and particular the amplitude he χiχ Ne σ vertex operators on the upper half plane. the S-matrix elements computed from (4.11) e = 1 by expanding (4. In particular the S-matrix to those computed from (3.10) e.9) holds only to quadratic order in χ (e χ ). Comparing this with the to e corresponding computation for the D-q-brane we see that the e σ vertex operator is identical to the σ vertex operator with Xk replaced by Z. whereas that for χ e is given by 2 cos( 2 U )e . we can conclude that these particular amplitudes in the compactified string theory are indeed identical to those on the D-q-brane. a χ i j the S-matrix element involving χ . A. and also to S-matrix elements involving higher level states. but the equality already established for the amplitudes hσ N i and hχ i χ j σ σ i in the two theories guarantees that the overall normalisation factors also agree in the two theories.6) with the fields ψ. i ik·Xk e . and an arbitrary number of e σ quanta for (4. (3. p gq e ξi = √ χ ei .7) e=1+ ψ gq e σ . The χ i vertex operator on the D-q-brane. functions h 2 cos( 2 U i (x1 )) 2 cos( 2 U j (x2 ))i and h∂X⊥ −2 both being equal to δij |x1 − x2 | . Sen / Nuclear Physics B 584 (2000) 300–312 If we identify gq2  √ d−q−1 2 = g¯ . 2π 2 (4. as the action is obtained by compactifying a p-adic string theory [41].6) (and hence (3. χ and an arbitrary number of σ quanta in (3. we have shown that the amplitudes hχ i χ j σ N i computed from (3. Similarity of (4.308 D.9) have direct string theory interpretation.6) and (4. In particular we can define the analogues of Eqs.8)) shows that the Smatrix elements computed from the action (4. 8 Since the similarity of (3.8). .

2 (4. and using Eq. Sen / Nuclear Physics B 584 (2000) 300–312 309 We start with the solitonic q-brane. and consider a generalisation of the expansion (3. · · · ir } for which no two in the set are equal. We shall now compare the S-matrix elements computed from the action (4. Ghoshal. Vi1 ···ir = ∂X⊥ (4. ir } are all different. In this case p ∂ i1 · · · ∂x ir F (d−q−1)(x⊥ ) x⊥ ⊥ p−1 = pr F (d−q−1) (x⊥ ) ∂x i1 · · · ∂x ir F (d−q−1) (x⊥ ). For arbitrary number of external tachyon legs.15) with two or less external ξ -legs to that computed directly in the D-q-brane. in the ψ = 1 background.  1 p− 2 2k ξ i1 ···ir = p1−r ψ p−1 ξ i1 ···ir + O ξ 2 .13) ⊥ ⊥ Substituting (4. (4.12) P where 0 above denotes sum over those indices {i1 ..16) as long as the indices in the set {i1 ..14) The action involving these fields is given by  Z 1 1 p2 1 1 q+1 ψ p+1 xk − ψp− 2 2k ψ + d Sq (ψ. ⊥ (4.3). We take this to be i1 ir ik·Xk · · · ∂X⊥ e .ir }    1 − ψ p−1 ξ i1 ···ir ξ i1 ···ir +O ξ 3 . A.. (4.4): φ(x) = F (d−q−1)(x⊥ )ψ(xk ) + d−q−1 X X0 ∂ r=1 {i1 . . (4. ξ ) = 2 gq p − 1 2 p+1 d−q−1 X X0  1 1 − Cr ξ i1 ···ir p− 2 2k +r−1 ξ i1 ···ir 2 r=1 {i1 .16) In order to compare this with the world-volume theory on the D-q-brane.ir } i x⊥1 · · · ∂x ir F (d−q−1)(x⊥ )ξ i1 ···ir (xk ) + · · · . . is k 2 = 2(1 − r) .13) we get  1 p − 2 2k ψ = ψ p + O ξ 2 .D... (4. one can generalise the argument given for external χ i -legs. The key ingredient of this argument is that the . From this we see that the mass-shell constraint for ξ i1 ···ir . For two external tachyons and two external ξ we can do this explicitly and verify that it agrees with the corresponding computation on the D-q-brane. .12) into the equation of motion (2.15) Cr is a normalisation constant which can be absorbed into the definition of ξ i1 ···ir . we need to first identify the vertex operator corresponding to the mode ξ i1 ···ir . (4. The computation is identical to the one discussed earlier.17) This describes a physical state satisfying the same mass shell constraint as Eq..

Thus one might hope that the full tachyon effective action in bosonic string field theory is related in some way to the tachyon effective action in p-adic string theory. then by switching on Wilson lines corresponding to the gauge fields followed by a T-duality transformation. Indeed. Thus by examining the tree level tachyon amplitudes in the world-volume theory we shall not discover the existence of the other modes. In this direction. Furthermore if there are massless gauge fields in the spectrum of open strings in the p-adic string theory. [26] that it is possible to assign Chan–Paton factors to the open string states of a p-padic string theory. Ref. Ideas developed in Ref. .1) α 0 2 + 1 e−c α 2 − 2 φ = gφ . . [40] attempted to generalise the p-adic string amplitudes to external vector states. If these modes are present they will give rise to new degrees of freedom on the solitonic q-brane. there is evidence of close relationship between tachyon amplitudes in the p-adic and ordinary bosonic string theory [25. we would like to point out that even if the tachyon potential in the p-adic string theory is totally unrelated to that in the ordinary bosonic string theory. To lowest order in the level truncation scheme [13. • It is natural to ask if this analysis has any relevance to the ordinary bosonic string theory. . it can be regarded as a toy model which nicely illustrates the features expected of the full bosonic string field theory action. we can produce static configuration of D-q-branes separated in space. and will have to be taken into account in comparing the world-volume theory on the D-q-brane with that on the solitonic q-brane. Firstly. . Comments • We have shown that one can get a consistent truncation of the world-volume theory on a D-q-brane in p-adic string theory by keeping only the tachyonic mode. Besides. .3) admit such solutions. and compare with Eq. the tachyon equation of motion in open bosonic string field theory may be written as [20] i h  0 ¯ 2. (5. cos( 2 U ir )e ik·Z 2) cos( 2 U identical to that between the vertex operators ( √ √ √ 0 0 0 and ( 2)r cos( 2 U i1 ). This suggests that there may be other (massless and massive) modes living on the world-volume of the space-filling D-(d − 1)-brane as well. 5. and a proper understanding of the closed string sector of the theory. It will be interesting to examine if the equation of motion (2. Ghoshal.38]. . • It has been shown in Ref.310 D. This will require careful analysis of the cylinder amplitude. we cannot resist the temptation to point out some apparent similarities between the equation of motion (2.14]. .8) describing the tension of a solitonic q-brane. This shows the existence of multiple D(d − 1)-branes. (2. and if there is a T-duality transformation relating the D-(d − 1)-brane to D-q-brane.37. [42] may be useful in this context. cos( 2 U ir )eik ·Z of the compactified string theory.3) and that in the open bosonic string field theory [43]. A. . Sen / Nuclear Physics B 584 (2000) 300–312 i0 i0 0 i1 ir ik·Xk r ik ·Xk 1 two point function of ∂X⊥ · · · ∂X⊥ e and ∂X is √ √ ⊥ r· · · ∂X√⊥ e i for the D-q-brane 1 ). • It will also be of interest to compute the tension of a Dirichlet q-brane in the padic string theory independently. inspite of the fact that there are no poles in the tachyon S-matrix elements corresponding to these states.

JHEP 9808 (1998) 012. JHEP 9808 (1998) 010. [10] A. Lett. [15] A. [7] E. [6] A. [16] N. Phys.D. Descent relations among bosonic D-branes. hepth/0002211. Acknowledgement We would like to thank Sunil Mukhi for collaboration in the initial stages and for many useful discussions. Samuel. JHEP 9809 (1998) 023. Phys. then this equation. so that φ = 0 is the vacuum without any D-brane. Sen. Nucl. Frau. B 564 (2000) 60.G. Phys. Nucl. Gallot. Sen. Potting. A 14 (1999) 4061. J. B. Phys. Sen. If we drop the first and the third terms on the left hand side of Eq. A. Ghoshal. [3] C. K-theory. hep-th/9805019. hepth/0001084. Int.M. Berkovits. Sen. . and matrix theory. Vortex pair creation on brane–antibrane pair via marginal deformation. JHEP 9812 (1998) 021. Stable non-BPS D-branes in type I string theory. hep-th/9404008. Majumder. P. hep-th/9808141. hep-th/9812135. V.W. Adv. Kostelecky. hep-th/9811237. [9] P. Sen. B 422 (1994) 417. hep-th/9810188. hepth/9902105. Exact solution of a boundary conformal field theory. Expectation values Lorentz invariance.A. Stable non-BPS bound states of BPS D-branes. Phys. (2. Maldacena. A. JHEP 9812 (1998) 019. S. [8] A. Type IIA D-branes. Zwiebach. Callan. Lett. Ludwig. hepth/9805170. B 381 (1996) 89. D 50 (1994) 622.1) by hand. Tachyon condensation in superstring field theory. B 545 (1999 ) 233. [13] V. [4] J. so we shall not pursue this matter any further. 2 (1999) 1373. Sen. Math. Thorlacius. Strigazzi. reduces to Eq. [14] V. Witten.3) for p = 2. [11] M. D-branes and K-theory. g¯ is open string coupling constant after suitable normalisation. Kostelecky. Sen / Nuclear Physics B 584 (2000) 300–312 311 where c = ln(33 /42 ). B 207 (1988) 169. [17] N. and φ = −1/g¯ denotes the D-brane. Tachyon condensation on the brane antibrane system. hep-th/9605088. after suitable rescaling of x and φ. hep-th/9912249. but it is not inconceivable that some exact relation between bosonic string field theory and p-adic string theory will be discovered in the future.A. Lerda. R. Mod. Nucl. Recknagel. I. Schomerus. Boundary deformation theory and moduli spaces of D-branes. BPS D-branes on non-supersymmetric cycles. Of course there is no justification for dropping these terms. hep-th/9812031. L. Free fermion representation of a boundary conformal field theory. References [1] A. hep-th/0003124. Phys. A. [2] A. [5] A. Phys. and CPT in the open bosonic string. and 0 φ is related to the original tachyon field T by a field redefinition T = e−c α 2/2 φ + g¯ −1 . Sen. Phys. Horava. (5. A. hep-th/9903123. Rev. The tachyon potential in open Neveu–Schwarz string field theory. Berkovits. J. Tachyon condensation in string field theory. Theor. Polchinski. B.R. The static tachyon potential in the open bosonic string theory. hep-th/9402113. L. Zwiebach. Sen. SO(32) spinors of type I and other solitons on brane-antibrane pair. [12] J. Klebanov. A.

Nonarchimedean strings. Commun. Adelic string amplitudes. Sen / Nuclear Physics B 584 (2000) 300–312 [18] N.H. 60 (1988) 484. Ghoshal.L. Chekhov. 130 (1990) 130. Brekke. [27] P. Martinec. Lett. [35] I.H.O. R. Chekhov. E. H.H. Phys. Okada. Ubriaco. Lett. Moeller. [28] P. Phys. P. 4 (1987) L83. Ann. [30] B. Y. Phys. Phys.O.A. P. Kutasov.312 D. Recknagel. Harvey. Y. Taylor.A. 133 (1993) 1. [34] L. Zabrodin. Zabrodin. Quant. Harvey. [25] P. On adelic formulas for the p-adic string.O.B.O.G. A. Parisi. Phys. Frampton. Nucl. p-adic strings. Phys. [19] J. Math. Freund.G. E. Noncommutative geometry and string field theory. [24] P. B 197 (1987) 101. Witten. Freund. [23] A. Math. [33] Z. B 302 (1988) 365.D. Multiloop calculations in p-adic string theory and Bruhat–Tits trees.O. Warner. Phys. hep-th/9406125. [20] R. Mod. Tatar. Phys. Lumps and p-branes in open string field theory. Roggenkamp. W. Phys. Kraus. Lett. D 37 (1988) 3077. Strominger. N. Rev. Marshakov. Freund. Mod. Level truncation and the tachyon in open bosonic string field theory. Class.G. Commun. Frampton. Gopakumar. Rev. Y. [26] L. hep-th/0003101. Phys. Volovich. Grossman. Phys. On p-adic functional integrals.P. B 199 (1987) 191. H. [41] L. Hlousek. A. M. Lett. Lagrangian formulation of open and closed p-adic strings. hepth/0002117. p-Adic string compactified on a torus. 189 (1989) 370. [42] R. Exact solution of a massless scalar field with a relevant boundary interaction. Phys. [40] A. D. D-branes as unstable lumps in bosonic open string field theory. Frampton. On the relevance of tachyons. [38] L. E. Zhang. hep-th/0003031. [32] R. B 209 (1988) 229. [39] P. Lett.V. Saleur. Lett. Phys. Noncommutative solitons. Olson. Okada. Brekke.R.H. Mihailescu. On relevant boundary perturbations of unitary minimal models. [43] E. Witten. A. Spokoiny. B 213 (1988) 260. Frampton. 123 (1989) 463. 125 (1989) 675. p-adic string. Nishino. Schomerus. . A. Zinoviev.V. Math. Stability analysis of p-adic string solitons. Lett. Zabrodin. Mironov. Witten. [36] B. Phys. D.V. [29] P. A. Nucl. hep-th/0003160. p-adic string theory. B 208 (1988) 401. and references therein. Y. hep-th/0002237.M. A 3 (1988) 639. Phys. The p-adic string N point function. Non-archimedean strings and Bruhat–Tits trees. p-adic numbers in physics. M. Lett. Okada.G. B 430 (1994) 577. Spector.O. M. Commun. A. S. Lett. B 199 (1987) 186. Phys. hep-th/0003110. Olson. D. Fendley. Quantum geometry of non-archimedean particles and strings. [21] P. Effective scalar field theory of p-adic string.J. the Weyl conjectures and anomalies. M. [22] J. New p-adic string amplitudes. [31] G. Minwalla.V. Grav. V. B 268 (1986) 253. P. A 5 (1990) 265. De Mello Koch. Rep. Jevicki. Phys. Phys. B 242 (1990) 354. Freund. Lett. [37] A. Phys. Nucl. Non-archimedean string dynamics.V.

This new construction. José Santiago b.V.+h.3 2 1 3 ( 0 0 ) 0 0 2 4 1 .ox.ox. Granada. Avenue Fuente Nueva. Stavros Mouslopoulos a. Department of Physics.ross@physics.ox. 04.V.10.2 . PII: S 0 5 5 0 . 11. Antonios Papazoglou a. the hierarchy between the Planck scale and the electroweak scale.50. Ross a.papazoglou@physics.mouslopoulos@physics.  2000 Elsevier Science B. Graham G. The study of the phenomenology of this model reveals an anomalous first KK state which is generally much lighter than the remaining tower and also much more strongly coupled to matter. The model suggests a further exotic possibility if one drops the requirement of solving the hierarchy problem.elsevier.uk 5 jsantiag@ugr. has the advantage that the SM fields are confined on a positive tension brane.e. Furthermore. apart from providing a solution to the Planck hierarchy problem.uk 4 g.Nuclear Physics B 584 (2000) 313–328 www.uk 2 s. PACS: 11. these models make 1 i.nl/locate/npe A three three-brane universe: new phenomenology for the new millennium? Ian I. 1 Keble Road.Kk. Kogan a. 18071.5 a Theoretical Physics.8 . All rights reserved.kogan@physics. There remains the problem of the hierarchy between the weak and gauge unification scales [6]. UK b Department of Theoretical Physics. In this case gravity may result from the exchange of the ordinary graviton plus an ultralight KK state and modifications of gravity may occur at both small and extremely large scales.uk 3 a. revised 7 April 2000.. Oxford University.3 . Oxford. Granada University. accepted 17 April 2000 Abstract We consider an extension of the Randall–Sundrum model with three parallel 3-branes in a 5-dimensional spacetime. Spain Received 14 February 2000.ac. OX1 3NP.27.+d 1.ac. 0550-3213/00/$ – see front matter  2000 Elsevier Science B.ac.ox. Bounds on the parameter space of the model can be placed by comparison of specific processes with the SM background as well as by the latest Cavendish experiments. All rights reserved. there has been considerable interest in theories in which the SM fields are localized on a 3-brane in a higher dimensional spacetime.1 . Depending on the dimensionality and the particular form of the geometry of this space.es 6 I.4 . Introduction Recently. the long standing (Planck) hierarchy problem 6 can find alternative resolutions.ac.

The higher dimensional 2 = M n+2 V . Here and further a brane with positive cosmological constant is called positive or ‘+’ brane and a brane with negative cosmological constant is called negative or ‘−’ brane. namely the ratio of the large compactification radius to the electroweak scale.314 I. Arkani-Hamed.I. Kogan et al. Taking the size of the new n dimensions to be sufficiently large and identifying the (4 + n)-Planck scale with the electroweak scale. In this the compactification radius need only be some 35 times larger than the Planck length. 1). Experimental and astrophysical constraints demand that n > 2. Even for six extra dimensions this must be greater than 105 . Randall–Sundrum model with two branes at the orbifold S 1 /Z2 fixed points. The possibility that gravity be trapped on a brane was suggested by Gogberashvili [5]. / Nuclear Physics B 584 (2000) 313–328 dramatical phenomenological predictions which can be directly confronted with current and future accelerator experiments as well as cosmological observations. M & 30 TeV and allow new dimensions even of submillimeter size. The phenomenology of this model has been extensively explored in Refs. Antoniadis. This last requirement forces the one of the two branes to have negative tension. [11. . but rather a low dimensional (3 + 1 in the case of our universe) brane dates back to early eighties when independently Akama [1] and Rubakov and Shapochnikov [2] suggested models of our universe as topological defects and later Visser [3] and Squires [4] described how particles can be gravitationally trapped on the brane. Randall and Sundrum proposed [10] an alternative scenario where they assumed one extra dimension along which the geometry is nonfactorizable. The 5D spacetime is essentially a slice of AdS5 and the tensions of the two 3-branes are chosen so that the 4D spacetime appears flat.12]. However. where Planck scale M is then related to the 4D Planck scale by MPl n Vn is the compactification volume. an hierarchy between the electroweak and the Planck scale is introduced. Dimopoulos and Dvali [7–9] proposed that we live on a 3-brane in a 3 + 1 + n space with fully factorizable geometry. Assuming that the fundamental mass scale on the positive brane is of the order of MPl we can readily get a mass scale on the negative brane of the order the electroweak scale. The idea that a multidimensional universe may not be of KK type. In the light of this new problem. Their construction consists of two parallel 3-branes sitting on the fixed points of an S 1 /Z2 orbifold (see Fig. An exponential “warp” factor in the metric then generates a difference of the mass scales between the two branes that could be O(1015 GeV) although the size of the orbifold is only of the order of Planck length. thus solving the hierarchy problem. a new hierarchy must now be explained. The KK tower of spin-2 graviton Fig. 1.

[21–24]. The model has three parameters. This statement is not true for more general models with nontrivial bulk stress–energy tensor [36. gravity is effectively confined to the brane by the steep “warp” factor generated by the tension dominating the brane. Of course in realistic cosmologies the energy density of the universe must be dominated by matter. Interesting cases where the branes are not flat or the extra dimension(s) is not compact have also been considered in Refs. However it puts extra constraints on parameters of the model and we need to make some fine tuning to avoid potential cosmological problems. . 2). On a brane with a positive tension (as for example in the single brane scenario [18])..I. [16–20]. i. The purpose of this paper is to formulate a new model of a “brane world” in which we live on a positive brane thus avoiding these cosmological problems.e. The model consists of two positive branes located at the fixed points of a S1 /Z2 orbifold with one negative brane which can move freely in between (see Fig. brane tension Λbr and a matter energy density ρm and the correct expression for Hubble constant squared can be obtained by cancelling the leading (Λbr + ρm )2 term with the term Λ2B √ coming from the negative bulk cosmological constant ΛB so that H ∼ Λbr ρm . These papers showed that the Hubble parameter H governing the expansion of the scale factor on the brane has a different behavior than derived from the usual 4-d Friedmann equations. the warp factor and the x factor — which effectively measures the distance between one of the positive branes and the negative brane. the bulk curvature. Kogan et al. with a single or different cosmological constants between them.37] in which it was shown that one may have real Hubble constant without antigravity. For a viable theory the “brane world” must reproduce correct gravity and cosmology of our universe. i. parallel or intersecting. Several models have been constructed since then [13–15] that extend the original RS model to multibrane configurations. / Nuclear Physics B 584 (2000) 313–328 315 resonances starts from the TeV scale with TeV spacing giving rise to characteristic signals in high energy colliders.e. A generic characteristic of these models is the presence of negative tension branes which are necessary for the branes to be flat. Because we are living in the universe with a real Hubble constant and without any noticeable effects of antigravity the negative brane as a model of our world must be excluded in the original RS model. The RS model corresponds to the limiting case x = 0. when the negative brane hits one of the positive branes. but negative energy density. In the papers [25–29] (see also [30–35]) it was shown that for the Randall–Sundrum construction the full energy density ρ is a sum of a vacuum energy density. or real Hubble constant. but imaginary Hubble constant. antigravity. It is easy to see that the two-brane RS model is nothing but the limiting case of our three-brane model. The phenomenology of these models is generally complicated and little has been written about them. the Hubble parameter is proportional to the energy √ density on the brane instead of the familiar dependence H ∼ ρ. In particular. An important observation about the cosmology of the “brane world” was made in Refs. In the RS picture the negative cosmological constant of the bulk is used to cancel the cosmological constant or tension on the brane. From this picture one can immediately see that in these simplest models life on a negative brane is impossible — we either have normal matter with positive energy density..I.

Kogan et al. 2.316 I. Surprisingly enough this situation is not excluded experimentally! Thus one may have “bi-gravity” — in all experimentally analyzed regions gravitational attraction is due to an exchange of two particles — the massless graviton and ultralight first KK mode. This configuration can be obtained from a configuration with only ‘−’ branes at the orbifold points and several moving ‘+’ branes. An unusual possibility arises if we relax the requirement that the Planck hierarchy problem is solved. One of the striking predictions of our model is the fact that the first KK mode can be very light and strongly coupled compared to the rest of the KK states. Moving a ‘+’ brane towards one of the ‘−’ branes we can transform it into a ‘+’ brane. Let us note that each moving brane on an orbifold must be counted with a double charge because of it mirror image. Fig. so that the phenomenology is determined mostly by it. In this case this light mode can be so light that the corresponding wavelength can be by order of 1% of the observable size of the universe while the second KK mode is in submillimeter region. At the same time it is necessary to say that so far the full string/Mtheory description of these multibrane configurations is missing and we can not exclude a possibility that there are moving negative branes. In the limiting case when x → 0 we have a RS configuration.I. Only at scales larger than 1026 cm will the first KK mode decouple leading to a much smaller gravitational coupling beyond this length scale. 7 We are grateful to Andre Lucas for interesting discussion on this subject. 3. so one can get a configuration of the type + + − or with more than one moving ‘+’ brane — the only constraint is the that the sum of all “charges” is zero. 7 It is unclear if one may have moving ‘−’ branes. + + − model which is a RS configuration on Figure 1 with an insertion of a moving ‘+’ brane. In this paper we shall discuss in details the + − + model of Fig. In a heterotic M-theory for example [38] one can think about boundaries of eleven-dimensional space after compactification on a suitable Calabi–Yau manifold as negative branes where a fundamental five-brane in M-theory after wrapping on 2-cycles may play the role of a ‘+’ brane. . Let us note that so far it is unclear what are the selection rules on branes at orbifold fixed points and moving branes. / Nuclear Physics B 584 (2000) 313–328 Fig. + − + model with two ‘+’ branes at the fixed points and moving ‘−’ branes. 2.

In Section 4 we discuss the phenomenology when the first KK mode has a mass in the meVTeV region. we discuss the possible generalization of our model and questions for future investigation. A straightforward calculation gives us the following differential equations for σ (y): . but the gap may be very small — it is given by a tunneling factor. This effect takes place not only in + − + model but in + + − and other models with more than two ‘+’ branes (in which case there may be more than one light mode). The action of this setup is: ZL2 Z S= 4 d x dy √  X −G − Λ + 2M 3 R − −L2 Z p b(i) . L1 and L2 where L0 and L2 are the orbifold fixed points (see Fig.I. In conclusion. d 4 x Vi −G (1) i y=L i bµν is the induced metric on the branes and Vi their tensions. The 5th dimension has the geometry of an orbifold and the branes are located at L0 = 0. In Section 5 we discuss the unusual “bi-gravity” scenario when the Compton wavelength of the first KK mode is by order of 1026 cm while the Compton wavelength of a second KK mode is less than 1 mm. In the next section we construct the model and discuss the spectrum of KK excitations. / Nuclear Physics B 584 (2000) 313–328 317 The reason the anomalously light KK mode exists is due to the fact that with more that one ‘+’ brane there will be a bound state on each of them when they have infinite separation. The metric ansatz with this property is the following: ds 2 = e−2σ (y)ηµν dx µ dx ν + dy 2. 2. The organization of the paper is as follows. 2). The 3-brane model Our model consists of three parallel 3-branes in an AdS5 space with cosmological constant Λ < 0. One superposition is the true ground state while the other configuration has non-zero mass.I. Other models will be considered in separate publications. Firstly we consider the branes having no matter on them in order to find a suitable vacuum solution. (3) Here the “warp” function σ (y) is essentially a conformal factor that rescales the 4D component of the metric. Kogan et al. In this paper we consider only the + − + configuration which has the interesting “bi-gravity” possibility. At finite distances there is a mixing between the two localized states. The notation is the where G same as in Ref. [10]. (2) Vi √ ΛGMN + RMN − GMN R = − 2 4M 3 −G i (i) At this point we demand that our metric respects 4D Poincaré invariance. In Section 3 the first and subsequent KK modes are considered in more detail and their masses and couplings as a function of the “warp” factor and the parameter x are discussed using both analytical and numerical methods. The Einstein equations that arise from this action are: ! p X b(i) (i) µ ν 1 1 −G b Gµν δM δN δ(y − Li ) .

318 I.I. Kogan et al. The solution of these equations consistent with the orbifold geometry is precisely: . σ 00 = 12M 3 i p where k = −Λ/(24M 3 ) is a measure of the curvature of the bulk. X Vi δ(y − Li ). / Nuclear Physics B 584 (2000) 313–328 (σ 0 )2 = k 2 .

.

 σ (y) = k L1 − .

|y| − L1 .

V1 = Λ/k < 0. we find the 4D Planck mass is given by  M3  1 − 2e−2kL1 + e−2k(2L1−L2) . To determine the phenomenology of the model we need to know the KK spectrum that follows from the dimensional reduction.. In −1 this case all the parameters of the model L−1 1 . [10] and then integrate over the 5th dimension. Furthermore. [37. M.39–41]. 2L1 − L2 ≈ 35k −1 . Thus we take k ∼ O(M) in order not to introduce a new hierarchy. This is determined by considering the (linear) fluctuations of the metric of the form:   2 (9) ds 2 = e−2σ (y)ηµν + 3/2 hµν (x. k can be taken to be of the same order. (7) k The above formula tells us that for large enough kL1 and k(2L1 − L2 ) the three mass scales MPl . (4) (5) (6) with the requirement that the brane tensions are tuned to V0 = −Λ/k > 0. with the additional restriction k < M so that the bulk curvature is small compared to the 5D Planck scale so that we can trust our solution. if we put matter on the third brane all the physical masses m on the third brane will be related to the mass parameters m0 of the fundamental 5D theory by the conformal (warp) factor 2 = MPl m = e−σ (L2 ) m0 = e−k(2L1 −L2 ) m0 . While the dilaton may lead to additional observable phenomena its inclusion should not change the phenomenology of the graviton KK modes. If we consider massless fluctuations of this vacuum metric as in Ref. We expand the field hµν (x. y) = ∞ X (n) h(n) (y). y) dx µ dx ν + dy 2 . V2 = −Λ/k > 0. M Here we have ignored the dilaton mode that could be used to stabilize the brane positions L1 and L2 as discussed in Refs. The function Ψ (n) (y) will obey a second order differential equation which after a change of variables reduces to an ordinary Schrödinger equation:   2 1 2 b(n) (z) = mn Ψ b(n) (z) (11) − ∂z + V (z) Ψ 2 2 . L2 and k are of the order of Plank scale. i. µν (x)Ψ (10) n=0 (n) (n) (n)α where (∂κ ∂ κ − m2n )hµν = 0 and fix the gauge as ∂ α hαβ = h α = 0.e. (8) Thus we can assume that the third brane is our universe and get a solution of the Planck hierarchy problem arranging e−k(2L1−L2 ) to be of O(10−15 ). y) in graviton and KK states plane waves: hµν (x.

2 2g(z) 8[g(z)] The new variables and wavefunction in the above equation are defined as:   y ∈ [L1 . the difference generating the hierarchy between the first and the third brane. z1 . For the KK modes the solution is given in terms of Bessel functions. where z1 = z(L1 ). (16) J2 Y2 Ψ = B1 B2 B k k k The boundary conditions (one for the continuity of the wavefunction at z1 and three for the discontinuity of its first derivative at 0. This is a quantum mechanical problem with δ-function potentials of different weight and an extra 1/g 2 smoothing term (due to the AdS geometry) that gives the potential a double “volcano” form. 2ekL1 − e2kL1 −ky − 1 /k. − e      kL 2kL +ky 1 1 − 1 /k. −L1 ]. / Nuclear Physics B 584 (2000) 313–328  15k 2 3k  − δ(z) + δ(z − z2 ) − δ(z − z1 ) − δ(z + z1 ) . L2 ]. in order to have a non-trivial solution. L1 ].I. For y lying in the regions A ≡ [0. It is given by A b(0) = . L2 ]. − 2e − e V (z) = with b(n) (z) ≡ Ψ (n) (y)eσ/2. (15) Ψ [g(z)]3/2 The normalization factor A is determined by the requirement Zz2  (0) 2 b (z) = 1. In the specific case where L1 = L2 /2 (and with zero hierarchy) the potential and thus the zero mode’s wavefunction is symmetric with respect to the second brane. y ∈ [−L1 . L1 ] and B ≡ [L1 . y ∈ [−L2 . When the second brane moves towards the third one the wavefunction has a minimum on the second brane but different heights on the other two branes.      eky − 1/k. z≡  −ky  − 1 /k. The change of variables has been chosen so that there are no first derivative terms in the differential equation. Ψ 319 (12) (13) (14) and the function g(z) as g(z) ≡ k{z1 − ||z| − z1 |} + 1.I. leads to the vanishing determinant: . z2 ) result in a 4 × 4 homogeneous linear system which. 0]. An interesting characteristic of this potential is that it always gives rise to a (massless) zero mode which reflects the fact that Lorentz invariance is preserved in 4D spacetime. Kogan et al. dz Ψ 0 chosen so that we get the standard form of the Fierz–Pauli Lagrangian. we have:       r     g(z) mn mn A2 A1 (n) A b g(z) + g(z) . y ∈ [0.

.

  .

.

J1 m Y1 mk 0 0 .

k   .

.

.

m m .

Y1 k g(z2 ) .

0 0 J1 k g(z2 ) .

.

(17) .

J m g(z ) Y m g(z ) J m g(z ) Y m g(z ) .

. = 0.

1 k 1 1 k 1 1 k 1 1 k 1 .

.

   .

.

J2 m g(z1 ) Y2 m g(z1 ) −J2 m g(z1 ) −Y2 m g(z1 ) .

k k k k .

b SM fields) with the minimal gravitational coupling of the SM Lagrangian d4 x −GL( respect to the metric. this is simply given by reversing the sign of the graviton wave function for y > L1 (it has one zero at L1 ). the separation of the second and third brane x = k(L2 − L1 ) and the hierarchy factor w = e−k(2L1−L2 ) . [42] by expanding R b G. For each mass we can then determine the wave function with normalization Zz2  (n) 2 b (z) = 1.I. / Nuclear Physics B 584 (2000) 313–328 where we have suppressed the subscript n on the masses mn . (21) . c0 M b(n) (z2 )g(z2 )3/2 Ψ 1 = . 3.320 I. L1 = L2 /2. In the symmetric case. the first KK state. Kogan et al. In the case x  1 we may obtain a reliable approximation to its mass by using the first terms of the Bessel power series. a large value on the third and a zero very close to the first brane. The interaction of the KK states to the SM particles is found as p in Ref. From now on we shall use a convenient choice of parameters: the measure of the curvature of the bulk k. cn M 3/2 (19) (20) For the zero mode the normalization constant A is M 3/2 /MPl which gives the Newtonian gravitational coupling suppression c0 = MPl . This is essentially the mass quantization condition which gives the spectrum of the KK states. After the rescaling due to the “warp” factor we get: Lint = − g(z2 )3/2 X (n) b (z2 )h(n) Ψ µν (x)Tµν (x) M 3/2 n>0 =− XΨ b(n) (z2 )g(z2 )3/2 (n) A (0) h (x)T (x) − hµν (x)Tµν (x) µν µν M 3/2 M 3/2 (18) n>0 with Tµν the energy momentum tensor of the SM Lagrangian. dz Ψ 0 From the form of the potential we can immediately deduce that there is a second “bound” state. The first KK mode has mass given by m1 = 2kwe−2x . The first and subsequent KK modes: masses and coupling constants As discussed above the KK spectrum has a special first mode which for all x significantly different from unity has very different behaviour compared to the other KK states. Thus the coupling suppression of the zero and KK modes to matter is respectively: A 1 = 3/2 . When the second brane moves towards the third this symmetry is lost and the first KK wave function has a very small value on the first brane.

x = 5 and k = 1017 GeV. Again solving numerically. the positive brane disappears and we obtain the original Randall–Sundrum model. being dominantly a bound state of the volcano potential on our brane. / Nuclear Physics B 584 (2000) 313–328 321 The normalization integral can also then be done analytically and the coupling suppression is found to be independent of x and equal to c1 = wMPl . Numerically we find out that the mass of the second state and the spacing 1m between the subsequent states have the form: m2 ≈ 4kwe−x . Its mass rises above 100 GeV and the coupling is no longer constant but has a small dependence on x. 4. A more complete discussion will appear elsewhere.7 MeV with c1 ≈ 630 GeV. The analysis is readily generalized to include q q. we consider the case w = e−35 . the above approximations break down and the first mode is not so different from the others. (22) The reason for this is readily understood because. Using the Feynman rules of Ref. In this case the spectrum starts from several hundreds of GeV up to some TeV depending on the choice of k. m2 ≈ 1.6 GeV with c2 ≈ 52 500 GeV. The spacing only approaches a constant for high enough levels when the Bessel arguments become much greater than one. [42] the contribution of the KK modes to e+ e− → + µ µ− is given by . ¯ gg initial and final states. Kogan et al. 1m ≈ εkwe −x (23) .I. The masses of the other KK states in the above region are found to depend in a different way on the parameter x. we find the couplings of the higher modes are suppressed relative to the lowest mode by a factor proportional to ex . (25) When the second brane approaches the third brane. Phenomenology In this section we will present a brief discussion of the phenomenology of the KK modes to be expected in colliders. concentrating on the simple process e+ e− → µ+ µ− . It has large spacing between the KK states and TeV coupling suppression. (24) where ε is a number between 1 and 2. To illustrate the difference between the first and second modes at large x. it is largely localized on it. The masses and coupling suppressions of the first two modes are m1 ≈ 5.I. In the extreme case where the positions of the second and the third brane coincide.

2 s 3 .

.

D(s).

(26) 1280π where D(s) is the sum over the propagators multiplied by the appropriate coupling suppressions:  σ e + e − → µ+ µ− = .

In the limit of very low x (x . 5) however there are always KK resonances in the range of energies of collider experiments. Of course if k is raised the KK modes become heavier and there will be a value for which the lightest KK mode is above the experimental limits.094 (in the case where the KK is heavy enough that can decay and neutrinos) and 240π to all SM particles). Note that the bad high energy behaviour of this cross section is expected since we are working with an effective — low energy non-renormalizable theory of gravity. To be definite we will take k = 1017 GeV. 5. that can decay only to massless gauge bosons 71 ≈ 0. 5. The theory that is valid above this scale is supposed to give a consistent description of quantum gravity. we have a widely spaced discrete spectrum (from the point of view of TeV physics) close to the one of the Randall–Sundrum case with cross section at a KK resonances of the form σres ∼ s 3 /m8n .039 (in the case that the KK where β is a dimensionless constant that is between 320π is light enough. In the case that x is small. and x is already excluded. e+ e− → µ+ µ− . 0. x .322 I.M.1) the mass of the first mode is above current experimental energies (m1 ∼ 200 GeV) and the phenomenology is that discussed in [11].1 .. energy of e+ e− . 4. i. Thus this range of k. D(s) = X n>0 1/cn2 s − m2n + iΓn mn (27) and s is the C. In this case the contribution of the KK modes gives an excess of ∼ 1000% over the SM contribution which would have been seen either by direct scanning if the resonance is near the energy at which the experiments actually run or by means of the process e+ e− → γ µ+ µ− which scans a continuum of energies below the center of mass energy of the experiment. For the rest of the discrete spectrum region (0. w = e−35 . The decay rates of the KK states that are present in the above formula are given by: Γn = β m3n . Kogan et al. cn2 (28) 39 ≈ 0. Our effective theory is valid up to an energy scale Ms (which is O(TeV)). smaller than 0. which acts as an ultraviolet cutoff. x . . Due to the uncertainty in the collision energy the narrow KK peak will be reduced as shown in Fig.5 MeV.I.e. w. The details of computation of the total cross section depends on the KK spectrum. This means that the summation in the previous formula should stop at the KK mode with mass near the cutoff. / Nuclear Physics B 584 (2000) 313–328 Fig. Since this is unknown we are only able to determine the contributions of the KK states with masses less than this scale.

To summarize.3 GeV and ∼ 200 GeV). In this case we substitute in D(s) the sum for n > 2 by an integral over the mass of the KK excitations. / Nuclear Physics B 584 (2000) 313–328 323 Fig. Kogan et al. x . 6 where three curves are shown. The first state is singled out because of its different coupling. the smaller one is the result of the smoothing by the convolution of the theoretical curve and the experimental resolution of the beam. w = e−35 and x = 1. The SM background lies below them. 0.e. In fact at these values of x the biggest contribution comes from this state (the coupling of the rest being very small). the lowest one corresponds to the SM alone. . Finally. A quantitative estimate is given in Fig. However as k decreases it reintroduces the hierarchy problem which the warp factor is designed to eliminate.1 region. 5 for the preferred value of w (this window corresponds to a value of the mass of the first KK mode between ∼ 0. s − m2 + i (29) √ where the value of the integral is ∼ iπ/(2 s) with the principal value negligible in the √ region of interest ( s  Ms ) and we have considered constant coupling suppression c for the modes with n > 2. from the process e+ e− → µ+ µ− we can exclude a window of the form 0. 1/c12 1 + D(s)KK ≈ s − m21 + iΓ1 m1 1m c2 ZMs dm m2 1 . However if we allow the value of w to decrease.. e+ e− → µ+ µ− Resonance of the third mode for k = 1017 GeV. the second takes into account also the tower of KK in the case of w = e−35 and the last one is also with the KK tower but this time with w = e−36 which as can be seen from the plot is already excluded. Both the the very low x limit and the continuum limit are allowed.I. i. For values of x greater than x ∼ 5 the spacing in the spectrum is so small that we can safely consider it to be continuous. Thus we can exclude in this continuum limit values of w ≈ e−36 and below. For w ≈ e−35 the contribution of the KK tower is negligible compared with the SM background.1 . the coupling of the first √ state increases enough to give noticeable contribution to the cross section (∼ 30% at s = 189 GeV for w ≈ e−36 ). 5. we can also exclude values of the hierarchy greater than ∼ e−36 except for the 0 6 x . The higher curve is the theoretical curve.I.

The fact that gravity is Newtonian at least up to millimeter distances implies that the corrections to gravitational law due to the presence of the KK states must be negligible for such distances. (31) V (r) = − 2 cn MPl r n>0 The contribution to the above sum of the second and higher modes is negligible compared with the one of the first KK state. σ (e+ e− → γ + γ ) c1 α c1 (30) and so we see that this process may be phenomenologically very important. w gives a contribution O(1). the condition for the corrections of the Newton law to be small for millimeter scale distances is:   − ln w 1 [GeV] . The gravitational potential is the Newton law plus a Yukawa potential due to the exchange of the KK massive particles (in the Newtonian limit): ! X MPl 2 1 M1 M2 1+ e−mn r . The resonance we see is the Z peak. the medium one for w = e−35 and the lower one is the SM contribution (one loop calculation). 6. We have considered as an example the process e+ e− → µ+ µ− but it may be possible to obtain similar or even stricter bounds could be obtained from other processes like e+ e− → γ + missing energy. / Nuclear Physics B 584 (2000) 313–328 Fig.I. Thus. let us only mention here that by dimensional analysis the ratio of this cross-section to two-photon electron–positron annihilation s s σ (e+ e− → γ + light KK mode) ∼ 2 ∼ 102 2 . for example. so we can safely say that x ≈ 15 is the maximum brane separation allowed. (32) x < 15 − ln 2 kw The logarithm for any reasonable choice of the parameters k. Kogan et al. A further bound on the parameters of our model can be put from the Cavendish experiments. . The upper curve is for w = e−36 .324 I. e+ e− → µ+ µ− Cross section in the continuum limit for k = 1017 GeV and x = 11. because they have larger masses and coupling suppressions. We shall present the detailed cross-section of the process e+ e− → γ + light KK mode in another publication.

Bi-gravity Equations (21) and (24) show that. However our example does illustrate how gravity may be quite different from the form that is usually assumed. For m1 = 10−31 eV we have m2 ≈ 10−2 eV. / Nuclear Physics B 584 (2000) 313–328 325 Fig. the lightest KK mode splits off from the remaining tower. This leads to an exotic possibility in which the lightest KK mode is the dominant source of Newtonian gravity! Cavendish experiments and astronomical observations studying the motions of distant galaxies have put Newtonian gravity to test from submillimeter distances up to distances that correspond to 1% of the size of observable universe. In the context of the graviton KK modes discussed above this constrains m < 10−31 eV or m > 10−4 eV.. for length scales less than 1026 cm gravity is generated by the exchange of both the massless graviton and the first KK mode. w. wM.I. although the “warp” factor. 7. Combining the above results the allowed region of the parameters of our model for k = 1017 GeV and w = e−35 are shown in Fig. Our exotic scheme corresponds to the choice m1 ≈ 10−31 eV and m2 > 10−4 eV. may still be small (i. This implies (taking into account the different coupling suppressions of the massless graviton and the first KK state). Excluded regions for k = 1017 GeV and w = e−35 . Kogan et al. 5. the fundamental scale M  MPlanck ). 7. searching for violations of the weak equivalence principle and inverse square law. for large x. is now the Planck scale so. the phenomenology of the continuum of the KK states is similar to the case of Refs.e. we do not now solve even the Planck hierarchy problem. Also the presence of the ultralight first KK state will give deviations from Newton’s law as we probe cosmological scales (of the order of the observable universe). This comfortably satisfies the bound m > 10−4 eV coming from Cavendish experiments. Using equations (33) and (21) and assuming as before that k ≈ M. [7. In this case. (33) 1 + ≈ 2 2 2 w (wM)2 MPl M We see that the mass scale on our brane. that the gravitational coupling as we measure it is related with the parameters of our model as:   1 1 1 1 = H⇒ MPl ≈ wM. we find that m1 = 2ke−2x ≈ MPlanck e−2x . Here we should note that since the coupling of the second mode is always smaller than the one of the first mode. The .I. According to this picture deviations from Newton’s law will appear in the submillimeter regime as the Yukawa corrections of the second and higher KK states become important.8] and thus does not conflict to experiment.

/ Nuclear Physics B 584 (2000) 313–328 Fig. and G. 8117781027.S.M.I. Our world is confined to a positive tension brane which makes this construction the minimal realistic model of the RS class. one associated with the graviton and one with the first KK mode..’s work is supported by the Hellenic State Scholarship Foundation (IKY) No.’s work is supported by M. 8.. It is clearly of interest to explore the cosmological consequences of such a scheme. Conclusions In this paper we discussed a model of a three 3-brane universe with two positive and one intermediate negative tension brane. would like to thank Peter B.E. for a kind hospitality during the course of this project. A.P. Bounds on the parameter space of this model were placed by comparison with data from collider and Cavendish experiments. . Exclusion regions for the bi-gravity case and correlation of the first two KK states. the phenomenology of the model needs further investigation and it may be that missing energy processes can put stricter bounds on the parameter space. and J. A.326 I.R is supported in part by PPARC rolling grant PPA/G/O/1998/00567. S. Acknowledgements S. phenomenological signature of this scenario is that gravitational interactions will appear to become weaker by the factor w for distances larger than 1026 cm! 6. The work of I. Oxford University. the latter is has relatively small mass and large coupling.S. J. Renton and Peter Richardson for useful discussions. However.S. The novel feature of this description is that gravity is modified at both large and small scales.G. AEN 96-1672) and Junta de Andalucía (FQM 101). the EC TMR grant FMRX-CT-96-0090 and by the INTAS grant RFBR-950567.C. would like to thank the Department of Theoretical Physics.M. Compared to the remaining tower of KK states. Kogan et al. 8017711802. In particular at large scales the strength of the gravitational force will be reduced by the warp factor. J. All authors would like to thank Pavel Kogan for the first three figures and Subir Sarkar for informative discussions. We also explored the possibility of “bi-gravity” in which observable gravity is due to the exchange of both the ordinary graviton and the first ultralight KK state.P. CICYT (Spain. Due to the presence of two positive branes there are now two “bound” states.’s work is supported by the Hellenic State Scholarship Foundation (IKY) No.I.K.

Nakanishi. Randall. Phys. Phys. hep-th/9910149. R. B 462 (1999) 34. Wise. N. Hatanaka. Kehagias. Li. hep-th/9907209. [4] E. Phys.A. hep-ph/9911406. Lukas. J. Rev. Horava. [7] N. [21] A. Chang. Graesser.I. Stelle. D 59 (1999) 086004. hep-ph/9909255. D 61 (2000) 024008. Arkani-Hamed.H. Rizzo. 1983. Cline. G. Davoudiasl. Olive. Sundrum. hep-th/0001113. Kim. [30] C. Phys.B. Japan. Witten. Phys. Kraus. T. hep-th/9902071. Lett. Randall. Phys. M. . hep-th/9906064. Deffayet.A. Ghilencea. M. Squires. hep-ph/9912266. Dimopoulos.S. Lett. Lett. Sundrum. L. Rev. [22] A. Dvali. Binétruy. Dimopoulos. D. D 60 (1999) 123506. [34] E. 83 (1999) 4245. S. Rubakov. [2] V. D 60 (1999) 107505. Phys. K. Phys. Y. [19] J. Cline. Kim. [11] H. Phys. Csáki. Pospelov. Shiromizu. Sasaki. D 59 (1999) 086001. 1982. Kanti. 267. [37] C. Randall. hep-th/9905012. [16] T. Hewett. [12] S. Grojean. Langlois. A. Wasserman. [38] P. Dvali. [13] I. Nihei. [39] W. Rev. hep-th/9908104. B 167 (1985) 286. [9] I. Phys. E. Nara. Waldram. hep-ph/9910498. B 475 (1996) 94. D 60 (1999) 086001. G. Phys. Lett. I. [29] H. Nucl. Lett. Arkani-Hamed.A. Phys.I. gr-qc/9912002.). Rev. [15] C. Servant. Phys.D. M. Kiritsis. Gogberashvili.L. Waldram. R. B 465 (1999) 81. Oda. M. Shirman. Shaposhnikov. Rev. Phys. Sakamoto. Graesser. Lett. [10] L. Rev. Lett. K. G. M. Kogan et al. Maeda. Kaloper. G. [32] A. Akama for sending manuscript of his paper before it became available from xxx archive. [31] P. B. [33] T.E. Phys. Csáki. Ovrut. Kaloper. [17] N. Lett. [28] J. Lett. Phys. Flanagan. J. G. Terning. D. J. M.J. Lykken. hep-ph/9909523. Gauge Theory and Gravitation. Nucl. Dvali. Phys. Randall. JHEP 9911 (1999) 022. Goldberger. Proceedings of the International Symposium. Rev.G. S. C. [36] P. in: K. Terning. hep-ph/9904383. B 429 (1998) 263. Kaloper. N. D 59 (1999) 101303.I. [6] D. S. B 125 (1983) 136. [20] N. H. Kolda. hep-ph/9908369. K. I. 83 (1999) 3370. Lukas.A. Visser. Arkani-Hamed. Yamaguchi. Lett. M. p. hep-th/9909053. B 468 (1999) 31. D. hep-th/9910081. Phys.M. Phys. References [1] K. Ida. Servant. Tachibana. E. N. [14] H.D. hep-th/991123. S. [3] M. Springer-Verlag. [27] C. Phys. Grojean.G. Kogan. S. [25] T. Antoniadis. Akama. Rev.H. [35] D. B 460 (1996) 506. [26] N. / Nuclear Physics B 584 (2000) 313–328 327 Note added in proof We would like to thank Prof. hep-th/9909076. [24] P. Takenaga. Dimopoulos. Ovrut. [5] M. C. Rev. [18] L. K.B. Ross. hep-th/9908076. L. [8] N. C. M. hep-th/9909048. Rev. Rev. hep-ph/9911457. Kaloper. Tye. gr-qc/9910076. H. Phys. Linde. Dimopoulos. Kikkawa. Lett. B. Arkani-Hamed. [23] N. B 159 (1985) 22. 83 (1999) 4922. G. Dvali. B 436 (1998) 257. Nariai (Eds. G. J. Csáki. M. D 60 (1999) 123506.

/ Nuclear Physics B 584 (2000) 313–328 [40] O. De Wolfe. Sundrum. A. S. Karch. Han.I. hep-th/9910202. D. Phys. Freedman.D. Zhang.Z.A.J. Luty. Rev. R. hep-th/9909134.S. Gubser. [42] T. Kogan et al.328 I. Lykken. . R. [41] M. J. D 59 (1999) 105006.

we establish a systematic method under the α 0 expansion to derive the constraints on the effective Lagrangian imposed by the compatibility of the two descriptions where the form of the field redefinition is not assumed. PII: S 0 5 5 0 .-q Keywords: String theory. Faculty of Science. Japan Received 29 February 2000.ac.elsevier.15. Osaka University. fax: +81-6-68505741 1 seiji@hep-th. tel.V. accepted 30 May 2000 Abstract It was argued that there are two different descriptions of the effective Lagrangian of gauge fields on D-branes by non-commutative gauge theory and by ordinary gauge theory in the presence of a constant B field background. All rights reserved. revised 25 May 2000.: +81-6-6850-5733. Toyonaka.s. 11. University of Tokyo.-w.1 a Department of Physics. In some situations or limits.V. Tokyo 113-0033. In this paper we resolve this puzzle to observe the necessity of gauge-invariant but B-dependent correction terms involving metric in the field redefinition which have not been considered before.u-tokyo. however. E-mail: okawa@het.osaka-u.jp 0550-3213/00/$ – see front matter  2000 Elsevier Science B.  2000 Elsevier Science B. it was found in the previous works that the two descriptions are incompatible under the field redefinition which relates the non-commutative gauge field to the ordinary one found by Seiberg and Witten. Introduction In the light of the recent developments in superstring and M-theory brought by introducing D-branes.phys. such as scalar fields on the worldvolume of the D-branes representing their transverse positions and gauge fields describing internal degrees of freedom. Japan b Department of Physics.jp.3 2 1 3 ( 0 0 ) 0 0 3 6 0 .phys.sci.Nuclear Physics B 584 (2000) 329–358 www. Non-commutative geometry 1.nl/locate/npe Constraints on effective Lagrangian of D-branes from non-commutative gauge theory Yuji Okawa a. the effective Lagrangian describing such collective coordinates is approximated or even supposed to be exactly described by the dimensional reduction of super Yang–Mills theory from ten dimensions to the worldvolume ∗ Corresponding author. Seiji Terashima b.6 .∗ . Graduate School of Science. PACS: 11. All rights reserved.ac. it would be impossible to underestimate the importance of understanding the dynamics of collective coordinates of D-branes. With the problem resolved. Osaka 560-0043. D-branes.25. In the case of bosonic string theory.

The resulting Lagrangian is believed to coincide with the one obtained from the string S-matrix at least for tree-level processes. Okawa. For example. Actually a method towards this goal was developed to some extent in [17] where the problem of whether it is possible to include two-derivative corrections 3 to the DBI 2 For a recent review of the Dirac–Born–Infeld theory see [16] and references therein.330 Y. Actually. S. Since the perturbative interactions between D-branes and those between D-branes and elementary excitations of strings are completely defined by the open string sigma model with the Dirichlet boundary condition. then construct the effective Lagrangian such that it reproduces the S-matrix correctly. Terashima / Nuclear Physics B 584 (2000) 329–358 dimensions of the D-branes [1. For example. it was shown in [12] that the Dirac– Born–Infeld (DBI) Lagrangian [13–15] 2 satisfies the compatibility in the approximation of neglecting derivatives of field strength and its particular form was essential for the compatibility. the B-dependence in the two descriptions is totally different and it turned out that it is possible to constrain the form of the effective Lagrangian by the compatibility of the two descriptions. .2]. the matrix model of M-theory [3] is based on the description of D0-branes in terms of super Yang–Mills theory and the most typical case of the AdS/CFT correspondence [4]. however it is not clear how we can obtain the constraints for other terms imposed by the compatibility in a systematic way so as to study how powerful and useful this approach will be. however. is based on that of D3-branes. we mean terms with n derivatives acting on field strengths (not on gauge fields). The proof of the equivalence of the two descriptions for the DBI Lagrangian in [12] was beautiful. it is in principle possible to calculate systematic corrections of the effective Lagrangian to the Yang–Mills theory. we should calculate the S-matrix of the scattering processes of the gauge fields on D-branes in string theory. namely the correspondence between AdS5 and the four-dimensional super Yang–Mills theory. It is impossible to derive the DBI Lagrangian from the gauge invariance alone so that this shows that the requirement of the compatibility does provide us with information on the dynamics of the gauge fields. Recently it is argued that the effective Lagrangian of the gauge fields on D-branes is described by non-commutative gauge theory [5–11] in the presence of a constant background field of the Neveu-Schwarz–Neveu-Schwarz two-form gauge field which is usually referred to as B field. Another way to calculate the effective Lagrangian is to calculate the beta function of the open string sigma model with Dirichlet boundary condition and to look for a Lagrangian whose equation of motion coincides with the condition that the beta function vanishes. However. This is our basic motivation of the present paper and we will present a method to obtain the constraints systematically in the α 0 expansion. the complexity of the calculation will necessarily increase if we proceed to higher orders in the expansion with respect to α 0 and the string coupling constant gs in the S-matrix approach and to higher loops in the beta function approach so that it would be helpful if other complementary approaches to the effective Lagrangian are available. if we want to obtain the effective Lagrangian of gauge fields on D-branes. 3 By n-derivative corrections to the DBI Lagrangian. It is also possible to describe it in terms of ordinary gauge theory.

it would be dangerous to develop discussions based on the equivalence of the two descriptions. Terashima / Nuclear Physics B 584 (2000) 329–358 331 Lagrangian satisfying the compatibility was discussed and the most general form of the two-derivative corrections up to the quartic order of field strength. to clarify what we assume when deriving the constraints. the one in terms of ordinary gauge theory and the one by non-commutative gauge theory. The field redefinition which preserves the gauge equivalence relation found in [12] and further discussed in [18] should be modified in general and suffered from gauge-invariant but B-dependent correction terms involving metric. If it is the case. Without resolving the puzzle. In particular. We then derive the F 4 terms in the DBI Lagrangian without assuming the form of the field redefinition which relates the ordinary gauge field to the non-commutative one in Section 3. We will argue that the form of the field redefinition should not be assumed as input when constraining the form of the effective Lagrangian and can be rather regarded as a consequence of the compatibility of the two descriptions. Section 5 is devoted to conclusions and discussions. we do not think that it is the case. Furthermore the problem may not be limited to the case of bosonic string theory. It is most likely that we had made too strong assumptions so that we only obtained a limited class of Lagrangians which excludes that of bosonic string theory. we first review the two descriptions of the effective Lagrangian of the gauge fields on D-branes in the presence of a constant B field. In Section 2. We could jump into the problem of two-derivative correction terms to resolve the puzzle. F 4 . our result will show that such terms must exist in the case of bosonic string theory. the methods which are currently available such as the one in [17] do not fulfill our purpose to derive the constraints correctly. Does this mean that the equivalence of the two descriptions in the presence of a constant B field fails in bosonic string theory? In the light of the argument in [12]. We extend our consideration to two-derivative corrections to the DBI Lagrangian in Section 4 where the discrepancy in the case of bosonic string theory is resolved by generalizing the form of the field redefinition. namely. S. we will first determine the F 4 terms which coincide with those in the DBI Lagrangian correctly without assuming the form of the field redefinition in order to show that the idea presented in this paper is useful to constrain the effective Lagrangian.Y. It will turn out that the assumption which is not satisfied in bosonic strings is the one on the form of the field redefinition which relates the ordinary gauge field to the noncommutative one. We then apply it to twoderivative corrections to resolve the puzzle and the generalization to other cases would be straightforward. in the α 0 expansion was derived. The organization of this paper is as follows. . however. This argument is essential in resolving the puzzle in the case of bosonic string theory as we will see. Okawa. Therefore we have to reconsider the assumptions which have been made and find out the correct set of the assumptions from which we should derive the constraints using the problem of whether the puzzle in bosonic strings is resolved as a touchstone of the validity of our approach. However there was a puzzle that the form of the two-derivative terms which is consistent with the compatibility disagreed with the effective Lagrangian derived from bosonic string theory although it was consistent with superstring theory.

2) Sint = −i dτ Ai (x)∂τ x i ∂Σ to the action (2. we see that a constant B field can be replaced by the gauge field 1 Ai = − Bij x j . Review of the two descriptions in the presence of B Let us first review the two descriptions of the effective Lagrangian of D-branes in the presence of a constant B field background Bij . the boundary condition of open strings is modified and is no longer the Neumann one along the D-brane. Thus the propagator in the sigma model is also modified so as to satisfy the new boundary condition.6) where the worldsheet is mapped to the upper half plane. Thus we conclude that there exists a definition of a gauge field in the effective Lagrangian such that the effective Lagrangian depends on B and F only in the combination B + F when we turn on a constant B field.2). with constant metric gij . for simplicity. The explicit form of the propagator evaluated at boundary points is [13–15] i x i (τ )x j (τ 0 ) = −α 0 (G−1 )ij log(τ − τ 0 )2 + θ ij (τ − τ 0 ).332 Y. Okawa. In the presence of a constant B field.5) This is the first description of the effective Lagrangian in terms of ordinary gauge theory. 2 (2. The worldsheet action describing this system is Z  1 d 2 σ gij ∂a x i ∂ a x j − 2πiα 0 Bij  ab ∂a x i ∂b x j S= 0 4πα Σ Z Z i 1 2 i a j d σ gij ∂a x ∂ x − dτ Bij x i ∂τ x j . A background gauge field couples to the string worldsheet by adding Z (2. the gauge transformations and its field strength are defined by δλ Ai = ∂i λ. In this paper. namely.1). we concentrate on the effective Lagrangian of a gauge field on a single D-brane in flat space–time. S. Fij = ∂i Aj − ∂j Ai . Terashima / Nuclear Physics B 584 (2000) 329–358 2. let us examine the propagator in (2. This gauge field is an ordinary one. δλ Fij = 0.4) (2. 2 whose field strength is Fij = Bij .7) .1) and (2. (2. (2. τ and τ 0 are points on the boundary and Gij = gij − (2πα 0 )2 (Bg −1 B)ij .3) (2. (2.1) = 4πα 0 2 Σ ∂Σ where Σ is the string worldsheet with Euclidean signature and ∂Σ is its boundary.1). Comparing (2. To derive the second description in terms of non-commutative gauge theory.

10) exp − 2 n>m i in (2. the θ -dependent part can be factorized as the right-hand side of (2. To distinguish the gauge field in this description from that in the preceding one.θ=0 where Pn ’s are polynomials in derivatives of x and x are coordinates along the D-brane. Now consider the θ -dependence of correlation functions of open string vertex operators which are given by + * k Y  n Pn ∂x(τn ). let us denote the coupling constant in the presence of B as Gs . the metric which appears when contracting Lorentz indices is modified to Gij instead of gij corresponding to the modification in the propagator.9). Therefore. It corresponds to modifying the ordinary product of functions to the associative but non-commutative ∗ product defined by . (2. Okawa. Finally. (2.9) G. Terashima / Nuclear Physics B 584 (2000) 329–358 333 ij 1 1 B θ = −(2πα ) . . Secondly. Since the second term in the propagator does not contribute to contractions of derivatives of x. .Y. The string S-matrix can be obtained from these correlation functions by putting external fields on shell and integrating over the τ ’s. The first one is the coefficient in front of the log term is no longer the metric (g −1 )ij . . e × n=1 . .8) g + 2πα 0 B g − 2πα 0 B There are two important modifications here. .9). ∂ x(τn ). the S-matrix and the effective Lagrangian constructed from it have a structure inherited from this form. The second one is the appearance of the term proportional to the step function (τ ) which is 1 or −1 for positive or negative τ . ∂ 2 x(τn ). S. eip ·x(τn ) 0 2 ij n=1  G. since the coupling constant can depend on B. . let us go on to the most important modification related to the appearance of the θ -dependent factor ! i X n ij m p θ pj (τn − τm ) (2. First.θ ! i X n ij m = exp − p θ pj (τn − τm ) 2 n>m i + * k Y  ipn ·x(τn ) 2 Pn ∂x(τn ). constructed from the one L in the absence of B as follows. So we can see how the effective Lagrangian is modified when we turn on the constant B field. ˆ The Lagrangian Lˆ is let us rename it to Aˆ and denote the Lagrangian in terms of Aˆ as L.

  .

11) f (x) ∗ g(x) = exp θ ij i j f (x + ξ )g(x + ζ ). (2. ∂ ∂ i .

.

Corresponding to the modification of the product. 2 ∂ξ ∂ζ ξ =ζ =0 in the momentum-space representation. gs by Gs and ordinary multiplication by the ∗ product. Now the B-dependence of the effective Lagrangian ˆ gij by Gij . the gauge transformations and the definition of field strength are also modified as follows: . in this description can be obtained through the following replacements: A by A.

F ).14) We have seen that there are two different effective Lagrangians of the gauge field on the D-brane which reproduce the S-matrix of string theory in the presence of a constant B. . The field redefinitions of this kind preserve the ordinary gauge invariance. ∂k Fij .17) 2 1 (2. Thus we do not expect that the two gauge fields Ai and Aˆ i coincide: they would be related by a field redefinition. ˆ = A(A A(A) + δˆλˆ A(A) (2. Terashima / Nuclear Physics B 584 (2000) 329–358 δˆλˆ Aˆ i = ∂i λˆ + i λˆ ∗ Aˆ i − i Aˆ i ∗ λˆ . 2 However we should emphasize here that we do not assume the explicit form of the field redefinition which relates Aˆ i to Ai when we derive constraints on the form of the effective 4 Solutions to the gauge equivalence relation were further discussed in [18].15) These are our fundamental assumptions and we will consider constraints on the form of the effective Lagrangian imposed by the compatibility of them in what follows.2) can be summarized as follows. however.1) and the interaction (2.334 Y.18) λˆ = λ + θ kl ∂k λAl + O(θ 2 ). (2. ˆ δˆλˆ F (2. namely it satisfies ˆ ˆ + δλ A). Okawa.12) bij = ∂i Aˆ j − ∂j Aˆ i − i Aˆ i ∗ Aˆ j + i Aˆ j ∗ Aˆ i . There exists a definition of a gauge field Ai such that the Lagrangian in terms of it respects the ordinary gauge invariance and it depends on B only in the combination B + F. F bij = i λˆ ∗ F bij − i F bij ∗ λ. 2. where fi (∂. a perturbative solution with respect to θ was found by Seiberg and Witten [12].16) ˆ Whether there exists a field redefinition which satisfies (2. is a nontrivial question. and so on. F ) denotes an arbitrary gauge-invariant expression made of Fij . (2. What we have learned from the action (2. There exists a definition of a gauge field Aˆ i such that the Lagrangian in terms of it respects the non-commutative gauge invariance and it depends on B only through Gij . ∂k ∂l Fij . It is not surprising that there are different descriptions of the effective Lagrangian since the S-matrix is unchanged under field redefinitions in the effective Lagrangian so that the construction of the effective Lagrangian from the S-matrix elements is always subject to an ambiguity originated in the field redefinitions. S. The field redefinition which relates Aˆ i to Ai must preserve the gauge equivalence relation. (2.16) with infinitesimal λ and λ. Usually we consider field redefinitions of the form Ai → Ai + fi (∂.13) (2. Its explicit form for the rank-one case is given by 4 1 Aˆ i = Ai − θ kl Ak (∂l Ai + Fli ) + O(θ 2 ). 1. Gs and θ ij in the non-commutative ∗ product. However they will not work in this case because the gauge transformation of Aˆ i is different from that of Ai .

This is an important difference from the previous works such as [12] or [17].15) alone.15) and regularization schemes in the sigma model. Terashima / Nuclear Physics B 584 (2000) 329–358 335 Lagrangian in the present paper. we present the following formulas for convenience which will be used repeatedly: (G−1 )ij = (g −1 )ij + (2πα 0 )2 (g −1 Bg −1 Bg −1 )ij + O(α 0 4 ). 3. (3. our basic standpoint is that the effective Lagrangian we discuss in this paper is constructed so as to reproduce the S-matrix elements correctly and it is not necessary to consider its relation to the background field in the sigma model in what follows. We arrived at the assumptions (2. Since we study the effective Lagrangian in the α 0 expansion.1) and (2. The form of the field redefinition is rather regarded as a consequence of the compatibility of the two descriptions in terms of ordinary and non-commutative gauge theories as we will see in the next section.2) at quantum level. We mentioned the ambiguity related to field redefinitions in constructing effective Lagrangian from S-matrix elements.2) at classical level. Okawa.1) (3.Y. 0 2 θ = −(2πα ) (g ij −1 Bg −1 ij 04 ) + O(α ). Therefore it is not clear whether there is an appropriate regularization corresponding to the non-commutative gauge field Aˆ i in the second assumption of (2. In this sense. the action after turning on B is automatically invariant under the non-commutative gauge transformation (2. the non-commutative gauge transformation suffers from α 0 corrections before taking the zero slope limit. What we assume is the two assumptions (2.15) from the properties of (2.15) although the form of the field redefinition is not assumed. Determination without assuming the form of the field redefinition Let us now proceed to see how the F 4 terms in the DBI Lagrangian are determined by the assumptions (2.15) as assumptions although we can argue that they are plausible in the following way. If the effective action before turning on B is invariant under the ordinary gauge transformation and the B-dependence can be made only through Gij . Determination of F 4 terms revisited 3.15) where no zero slope limit is taken.1. we can also understand the origin of the ambiguity in the point of view of the sigma model to be coming from degrees of freedom to choose different regularization schemes as was discussed in [12].12) at least for the case where the rank of the gauge group is greater than one. However it is necessary to regularize the theory to define composite operators such as (2. Before proceeding. Gs and θ ij . The case with the rank-one gauge theory may be slightly subtle but it would be naturally expected that it holds in this case as well. At any rate. In the case of string theory. we should make a comment on the relation between our assumptions (2. S. The description in terms of the ordinary gauge field Ai will be derived from a Pauli–Villars type regularization while the description in terms of the non-commutative gauge field Aˆ i will be derived from a point-splitting type regularization. However if we take the simple point-splitting regularization discussed in [12] in which we cut out the region |τ − τ 0 | < δ and take the limit δ → 0. we regard (2.2) .

≡ gs Here we omitted a possible overall factor including an appropriate power of α 0 . (3.336 Y.7) Tr(B + F )2 + O(α 0 ) . (3.4). b (3.9) (3.6) to simplify the expressions unless the other metric Gij is explicitly used. on the other hand.7) and (3. we can only determine the normalizations of Gs and Aˆ i as Gs = tgs + O(α 0 ). we only need to multiply an appropriate power of α 0 to the Lagrangian to make the action dimensionless and a numerical constant which depends on the convention. Since the discussions presented in this paper do not depend on the dimension of space–time on which the gauge theory is defined.5) Since Lorentz indices in most of the expressions in what follows are contracted with respect to the metric gij .15) imply that we can describe the system in two different ways when we turn on B as follows: √  det g  (3. The normalizations of Ai and Aˆ i and hence that of Gs are already determined . (3. if we want to supply the overall factor.11) (3. √ Aˆ i = t Ai + O(α 0 ). Now the assumptions (2. (3. namely. In the second line of (3.8) when B vanishes that Gs = gs + O(α 0 ). we made g −1 implicit as Ai Bi ≡ (g −1 )ij Ai Bj .10) In the rank-one case. Terashima / Nuclear Physics B 584 (2000) 329–358 i f ∗ g = fg − (2πα 0 )2 (g −1 Bg −1 )kl ∂k f ∂l g + O(α 0 4 ). Aˆ i = Ai + O(α 0 ).12) from the consideration at the lowest order in α 0 alone since there is no interaction in the F 2 term. Okawa. S. (3. it follows from the comparison between (3.8) (G ) Fj k ∗ (G−1 )kl F L(F ) = Gs In the case of higher-rank gauge theory. And Tr denotes the trace over Lorentz indices as can be seen from the third line of (3. we will adopt this convention together with ∂ 2 ≡ (g −1 )ij ∂i ∂j . the dimension of worldvolume of the D-brane. L(B + F ) = gs √  det G  −1 ij b ˆ bli + O(α 0 ) .4) Tr F 2 + O(α 0 ) .4).3) 2 The lowest order term of the effective Lagrangian of a gauge field on a D-brane in the α 0 expansion is the F 2 term: √  det g  −1 ij (g ) Fj k (g −1 )kl Fli + O(α 0 ) L(F ) = gs √  det g  Fij Fj i + O(α 0 ) = gs √  det g  (3.

2.15) since if we rescale Ai or Aˆ i then the B-dependence does not take the combination bij does not take the B + F for the description in terms of Ai and the field strength F form (2. .17) kia = 0.13) reduces to up to total What is less trivial is the question whether Tr(B + 2 derivative. namely. In general. However the calculation for the determination of t is slightly messy so that we will defer it to Appendix A and proceed assuming t = 1 in this section for the sake of brevity which will be justified in Appendix A.16) gives a contribution to the three-point amplitude of order O(B. .13) anymore as for the description using Aˆ i . Terashima / Nuclear Physics B 584 (2000) 329–358 337 by (2. Therefore this contribution cannot .8). S. whether Tr F satisfies the initial term condition defined by F )2 Tr F 2 f (B + F ) = f (F ) + total derivative. Okawa. which are given by √ det G −1 ij b bli (G ) Fj k ∗ (G−1 )kl F Gs √ det g  ˆ = (∂i Aj − ∂j Aˆ i )(∂j Aˆ i − ∂i Aˆ j ) − 4(2πα 0 )2 Bkl ∂k Aˆ i ∂l Aˆ j ∂j Aˆ i Gs + 2(2πα 0 )2 (B 2 )ij (∂j Aˆ k − ∂k Aˆ j )(∂k Aˆ i − ∂i Aˆ k )  − 12 (2πα 0 )2 Tr B 2 (∂i Aˆ j − ∂j Aˆ i )(∂j Aˆ i − ∂i Aˆ j ) + O(α 0 4 ) . (3.15). (G−1 )ij F (3. the Lagrangian Lˆ on the noncommutative side reduces to the one L on the commutative side at the lowest order in α 0 . . (3.15) The F 4 terms in the DBI Lagrangian are determined by the consideration at the next order terms in the α 0 expansion of (3.16). More precisely. (3. N X (3. a ·x . (3. N.14) in [17]. bj k ∗ (G−1 )kl F bli = (∂i Aˆ j − ∂j Aˆ i )(∂j Aˆ i − ∂i Aˆ j ) + O(α 0 2 ). ζ a · k a = 0. It is verified that Tr F 2 satisfies this condition as follows: Tr(B + F )2 = Tr F 2 + 2 Tr BF + Tr B 2 = Tr F 2 + total derivative + const. It can be easily shown that no other term can produce the contribution of this form on the non-commutative side. which is the condition for a term to be qualified as an initial term of a consistent Lagrangian in the α 0 expansion.15). In this case. Therefore we can in principle determine the constant t from the assumptions (2.16) What is important here is the existence of the second term on the right-hand side of (3. which Let us first check that L(B + F ) and L( is necessary to be consistent with (2. k 3 ).18) a=1 the second term on the right-hand side of (3. ˆ F b ) coincide at the lowest order in α 0 . It gives a non-vanishing contribution to the three-point scattering amplitude of the gauge fields. a = 1. ζ 3 . if we represent the asymptotic fields in N -point scattering as asym a Ai (x) = ζia eik (k a )2 = 0.Y. .

Tr(G−1 F while there were two on the commutative side: Tr BF 3 → 2Bij ∂j Ak ∂k Al ∂l Ai − 2Bij ∂i Al ∂j Ak ∂k Al .20) There are several terms in Tr BF 3 and Tr BF Tr F 2 when we expand them as Fij = ∂i Aj − ∂j Ai . Moreover.23) Tr BF Tr F = Bij Fj i Fkl Flk 2 = 4Bij ∂j Ai ∂k Al ∂l Ak = −8Bij Ai ∂k Al ∂j ∂l Ak + total derivative = 8Bij ∂l Ai ∂k Al ∂j Ak + a term with ∂l Al + total derivative. 2 Tr(B + F )2 = (Tr F 2 )2 + 4 Tr BF Tr F 2 + O(B 2 ). but some of them which contain ∂ 2 Ai or ∂i Ai do not contribute to the S-matrix of the three-point scattering because of the on-shell conditions k 2 = 0 and ζ · k = 0.24) To summarize. (3.21) where we used k 1 + k 2 + k 3 = 0 and (k a )2 = 0. (3. we can extract the part which contributes to the S-matrix from Tr BF 3 and Tr BF Tr F 2 as follows: Tr BF 3 = Bij Fj k Fkl Fli = 2Bij ∂j Ak ∂k Al ∂l Ai − 2Bij ∂j Ak ∂k Al ∂i Al + terms with ∂ 2 A + total derivative. Okawa. They are Tr(B + F )4 and [Tr(B + F )2 ]2 :  Tr(B + F )4 = Tr F 4 + 4 Tr BF 3 + O(B 2 ).19) (3. ζ 3 . It is not difficult to show that the contributions to the S-matrix from Bij ∂j Ak ∂k Al ∂l Ai and Bij ∂i Al ∂j Ak ∂k Al are non-vanishing and linearly independent. This follows from the fact k 1 · k 2 = k 2 · k 3 = k 3 · k 1 = 0 which can be easily seen as 0 = (k 3 )2 = (k 1 + k 2 )2 = 2k 1 · k 2 . g and h are massless fields or their derivatives. 2 2 (3. S. it will be useful to observe that terms of the form f ∂i g ∂i h in general do not contribute to the S-matrix of three-point scattering where f . k 3 ). Thus the conclusion . k 3 ) contribution to the three-point amplitude. There are two terms on the commutative side which can produce the O(B. Another way to see this is to rewrite f ∂i g ∂i h as follows: 1 1 f ∂i g ∂i h = (∂ 2 fgh − f ∂ 2 gh − fg ∂ 2 h) + ∂ 2 (fgh) − ∂i (∂i fgh). Tr BF Tr F 2 → 8Bij ∂j Ak ∂k Al ∂l Ai . ζ 3 . b ∗ G−1 F b ) → −4(2πα 0 )2 Bkl ∂k Aˆ i ∂l Aˆ j ∂j Aˆ i . we have found that all the terms which contribute to the S-matrix of order O(B. Terashima / Nuclear Physics B 584 (2000) 329–358 be canceled and must be reproduced from the Lagrangian L(B + F ) on the commutative side.22) Having been equipped with this formula. (3. there was only one source.338 Y. On the non-commutative side. (3.

ζ 3 . (3. Since the ∗ product is non-commutative. Here we added the ˆ F b ) which were required by the existence of the corresponding F 4 terms b4 terms to L( F in L(B + F ) and the subscripts “arbitrary” there mean that the ordering of the four field strengths in each term is arbitrary.26) + O(α 0 4 ) + derivative corrections .27) LDBI (F ) = p 0 (p+1)/2 gs (2π) (α ) up to an overall factor and an additive constant. and ˆ F b) = L( √   det G −1 b −1 b 0 2 1 b )4 Tr(G−1 F Tr(G F ∗ G F ) + (2πα ) arbitrary Gs 2    1 −1 b 2 2 04 − Tr(G F ) arbitrary + O(α ) + derivative corrections . This coincides with the α 0 expansion of the DBI Lagrangian for a single Dp-brane.2. L(B + F ) √     1 det g 2 0 2 1 4 2 2 Tr(B + F ) − Tr(B + F ) Tr(B + F ) + (2πα ) = gs 2 8  + O(α 0 4 ) + derivative corrections . Thus we have succeeded in determining the F 4 terms in the DBI Lagrangian from the assumptions (2. 2 We can uniquely construct the Lagrangian L(F ) such that L(B + F ) generates the terms (3. Field redefinition We have seen that the two effective Lagrangians. S. Terashima / Nuclear Physics B 584 (2000) 329–358 339 derived from (2.15) without referring to the explicit form of the field redefinition which relates Aˆ i to Ai .28) (3. the following terms must exist in the Lagrangian L(B + F ): Lagrangian L( 1 (3.25) 2(2πα 0 )2 Tr BF 3 − (2πα 0 )2 Tr BF Tr F 2 . The fact that two effective Lagrangians produce the same contribution to the S-matrix does not mean that the two must coincide at off-shell level but . 3. k 3 ).25). Okawa. p 1 det(g + 2πα 0 F ).29) produce the same contribution to the S-matrix of order O(B.15) is that to reproduce the contribution to the S-matrix from the ˆ F b ).Y. However. 8 (3. there is no principle in determining the ordering for the rank-one case and we leave it arbitrary for now. which is given by √    1 det g 2 0 2 1 4 2 2 Tr F − (Tr F ) Tr F + (2πα ) L(F ) = gs 2 8  (3. We will derive its form in the next subsection. we have to specify the ordering of field strengths as in the case of the ordinary Yang–Mills theory.

2 Obviously the O(B) and O(B 2 ) parts of (3.340 Y. S. we have  2 1 1 (2πα 0 )2 Tr(B + F )4 − (2πα 0 )2 Tr(B + F )2 2 8 1 1 0 2 4 = (2πα ) Tr F − (2πα 0 )2 (Tr F 2 )2 2 8 1 + 2(2πα 0 )2 Tr BF 3 − (2πα 0 )2 Tr BF Tr F 2 2 1 + 2(2πα 0 )2 Tr B 2 F 2 − (2πα 0 )2 Tr B 2 Tr F 2 + total derivative + const.34) (3. where we used the facts that 1 Fj i ∂i Flj = Fj i ∂l Fij .16) if we assume Aˆ i = Ai .31) (2πα ) Tr(BF ) − (Tr BF ) = total derivative. (3.30) 4 where we used the fact that   1 0 2 2 2 (3.(3.35) Then the difference 1L can be rewritten using (3.. Terashima / Nuclear Physics B 584 (2000) 329–358 implies that the fields in the two effective Lagrangians can be related by a field redefinition.33) as   1L = 2(2πα 0 )2 Bkl Fj i ∂i Ak (∂l Aj + Flj ) + total derivative = 2(2πα 0 )2 Bkl ∂i Fij Ak (∂l Aj + Flj ) + total derivative. Let us first consider the difference in the O(B) part: 1 1L ≡ 2(2πα 0 )2 Tr BF 3 − (2πα 0 )2 Tr BF Tr F 2 2  − −4(2πα 0 )2 Bkl ∂k Ai ∂l Aj ∂j Ai = 2(2πα 0 )2 Bkl Flj Fj i Fik + 2(2πα 0 )2 Bkl Ak ∂l Fij Fj i + 2(2πα 0 )2 Bkl ∂k Ai ∂l Aj Fj i + total derivative = 2(2πα 0 )2 Bkl Fj i (Flj Fik + Ak ∂l Fij + ∂k Ai ∂l Aj ) + total derivative.30) do not coincide with those of (3. Let us see this explicitly for the case in hand. 2 and that Bkl Fj i (Fik ∂l Aj + ∂k Ai Flj ) = 0.36) .32) This must be reduced to the field redefinition which relates Aˆ i to Ai . (3.33) (3. (3. Okawa. We can make it manifest by noting the fact that   Bkl Fj i ∂i Ak (∂l Aj + Flj )   = Bkl Fj i ∂i Ak (∂l Aj + Flj ) + Ak (∂l ∂i Aj + ∂i Flj )   = Bkl Fj i (Fik + ∂k Ai )(Flj + ∂l Aj ) + Ak 12 ∂l Fij + ∂i Flj = Bkl Fj i (Flj Fik + Ak ∂l Fij + ∂k Ai ∂l Aj ). By expanding the O(α0 2 ) terms in L(B + F ).

and the relation between Aˆ i and Ai is 1 (3. the ambiguity must be fixed by the consideration at higher orders. (3. However since the normalizations of Ai and Aˆ i are already determined by (2. Okawa.39) The solution is given by 1 (3. We should now proceed to the reconsideration of the constraints on the two-derivative corrections to the DBI Lagrangian where the discrepancy was found in the case of bosonic string theory [17].17) found by Seiberg and Witten [12] if we express θ in terms of B. Terashima / Nuclear Physics B 584 (2000) 329–358 341 The fact that 1L does not contribute to the S-matrix and can be reduced to the field redefinition of Aˆ i is now manifest in this form since 1L is proportional to ∂i Fij and hence vanishes using the equation of motion. .37) it obeys that (∂i Aˆ j − ∂j Aˆ i )(∂j Aˆ i − ∂i Aˆ j ) = Fij Fj i + 4(2πα 0 )2 ∂i Fij 1Aj + O(α 0 4 ).38) Thus the appropriate field redefinition is determined by solving the equation 4(2πα 0 )2 ∂i Fij 1Aj = 1L.30). 2 This precisely coincides with the field redefinition (2. 2 up to gauge transformations.15) for the F 4 terms in the DBI Lagrangian. The O(B 2 ) part of the difference between (3.41) Aˆ i = Ai + (2πα 0 )2 Bkl Ak (∂l Ai + Fli ) + O(α 0 4 ).42) by a field redefinition of Aˆ i just as in the case of the difference in the O(B) part and we cannot determine how we should treat (3.15) the ordinary gauge invariance in the description in terms of Ai and the non-commutative gauge invariance in the description using Aˆ i so that the gauge equivalence relation (2. If we write Aˆ i = Ai + (2πα 0 )2 1Ai + O(α 0 4 ).42) 4 is proportional to the F 2 term so that it can be absorbed into the definition of Gs as follows:   1 0 2 2 04 (3.16) and (3. Our result is therefore consistent with the previous works.15) as we mentioned below (3. We did not have to know the form of the field redefinition in the determination of the F 4 terms and the form of the field redefinition was determined from the difference between the two effective Lagrangians at off-shell level.43). 4 Here it is also possible to take care of the difference (3. We will determine the O(α0 2 ) part of Gs in Appendix B from the consideration at order α 0 4 . However it is important to note that this form of the field redefinition should be regarded as a consequence of the assumptions (2. (3.40) 1Ai = Bkl Ak (∂l Ai + Fli ).16) must be satisfied. which justifies (3. This was expected since we assumed in (2. (3.12).15) in our approach.Y. S.43) Gs = gs 1 − (2πα ) Tr B + O(α ) . 1 (2πα 0 )2 Tr B 2 Tr F 2 . We have demonstrated how to constrain the effective Lagrangian of gauge fields on D-branes from the assumptions (2. (3.42) from the consideration at the order α 0 2 .

O(α 0 ) terms The two-derivative corrections to the DBI Lagrangian can first appear at order α0 compared with the F 2 term. Using the Bianchi identity. T3 = −2T2 . (4. The building blocks b and its covariant of Lagrangians in non-commutative gauge theory are field strength F derivatives defined by bj k + i F bj k ∗ Aˆ i . Thus ordinary gauge theory. terms of the forms ∂F ∂F . but F there are four terms at order α 0 : 5 bi F bj F bik ∗ D bj k .7) but the terms Tb1 and Tb2 do not coincide up to total derivative since 5 Lorentz indices on the non-commutative side should be regarded as being contracted using G although we ij will not write it explicitly in this subsection contrary to the conventions (3. bj k = ∂i F bj k − i Aˆ i ∗ F bi F D (4.1. T3 ≡ ∂k Fij ∂k Fj i . (4. . Lagrangians are made of field strength and its derivatives. terms of the forms D b b b b can be transformed to the D F D F terms using the integration by parts as in the case of b3 terms no longer vanish even for the rank-one case.3) T2 = −Fik ∂j ∂i Fj k + total derivative. Tb4 ≡ i F b2 ≡ D bj F bi F bik ∗ D bj k . Let us first survey possible terms at this order in both ordinary and non-commutative gauge theories. Okawa. The story is slightly different in non-commutative gauge theory. T2 ≡ ∂j Fik ∂i Fj k . There are three different ways to contract Lorentz indices: T1 ≡ ∂i Fik ∂j Fj k .1) Using the Bianchi identity. (4. it is sufficient to consider the ∂F ∂F terms. (4. Constraints on two-derivative corrections 4.6) b3 term by i to make it Hermitian. Tb1 ≡ D bij ∗ F bj k ∗ F bki .2) and the two remaining terms T1 and T2 coincide up to total derivative: T1 = −Fik ∂i ∂j Fj k + total derivative. (4.4) Thus any term at order α 0 can be transformed to T1 .5) and (3. T bk F bk F bij ∗ D bj i . where we multiplied the F b b the term T3 reduces to T2 as before. F ∂ 2 F and F 3 are possible.342 Y. the term T3 reduces to T2 . Tb3 ≡ D (4. S.5) bF bD bF b. At order α 0 . The F b terms At order α 0 . In ordinary gauge theory. Tb3 = −2Tb2 . Terashima / Nuclear Physics B 584 (2000) 329–358 4.6). However since the F ∂ 2 F terms can be transformed to the ∂F ∂F terms using the integration by parts and F 3 terms vanish for the rank-one case. F bD b 2F bD b2F b and F b3 are possible.

8) (4. we can proceed allowing a finite T1 term to be present in the Lagrangian without restricting the definition of the gauge field . eij F F (4.15) where field strengths in fi are accompanied by at least one derivative. We will choose {Tb1 . the first assumption of (2. namely. ∂ 2 F. We have found bases of O(α 0 ) terms for both ordinary and non-commutative gauge theories. field redefinitions of the form A˜ i = Ai + fi (∂F. It is possible to absorb T1 into the F 2 term by a field redefinition which is given by A˜ i = Ai + a(2πα 0 ) ∂j Fj i + O(α 0 2 ). Tb1 = −F bik ∗ D bj D bi F bj k + total derivative. The remaining three terms Tb1 .). the Lagrangian in terms of Ai also depends on B only in the combination B + F .15) does not determine the definition of the gauge field uniquely. For instance. ej i = Fij Fj i + 4a(2πα 0 ) ∂i Fij ∂k Fkj + total derivative + O(α 0 2 ). D bj ]F bj k + total derivative Tb1 − T bij ∗ F bj k + i F bj k ∗ F bij ) + total derivative bik ∗ (−i F = −F = −2Tb4 + total derivative. T b2 and Tb4 are not bi and D where D independent which can be seen as follows: b2 = −F bik ∗ [D bi . . (B + F (4.15). (4.13) It is important to notice that this field redefinition has the following property: e)ij = (B + F )ij + a(2πα 0 ) ∂ 2 (B + F )ij + O(α 0 2 ).11) Tb4 = F ij ∗ (F 2 2 which precisely corresponds to the ambiguity of the ordering of covariant derivatives when we construct a non-commutative counterpart of the term Fij ∂k ∂i Fj k . T The origin of the extra term Tb4 can be interpreted as an ambiguity in constructing noncommutative gauge theory from ordinary gauge theory for the rank-one case.15).14) because of the fact that ∂i (B + F )j k = ∂i Fj k for a constant B. We will next consider the properties of the bases with respect to their behavior under field redefinitions and to the relation to our assumptions (2. As can be seen from this example. do not spoil the first assumption of (2. Tb2 = −F 343 (4. .9) bj no longer commute.Y. This is characteristic of the rank-one theory and there is no such ambiguity in higher-rank cases where the ordering of field strengths or covariant derivatives is already determined in ordinary Yang– Mills theory.10) b4 } as a basis of O(α 0 ) terms in non-commutative gauge theory. Okawa. (4. For ordinary gauge theory our basis consists of T1 alone.12) (4.15). . Since the term T1 satisfies the initial term condition (3.14) This implies that if the effective Lagrangian in terms of A˜ i depends on B only in the form of B + F . both Ai and A˜ i satisfy the first assumption of (2. (4. This can be seen manifestly if we rewrite Tb4 as ib bij ∗ [D bj k ∗ F bki − F bki ∗ F bj k ) = 1 F bk . Terashima / Nuclear Physics B 584 (2000) 329–358 bik ∗ D bi D bj F bj k + total derivative. D bi ]F bj k . S.

For non-commutative gauge theory our basis consists of Tb1 and Tb4 . the term T given by e b F ij ˜ bj i + O(α 0 2 ).19) = i tr Tr F 3 .18) namely. k 3 ) and that the F 4 terms are determined by the requirement that the Lagrangian L(B + F ) should reproduce the contribution.15) so that we can select a definition of the non-commutative gauge field satisfying (2. It would be rather trivial that the existence of Tb4 in the effective Lagrangian is consistent with our assumptions (2.14). (B + F )ki 2 i i = tr Fij [Fj k . On the other hand the term Tb4 cannot be redefined away and it gives a non-vanishing contribution to the S-matrix at O(B) as we will see shortly.15) for convenience. Having understood that b4 is possible at order α 0 . Fki ] 2 2 (4.15) such that the term Tb1 vanishes in the Lagrangian. With this convention and the one for the ordinary gauge field we mentioned in the last paragraph. Fki ] + Bij tr[Fj k .2. bj F Aˆ i = Aˆ i + a(2πα 0 )D e =F b bij ∗ F bj i + 4a(2πα 0 )D bi F bk F bij ∗ D bkj ∗F ji + total derivative + O(α 0 2 ). ζ 3 .17) This field redefinition preserves the second assumption of (2. (4. S. there is no O(α 0 ) term in L(B + F ) and only the Tb4 ˆ F b ) at order α 0 . (4. (4.15) b4 is for the rank-one case since it vanishes in the commutative limit. which implies that term exists in L( Aˆ i = Ai + O(α 0 2 ).16) (4. which can be shown as follows:   i i tr Tr(B + F )3 = tr(B + F )ij (B + F )j k .22). no O(α 0 ) part in the field redefinition. 4. b4 is evaluated in the α 0 expansion as follows: The term T 1 bij ∂n F bj k ∂m F bki + O(α 0 4 ). let us develop a similar discussion for two-derivative the term T corrections. the term T 3 consistent for higher-rank cases as well since its commutative counterpart i tr Tr F . we have shown that the F the S-matrix of order O(B. However we will take an alternative approach that we choose a definition of the gauge field in terms of which the T1 term vanishes in the Lagrangian among the ones which satisfy the first assumption of (2. The result is . Okawa. As in the case of b2 term by a field redefinition b1 can be absorbed into the F ordinary gauge theory. Constraints on two-derivative corrections b2 term produces a non-vanishing contribution to In Section 3. Incidentally. where tr denotes the trace over color indices. satisfies the initial term condition (3.20) Tb4 = (2πα 0 )2 Bnm F 2 We can extract the part which gives a non-vanishing contribution to the three-point amplitude using the formula (3.344 Y.1. Terashima / Nuclear Physics B 584 (2000) 329–358 further.

J1 (B 2 ) = ∂n Fij ∂n Fj i Bkl Blk .23) The terms J4 . (4. which can be rewritten using (3. J3 (B) = 2Bni Fim ∂n Fkl ∂m Flk . ∂ 2 Fij = ∂i ∂k Fkj − ∂j ∂k Fki .25) . J4 = ∂n Fni ∂m Fim Fkl Flk . k 5 ) on the commutative side. J2 (B 2 ) and J3 (B 2 ) vanish if they are evaluated at on-shell configurations (3. b2 and b3 in this basis do not change under field redefinition and unambiguous [20]. ζ 3 . J6 = ∂ 2 Fij Fj i Fkl Flk .24) It is easily seen that the values of J1 (B 2 ). Terashima / Nuclear Physics B 584 (2000) 329–358 345 Tb4 = (2πα 0 )2 Bnm ∂i Aˆ j ∂n ∂j Aˆ k ∂m ∂k Aˆ i + terms with ∂ 2 A + total derivative + O(α 0 4 ). J2 = ∂n Fij ∂n Fj k Fkl Fli . On the other hand. 3 are given by J1 (B) = 2 ∂n Fij ∂n Fj i Bkl Flk . J3 = Fni Fim ∂n Fkl ∂m Flk . Explicit expressions of Ji (B) and Ji (B 2 ) for i = 1.21) The first term on the right-hand side of (4. (4. J6 and J7 contain the part ∂j Fj i so that they do not contribute to the S-matrix. Okawa. J5 = −∂n Fni ∂m Fij Fj k Fkm . J3 (B 2 ) = Bni Bim ∂n Fkl ∂m Flk . the first three coefficients reproduces the contribution from the term T b1 . J2 (B) = 2Bij Fj k ∂n Fkl ∂n Fli . (4. J7 = ∂ 2 Fij Fj k Fkl Fli . k 5 ). Any term of order O(∂ 2 F 4 ) can be transformed to the following form using the integration by parts and the Bianchi identity [19]: L= 7 X bi Ji . J2 (B 2 ) = ∂n Fij ∂n Fj k Bkl Bli . ζ 3 .15).18). Let us denote the O(B n ) part of Ji with F replaced by B + F as Ji (B n ) following [17].22). Therefore the term J3 (B) is the only one which contributes to the S-matrix of order O(B. Therefore our goal is to answer the question whether these coefficients are constrained by our assumptions (2.Y. Thus we do not need to consider these terms in the search for the term which b4 . On the commutative side. J5 .21) provides a non-vanishing contribution to the S-matrix of order O(B. S. 2. (4. This holds after replacing F with B + F since the part ∂j Fj i remains intact in the replacement.17) satisfying (3.22) i=1 where J1 = ∂n Fij ∂n Fj i Fkl Flk . We can also show that the terms J1 (B) and J2 (B) do not contribute to the S-matrix using the formula (3.22) as follows: J3 (B) = 4Bni ∂i Am ∂n ∂k Al ∂m ∂l Ak + terms with ∂ 2 A + total derivative = −4Bni ∂i ∂l Am ∂n ∂k Al ∂m Ak + a term with ∂l Al + terms with ∂ 2 A + total derivative = −4Bnm ∂i Aj ∂n ∂j Ak ∂m ∂k Ai + a term with ∂ · A + terms with ∂ 2 A + total derivative. (4. only terms of the form O(∂ 2 F 4 ) can produce the same form of the contribution after replacing F with B + F .

There is another comment on our result regarding the relation between the coefficients b4 and J3 (4. Following this terminology.26). We will get back to this point after discussing the issue on field redefinitions.27) we can add the term to the effective Lagrangian without violating the assumptions (2. we can add the terms J1 and J2 to the effective Lagrangian without violating the assumptions (2. however. 4 (4. Okawa. the coefficient in front of T as we discussed in the preceding subsection so that the coefficient in front of J3 is also arbitrary. In general it would become more difficult to satisfy the on-shell initial term condition when the number of field strengths minus the number of derivatives increases in the term under consideration. the coefficients in front of J1 and J2 are not constrained by the assumptions (2.26) In addition to J3 . We admit.15) at the same order of α 0 as f (F ). In general.15) since J1 and J2 satisfy the on-shell initial term condition. S. We will call (4. we can say that the terms J1 and J2 do not satisfy the initial term condition (3. However if we succeed in extending our consideration to higher-rank cases. The most general form of O(∂ 2 F 4 ) terms was derived in [17] from the requirement that ˆ F b ) coincide up to total derivative under the assumption that the field L(B + F ) and L( . The coefficient in front of J3 is correlated with that in front of Tb4 on the non-commutative b4 was arbitrary side following the relation (4.26). we can reasonably expect severe constraints on such higher-order terms.346 Y. However we do not expect that it holds at higher-order terms in the α 0 expansion because of the following argument. We should now clarify the relation between the result in this paper and that in [17]. As an example of promising methods we can refer to the one discussed in [21].15) at this order. if a term f (F ) satisfies the condition that f (B + F ) = f (F ) + total derivative using the equation of motion. Terashima / Nuclear Physics B 584 (2000) 329–358 The non-vanishing contribution to the S-matrix from J3 (B) takes the same form as that of Tb4 (4. However.15) since J1 (B) and J2 (B) do not contribute to the S-matrix at the order we are discussing. This result may seem discouraging in view of our motivation to obtain constraints on the effective Lagrangian. This provides no information on the effective Lagrangian for in front of T the rank-one case since Tb4 vanishes in the commutative limit. that the approach presented in this paper will not be practical in deriving the constraints on the higher-order terms and we need more efficient methods. it might be possible to obtain a prediction on a relation between the coefficient in front of the F 3 term and coefficients in O(D 2 F 4 ) terms. (4.27). If we note that the existence of the solutions to the on-shell initial term conditions was essential to our conclusion that there is no constraint on the O(∂ 2 F 4 ) terms.14) but satisfy the on-shell initial term condition (4. Thus our conclusion is that two-derivative corrections of the form O(∂ 2 F 4 ) are not constrained at all by the assumptions (2. To summarize.21) so that it is possible to reproduce the S-matrix from Tb4 by J3 (B) with the following normalization factor: 1 Tb4 ∼ − (2πα 0 )2 J3 (B).27) the on-shell initial term condition.

which is necessary to satisfy the assumptions (2.17). they are less interesting than J1 and J2 since they do not contribute to the S-matrix.27) but does not satisfy the initial term condition (3.29) The conclusion in this paper that no constraint is imposed on O(∂ 2 F 4 ) terms is trivially consistent with (4.15) in the preceding subsection.Y. 4 This was inconsistent with the O(∂ 2 F 4 ) terms in bosonic string theory derived from the string four-point amplitude [19] or from the two-loop beta function in the open string sigma model [20] which are proportional to 6 1 − J1 − 2J2 + J3 .26). J6 and J7 also satisfy the on-shell initial term condition. Therefore the terms J2 (B) and J2 (B 2 ). J1 (B 2 ). the O(α 0 2 ) part of L( L( 0 3 α which is manifest under our convention (4.29) and the difference between this conclusion and that in [17] implies that the relation between the two gauge fields Aˆ i and Ai in (2. does not affect O(α 0 3 ) terms under consideration if there are no O(α 0 ) terms in ˆ F ˆ F b ) cannot generate B-dependent terms of order b ).41). In particular. J5 . the discrepancy between (4. Furthermore.27) as a necessary condition for a term to be added to the effective Lagrangian without violating the assumptions (2.18). . In particular. The terms J4 . the ˆ F b ) because of the relation (4. We will construct a field redefinition which is relevant to bosonic string theory in the next subsection. The relation between Aˆ i and Ai must be in general modified if we add a term which satisfies the on-shell initial term condition (4.28) and (4. however. 4.28) − J1 + 2J2 + J3 .15) does not in general take the form of (2. As we have seen.14). J2 and J3 must appear in the combination that 1 (4. absence of J3 means that Tb4 is not allowed to exist in L( ˆ F b ) under our convention that Tb1 should be redefined Thus there are no O(α 0 ) terms in L( away. 6 This expression is slightly different from (4) in [20] but one of the authors was informed of a misprint in (4) of [20]: the last coefficient b3 should have sign +. 4 (4. The result was that the terms J1 .2 but we will not do that for completely general cases. Corrections to the field redefinition We presented the on-shell initial term condition (4. It would be sufficient to demonstrate it for some examples including the one which is relevant to bosonic string theory since the generalization is straightforward. An explicit form of the required field redefinition which relates Aˆ i to Ai when we add a term which satisfies the on-shell initial term condition to the effective Lagrangian can be derived in the same way as we did in Section 3.17) assumed in [17].3. S.15) as we have seen in the preceding section. the terms J1 and J2 are examples of such terms since J1 (B). Terashima / Nuclear Physics B 584 (2000) 329–358 347 redefinition is given by (2.29) shows that it is indeed the case for bosonic string theory. J2 (B) and J2 (B 2 ) are not total derivative although values of them vanish when evaluated at configurations satisfying the on-shell conditions (3. This simplifies the discussion since the O(α 0 2 ) part in the field redefinition (3. Let us first consider a case where only J2 exists in the O(α 0 3 ) part.18). Okawa.

17).32) as follows: Aˆ i → A˜ i → Ai .34) 0 3 bF b’s and F b’s in the O(α 0 3 ) term contracted using Gij as with an arbitrary ordering of D indicated by the subscript. Okawa. (4.16).36) .35) this does not take the form of (2. (4. J2 (B ) = −∂n Fni ∂j (B F + FB )j i + total derivative. Terashima / Nuclear Physics B 584 (2000) 329–358 which are necessary to realize the B-dependence of the form B + F when we add J2 . where (4.41) are in general possible. must b 2 term by the O(α 0 3 ) part of the field redefinition of Aˆ i . the two generates c2 (2πα 0 )3 (J2 (B) + J2 (B 2 )) from the F Lagrangians.32) preserves the gauge equivalence relation (2. 2 2 2 (4.30) (4.32). Since Bij = − 1 (gθg)ij + O(θ 2 ). (2πα 0 )2 (4. This example shows that α 0 corrections of O(B) to the field redefinition (3.31) It follows from a similar argument to the one used to determine the form (3.33) and ˆ F b) L( √  det G b ∗ G−1 F b) Tr(G−1 F = Gs    1 b )4arbitrary − 1 Tr(G−1 F b )2 2 + (2πα 0 )2 Tr(G−1 F arbitrary 2 8  04 b b b b b b + c2 (2πα ) (Dn Fij ∗ Dn Fj k ∗ Fkl ∗ Fli )G.348 Y. S. Let us decompose the field redefinition (4. are related by the field redefinition (4. Therefore it would be helpful to confirm that (4.41) that the field redefinition 1 Aˆ i = Ai + (2πα 0 )2 Bkl Ak (∂l Ai + Fli ) 2 1 (4. arbitrary + O(α ) . Its explicit be generated from the F form is easily derived if we rewrite J2 (B) and J2 (B 2 ) as follows: J2 (B) = −2 ∂n Fni ∂j (FBF )j i + total derivative. To summarize. L(B + F ) √    2 1 det g 1 Tr(B + F )2 + (2πα 0 )2 Tr(B + F )4 − Tr(B + F )2 = gs 2 8  + c2 (2πα 0 )3 ∂n (B + F )ij ∂n (B + F )j k (B + F )kl (B + F )li + O(α 0 4 ) .32) − c2 (2πα 0 )3 ∂j (2FBF + B 2 F + FB 2 )j i + O(α 0 4 ) 4 b2 term.

43). 7 Now the extension to cases where other Ji ’s except J3 exist in the effective Lagrangian would be straightforward. √     det G b 1 ˆ b (4.41) with satisfy L( the definition of Gs (3.15). S. A˜ i in this example. However there is no contradiction since it was assumed in [18] that Lorentz indices in a mapping from Ai to Aˆ i are contracted among derivatives of the gauge field and δθ ij alone while (g −1 )ij is used in our case (4. the field redefinition (3.41) maps a noncommutative gauge field to an ordinary gauge field but the B-dependence of the effective Lagrangian in terms of the resulting gauge field.32) preserves the gauge equivalence relation (2. Here Jˆi ’s are the non-commutative counterparts of Ji ’s with an arbitrary ordering of the fields.38) A˜ i = Ai − c2 (2πα 0 )3 ∂j (2FBF + B 2 F + FB 2 )j i + O(α 0 4 ). Therefore further B-dependent redefinition like (4. The form of the field redefinition (4. In general.Y. the non-commutative gauge field Aˆ i is mapped to an ordinary gauge field A˜ i which respects the ordinary gauge invariance while A˜ i is mapped to another ordinary gauge field Ai by the second part (4. A linear combination of the two Lagrangians is expressed in our basis {Tb1 . suffers from the ambiguity in the rank-one case discussed direction. This shows that (4. However if J3 exists the story becomes slightly complicated ˆ F b ) which accompanies J3 following the relation (4.39) T3 + (2πα 0 )2 − Jˆ1 + 2Jˆ2 + Jˆ3 + O(α 0 4 ) . Aˆ i = A˜ i + (2πα 0 )2 Bkl A˜ k (∂l A˜ i + F (4.38) since the difference between A˜ i and Ai is gauge invariant although it depends on B. Let us review acts on the O(α 0 ) term T them briefly. Kishimoto for clarifying this point.37) 2 1 (4.41) when it b4 .26). Here it is convenient to utilize the results of [17].37). b4 in L( because of the presence of T We have to consider the effects of the O(α 0 2 ) part in the field redefinition (3.32) although it is implicit under our convention (3. . does not take the form of B + F .40) ˆ F b ) = L(B + F ) up to total derivative under the field redefinition (3. Gs 8 2 (4. Terashima / Nuclear Physics B 584 (2000) 329–358 349 1 eli ) + O(α 0 4 ).1. It was shown in [17] that the two Lagrangians. 4 By the first part (4.38) is necessary to map it to the gauge field which satisfies the first assumption of (2.32) does not belong to the class of solutions to the gauge equivalence relation (2. Okawa.16) found in [18]. We presented the Lagrangians on the non-commutative side because we can uniquely construct their commutative counterparts while the other ˆ F b ). Tb4 } as follows: √    ˆ b4 + a(2πα 0 )2 Jˆ5 − 1 Jˆ6 + 1 Jˆ7 b1 + b T b ) = det G a T L(F Gs 8 2 7 We would like to thank I.5).16). L(B + F ) → L( in Section 4. L1 (F ) = Gs 4 and b) = Lˆ 2 (F √     1 ˆ 1 ˆ det G b 0 2 ˆ 04 T1 + (2πα ) J5 − J6 + J7 + O(α ) .

41) with the definition of Gs (3.41) with the definition of Gs (3. it follows that 1 1 1 (4.42) + O(α 0 2 ) terms different from those of L( b ) = L0 (B + F ) up to total derivative under the field redefiniwhich also satisfies Lˆ 0 (F tion (3. 8 To see that it is the case.14). It is difficult to see whether F (F ) satisfies the initial term condition by a direct calculation. we can obtain the answer by an indirect argument in the following way. (4. Terashima / Nuclear Physics B 584 (2000) 329–358    1 1 1 − b(2πα 0 )2 − Jˆ1 + 2Jˆ2 + Jˆ3 + 2Jˆ5 − Jˆ6 + Jˆ7 + O(α 0 4 ) .41) is the most general form of two-derivative ˆ F b ) = L(B + F ) up to corrections up to this order in the α 0 expansion which satisfy L( total derivative under the field redefinition (3. it was shown [17] that the condition that O(∂ 2 F 4 ) terms must be proportional to the combination that 1 1 (4. This result provides us with a good starting point for the case where J3 is non-vanishing. 4 4 is necessary to satisfy the initial term condition (3.43). S.350 Y.14). Namely.41) It was further argued in [17] that (4.44) − F (B + F ) = Bnm Fij ∂n Fj k ∂m Fki − F (F ) + total derivative.43).41) is the most general form reduces to the one whether there are solutions of the form O(∂ 2 F 4 ) to the initial term condition (3. 4 2 4 which implies that F (F ) does not satisfy the initial term condition (3. 4 4 4 (4.41) with the definition of Gs (3. Thus the question whether (4. Now that the only remaining possibility was denied.41) with the satisfy L( definition of Gs (3. the statement that there is no solution of the form O(∂ 2 F 4 ) to the initial term condition (3.43) F (F ) ≡ − J1 + 2J2 + J3 + 2J5 − J6 + J7 . it is helpful to notice that if there is another Lagrangian √  0 b ) = det G a T b4 b1 + b T Lˆ (F Gs  ˆ F b ) + O(α 0 4 ) .43).14) was shown and this implies that (4.20) and the fact ˆ F b ) = L(B + F ) up to total derivative under the field that (4. From (4. .41) with a = 0 satisfies L( redefinition (3. Regarding the latter question.14). the Lagrangian √  ˆ F b ∗ G−1 F b ) + b(2πα 0 )Tb4 b ) = det G Tr(G−1 F L( Gs    1 b )4arbitrary − 1 Tr(G−1 F b )2 2 + (2πα 0 )2 Tr(G−1 F arbitrary 2 8 8 The argument for proving this statement developed in Section 3 of [17] was incorrect as explained in the note added at the end of hep-th/9909132 v2. Okawa.43). however. then the difference L0 (F ) − L(F ) must be a solution of the form O(∂ 2 F 4 ) to the initial term condition (3.41) is the most general form of two-derivative corrections up to this order in the α0 expansion which ˆ F b ) = L(B + F ) up to total derivative under the field redefinition (3.14).

the form of the field redefinition (4. However there is no general argument that it persists to higher orders in the α 0 expansion. In particular. (4. we established the equivalence of the two descriptions for a larger class of Lagrangians. the coefficients in front of J1 . J5 .41). This corresponds to adding b(2πα 0 )3 Jˆ2 to (4. S. In particular it contains the effective Lagrangian of bosonic string theory and thus the . By applying this method to two-derivative corrections to the DBI Lagrangian. 5. J2 and J3 calculated in bosonic string theory are proportional to (4. we should modify the form of the field redefinition at order α 0 3 appropriately as in the preceding example where only J2 exists. We should keep such possibility of corrections in mind when we use properties of the field redefinition which relates the non-commutative gauge field to the ordinary one. 4 (4. As an interesting example of such cases. As we mentioned in the preceding subsection.41) are not only possible in principle but also realized actually in bosonic string theory.29) does not take the general form (4. Okawa.Y.45) 4 4 4 ˆ F b ) are related by the field redefinition (3. let us derive the form of the field redefinition which is relevant to bosonic string theory.41) in the absence of the O(α 0 3 ) terms. Conclusions and discussions We considered the constraints on the effective Lagrangian of the gauge field on a single D-brane in flat space–time imposed by the compatibility of the description by nonˆ F b ) with that by ordinary gauge theory L(B + F ) in the commutative gauge theory L( presence of a constant B field background. it would be important to note that corrections of O(B) ∼ O(θ ) modify ˆ ) introduced in [12] for more general descriptions of the the differential equation of δ A(θ system in terms of non-commutative gauge theory.41) at order α0 3 are not required. J6 and J7 which do not affect the S-matrix.15) alone without assuming the form of the field redefinition which relates the non-commutative gauge field Aˆ i to the ordinary one Ai . However we cannot make the O(α 0 3 ) terms vanish since (4.45) so that the form of the field redefinition is modified to 1 Aˆ i = Ai + (2πα 0 )2 Bkl Ak (∂l Ai + Fli ) 2 1 − b(2πα 0 )3 ∂j (2FBF + B 2 F + FB 2 )j i + O(α 0 4 ).46) If we further change the coefficients in front of J4 . it was found that the coefficients in front of J1 .29) [19.20]. J2 and J3 vanish [19] so that corrections to the field redefinition (3. Terashima / Nuclear Physics B 584 (2000) 329–358 351    1 1 1 − b(2πα 0 )3 − Jˆ1 + 2Jˆ2 + Jˆ3 + 2Jˆ5 − Jˆ6 + Jˆ7 + O(α 0 4 ) .46) itself is modified correspondingly. Thus the corrections to the field redefinition (3. If we and L(B + F ) constructed from L( want to change the coefficients in font of Ji ’s except J3 . For superstring theory. We presented a systematic method under the α0 expansion to derive the constraints based on the assumptions (2.

2. Regarding this issue. it is in principle possible to calculate the general form of the effective Lagrangian which satisfies the assumptions (2. Therefore if there are no solutions to these conditions in the two-derivative terms at higher order in α 0 . See also a related work [25]. there is no argument which justifies the assumption in the superstring case that the metric gij does not appear in the field redefinition which relates Aˆ i to Ai . It is therefore necessary to extend the method to apply it to more general cases where there are corrections to the field redefinition of the form (2.17) was presented in [21]. as far as we are aware of. S. In the superstring case the field redefinition of the form (2. We believe that we have elucidated the mechanism to constrain the effective Lagrangian of the gauge fields on D-branes using non-commutative gauge theory. It is sometimes said that the form of the field redefinition which relates Aˆ i to Ai can be determined by solving the differential equation derived from the gauge equivalence relation in [12] up to the ambiguities found in [18]. Since we presented a systematic method to obtain the constraints in the α 0 expansion. the simplified Seiberg–Witten map considered in [22. Furthermore our study implies that the number of the free parameters in the general form is equal to or less than the number of solutions to the initial term condition (3.17) can be consistent up to two derivatives [21] and may not be corrected.15) up to an arbitrary order in α 0 .17) such as the case of bosonic string theory.352 Y.23] and [24] may be useful because of its geometric nature although we should clarify its meaning for our approach. Even the form of the differential equation itself can be modified and the form of the field redefinition is hardly constrained by the gauge equivalence relation without the assumption. In addition. However our result clearly shows that it is no longer the case if we allow the metric gij to appear in the field redefinition as in the case of bosonic string theory. . However.27).14) plus that of nontrivial solutions to the on-shell initial term condition (4. for example. If we could succeed in such generalization. How to construct actions which are invariant under the simplified map was recently discussed in [26]. Terashima / Nuclear Physics B 584 (2000) 329–358 puzzle in bosonic strings found in the previous works was resolved and the equivalence in this case was first made consistent. Okawa. J2 and J3 . a general method to construct 2n-derivative terms to all orders in α 0 which satisfy the compatibility of the two descriptions in the approximation of neglecting (2n + 2)-derivative terms when the field redefinition takes the form of (2. They were not considered previously because it was assumed that the metric gij does not appear in the field redefinition which relates Aˆ i to Ai . this implies that the form of the two-derivative terms is in principle determined uniquely by the requirement of the compatibility up to the parameters in front of J1 . It should be emphasized that it was crucial to reach this observation that we did not assume the form of the field redefinition when we derive the constraints. However we admit that the approach adopted in the present paper is not practically useful to proceed to the higher orders to obtain the constraints or the explicit form of the terms as we mentioned in Section 4. In resolving the puzzle it was essential to observe that the gauge-invariant but B-dependent corrections to the field redefinition (2. however we showed that they must exist for the case of bosonic string theory. it would be expected that it will provide us with a new powerful method to study the dynamics of D-branes.17) are in general necessary for the compatibility. For developments in this direction.

3) ˆ F b ) (3. Okawa. 11-08864. gs ˆ ˆ i = √Ai . they are not related to the expansions with respect to α0 and gs once extracted. An issue related to this kind of extension was discussed in [27]. Y. A safer approach is to make calculations after rescaling both fields such that their normalizations ˆ i: coincide. it will be worth investigating in view of the various important developments which have been made by the super Yang–Mills theory in the description of multi-body systems of D-branes. were supported in part by the Japan Society for the Promotion of Science under the Postdoctoral Research Programs No.8) can be rewritten as follows: the effective Lagrangians L(B + F ) (3. Sato and T. 11-01732 and No.1) If we define the field strengths of the normalized fields as Fij ≡ ∂i Aj − ∂j Ai . Since the normalizations of Ai and Aˆ i do not coincide when t 6= 1 as can be seen from (3. Acknowledgements We would like to thank K. Furthermore.12). We only considered the constraints on the effective Lagrangian at the lowest order in the expansion with respect to the string coupling constant gs in this paper. and S. Kawano for useful communications on non-commutative gauge theory.11).O.Y. Ohta. H. p p ˆ j + i Gs A ˆ i. namely the constant t in (3. Terashima / Nuclear Physics B 584 (2000) 329–358 353 it is interesting to combine our approach with consideration of supersymmetry and string dualities.O. We denote the normalized fields by Ai and A Ai Ai = √ . There seems to be no crucial obstruction to the extension of our approach to higher orders in gs although some modifications may be required. we should be careful when evaluating the S-matrix.7) and L( . Although we foresee possible complication originated in its non-Abelian nature which exists even on the side of ordinary gauge theory. Of course it might be the case that there are limitations of the description in terms of non-commutative gauge theory at nonperturbative level and it is important to investigate them.15) were derived from the action of the sigma model. M. ˆ j − ∂j A ˆ i − i Gs A ˆi ∗A ˆj ∗A b Fij ≡ ∂i A (A. Yoneya for valuable discussions and comments.2) (A. It would be interesting if we could obtain some non-perturbative information on the dynamics of D-branes from them. by the assumptions (2. although the assumptions (2. S.15). It would be interesting if we could obtain some insight into the non-Abelian generalization of the DBI Lagrangian [28]. A Gs (A. also thanks N. Hashimoto.T. Determination of the O(α 0 0 ) part of Gs In this appendix. It will probably provide us with further constraints. This works of Y. Kajiura and T. respectively. we determine the O(α 0 0 ) part of Gs . Another important extension of our approach is to consider higher-rank gauge theory. Appendix A.

= (A.11). S.10) + O(α ) + derivative corrections . The result is as follows: L(B + F )    2 4  p p B B 1 0 2 Tr √ + F = det g Tr √ + F + (2πα ) Gs gs gs 2 gs   2 2   B 1 Tr √ + F − + O(α 0 4 ) + derivative corrections 8 gs √   √ det g 2 0 2 1 Tr(B + F )4 Tr(B + F ) + t (2πα ) = gs 2    1 2 2 04 − Tr(B + F ) (A. 8 where we used t defined by (3.8) (A. The O(B) part of this Lagrangian coincides with that of (A. Terashima / Nuclear Physics B 584 (2000) 329–358 L(B + F ) = p p  B det g Tr √ + F gs 2 + O(α 0 ) det g Tr F2 + total derivative + O(α 0 ). √ ˆ F b ) = det G Tr(G−1b L( F ∗ G−1b F ) + O(α 0 ).1.5) ˆ i coincide at the lowest It is clear from these expressions that the normalized fields Ai and A 0 order in α : ˆ i = Ai + O(α 0 ). the F 4 terms can be determined in this case as well.6) Following the procedure presented in Section 3. The evaluation of the Lagrangian on the non-commutative side in the α 0 ˆ i by expansion is given in terms of the normalized field A √ det G Tr(G−1b F ∗ G−1b F) p p  ˆ ˆ i )(∂j A ˆ i − ∂i A ˆ j ) − 4(2πα 0 )2 Gs Bkl ∂k A ˆ i ∂l A ˆ j ∂j A ˆi = det g (∂i Aj − ∂j A ˆ k − ∂k A ˆ j )(∂k A ˆ i − ∂i A ˆ k) + 2(2πα 0 )2 (B 2 )ij (∂j A  1 ˆ j − ∂j A ˆ i )(∂j A ˆ i − ∂i A ˆ j ) + O(α 0 4 ) . √ 2 2   p  det g  B 2 2 Tr(B + F ) = det g gs Tr √ + F gs gs p   √ 2 2 = det g gs (Tr F ) + 4 gs Tr BF Tr F2 + O(B 2 ) . but it may not the .7) up to field redefinition since both produce the same S-matrix.354 Y. − (2πα 0 )2 Tr B 2 (∂i A 2 The relevant terms on the commutative side are √  4 p B det g 4 Tr(B + F ) = det g gs Tr √ + F gs gs p   √ 4 3 = det g gs Tr F + 4 gs Tr BF + O(B 2 ) .9) ˆ F b ) and L(B + F ) should produce the same The requirement that both Lagrangians L( 3 3 S-matrix of order O(B. A (A.1. k ) determines the form of the Lagrangian L(B + F ) in a completely parallel way to Section 3. Okawa. ζ .4) (A. (A.7) (A.

4) where the product between the field strengths on the right-hand sides of these expressions b4 terms in (3.10) expanded with respect to B is given by    4 2 2   p p B B 1 1 0 2 Tr √ + F − Tr √ + F det g (2πα ) Gs gs 2 gs 8 gs    p √ 1 1 Tr F4 − (Tr F2 )2 = det g (2πα 0 )2 t gs 2 8   √ 1 + t gs 2 Tr BF3 − Tr BF Tr F2 2    √ 1 (A. the ∗ b 4 terms of (3.7). (B.1). whereas the values of the Tr B 2 F 2 term evaluated at on-shell configurations satisfying (3.11) + t 2 Tr B 2 F2 − Tr B 2 Tr F2 + total derivative + const. (B.7) and (A. S.18) do not vanish so that if there is a difference in the term it cannot be redefined away. product between the field strengths in the F namely. 4 The O(B 2 ) part does not coincide with that of (A. (A. Terashima / Nuclear Physics B 584 (2000) 329–358 355 case for the O(B 2 ) part.1) and (B. The difference in the Tr B 2 Tr F 2 term can be absorbed by field redefinition as we explained in Section 3. In this appendix. The O(α 0 2 ) part of the Lagrangian (A. Determination of the O(α 0 2 ) part of Gs As we discussed in the last part of Section 3.3) 06 + O(α ). (3.29) does not affect O(α 0 4 ) terms. (B. The result is t = 1. Therefore the Tr B 2 F 2 terms in (A. b )4arbitrary = (2πα 0 )2 Tr(G−1 F bG−1 F bG−1 F bG−1 F b ) + O(α 0 6 ). we determine the value of c by the consideration at order α 0 4 . which determines the value of t. Now the F expansion using (3.2) Aˆ i = Ai + (2πα 0 )2 Bkl Ak (∂l Ai + Fli ) + (2πα 0 )2 Tr B 2 Ai + O(α 0 4 ).29) are evaluated in the α 0 is the ordinary one. not the ∗ product.Y. 2 8 where c is an undetermined constant. the consideration at order α0 2 is not sufficient to determine the O(α 0 2 ) part of Gs and that of the field redefinition which relates Aˆ i to Ai uniquely but allows the following ambiguity:   c−1 (2πα 0 )2 Tr B 2 + O(α 0 4 ) .12) Appendix B.1) Gs = gs 1 + 4 1 c (B. .3 so that it is irrelevant to the determination of t. (2πα 0 )2 Tr(G−1 F 0 2 (2πα )  b )2 2 Tr(G−1 F arbitrary 0 2  = (2πα ) Tr(G −1 b −1 b FG F) 2 (B.3).2) as follows: . Okawa.11) must coincide. We should first note that whatever ordering of the field strengths we choose.

5) are reproduced by the following terms in L(B + F ): √   3 1 1 (2πα 0 )4 det g 1 Tr(B + F )6 − Tr(B + F )2 Tr(B + F )4 + Tr(B + F )2 . However. (B.5) gives the complete form of them on the non-commutative side. although we believe that there is none. even if there exist such solutions.9).27) is possible in the F 6 terms.5) + total derivative + O(α 0 6 ) . 4 8 (B.8) By comparison.6) Tr(B + F )2 Tr(B + F )4 = Tr F 2 Tr F 4  Tr(B + F )  2 3 + 2 Tr BF Tr F 4 + 4 Tr F 2 Tr BF 3 + O(B 2 ). To show that this is the unique structure of the F 6 terms consistent with the assumptions (2. Okawa. there are three sources for O(BF 5 ) terms on the commutative side. we can see that the O(BF 5 ) terms in (B.27) under our normalization convention. = (Tr F 2 )3 + 6 Tr BF (Tr F 2 )2 + O(B 2 ).356 Y. we must verify that no solution to the on-shell initial term condition (4.9) These are precisely the terms needed to take the form of the DBI Lagrangian (3.15). On the other hand. (B. S. which are Tr(B + F )6 = Tr F 6 + 6 Tr BF 5 + O(B 2 ). Thus the argument which has been made so far is sufficient for the determination. Since no other terms can produce O(BF 5 ) terms. Now let us compare the O(B 2 ) part of (B. gs 3 8 96 (B.10) . the resulting ambiguity does not affect the determination of the O(α 0 2 ) part of Gs since solutions to the on-shell initial term condition by definition do not contribute to the B-dependent part of the S-matrix. (B.7) (B. Terashima / Nuclear Physics B 584 (2000) 329–358 √    1 det G 1 0 2 −1 b 4 0 2 −1 b 2 2 (2πα ) Tr(G F )arbitrary − (2πα ) Tr(G F ) arbitrary Gs 2 8 √  1 det g 1 (2πα 0 )2 Tr F 4 − (2πα 0 )2 (Tr F 2 )2 = gs 2 8   1 1 1 0 4 5 4 3 2 2 2 Tr BF (Tr F ) + (2πα ) 2 Tr BF − Tr BF Tr F − Tr BF Tr F + 4 2 16  c−1 Tr B 2 Tr F 4 + (2πα 0 )4 2 Tr B 2 F 4 + 8  1 1−c 2 2 2 2 2 2 Tr B (Tr F ) − Tr B F Tr F + 2 32  (B. √  1 1 1 (2πα 0 )4 det g Tr B 2 (Tr F 2 )2 2 Tr B 2 F 4 − Tr B 2 Tr F 4 − Tr F 2 Tr B 2 F 2 + gs 8 2 32 + 2 Tr BFBF 3 + Tr BF 2 BF 2 − Tr BF Tr BF 3  1 1 2 2 2 − Tr F Tr BFBF + Tr F (Tr BF ) .

P. Nucl. (B. Phys. Connes. Susskind. Nappi. 2 (1998) 008. B 485 (1997) 85. F.G. Phys. (B. Terashima / Nuclear Physics B 584 (2000) 329–358 357 with the corresponding one on the non-commutative side. Callan. Adv. M. Phys. Krogh. Phys. High Energy Phys. Phys. Witten. Okuyama. T. Fischler. High Energy Phys.R.17). D. Nucl. Seiberg. B 568 (2000) 447. Banks. Rev. Yost. Shenker. A. Phys. the value of the constant c is determined. (B. T. S. D 55 (1997) 5112.A. Phys. C. Maldacena.Y.M.R. 9 (1999) 032. Sheikh-Jabbari. High Energy Phys. The result is c = 0. Y. M. there is the following contribution from the F √ (2πα 0 )4 det g  Tr BF 2 BF 2 + 2Ak Bkl ∂l Fij (FBF )j i + Ak Bkl ∂l Fij An Bnm ∂m Fj i gs  + 2Bkl Bnm An (∂m Ai + Fmi ) ∂l Aj ∂k Fj i + O(B 3 ) + total derivative . Cheung. V. N.S. Phys. C. A. Theor. Douglas. . Phys. Hull. This is an interesting problem itself since it is related to the O(θ 2 ) part of (2. Pouliot. √  (2πα 0 )4 det g 2 Tr BFBF 3 − Tr BF Tr BF 3 gs 1 1 − Tr F 2 Tr BFBF + Tr F 2 (Tr BF )2 4 8 − 2Ak Bkl ∂l Fij (FBF )j i − Ak Bkl ∂l Fij An Bnm ∂m Fj i  − 2Bkl Bnm An (∂m Ai + Fmi ) ∂l Aj ∂k Fj i . Nucl.12) To complete the argument of the O(B 2 ) part at order α 0 4 . Witten. C. Lett. Lett. A. High Energy Phys.M. B 288 (1987) 525. 6 (1999) 030. Nucl. J. C.R. Schomerus. E. Ardalan. E. J. Tseytlin. J. B 433 (1998) 29. Schwarz. B 576 (2000) 578.H. Nucl. Fradkin. Shenker. 2 (1998) 231.H. J. Douglas. J. Phys. References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] E. B 163 (1985) 123. 2 (1998) 003. it is necessary to verify that the difference. W. P. High Energy Phys.E. Okawa. B 550 (1999) 151. Phys. Kabat. Douglas.13) does not contribute to the S-matrix and is absorbed by a field redefinition of Aˆ i at order α 0 4 .5) at order α 0 4 . S. Chu. However it is easily seen that it is irrelevant to the determination of the constant c at any rate because none of the terms in (B. C. Nucl. K. M. Lovelace. S. H. Ho. Kawano. 2 (1999) 016.13) have the structure of the contraction Tr B 2 = Bij Bj i which was relevant to the determination.11) By the comparison with respect to the terms Tr B 2 Tr F 4 and Tr B 2 (Tr F 2 )2 which obviously contribute to the S-matrix. Arfaei. M. L. In addition to the O(B 2 ) part b2 term: of (B. Math.R. S. B 528 (1998) 185.A. J. M. B 460 (1996) 335. Nucl.

A. Nucl. hep-th/9909081. Andreev. T. Phys. A. N. Callan. L. S. hep-th/9907211. Andreev. hep-th/9908105. J. O. 11 (1999) 024. B 501 (1997) 41. Okawa. J.D. I. Tseytlin. hep-th/9912293. Schiappa. Y. Okawa. B 566 (2000) 348. C. Tseytlin. Abouelsaood. L. B 481 (2000) 125. . Terashima / Nuclear Physics B 584 (2000) 329–358 A. 3 (2000) 016.A. 2 (2000) 029. A. A. Phys. Andreev. Cornalba.R. hep-th/9909176.D. Nappi. S.G. Cornalba. Cornalba. B 311 (1988) 205. O. L. Phys. Phys. Phys. Nucl. Asakawa. Phys. Nucl. Tseytlin. Kishimoto. Lett. C. A. S. Nucl. Tseytlin. J.358 [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] Y. K. High Energy Phys. High Energy Phys. Okuyama. R. Lett.A. Terashima.A. Mod. B 280 (1987) 599. Yost. O. High Energy Phys.A. A 3 (1988) 1349. Ishibashi.

Berkeley.50. Introduction There has been renewed interest in Kaluza–Klein theories over the past two years. E-mail: grojean@thsrv. The number of allowed brane tension values is related to the number of negative stationary points of the scalar bulk potential and. mainly due to the realization that localization of matter [1. CA 94720.Nuclear Physics B 584 (2000) 359–386 www. the brane tension for which gravity is localized without singularities is quantized. USA b Department of Physics.V.+e. USA c Theoretical Physics Group. These theories ∗ Corresponding author. CA 94720.Kl Keywords: Brane universe.d a Theory Division T-8.65. However. SA2 8PP. Robert Oppenheimer fellow. PACS: 04. Los Alamos National Laboratory.3 2 1 3 ( 0 0 ) 0 0 3 9 0 . and find that fine-tuning is generically reintroduced at the singularities in order to retain a static solution. Hollowood a.∗ . Christophe Grojean b. Berkeley. 0550-3213/00/$ – see front matter  2000 Elsevier Science B. PII: S 0 5 5 0 . We demonstrate that singularities are generic in the self-tuned solutions compatible with localized gravity on the brane: we show that localized gravity with an infinitely large extra dimension is only consistent with particular fine-tuned values of the brane tension. we speculate that the presence of additional fields may restore self-tuning. University of Wales. Los Alamos. Self-tuning.10.lbl. USA d Department of Physics.nl/locate/npe General properties of the self-tuning domain wall approach to the cosmological constant problem Csaba Csáki a. Localized gravity. 11.elsevier. Swansea.2] and localization of gravity [3] may drastically change the commonly assumed properties of such models. Joshua Erlich a . UK Received 1 May 2000. in the case of an oscillatory potential. NM 87545.V. Timothy J.+h.4 . 04. accepted 20 June 2000 Abstract We study the dynamics of brane worlds coupled to a scalar field and gravity. We also examine a resolution of the singularities. Cosmological constant 1.gov 1 J. University of California.c. and find that selftuning of the cosmological constant is generic in theories with at most two branes or a single brane with orbifold boundary conditions. Lawrence Berkeley National Laboratory. All rights reserved. All rights reserved.1 .  2000 Elsevier Science B.

by proposing a mechanism for the cancellation of all order Standard Model (SM) loop contributions and the leading (tree-level) gravity contribution to the effective 4D cosmological constant. the Planck scale in the bulk for instance. The necessity of singularities in the bulk is here only dictated by the phenomenological requirement of localized gravity. which is coupled to the brane tension Tbr (assumed to be the full loop-corrected SM vacuum energy density).6]) may be circumvented. The mechanism is based on a single 3-brane (to which the SM fields are localized) embedded into a 4+1 dimensional space–time. we naturally obtain 3 M5 ∼ 108 GeV and yc ∼ 1 mm (1. The authors of [7. . O(M 4 ) + O(Tbr ) + O(Tbr2 M −4 ) + · · · | {z } | {z } | {z } tree-level SM loops (1. is needed to cancel the quadratic divergences which should bring Tbr near M54 . which is still forty orders of magnitude too large. Csáki et al.360 C.2) which is phenomenologically safe. 3 The fact that the electroweak scale on the brane is five orders of magnitude below the fundamental scale. M . the next possible term is of the order of 10−84 M44 . M4 = κ4 . If the tension. such as low energy supersymmetry for instance. on the brane: in particular it is possible. to find solutions to the Einstein equations with a flat Minkowski metric on the brane in which case the brane cosmological constant is zero. A particularly simple scenario has been recently suggested in [7. while the SM contribution to the 4D cosmological constant would be of the order of 10−64 M44 . The essential new ingredient is a bulk scalar field φ. for some recent proposals for a 4D adjustment mechanism. the self-tuning mechanism cancels this term.1) higher order where M is the fundamental scale of gravity.8] then showed that one can find static solutions to the classical equations of motion for a vanishing bulk potential of the scalar field. but for arbitrary values of the brane tension Tbr . However. associated to the vacuum energy of the SM fields. since using extra dimensions the no-go theorem of Weinberg for adjustment mechanisms ([4]. 5 is the gauge hierarchy problem in this model and a mechanism. which has been interpreted as the boundary of the extra dimension (see also [10.8] to at least improve on the cosmological constant problem (see also [9] for an earlier mechanism involving extra dimensions). for a non-vanishing tension. all the solutions found in [7.11] for a discussion on singularity in a braneworld context). However. Since the bulk is effectively compactified.8] which localize gravity involve a naked space–time singularity. by Tbr = κ42 /κ54 . it complements the singularity theorem of 2 The cosmological constant problem concerns the brane cosmological constant that governs the expansion of our universe and it has to be distinguished to the vacuum energy density. is of the order of the electroweak scale. see also [5. which singularity is found to be yc ∼ κ5−2 Tbr−1 where M5 = κ5 −1 is related to the 4D Planck scale. / Nuclear Physics B 584 (2000) 359–386 clearly open new approaches to the cosmological constant problem 2 . or tension. we have to worry whether the size of the extra dimension is compatible with our phenomenological knowledge of gravity which has been tested up to millimeter distances. It is worth noticing that the self-tuning mechanism eliminates twenty orders of magnitude and thus should be considered on equal footing to the Randall–Sundrum solution to the hierarchy problem. Tbr ∼ TeV4 . However. The proper distance from the brane to the −2/3 is the 5D Planck scale. Finally.

We show that for the 5D scalar-gravity system the only possibility for such solutions is when space–time asymptotes to five-dimensional anti-de Sitter space (AdS5 ) far from the brane (therefore producing solutions of the sort considered in [18]). and present an example for both methods using the exactly solvable cases of Section 3. Csáki et al. and all the degrees of freedom are transparent without having to make a particular ansatz for the solution. . In Section 5 we present a no-go theorem for self-tuning branes that would localize gravity without singularities. the exponential bulk potential is in fact self-tuning. After a careful analysis we show that such solutions are always fine-tuned. Several works [13–15] have studied the self-tuning mechanism. by relaxing an ansatz for solutions made in [8]. In Section 3 we present a class of exactly solvable systems. which appears as a maximal index. Standard Model corrections are expected to occur at the weak scale. (ii) to overcome the singularity problem by considering a more general bulk geometry. whereas self-tuning works up to the larger 5D Planck scale. In Section 4 we compare a perturbative and a numerical method to solve the equations for the most general bulk potentials. we thus obtain a quantization of the brane tension as in a brane world construction from supergravity [19. they occur only for particular isolated values of the brane tension Tbr . We also find that. We show that the number of allowed brane tensions is related to the number of negative stationary points of the bulk potential.C. given by an exponential bulk potential. In the case of an oscillatory bulk potential. We conclude in Section 7. We explain that the self-tuning behavior is expected to occur for a generic bulk potential. that gravity is localized to the brane (so as to reproduce 4D gravity for the observer on the brane).21–28].20]. in the generic solutions (except for a vanishing bulk potential) selftuning does have the more restricted interpretation that there is at least a finite region for the brane tension for which static solutions can be found (but not necessarily for all values of Tbr ). In this paper. / Nuclear Physics B 584 (2000) 359–386 361 Hawking and Penrose under which generic initial conditions lead to singular solutions of Einstein’s equations [12]. we examine the general properties of the solutions to the coupled 5D scalar-gravity system in detail. The advantage of this method is that it results in first order ordinary differential equations. After presenting our general results and methods on the solutions in the presence of a generic bulk potential we ask the following question: are there bulk potentials such that the self-tuning solutions avoid naked space–time singularities for a range of values of the brane tension in such a way. In Section 6 we comment on the effect of resolution of the singularities on self-tuning. we investigate the solutions in the presence of a general bulk potential for the scalar field. which includes all the models presented in [7. that is. The paper is organized as follows: in Section 2 we present our method for searching for solutions of the coupled scalar-gravity system using the superpotential formalism of [18. However. More precisely. We analyze these models in detail and find that all of these solutions necessarily involve a naked singularity or lead to an infinite Planck scale on the brane.8]. The motivation to include a scalar bulk potential are twofold: (i) to give a mass to the scalar field and thus evade the cosmological problems associated to a massless scalar field coupled to gravity and/or those associated to an unstabilized extradimension. First.

2) The unconventional normalization of the action (2.1) and the warp factor in (2. Self-tuning and scalar fields In this section we demonstrate that the 4D cosmological constant of branes coupled to scalar fields is generically self-tuned to zero. / Nuclear Physics B 584 (2000) 359–386 2. A00 (y) = φ 0 (y)2 + f [φ(y)] T δ(y). a new ingredient here is the localization of SM fields) 4 .7) φ 0 (y) = ∂φ(y) (2.5) where the primes denote derivatives with respect to y. A nice feature of the case with vanishing bulk potential is that the flat solutions are the only 4D maximally symmetric solutions [7]. and allow it to be arbitrary.1) with the ansatz (2. If φ(y) is monotonic in the bulk then the bulk equations of motion can be written in a first order form [18. φ 00 (y) − A0 (y)φ 0 (y) = 2 ∂φ ∂φ 2 2 (2.1) − dD−1 x |g| 2 (D − 1)κD where GMN is the D-dimensional metric and gµν is the pullback of the metric onto the flat domain wall at x D−1 ≡ y = 0.   Z  1 p D−2 D M ∂M φ∂ φ + V [φ] S = d x 2 |G| R − D−1 2κD Z p D−2 f [φ] T . 1 ∂V [φ] ∂f [φ] + T δ(y). (2.4) (2. T is the non-canonically normalized brane tension and includes standard model vacuum contributions. For now we will not be concerned with the origin of the coupling f [φ] to the brane. We look for static solutions with the metric ansatz 5 .28] introducing an auxiliary field W [φ].2) will be convenient in what follows. Csáki et al. as in [7.8) A0 (y) = W [φ(y)]. Bulk indices will be denoted by capital Latin indices and brane indices by Greek indices. (2.28]. 4 Our conventions correspond to a mostly positive Lorentzian signature (− + · · · +) and the definition of the curvature in terms of the metric is such that a Euclidean sphere has positive curvature. 5 We will not look for non-flat solution on the brane such as de Sitter or anti-de Sitter 4D configurations. . 2 + dy 2. The equations of motion which follow from the action (2. ds 2 = e−2A(y)/(D−1)dxD−1 (2.362 C. 0 0 A (y) = φ (y) − V [φ(y)]. it is a dynamical question to know which solution is preferred by stability.6) V [φ] = ∂φ ∂W [φ(y)] .2) are [18.8]. (2. We begin with the action for a brane coupled to gravity and a real scalar field (such dilatonic domain walls have been extensively studied from a supersymmetric point of view [29–31].   ∂W [φ] 2 − W [φ]2 .3) (2. When they exist simultaneously with flat Minkowskian solutions.21–26.

8) must then be supplemented by boundary conditions at the brane.C. . Csáki et al. / Nuclear Physics B 584 (2000) 359–386 363 Because of the relation (2.5) can be written. But the equations of motion (2. W [φ] will be called superpotential even though no supersymmetry is involved. The boundary conditions due to the delta-function terms in (2. φ± and A± on the right (+) and left (−) hand side.6)–(2.3) and (2. We will use the first order formalism in order to construct solutions in the bulk on both sides of the brane: W± .6).

∂f [φ(y)] .

.

1φ = 1 ∂φ(y) ∂φ(y) . ∂W [φ(y)] 0 =T .

y=0 .

9) 1A0 = 1W [φ(y)] = T f [φ(y)]. (2.

if there are N branes then there are 3N + 3 − 1 = 3N + 2 free parameters and 4N constraints. and T can be chosen arbitrarily large such that (f T . Hence. there can be up to two branes without fine-tuning. Hence. Thus. Hence. if f [φ] = f 0 [φ] then the range of T often extends to infinity. The additional constraints from the orbifold condition are A(y) = A(−y) and φ(y) = φ(−y). If the space is compactified .6)– (2. as argued above. In addition. where 1F indicates the jump of a discontinuous function F at y = 0. There is generically a (finite) line of solutions for a region of scalar-brane couplings f [φ] T . one overall constant shift in A(y) is not relevant because A(y) enters into the equations of motion only through its derivatives. In the generic case. there is in fact a continuous set of static solutions for a given boundary condition specified by f [φ] and T . leaving (1 − N)/2 free parameters in the solution. Once again. Hence. there is a range of “brane tensions” T such that static solutions exist. The reason is that if W [φ]  V [φ] asymptotically when W is large. A count of free parameters in the solutions to the equations of motion immediately demonstrates that given f [φ(y)]. which implies that W+ [φ] = −W− [φ]. In the absence of orbifold type boundary conditions. However. φ(y) and A(y) must be continuous across the boundary. With orbifold boundary conditions. where W± [φ] are the solutions for W on the two sides of the brane. then there are naively six free parameters: one from the solution on each side of the brane of each of equations (2. F (0+ ) − F (0− ). 2W 0 ). there is generically self-tuning in this case. However. as well. Furthermore. there are four boundary conditions at the brane. continuity of A(y) and φ(y) is then guaranteed. there are 3(N − 1)/2 + 3 − 1 = (3N + 1)/2 free parameters and 4(N − 1)/2 + 2 = 2N constraints. in this case there are two free parameters (there would be three but a constant shift in A is not relevant) and two boundary conditions. That leaves five free parameters and four boundary conditions. this is what we mean by self-tuning: Given an arbitrary scalar-brane coupling f [φ] (possibly satisfying some constraints).y=0 .8). only if there is a single brane at the orbifold fixed point will self-tuning occur. If we do not impose orbifold boundary conditions in addition to those above. and there is generically a solution for a region of couplings f [φ] T . where image branes are included in N and there is assumed to be a brane at the orbifold fixed point (which contributes 1 to N ). then W 0 ' W there. the tension T is generically not fine tuned. Orbifold boundary conditions are more constraining. f 0 T ) ∼ (2W. leaving 2 − N free parameters in the solution for a given set of boundary conditions.

The equation for the superpotential can be solved without introducing a prepotential. The challenge is finding a class of solutions to the nonlinear equation (2.2) then translates into a differential equation for w[φ]: V 0 ω2 + 1 ω . In what follows we will concentrate on the case of a single brane coupled to a scalar field. (2. we will denote by a prime a derivative of V . • an exponential bulk potential: V 0 [φ]/V [φ] = const 6= 0. there are two free parameters in the solution but four boundary conditions. 8].3) 2V ω2 − 1 Exact solutions for w[φ] can be found when (3. [7. 3. Let us first find the exactly solvable models. / Nuclear Physics B 584 (2000) 359–386 on a circle with orbifold boundary conditions. but we will not restrict ourselves to the ansatz A0 [φ] ∝ φ 0 made in [8]. . 7 From now. when V 0 [φ]/V [φ] is a constant. Our discussion of the exact solutions with a vanishing bulk potential is similar. W or ω with respect to φ.6) is immediately satisfied for any prepotential w[φ].6) becomes simply W 0 [φ]2 = W [φ]2 . (3.6) for a given V [φ]. although some of the exact solutions exhibit non-generic behavior. (3. (3. If (in a region where 6 V < 0) we write W [φ] and W 0 [φ] as 7   1p 1 .4) 6 The case of a positive bulk potential can be studied in a very similar way up to some changes of sign in Eqs.e.1) W= −V [φ] w[φ] + 2 w[φ]   1p 1 −V [φ] w[φ] − . ω0 = ω − 3. Csáki et al.1. i. (3.364 C. or a derivative of φ or A with respect to y. Integrable bulk potentials In this section we study some special cases which were also partially discussed in [7.. • a negative bulk cosmological constant: V [φ] = Λ < 0.1) and (3. Namely.8] but it is a worthwhile exercise to repeat the discussion in terms of a superpotential. The consistency of (3. In agreement with the parameter count in the previous section we will find that there is no fine-tuning in these theories. We distinguish three cases: • a vanishing bulk potential: V [φ] = 0. (3.3) is separable. as well [18]. f . Vanishing bulk potential This case has been extensively studied by Refs. then a parameter count for the case of branes at the two orbifold fixed points demonstrates that a fine-tuning is necessary in this case.1). Eq. Indeed.2) W0 = 2 w[φ] then (2.2). (3.

In addition.5) where c is a constant of integration and  is a sign. f 0 ∼ O(M5 ). namely φ0 . • If + − = −1: the solutions with singularities exist only if f [φ0 ] T > 0 and −1 < f 0 [φ0 ]/f [φ0 ] < 1 but do not require any fine-tuning. As a result of this symmetry. only f 0 [φ]/f [φ] is relevant.6) need: d+ > 0 and d− > 0. d+ d− (3. the fact that a specific form had to be chosen for f [φ] is a result of two non-generic features of the model: First. (3. f 0 [φ]T ).9) while the jump equations are + c+ − c− − = f 0 [φ0 ] T . Thus the continuity requires φ(0) ≡ φ0 = −+ ln d+ = −− ln d− . As a result. to make sense Eq. it turned out that the left hand sides of the two boundary conditions had the same form up to a sign. Eqs. self-tuning occurs for T to within O(M5 )  O(MEW ) (where O(MEW ) is the expected Standard Model contribution to the tension) is generic. W [φ] = ceφ .8) A(0) ≡ A0 = e+ − ln d+ = e− − ln d− . one of the free parameters.. Csáki et al.6) A(y) = − ln(d − cy) + e. (3. we conclude that the Planck scale on the brane. and is what we mean by self-tuning.11) From the expression of the warp factor. (2. is finite iff singularities are encountered on both sides κD−1 of the brane. both of them can take different values on the two sides of the brane (the constants of integration relative to the right (left) hand side of the brane will be denoted with a + (−) subscript). which is quite non-generic. does not appear on the left hand side of the boundary conditions for the derivatives φ 0 and A0 .10) (3.7). Moreover. / Nuclear Physics B 584 (2000) 359–386 365 with two branches of solutions. and given one solution (f [φ]T .C. But the fact that given f. (2. (3. (3. . With this form of the superpotential. If we choose f [φ] = Ceφ for the case  = + = − then it is clear that the boundary conditions can be satisfied for any T . The fact that T is completely arbitrary is most likely unique to the case of vanishing bulk potential.7) where d and e are some constants of integration that can also differ on the sides of the branes. d+ d− c− c+ − = f [φ0 ] T . There are several important comments to make about this case.e. −2 −2 R = κD dy e−(D−3)A/(D−1). (3. there is an infinite set of solutions with the same f [φ] and arbitrary T . First of all. there is a symmetry φ → φ + const.8) can be easily solved as φ(y) = − ln(d − cy). (3. i. c+ > 0 and c− < 0: • If + − = 1: the consistency of the jump equations requires that f 0 [φ0 ] = f [φ0 ] and then it is possible to find solutions with or without singularities for any value of the brane tension but the singular solutions correspond to f [φ0 ] T > 0.

.15) where c is a constant of integration.8] the required exponential form of f [φ] might be natural from a stringy perspective where φ is interpreted as the dilaton.e. dy 2 (3. Plugging this expression for the prepotential into the remaining equations of motion.16) .  φ(y) = − ln c ϑ(y) . we will present our method on this rather simple example before proceeding to the somewhat more complicated case of an exponential bulk potential in the next subsection. ω = c eφ .366 C. dφ dφ 1 √ = −Λ ω(1 − ω−2 ) . Csáki et al. We will denote by V [φ] = Λ < 0 the cosmological constant. Furthermore.13) (3. The equations of motion are: dω =ω . / Nuclear Physics B 584 (2000) 359–386 Even given the constraint on f [φ] from the shift symmetry in φ in this case. as pointed out in [8]. for special discrete values of the brane tension. (3. Bulk cosmological constant Even though the case of a (negative) bulk cosmological constant can be studied without introducing a prepotential. This feature will turn out to be generic except in theories which admit solitonic solutions for fine-tuned boundary conditions. the solutions with singularities on either side of the brane must be chosen in order to have a finite gravitational coupling (assuming that the space–time can be cutoff at the singularity in a consistent way). i.12) (3.14) The differential equation for the prepotential can be easily integrated. as explained in [7. dy 2 dA 1 √ = −Λ ω(1 + ω−2 ) .2. (3. 3. we obtain.

.

.

.

2 .

.

.

.

yc− > 0). Conversely. also appears in the expression of the function ϑ(y) defined. (3. On the right (left) hand side. yc . Near this singularity. near infinity. y = yc . (3. ruining any phenomenological relevance. the warp factor blows up and the Planck scale on the brane diverges. if yc+ . on the two different branches of solutions. by √ √ −Λ −Λ (y − yc ) or ϑ(y) = − coth (y − yc ) . exp(−2A/(D − 1)). then the transverse dimension will be infinitely large. where a is a constant of integration and another constant of integration.17) A(y) = − ln ϑ(y) + ln 1 − ϑ (y) + a. a positive (negative) value of yc will correspond to a solution involving a singularity at a finite proper distance from the brane. however.18) ϑ(y) = − tanh 2 2 The nature of the solution depends on the sign of yc on each side of the brane. goes to zero so the singularity appears as an horizon in the bulk. is negative (respectively. the value of yc on the right hand side of the brane. . the warp factor.

/ Nuclear Physics B 584 (2000) 359–386 367 The values of the constants of integration are constrained by the continuity and jump √ equations (with ϑ± = tanh −Λyc± /2).26) is precisely the one appearing in the Randall–Sundrum model when the scalar coupling to the brane is a constant: f [φ0 ]2 − f 0 [φ0 ]2 = − Λbk = − D−1 2 2 κ T .25) The continuity conditions are automatically satisfied but the consistency of the jump equations requires a fine-tuning. Λbk = D−2 D−2 Λ and Tbr = f [φ0 ]T . i. f [φ0 ] (3. because of this shift symmetry we lose the appearance of a free parameter to adjust in the jump equations. which leads to a more restricted set of boundary conditions than for a generic bulk potential V [φ]. in the theory.29) f 0 [φ0 ] = 0 and f [φ0 ] T = 2 −Λ. .26) T2 As already discussed in the case with a vanishing bulk potential. 2 2 2(D − 1)κD (D − 1)κD (3. provided that f 0 [φ0 ] 6= ±f [φ0 ]. φ0 ≡ − ln(c+ ϑ+ ) = − ln(c− ϑ− ).23) T2 > f [φ0 ]2 − f 0 [φ0 ]2 Moreover.C. this fine-tuning is a consequence of the translational symmetry. (3. ϑ± = ±1.24) just as in the case of a vanishing bulk potential. The fine-tuning (3. c+ = −c− and a+ = a− . singularities will exist on both sides of the brane iff f [φ0 ] T > 0 and −1 < f 0 [φ0 ] < 1. (3.22) ϑ+ − (3.20].27) where Λbk and Tbr are the canonically normalized quantities [19.28) The solution constructed by Randall and Sundrum that localizes gravity with an infinitely large extra dimension corresponds to a limit of the singular solution where the singularities are pushed to infinity. 2 | + a+ A0 ≡ − ln |ϑ+ | + ln |1 − ϑ+   2 2 1 + ϑ+ 1 + ϑ− 1√ (3. 2 −Λ (3.. It is interesting to look at the Z2 symmetric solution for which there are the additional constraints: yc+ = −yc− .20) = f [φ0 ] T . 2 ϑ+ ϑ− (3.21)  2 2 1 − ϑ+ 1 − ϑ− 1√ −Λ − = f 0 [φ0 ] T . The jump equations then require √ (3.19) 2 = − ln |ϑ− | + ln |1 − ϑ− | + a− . φ → φ + const. ϑ− A solution to these equations can be found. 8(D − 2) D br (3. Csáki et al.e. (3. for any value of the brane tension such that −4Λ . 4Λ . As before.

the expressions for the scalar field and the warp factor simply become √ (3.27).34) The sign of a is not fixed and can be chosen independently on the two sides of the brane. The first differential equation can be easily solved to express φ in terms of ω: . We will concentrate on negative potential while the case of positive potential requires some minimal changes. is fine-tuned to (3. Tbr .33) (3. In this limit. as we will discuss.368 C.e. dy 2 dA 1 bφ = ae ω(1 + ω−2 ). (3. 3. the coupling between the brane and the scalar field vanishes and the canonically normalized brane tension. the equations of motion are simply: ω2 + 1 dω =ω −bω 2 . Exponential bulk potential The last case that can be solved analytically with our method involves an exponential potential for the scalar field in the bulk. / Nuclear Physics B 584 (2000) 359–386 i.3. dφ ω −1 dφ 1 bφ = ae ω(1 − ω−2 ). So the bulk potential will be parametrized by two real numbers a and b: V [φ] = −a 2e2bφ . Csáki et al.32) (3.30) φ = φ0 and A = −Λ |y|.. dy 2 (3.31) On each side of the brane.

.

35) e−bφ = e−bc |ω|−b/(1+b) . (3.b2 /(b2 −1) .

1 + b − (1 − b)ω2 .

c.36) (3. ω˜ (1 + b − (1 − b)ω˜ 2 ) b2 /(b2 −1) . and the constant of integration. Different kinds of solutions can be obtained depending on the value of b and of the range of integration for the variable ω: ˜ . can take any values compatible with the continuity conditions. the parameter a can differ by a sign while the initial bound of integration. ω0 .37) On the two sides of the brane. (3. where c is a constant of integration. A Z2 symmetric solution will correspond to two different choices of sign for a but the same values for ω0 and c. This last result can be used to obtain a parametric representation of A and y as functions of ω: 2 y(ω) = − e−bc a Zω ω0 Zω A(ω) = A0 − dω˜ ω0 |ω| ˜ −b/(1+b) |1 + b − (1 − b)ω˜ 2| dω˜ (1 + b − (1 − b)ω˜ 2 ) (1 + ω˜ 2 ) .

Fig.37) where on the right (left) hand side of the brane. ω0 . / Nuclear Physics B 584 (2000) 359–386 369 • b < −1: in that case. 1. y reaches a finite value yc while the warp factor goes like: A ∼ − ln |y − yc |. We have drawn Z2 symmetric solutions for different values of the tension on the brane: when the tension increases.38) from where it becomes evident that the singularity free solutions do not localize gravity because the Planck scale on the brane diverges. i. ω = √ ± (1 + b)/(1 − b) and ω = 0 are integrable while those appearing in dA/dω are non-integrable. The parameter ω will range from ω0 to 0− . to +∞. And the range of integration goes from a positive initial value.C. In both cases. (3.36). The behavior of the warp factor near the singularity insures that the −2 −2 R = κD dy e−(D−3)A/(D−1). the jump conditions require that the scalar coupling to the brane satisfies: |(df/dφ)/f | < 1 on the brane. (3. In that case. e−2A/(D−1) . blows up faster and faster for the the singularity-free solutions.36). All the solution are singular with a horizon or a curvature singularity. dy/dω has an integrable singularity at ω = +∞ but a nonintegrable singularity at ω = 0 while the singularities of dA/dω are both nonintegrable. . is finite. Planck scale on the brane. ω near 0− : A ∼ − ln |y|. a has to be chosen positive (negative) and ω0 is a negative initial bound of integration. y∼yc (3.. It is easy to find the asymptotic behavior of this solution for large |y|. |y|∼∞ (3.39) which indicates that the singularity is a horizon where the metric on the brane vanishes.37) with a positive (negative) parameter a on the right (left) hand side of the brane. (3. Shapes of the warp factor for singular and non-singular solutions to the equations of motion with an exponential bulk potential. − Solutions with singularity will still be given by Eqs.e. So we can find solution with or without bulk singularity at finite proper distance. − Solutions without singularity will be given by (3. the horizon becomes closer and closer to the brane for the singular solutions while the warp factor. Csáki et al. κD−1 • −1 < b < 1: in that case the singularities of dy/dω at ω = ±∞.

unless the brane tension is fine-tuned as in the Randall– Sundrum model. / Nuclear Physics B 584 (2000) 359–386 • 1 < b: this case is quite similar to the first case because the singularity of dy/dω at ω = 0 is integrable but the singularity at ω = +∞ is non-integrable while the two singularities of dA/dω are both non-integrable. which we then show to break down for the interesting case of perturbing around a solution with localized gravity.37) where on the right (left) hand side of the brane. This point surely deserves further scrutiny which the next sections will be devoted to. i.36). It is easy to find the asymptotic behavior of this solution for large |y|. . y reaches a finite value yc while the warp factor goes like: A ∼ − ln |y − yc |.. First. (3. Csáki et al.36). y∼yc (3. In that case. (ii) solutions without singularity at a finite proper distance but associated to a bulk geometry that decouples gravity on the brane. So we can construct solutions with or without singularity. The parameter ω will range from ω0 to +∞.40) from where.37) with a negative (positive) parameter a on the right (left) hand side of the brane. |y|∼∞ (3. 4. ω near +∞: A ∼ − ln |y|. • b = ±1: in these two cases.e. Perturbative and numerical methods In this section we give two methods for finding approximate solutions to the coupled equations of motion in the bulk that satisfy the boundary conditions at the brane. it becomes evident that the singularity free solution does not localize gravity. a has to be chosen negative (positive) and ω0 is a positive initial bound of integration. 1 illustrates the different types of solutions. And the range of integration goes from a negative initial value. However it seems impossible to find a singularity free solution that localizes gravity on the brane.41) which indicates that the singularity is a horizon where the metric on the brane vanishes. − Solutions with singularity will still be given by Eqs. (3. The main result of the analysis of these integrable bulk potentials is that we find two kinds of solutions: (i) solutions with a horizon in the bulk at a finite proper distance from the brane and with a finite lower dimensional Planck scale.370 C. T . ω0 . we present a perturbative method. we can construct singularity free solutions with a blowing up warp factor as well as singular solutions with a horizon at a finite proper distance. − Solutions without singularity will be given by (3. Fig. to 0− . Both kinds of solutions do not require any fine-tuning and can be constructed for a range of brane tension. (3. Then we give a systematic numerical method for solving the equations. once again.

1) around W0 we obtain a differential equation for δW of the form δW 0 W0 = 0. (4.9) for the perturbed tension T + δT . However. we obtain Z dA0 = CeA0 (y).6) where 1 (A00 /φ00 ) denotes the jump in A00 /φ00 at y = 0. Once C± is determined from (4. W00 [φ] (4. (4. we can simply integrate the equations δφ 0 (y) = δW 0 [φ0 (y)].1. The values of C± are then given by C± = 0 (0) f 0 [φ0 ] − f [φ0 ]A00∓ (0)/φ0∓ 1(A00 /φ00 ) δT .6).5) In order to satisfy the jump equations (2. Let us now assume that we have found a static solution to the equations of motion φ0 (y) and A0 (y). when the unperturbed solution asymptotes to AdS space thereby localizing gravity to the brane (that is for A(y) ∼ . we first need to find how one can change the superpotential W around W0 [φ] so as to leave the bulk potential unchanged: V [φ] = W00 [φ]2 − W0 [φ]2 . δA0 (y) = δW [φ0 (y)] (4. A perturbative method As emphasized in the previous section. Csáki et al. and assume that we have chosen one of the integration constants such that A0 (0) = 0.1) which should be invariant under W → W + δW . Thus.3) where the arbitrary constant C yields the extra degree of freedom needed to find a solution for any value of T . This is a solution for a particular value T of the brane tension. φ00 (y) (4.7) to obtain the perturbed solutions φ0 (y) + δφ(y) and A0 (y) + δA(y).4) δW [φ(y)] = C exp dy dy From (4. we need to choose C differently on the two sides of the brane. The superpotential variation can be expressed in terms of the unperturbed background solution using the equations of motion dφ0 /W00 [φ0 ] = dy and W0 [φ0 ] = dA0 /dy. δW W0 which is solved by δW [φ] = C exp (4. Linearizing (4. one has an additional degree of freedom if one picks a different superpotential function on the two sides of the brane. / Nuclear Physics B 584 (2000) 359–386 371 4.2) Z W0 [φ] dφ. in the most interesting case.C. This method always results in a perturbed solution.3) we can see that the change in the derivative of the superpotential is given by δW 0 [φ(y)] = C A00 (y) A0 (y) e . the key to finding the self-tuned solutions is to utilize the fact that one can transform the superpotential without changing the bulk potential for the scalar field. In order to find the solution obtained by perturbing the brane tension as T → T + δT . W+ [φ] and W− [φ]. (4.

11) 8 We will see in Section 5 that. which shows that in this case δW [φ(y)] ∝ e(D−1)|y|/RAdS . the superpotential.9) there is a second solution. This can be done by noting that once φ0 is fixed.2. Csáki et al. the superpotential functions must be such that they reproduce the correct value of the bulk potential at the brane: W+0 − W+2 = V0 . and in addition we can pick the value of the scalar field at the brane φ(0) ≡ φ0 arbitrarily. 2 W−0 − W−2 = V0 . W± = ± f0 T + f0 T 2 + 2 0 2 2 f0 − f02 s 1 0 1 4V0 0 . But first we give a numerical method that can be used for any choice of the bulk potential.8) where W± refers to the values of the superpotential functions to the right and left of the brane at φ0 .10) are both simultaneously flipped. one can easily see that the perturbative method presented here always breaks down.8) numerically for any brane tension and potential in the bulk. and thus the linearized approximation breaks down. W [φ]. (4.8) and (4.9). Our strategy is the following: we first determine the superpotential functions W+ [φ] and W− [φ] to the left and the right of the brane numerically such that the boundary conditions arising from the coupling to the brane are satisfied.6)–(2. 4. W+0 − W−0 = f00 T .372 C. one has to first make sure that the boundary conditions that one imposes do satisfy the jump equations (2. In addition. will be a 0 [φ] will keep the sign of W 0 . Once the value of W± and W±0 are fixed. unless the brane tension is fine-tuned. monotonic function of φ and thus W± ± . (4.9) where V0 = V [φ0 ]. Therefore the perturbed values of δφ and δA grow exponentially.9) together are enough to determine the values of both W± and W±0 at the branes. Eqs. etc. This can be seen by inspecting (4. / Nuclear Physics B 584 (2000) 359–386 (D − 1)|y|/RAdS for large values of y). (4. the jump equations are just given by W+ − W− = f0 T .4). the brane tension T . the coupling of the scalar to the brane determined by the function f [φ]. where the signs in front of the square roots in (4. We will examine the case of localized gravity in detail in Section 5. They are given by the expressions: s 1 1 0 4V0 .8) and (4. In order to find a numerical solution to these equations. (4. f0 = f [φ0 ]. Thus the input functions are the bulk potential V [φ].10) W± = ± f0 T + f0 T 2 + 2 0 2 2 f −f2 0 0 Due to the quadratic nature of Eqs. 2 (4. one can numerically integrate the equation 8 q W±0 [φ] = sgn(W±0 ) V [φ] + W± [φ]2 (4. A numerical method Next we present a method for solving the system (2.

(3. for a fixed value of the scalar field on the brane. The main conclusions are as expected: the perturbative method works well far from the singularities.7). 2. here the singularity of the perturbed solution remains at the same place where the singularity of the unperturbed solution was. and does not capture the essential feature of the self-tuning solutions: whereas in the self-tuning mechanism.3.13) Fig.1 works. we conclude that this method is not very efficient in capturing the basic properties of the self-tuning solutions. Therefore. An example for the perturbative method In this section we test how well the perturbative method described in Section 4. with the choice + = 1. with d± = e∓φ0 . However it gets worse as we approach the singularity itself. The initial unperturbed solution is given by the dashed curve. and compare the two results. The curve that blows up at smaller values of y corresponds to the exact solution. / Nuclear Physics B 584 (2000) 359–386 373 to obtain the superpotential functions to the left and the right of the brane that satisfy all boundary conditions. while we pick δT = 0. − = −1. and we choose f [φ] = eφ/2 . We will compare the analytic solution of the model with a vanishing bulk potential to the perturbed solution around a different analytic solution.12) to the left and the right of the brane to obtain the numerical solutions for φ(y) and A(y). 4 1 c− = − T e3φ0 /2 . the place of the singularity adjusts itself with respect to the value of the brane tension. 3 c+ = T e−φ0 /2 . the shift of the singularity with a variation of the tension is missing. and for the unperturbed solution we choose T = 1.1. The analytic solution is given by (3. Once W± [φ] are numerically known. 4. (4. Csáki et al. The example we consider is the vanishing bulk potential discussed in Section 3. A0 (y) = W [φ(y)] (4.1 for the perturbed solution. The singularity of the perturbed solution appears at the same distance from the brane as the singularity of the initial solution. 4 e± = ∓φ0 . We will show an example for this below for the case when the exact solution is known. .C. we can simply integrate the equations φ 0 (y) = W 0 [φ(y)]. φ0 = 1. The exact solution versus the perturbed solution for the case of a vanishing bulk potential.6).

However. 4 1 W− = − eφ0 /2 T . (4.17) φ0 to the left and right of the brane. one should use the numerical method rather than the method based on perturbations. δT = 0. T = 1 and + = 1. (3. Finally. since it becomes unreliable close to the singularities. The perturbed solution from (4. φ0 = 1. The perturbed solution obtained for T = 1. suggesting that the numerical method works very well around the singularities. we analyze the same example as above (the case with vanishing bulk potential) using the numerical method. and should be the preferred method of looking for solution in the absence of exact solutions. We are looking for a numerical solution to the case analyzed perturbatively above. δφ− (y) = . that is vanishing bulk potential.1 compared to the exact solution for T = 1. that the perturbative method in general does not do a good job in finding the solutions. the perturbative solution nicely follows the exact solution away from the singularity.374 C. f [φ] = eφ/2 .4) and (3.4.1 can be seen in Fig.14) δφ+ (y) = − T d+ T d− where δφ± has been normalized to zero on the brane. Indeed.8) we find the starting values of the superpotential to the left and the right of the brane: 3 W+ = eφ0 /2 T . the numerical solution should not have these problems. 2. the values for φ(y) can be obtained by numerically inverting the integral Zφ y= dφ W 0 (φ) (4. The numerical solution obtained this way overlayed on the exact solution of Section 2 can be seen in Fig. in particular the place of the singularity is incorrectly predicted to coincide with the singularity of the unperturbed solution.7) is given by     d + − c+ y d − − c− y δT δT ln ln . From Eqs. and find that the exact and numerical curves are virtually indistinguishable. As mentioned above. Csáki et al. but deviates from it close to the singularity. 4. / Nuclear Physics B 584 (2000) 359–386 where A has been normalized to zero on the brane.15) Numerically integrating the equation W±0 [φ] = ±W [φ] (4. An example for the numerical method We have shown above.16) with the boundary conditions W± [φ0 ] = W± one obtains the numerical values for W± (φ). Therefore. . One can see that the two curves are virtually indistinguishable. 3. we suggest that in order to analyze potentials for which no exact solutions can be found. 4 (4. − = −1.

5.1. A no-go theorem We would like to reexamine in this section the count of free parameters versus finetuning parameters needed to preserve a static Poincaré invariance on the brane.2) is equivalent to having a massless normalizable bound state. since in those models the A00 > 0 condition is not satisfied. this condition is: Z 1 1 = (5.2) dy e−(D−3)A(y)/(D−1) < ∞. (5.C. Csáki et al. is related to the warp factor. The fact that the two curves are indistinguishable shows that the numerical method works extremely well even close to the singularities.5) ψ0 ∝ Ω (D−2)/2 with 9 We do not consider the recently proposed possibility that gravity might be quasi-localized to the brane [33– 36]. Ω. In terms of the warp factor.4) Then the condition (5. . 2 2 κD−1 κD It is convenient to introduce a transverse coordinate z for which the bulk metric is conformally flat:  2 + dz2 . by the two identities: Ω(z) = e−A(y)/(D−1) and Ω 2 (z) dz2 = dy 2. (5. The numerical solution overlayed on the exact solution for the case of a vanishing bulk potential. (5. / Nuclear Physics B 584 (2000) 359–386 375 Fig. which is interpreted as the graviton on the brane: Z dz |ψ0 |2 < ∞. therefore it is not possible to generate those backgrounds from a single scalar field.1) The graviton zero-mode is localized on the brane 9 precisely when the effective Planck scale is finite on the brane [28]. Localized gravity without singularities 5.3) ds 2 = Ω 2 (z) dxD−1 where the conformal factor. (5. Consider the general D-dimensional background preserving PoincaréD−1 2 ds 2 = e−2A(y)/(D−1)dxD−1 + dy 2. with the restriction that the solution corresponds to an infinitely large extra-dimension (without singularities) in the bulk and localizes gravity on the brane. 3. e−2A(y(z))/(D−1).

According to the equations of motion. at infinity. The equation ∂φ dW = . In our study. In that case. (5. |y|∼∞ (5. Csáki et al. we easily deduce that A has to satisfy: d2 A/dy 2 > 0. W [φ]. the value of the parameter α is constrained by the fact that the background is created by a scalar field coupled to gravity. and corresponds to α = (D −2)/2. in order to extend the range of the transverse coordinate from y = −∞ to y = +∞.8) 2 So the only background for the scalar field coupled to gravity that localizes gravity without singularity (with an infinitely large extra dimension) is asymptotic. to the horizon of an anti-de Sitter space. (5.376 C. ∂y dφ (5. in the z coordinate.7) α> 2 Furthermore. In other words. This is not to say that there is no self-tuning in these models.5) then requires: α > 1/2. which translates. only that the nonsingular solutions require fine-tuning.10) is similar to an RGE with W 0 playing the role of the β-function.6) The localization of gravity (5. it is worth noticing that an upper bound on α comes by the requirement 10 of a geometry without singularity at a finite R proper distance. the β-function dW/dφ must have zeroes. From the equations of motion. the fact that A is asymptotically linear means that φ becomes constant and we will denote by φc− and φc+ the asymptotic values of φ at y = −∞ and y = +∞. / Nuclear Physics B 584 (2000) 359–386 Let us assume that the behavior of ψ0 at infinity is a power law: ψ0 ∝ z−α . the values φc− and φc+ at infinity must be some roots of dW/dφ: . the warp factor is exponentially decreasing with the proper distance to the brane: α6 A ∼ (D − 1)|y|/RAdS. respectively. Indeed the proper distance from the brane to infinity is given: l∞ = dz Ω that diverges iff: D−2 . In order for φ to approach a constant (fixed point) at infinity. z∼∞ (5. as in the RS model.9) The aim of this section is to show that such a background necessarily requires a finetuning between the brane and the bulk. The previous asymptotic behavior has a nice interpretation in terms of the superpotential. in a lower bound on the value of α: D−2 .

.

dW .

.

dW .

.

(5.11) dφ . = 0 and =0.

φc− dφ .

. the curvature always vanishes at infinity. However quadratic invariants such as RM1 M2 M3 M4 R M1 M2 M3 M4 will be singular at infinity as soon as α > (D − 2)/2.φc+ 10 For a power law conformal factor.

∂y (5. W [φc− ] < 0 and W [φc+ ] > 0. otherwise the conformal infinity of AdS would be reached without localized gravity [19. A has to be linearly increasing in |y|. Given that ∂A = W [φ]. / Nuclear Physics B 584 (2000) 359–386 377 Furthermore. which according to (5. φc− and φc+ must be dynamically reached at y = −∞ and y = +∞. Csáki et al.10) is possible iff: .C.12) we conclude that φc− and φc+ must satisfy. at infinity.20].13) Finally. (5.

.

d2 W .

.

d2 W .

.

14) dφ 2 . (5. > 0 and < 0.

φc− dφ 2 .

Finally. this differential equation evaluated at infinity gives: V [φc− ] = −W [φc− ]2 < 0 and V [φc+ ] = −W [φc+ ]2 < 0.φc+ or at least a similar condition for the first non-vanishing higher order derivatives at φc− and φc+ . the last equation of motion that relates the superpotential. 4. W . to the scalar potential in the bulk. partially fixes the possible values of φc− and φc+ . (5. Pictorially. V . the previous conditions are summarized in Fig. Indeed.15) while a differentiation with respect to φ gives: .

.

  dW d2 W dV .

.

dV .

.

dφ dφ dφ 2 dφ . dV =2 − W . thus = 0 and = 0.

φc− dφ .

4. Asymptotic behaviors of W [φ] leading to a singularity free bulk geometry localizing gravity 0 − 0 + on the brane. . while the localization of − + gravity requires Wc < 0 and Wc > 0.16) More information on W can be obtained by considering higher order derivatives of the differential equation between W and V . the 00 − 00 + dynamics of the equations of motion require W c > 0 and W c < 0. Indeed it is easy to prove by induction the following relation: Fig. The absence of singularities is equivalent to the conditions: W c = W c = 0.φc+ (5.

25) W c = (2) ± ± 2 nW c −Wc Of course.19) and then there are four different branches at each asymptotic point: q (5. the compatibility between the gravity localization requirement (5. it is worth noticing that the uniqueness of a superpotential W reaching a given asymptotic point follows from our requirements of a localized gravity without . (5. (5.22) of localized gravity.13) and the dynamics of the differential equation (5. it is important that ± no solution to the equation nW (2) c − Wc± = 0 can be found. Wc+ = q −Vc+ . we obtain a quadratic equation for W (2) c : ±2 ± ± 2W (2) c − 2Wc± W (2) c − V (2) c = 0.23) (5.15) and (5.14) can be fulfilled iff ± V (2) c > 0. W c =− 2 2 q p + 2V (2) c − Vc+ −Vc+ (2) + − . the roots of this equation would satisfy:   ± 2 1 V (2) c = 1− .22) and only one out of the four branches is retained: φc− : φc+ : q − Wc = − −Vc− .17) becomes linear in W (n) c and allows to compute ± W (n) c recursively in terms of lower derivatives:  ± P n−1 (k) ± (n−k+2) ± −W (n−k) ± +2(n−1)W (2) ± W (n−2) ± V (n) c − n−1 c c c c k=3 2(k−1)W c W (n) ±  .26) n n Vc± which is incompatible with the physical requirements (5. / Nuclear Physics B 584 (2000) 359–386 V (n)   n X  n−1 = 2 W (k) W (n−k+2) − W (n−k) . k−1 (5. However. for this expression to make sense for any integer n > 2. we ± will denote by W (n) c the values of W (n) at φ = φc± . q p − − 2V (2) c − Vc− −Vc (2) − + .24) ± At higher order.21) W c = 1 2 2 with 1 = ±2 = ±1. in addition. q p ± 2V (2) c − Vc± −Vc± (2) ± + 2 . (5.378 C.20) Wc± = 1 −Vc± . (5. However. by evaluating ± (5.18) The superpotential will be real-valued provided that: ± V (2) c > Vc± /2. At this stage.17) k=1 where V (n) denotes the nth order derivative of V and similarly for W (n) . Csáki et al. (5.17) at φ = φc± . the relation (5. At the second order. (5. W c = 2 2 (5.

then the jump equation becomes a differential equation that can be integrated on this interval and we obtain that f [φ] has to be proportional to W+ [φ] − W− [φ].15). a scalar field V in the bulk could be constructed such that the ± ± equation nW (2) c − Wc± = 0 has some solution in which case W (n) c would not be determined. And the jump conditions are given by: W+ [φ0 ] − W− [φ0 ] W+0 [φ0 ] − W−0 [φ0 ] = = T. T ∝ br f [φ0 ]T . Whether this is a solution to the cosmological constant problem or not. . but in that case there is only one possible value for the brane tension. if there exist a continuum interval of solutions for φ0 . Otherwise. the physical brane tension. Generically. (5. Finally. Indeed. n! (5. we can solve the equations of motion on the two sides of the brane. beside the fine-tuning requires on f [φ].16) and (5. We will explicitly describe an example in the next subsection. T . is not fine-tuned since the value of φ0 can vary continuously. (5. 11 In all the other cases. leading to a continuum of solutions. while nC is the number of critical points of V as defined above. the whole expression of W reaching the asymptotic points φc± can be uniquely reconstructed from its derivatives through a Taylor expansion: y <0: y >0: W− [φ] = W+ [φ] = ∞ X 1 (n) − W c (φ − φc− )n . we need to show that there is no loophole in the above argument due to the fact that we have used the superpotential 11 Whereas the non-canonically normalized brane tension. we usually find different values of the brane tension and the number of values of T that allows an infinitely large extra dimension with localized gravity is related to the number of critical points of the bulk potential V satisfying (5. n! (5. The upper bound has been semiquantitatively corrected by taking into account the average number. This issue deserves further analysis in future work. of the scalar field on the brane while the second one fine-tunes the value of the brane tension. It is worth noticing that in the case of an oscillatory bulk potential.28) n=0 ∞ X n=0 From the expression of the superpotential. we will obtain discrete solutions for φ0 and T . The multiplicity factor comes from the fact that the scalar field can asymptote either the same critical point or two adjacent ones on the two sides of the brane.29) The first equation will fix the value. is fine-tuned.22): nT 6 (3nC − 2)nS .C. To complete the proof of the no-go theorem presented above. using different critical asymptotic points. we will obtain only discrete values of φ0 . / Nuclear Physics B 584 (2000) 359–386 379 singularity. of solutions to the jump equation for φ0 . depends whether SM loops will modify T or Tbr .30) where nT stands for the number of values of T such that gravity is localized without a singularity. nS . Csáki et al. φ0 . It may also happen that some values of T are degenerate. Moreover. φc± . given two asymptotic points.27) 1 (n) + W c (φ − φc+ )n . f [φ0 ] f 0 [φ0 ] (5. we will obtain an infinite number of discrete values for the brane tension that is as quantized.

Csáki et al. However. we will only be concerned with static solutions and for our purposes any instabilities will not be relevant. as long as at φ∗ one is smoothly switching over to another branch of W . (2. The subtlety that one might worry about is that in the proof above we have implicitly assumed that φ is monotonic. we can construct four types of solutions that localizes gravity . all the critical points will actually happen at V 0 = 0. where W can be continued at both sides of φc . 5) is V (φ) = −φ 6 + 11φ 4 − 7φ 2 + 1. and therefore one can not circumvent the no-go theorem by gluing non-monotonic φ’s together. but instead one must define separate superpotential functions Wi for the regions between yi and yi+1 . / Nuclear Physics B 584 (2000) 359–386 formalism. by writing the second order equations (2. one can never switch off the ordinary branch. As in some of our previous illustrative examples the potential (5.380 C. However.31) is unbounded from below. and one can solve the equations in terms of the superpotential. 5.3). Thus the critical points at infinity must belong to an “ordinary branch” described above. and so the theory might be unstable to quantum fluctuations.32) W [φ] = φ 3 − φ reaches φc = ±1/ 3 on the negative (positive) branch. The reason is that in order to have localized gravity with an infinitely large extra dimension.5) that V 0 → 0. In particular. (5. the bulk potential (5. Since the only possibility for switching over to another branch is at W 0 = 0. (5. √ W [φ] = −φ 3 + φ reaches φc = ±1/ 3 on the positive (negative) branch. According to our no-go theorem.31) has two negative stationary points at √ φ = ±1/ 3. since otherwise φ is monotonic. For these superpotentials that are not globally defined it is then possible to have W 0 (φ∗ ) = 0 without satisfying V 0 (φ∗ ) = 0.33) Depending on which critical point we want to asymptote at infinity. This by itself however may not be a problem. and therefore we find from the second order equation (2. Next we show that such possibilities do not get around the nogo theorem presented above. The bulk potential in this example (Fig. there √ are isolated superpotentials solving the equations of motion with critical points at ±1/ 3. we need φ 0 → 0. it is impossible to continue the solution beyond φ∗ .4) in terms of first order equations involving W . and at those points V 0 = 0.31) which is generated by the superpotential W [φ] = φ 3 − φ. Of course this switch-over can only happen at a point where φ 0 = 0. However. if φ is not monotonic (that is if φ 0 = 0 at a finite value of y) one does not have a globally defined superpotential function W (φ). As emphasized in Fig. which is well-defined around φ∗ . 5.25) only a finite number of derivatives are non-vanishing thus the superpotentials are polynomial and take the form √ (5. once we are on an “ordinary branch” which can be globally defined. φ 00 → 0 for |y| → ∞.2. where φ 0 (yi ) = φ 0 (yi+1 ) = 0. Here these superpotentials are very simple because from (5. A numerical example In this section we demonstrate the ideas of the previous sections in a numerical example.

The physically interesting region. / Nuclear Physics B 584 (2000) 359–386 381 Fig. is emphasized.C.31). √ √ • For (φc+ = 1/ 3. φc− = 1/ 3 ): the solution is √ . near the two negative stationary points. Csáki et al. The scalar bulk potential (5. 5.

.

 1 φ(y) = √ tanh 3 .

y − yc± .

√ • For (φc+ = −1/ 3. three values for the brane tension are possible and they are just the opposite of the ones obtained in the first case. all other solutions will either decouple gravity on the brane or involve a horizon at a finite distance. 6. For instance in the case of an exponential coupling. Note that in agreement with the no-go theorem of the previous section there is a unique solution in this branch of solutions with stationary points at the positions of the minima of V [φ]. φc− = −1/ 3 ) or (φc+ = −1/ 3. we will generically obtain a finite number of discrete values for φ0 . there is no discontinuity in the √ superpotential and the brane tension has to vanish. The solutions are plotted in Fig. Besides the previous fine-tuned solutions with an infinitely large extra dimension and localized gravity. (5. for an exponential coupling.36) which admits three solutions whatever the value of b is. the values of φ0 will satisfy 3φ02 − 1 φ0 (φ02 − 1) = b. .√ • For (φc+ = 1/ 3. As a result of the uniqueness of . The continuity conditions imply that yc+ = −yc− and a+ = a− . φc− = −1/ 3 ): this case is analogous to the first one and. φc− = 1/ 3 ): in those cases. We would like now to examine numerically the self-tuning of these singular solutions. To do that we scan for solutions to (2.34) (5. Except in the degenerate case where f [φ] ∝ W+ [φ] − W− [φ] that has been discussed in footnote 11.35) where yc± and a± are four constants of integration to be determined by the continuity and jump conditions. A(y) = 18 9 (5. And thus there exist three values for the brane tension that √ √ will lead to a localized √ gravity. f [φ] = a ebφ . The jump equations will depend on the precise form of f [φ].6) by fixing the value of W [φ] at φ = 0. 3 √ √  2  1 tanh2 3 y − yc± + ln cosh 3 y − yc± + a± .

T can be rescaled by an amount given by the intersection of the set of solutions (2W. Also in agreement with the results of previous sections. Solutions for W [φ]. Then. f 0 T ).2 the boundary conditions can be expressed in terms of W [φ] and W 0 [φ] at the boundary. in this case φ∗ ' −0. Hence we have demonstrated the self-tuning of this model. 2W 0 [φ0 ]) can be equated with a certain (orbifold) boundary condition (f [φ0 ] T .382 C. Csáki et al.385. If f is O(1) at some matching point φ0 . / Nuclear Physics B 584 (2000) 359–386 Fig. That means that for a given (f [φ0 ]. As noted in Section 4. As is generically expected. given one such (f [φ0 ] T . the solitonic solution. 6. hence the theory is self-tuning. and T . we note that except for the solitonic solution. 2W 0 ) with a line through the origin and (f T . there is a fine-tuned region near W. Given any boundary condition parametrized by f [φ] and T . 7 that if f. the solitonic solution is the unique regular solution for φ(y) and A(y). However. there are no solutions (in this branch) with 0 < |W [0]| < W [φ∗ ] ' 0. Fig. W 0 [φ]) then no fine-tuning is necessary in order to have a static solution with arbitrary boundary conditions in that range. one can see from Fig. the caveat is that because of the isolated solitonic solution. fine-tuned.577.31. regular solution which approximates the Randall–Sundrum solution for large |y|. f 0 are O(1) = O(M5 ). then T can be rescaled by O(M5 ) without eliminating a solution. which in this case approximates the Randall–Sundrum solution for large |y|. f 0 [φ0 ] T ). In other words. all other solutions span an infinite range in φ. All other solutions are singular. 7 is a parametric plot of W [φ] versus W 0 [φ] for several solutions of W [φ]. It is intended to illustrate the fact that there is a space-filling region for W > 0. where φ∗ is the value of the field at the position of the local minimum of V [φ]. f 0 [φ0 ]) in a certain range of f 0 /f there is a range of tensions T for which there is a solution. In order to determine the range of boundary conditions at the brane which could be satisfied.385. φ(y) and A(y) in the bulk potential 5. Hence. if there is a space-filling range of solutions (W [φ]. a given (2W [φ0 ]. W 0 ∼ O(1). f 0 [φ0 ] T ) there is a continuous set of solutions around that point. There is a unique.

Einstein’s equations are not satisfied at the singularity but require a singular stress-energy tensor located at the singularity that may correspond to a brane: the introduction of this brane.17. is what we mean by resolution of the singularity.C. which is the case studied in [32]. Furthermore. 6. then once again self-tuning is natural. is quite non-generic. As pointed out in [13–17. [13–15] for a discussion on vanishing 4D effective vacuum energy.17. / Nuclear Physics B 584 (2000) 359–386 383 Fig. which specify the boundary conditions at the wall. This phenomenon is a result of the existence of solitonic solutions. This is generic in any region where the solutions satisfy W [φ]  V [φ]. are satisfied there. There is a shift symmetry in the scalar field which makes the boundary conditions more constraining. and is non-generic. is O(MEW /M5 )  1 then a fine tuning is reintroduced because of the uniqueness of the solitonic solution. 32] 13 the singularities contribute to the effective 4D energy density when integrated over the extra dimension. the case of zero bulk potential. which may be natural from a stringy perspective [7. Resolution of singularities and fine-tuning In this section we reconsider the resolution of singularities 12 as proposed in [16.8]. There is still a large region of parameter space where the theory is selftuning. Csáki et al. if we do not require orbifold boundary conditions. Parametric plot of several solutions for (W [φ]. we study the generic case. . there is a very large range for T for which there are solutions. 32]. W 0 include the line W = W 0 . As we have seen. and two of the boundary conditions turn out to have the same form. Note also that asymptotically the parametric plots of W vs. as well. 7. for instance. and propose a new resolution of the singularity which may restore self-tuning. This also implies that for any solution f ∼ f 0 . or boundary conditions. and extends the range of T over which self-tuning occurs possibly to ±∞ if f [φ] = f 0 [φ].32] is to add a brane at each of the singularities such that the equations of motion. W 0 [φ]). The idea of [16. but whether that region is natural or not requires exploration. 13 See also Refs. This behavior for large values of W is common. Hence. 12 As they stand in our solutions. and the contribution of the singularities is essential for vanishing of the 4D cosmological constant.

In [32] it is proposed that the space–time is either periodically continued or cut-off at the singularity. in that case a fine-tuning at each of the singularities is required. The singularities are generically horizons (cf. More precisely. Hence. 8. without orbifold boundary conditions there are 3 + 3 − 1 = 5 free parameters and 4 + 2 + 2 = 8 boundary conditions.384 C.3) because the warp factor e−2A(y)/(D−1) vanishes there. Although the details of any continuation of the space–time beyond the singularity will depend on quantum gravitational dynamics. following the counting in Section 2. The Penrose diagram for this scenario is illustrated in Fig. Hence. Csáki et al. a fine-tuning is required. but 2 + 4 = 6 boundary conditions. In order to answer the question of whether or not the theory is self-tuning we must better understand the continuation past the singularities. One might suspect that additional fields would add additional degrees of freedom which could help in self-tuning. the addition of a scalar field with second derivatives in its equation of motion would contribute two additional free parameters on each side Fig. then there are 3 + 3 − 1 = 5 free parameters from the bulk fields on both sides of the singularity. if there are orbifold boundary conditions. but only 3 − 1 = 2 free parameters. Section 3. we propose another scenario which does slightly better than the cut-off scenario. Although in the absence of a bulk potential the situation is modified because of the non-generic features mentioned above. If one imposes orbifold boundary conditions. / Nuclear Physics B 584 (2000) 359–386 However. there are generically two new boundary conditions at each singularity. Hence. then there are only half as many additional boundary conditions (2 + 2 = 4 boundary conditions in all). . and the system is overconstrained. In either case the system is overconstrained. Hence. addition of branes at the singularities adds new boundary conditions. although fine-tuning will still be required. but no additional free parameters. out of causal contact with our universe. 8. one fine-tuning is still required. However. Penrose diagram for causally disconnected regions glued together at the resolved singularity. In the case that the space–time is cut-off at the singularities on each side. as well. although the system is less constrained than the case in which the space– time is cut-off at the singularity. light does not cross the horizon and we can imagine a scenario in which there are bulk fields living beyond the singularity. the bulk fields across the horizon would provide an additional three free parameters which one might hope would help avoid fine-tuning. For example.

and have found that those case are more constrained than the generic case because of a shift symmetry in the scalar field in these models. Phys.C. We have reexamined the exactly solvable models. Rev. is supported in part by the US Department of energy under Contract DE-AC03-76SF00098 and in part by the National Science Foundation under grant PHY-95-14797. Phys. Sundrum.17. hep-ph/9905221. J. hep-ph/9803315. C. Lett. / Nuclear Physics B 584 (2000) 359–386 385 of a brane. Antoniadis.H. S. Martin Schmaltz and Anupam Singh for useful discussions. Dimopoulos. S. in agreement with a counting of free parameters in these models. hep-th/9906064. which could be used to restore self-tuning at the singularities or perhaps even produce nonsingular solutions. Robert Oppenheimer fellow at the Los Alamos National Laboratory. Finally. N. Randall. Lett. hepph/9804398. [2] I. and T.C. G.8]. Still. C. Whereas in [8] a fine-tuning was necessary in this case due to a particular ansatz for the scalar and graviton fields. Conclusions We have studied brane worlds coupled to a scalar field and have found that self-tuning is a generic feature of these models. Hence in a system of branes there are net. .G. C. B 429 (1998) 263.C. is a J. B 436 (1998) 257. and we presented a no-go theorem to this effect. we pointed out that the fine-tuning that is required in order to resolve the singularities in the spirit of [16. Dvali. at least naively. including the case of an exponential potential. Dvali. We provided perturbative and numerical techniques in order to calculate the self-tuned solutions. Pierre Binétruy. Tanmoy Bhattacharya. thanks the Theory Group at Berkeley for hospitality while this work was initiated. We thank Nima Arkani-Hamed. two of which were studied previously [7. which generically alleviates the need for fine-tuning of static solutions. [3] L. Arkani-Hamed. Dimopoulos. two additional free parameters from the extra scalar field. are supported by the US Department of energy under contract W-7405-ENG-36. Csáki et al. Acknowledgements We are extremely grateful to Chris Kolda for collaboration at early stages of this project. R. but only two boundary conditions (continuity and change in the derivative at the brane).. 7.C.E. Arkani-Hamed. a more detailed analysis is required in this case. Phys. Rev. We demonstrated that singularities in the self-tuned solutions are generic if gravity is to be localized. C. Lett. Phys. Lett. 83 (1999) 4690. In these models the dynamics of the scalar field provides additional degrees of freedom.32] for the case of zero bulk potential is generic. we showed that the more general solution is not fine-tuned. However. these theories are self-tuning. References [1] N. and we illustrated the major points of the paper via several numerical examples. G. 83 (1999) 3370.

hep-th/9411170.H. hep-th/0002161. Mod.M. B 341 (1994) 160. Phys. hep-th/0002160. Youm. M. J. hep-th/0001206. hep-th/0002164. Lett. Freedman. Cvetiˇc. R.T. Flournoy. Sfetsos. / Nuclear Physics B 584 (2000) 359–386 S. Cvetiˇc. hep-th/0001033. S.W. Dvali. Rubakov. Z. Csáki et al. Janssen. M. C. Hollowood. Rev. N. Irges. K. hep-ph/0003287. Lett. 71 (1993) 815. A. Sfetsos. Grojean. Nilles. C. Lett. V. Skenderis. S. E. Phys. M. Luty. Wetterich. C. O. hep-th/0002147. Gabadadze. Schulz. Phys B. R. B 468 (1999) 46.A. Phys. Phys. D. M. Gregory. JHEP 0002 (2000) 005. de Alwis. Shirman. Lond. Phys. JHEP 0004 (2000) 027. Townsend. Arkani-Hamed. Rev. A. hep-th/0003002. G. S. Roy. C. Lett. hep-th/0002130. B 479 (2000) 273. de Alwis. Bakas. to be published in Nucl. Erlich. Rev. hep-th/0004125. Lett. D 51 (1995) 5768. hep-th/9909070. hep-th/9402089. S. T. B 125 (1983) 139.386 [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] C. 84 (2000) 1090. hep-th/0002040. D 61 (2000) 101901. J. Kaloper. Gubser. Rubakov. Lett. Sibiryakov. Rev. . Weinberg. G. Lett.